Article Text

PDF

Society for Social Medicine annual meeting 2000
  1. M M GANNON
  1. MRC Institute of Hearing Research, Nottingham
    1. I GEMMELL
    1. MRC Social and Public Health Sciences Unit, Glasgow
      1. D A JONES,
      2. R R WEST
      1. University Of Wales College of Medicine, Cardiff
        1. B ALVES,
        2. A SHEIKH
        1. Department of Primary Health Care and General Practice, Imperial College School of Medicine
          1. C POPE
          1. Department of Social Medicine, University of Bristol
            1. C GOLDFRAD,
            2. K ROWAN
            1. Intensive Care National Audit and Research Centre, London
              1. J BARKER
              1. Office for Public Management, London
                1. A O'CATHAIN24-1,
                2. S J WALTERS24-1,
                3. J P NICHOLL24-1,
                4. K J THOMAS24-1,
                5. M KIRKHAM24-2
                1. 24-1 School of Health and Related Research, University of Sheffield, 24-2School of Nursing and Midwifery,University of Sheffield
                  1. K HOPAYIAN
                  1. School of Health Policy and Practice, University of East Anglia
                    1. J IBISON
                    1. Department of General Practice and Primary Care, St George's Hospital Medical School, London
                      1. S MARSHALL,
                      2. R HARDY,
                      3. D KUH
                      1. MRC National Survey of Health and Development, Department of Epidemiology and Public Health, Royal Free and University College London Medical School
                        1. S T BROOKES41-1,
                        2. T J PETERS41-1,
                        3. E WHITLEY41-1,
                        4. G DAVEY SMITH41-1,
                        5. M EGGER41-1
                        1. 41-1Department of Social Medicine, University of Bristol
                          1. W ANDERSON,
                          2. D FLORIN
                          1. Primary Care Programme, The King's Fund, London
                            1. R W MORRIS,
                            2. P H WHINCUP,
                            3. F C LAMPE,
                            4. M WALKER,
                            5. G WANNAMETHEE,
                            6. A G SHAPER
                            1. Department of Primary Care and Population Sciences, Royal Free and University College Medical School, London
                              1. A J LEE,
                              2. C J EVANS,
                              3. C M HAU,
                              4. M M BREE,
                              5. F G R FOWKES
                              1. Wolfson Unit, Department of Community Health Sciences, University of Edinburgh
                                1. J MUNRO,
                                2. J NICHOLL,
                                3. A O'CATHAIN,
                                4. E KNOWLES
                                1. Medical Care Research Unit, University of Sheffield
                                  1. A MOON,
                                  2. F THOMPSON,
                                  3. V LATTIMER,
                                  4. S GEORGE
                                  1. University of Southampton Health Care Research Unit
                                    1. S ABBOTT,
                                    2. S GILLAM
                                    1. The King's Fund, London
                                      1. J MINDELL
                                      1. Department of Epidemiology and Public Health, Imperial College School of Medicine, London
                                        1. P W ARMSTRONG
                                        1. Department of Health Sciences, University of East London
                                          1. S KENNEDY,
                                          2. N PAYNE,
                                          3. C SAUL,
                                          4. D LUFF,
                                          5. K MCKEE
                                          1. School of Health and Related Research, University of Sheffield
                                            1. D AUSTIN,
                                            2. E RUSSELL
                                            1. Department of Public Health, University of Aberdeen
                                              1. A MCPHERSON,
                                              2. S ZIEBLAND,
                                              3. A HERXHEIMER,
                                              4. R MILLER,
                                              5. S SHEPPERD
                                              1. DIPEx project, Department of Primary Health Care, University of Oxford, Institute of Health Sciences
                                                1. R W JAKES,
                                                2. K-T KHAW,
                                                3. N J WAREHAM
                                                1. Department of Public Health and Primary Care, University of Cambridge
                                                  1. C SANDERS,
                                                  2. J DONOVAN,
                                                  3. J CHARD,
                                                  4. P DIEPPE
                                                  1. Department Social Medicine/MRC HSRC, University of Bristol
                                                    1. J CRITCHLEY,
                                                    2. J REEVE,
                                                    3. S CAPEWELL
                                                    1. Department of Public Health, The University of Liverpool
                                                      1. R C WILSON,
                                                      2. P J SAUNDERS
                                                      1. Department of Public Health and Epidemiology, University of Birmingham
                                                        1. A C WILLIAMS,
                                                        2. J R SANDY
                                                        1. Division of Child Dental Health, University of Bristol
                                                          1. D M MCNAUGHTON,
                                                          2. G LIBBY,
                                                          3. P T DONNAN,
                                                          4. T M MACDONALD
                                                          1. Medicines Monitoring Unit, Department of Clinical Pharmacology, University of Dundee
                                                            1. A MCLEOD
                                                            1. MRC Social and Public Health Sciences Unit, University of Glasgow
                                                              1. P MCLOONE
                                                              1. MRC Social and Public Health Sciences Unit, University of Glasgow
                                                                1. S ZIEBLAND,
                                                                2. J JAY,
                                                                3. J ROBERTSON,
                                                                4. A NEIL,
                                                                5. G COWBURN
                                                                1. Division of Public Health and Primary Health Care, University of Oxford, Institute of Health Sciences
                                                                  1. A D COOK,
                                                                  2. A SHEIKH
                                                                  1. Department of Primary Health Care and General Practice, Imperial College School of Medicine, London
                                                                    1. H T O DAVIES,
                                                                    2. A V SHIELDS
                                                                    1. Department of Management, University of St Andrews, Fife

                                                                      Statistics from Altmetric.com

                                                                      abstracts of oral presentations

                                                                      An expert at your fingertips. Development of a computerised clinical guidance programme to help with the prophylactic oophorectomy decision

                                                                      Objective

                                                                      To develop a computerised Clinical Guidance Programme (CGP) to assist women and clinicians in the prophylactic oophorectomy (PO) decision and to undertake a preliminary evaluation. PO is undertaken in about 50% of abdominal hysterectomies as a prophylactic measure against ovarian cancer, but it can have a number of longer term health consequences that may be difficult to take into account during the clinical decision.

                                                                      Methods

                                                                      We have developed a CGP, based on a series of Markov models for each of the main health outcomes involved in the PO decision—ovarian cancer, breast cancer, coronary heart disease and major fracture due to osteoporosis. The results, using data drawn from a systematic literature review, are adjusted to reflect womens' own individual risk factors, health state valuations, attitudes to risk and predictions as to their likely intake of HRT. In conjunction with a facilitator, women record their major risk factors including smoking, hypertension and family history of breast and ovarian cancer. They complete the EuroQol 5D, and undertake a number of health state valuations using a time trade off method. Women may opt out of any of these tasks, in which case the CGP is run using average values from population surveys. Full access to the underlying data, calculations and sources of evidence is provided.

                                                                      Results

                                                                      Preliminary evaluation shows that the CGP administered by a researcher gives meaningful results and takes about 30 minutes to complete. Initial consultant reaction is favourable but almost all clinicians approached have been reluctant to use the CGP themselves. Using averaged values with “dummy patients” the CGP produces patient specific guidance in line with current published guidelines and current clinical practice. The CGP took approximately two person years to develop, but the model structure is generic and may be modified easily to numerous decisions.

                                                                      Conclusions

                                                                      This work indicates that development of a computerised CGP is possible. Consultants are wary of this new technology even though it is evidence-based and likely to offer the “best” decision by various criteria. Given the number of variables that should be considered in the decision and a desire for explicit and informed decision making, 30 minutes is not long—but ways of integrating a CGP into routine practice will need to be found.

                                                                      Benefits of adjuvant adenoidectomy in persistent OME (glue ear)

                                                                      Background

                                                                      From the mid-1980s, paediatric ENT operations encountered increasing scepticism in public health quarters. The 1992 Effective Health Care Bulletin, coinciding with fundholding, substantially reduced referral and intervention rates, and also compressed the district level practice variation. Variation in rates for adjuvant adenoidectomy remains high, suggesting uncertain clinical criteria and variable pressures of facilities and workload. TARGET (The MRC randomised Trial of Alternative Regimens in Glue Ear treatment) is a three arm multicentre trial not led by service providers. It was undertaken to address overall intervention policy by measuring broad family centred outcomes not used in previous trials. It has already reported benefits from combined surgery versus medical management controls over one year. This paper reports outline adenoidectomy results over two years.

                                                                      Methods

                                                                      Scores with norms were developed from parental questionnaires to a large sample of unaffected children and to trial children at +3, +6 and +12 months. These cover symptom areas (ear problems; respiratory infection; hearing difficulties), two behaviour factors (aggression and non-aggression problems), general health and parent quality of life. All children were aged 3.5–7.0 years and had met stringent criteria (chiefly 20dB hearing loss on two occasions, three months apart). In other first world healthcare systems, most would have been treated much earlier. Of 251 grommeted children, 127 were randomised to simultaneous adenoidectomy. The two year data include follow up also at +18 and +24 months.

                                                                      Results

                                                                      Highly significant additional gains were obtained for the two year average in weight, dB hearing level, parent reported hearing difficulties, respiratory symptoms, parent quality of life, and on the first principal component summarising all outcomes. Additional benefit in aggressive behaviour, and in general health was marginal, but was null for non-aggressive behaviours, ear symptoms and balance. Several of the adjuvant adenoidectomy effect sizes were about 0.5 standard deviation; limited effect modification suggests little scope for indicators of groups that could be targeted receiving greater benefit. Re-insertion of grommets was performed according to strict criteria reflecting hearing level, and its frequency within two years was significantly lower in those with adenoidectomy (15% versus 34%).

                                                                      Conclusions

                                                                      Overall policy conclusions await the analysis of the economic data being gathered. However, given some complementarity of effects from the two operations over time, these results rehabilitate grommet insertion plus adenoidectomy as a serious candidate for a cost effective treatment policy.

                                                                      Excess winter deaths, socioeconomic deprivation and housing conditions

                                                                      Background

                                                                      Seasonal patterns in mortality in the UK are known to be greater than in most other European countries. Research in the UK has failed to demonstrate a consistent association between excess winter deaths and area based measures of deprivation.

                                                                      Objective

                                                                      To explore the association between excess winter mortality (defined as the number of deaths in the three months December to February divided by the number of deaths in the rest of the year), socioeconomic deprivation and housing conditions.

                                                                      Methods

                                                                      Binomial regression analysis of postcode sector based excess winter mortality, deprivation and census housing variables incorporating information from the 1996 Scottish House Conditions Survey.

                                                                      Results

                                                                      Between 1988 and 1997 in Scotland, all cause excess winter mortality in people aged 65 and over was 41%. Excess winter mortality in this age group was positively associated with lack of home central heating, socioeconomic deprivation and fuel poverty (predicted total annual running cost/amount spent on fuel). Multiple regression analysis identified lack of home central heating (p=0.012) and socioeconomic deprivation (p=0.015) as the main predictors of all cause excess winter mortality. When cause specific mortality was examined, winter deaths from ischaemic heart disease (ICD9 410–414) were positively associated with fuel poverty (p=0.014). Excess winter deaths from respiratory disease (ICD9 460–519) were positively associated with lack of home central heating (p=0.036) and dampness (p=0.029). Excess winter mortality was 54% for deaths from respiratory disease and 42% for deaths from ischaemic heart disease. Deaths from all other causes demonstrated an excess winter mortality of 38%. Postcode sector based regression analysis showed no association between excess winter mortality from these other causes and socioeconomic deprivation or housing.

                                                                      Conclusion

                                                                      Housing condition, particularly availability of central heating and levels of fuel poverty are more closely associated with excess winter deaths than traditional measures of deprivation. These findings suggest that excess winter deaths in Scotland could be substantially reduced if adequate affordable heating was more widely available.

                                                                      The provision of information on the management and treatment of minor illness. Does it affect use of services or feelings of confidence in dealing with symptoms of minor illness?

                                                                      Background

                                                                      Patients are bombarded with information about the management of minor illness from government and professional organisations. It is assumed that the provision of this information will result in less demand for primary care services, and will enable people to engage in self care practices with more confidence.

                                                                      Study aims

                                                                      To investigate whether the provision of a patient information booklet influences the number and type of consultations sought with the primary care team. To investigate whether the provision of a patient information booklet influences patients' feelings of confidence in dealing with symptoms and signs of minor illness.

                                                                      Design

                                                                      Randomised controlled trial of two information booklets on the management of minor illness.

                                                                      Main outcome measures

                                                                      Use of services in 12 months following receipt of booklets and self reports of confidence in dealing with symptoms of minor illness.

                                                                      Setting

                                                                      20 Lothian general practices.

                                                                      Subjects

                                                                      Two random samples from participating practices—one of population from Community Health Index and one of users contacting out of hours services in the previous year (total n=9408).

                                                                      Intervention

                                                                      Participants were randomised to three groups. One group (n=3098) received a copy of “What Should I Do?”. The second group (n=2781) received “Healthcare Manual”. The control group (n=2712) were invited to participate in the study but received no booklet.

                                                                      Methods

                                                                      All groups were posted an invitation to participate in the study, a booklet if appropriate, and the opportunity to opt out of the study. At eight weeks participants were mailed a questionnaire ascertaining health status, use of services, confidence in dealing with minor illness and views of booklet. At one year, the number of health service contacts in the year before and year after the intervention was collected by a team of nurses examining the medical record of participants.

                                                                      Main results

                                                                      Eighty seven per cent of the sample did not opt out of the study, 53% responded to the questionnaire. Receipt of booklets had no significant effect on health service utilisation, although matched practices allocated to “Healthcare Manual” had relatively reduced consultation rates in comparison to matched practices allocated to “What Should I Do?”. The booklets did not impact on patients' confidence in dealing with minor ailments measured by their response to symptoms. Both booklets were well received and valued by patients.

                                                                      Conclusions

                                                                      These findings suggest that widespread distribution of booklets containing information on the management of minor illness in order to affect consulting behaviour is not advisable.

                                                                      All's fair in love and cardiology? Sex differences in risk factors, treatment and survival after acute myocardial infarction—a prospective observational study

                                                                      Introduction

                                                                      Coronary heart disease is the major cause of death of postmenopausal women in industrialised countries. Although acute myocardial infarction (AMI) affects men in greater numbers, the short-term outcomes for women are worse. In the longer term, studies suggest that mortality risk for women is lower or similar to that of men. However, length of follow up and adjustment for confounding factors have varied and more importantly, the association between treatment and outcomes has not been examined.

                                                                      Study objective

                                                                      To investigate the effect of sex differences in risk factors and hospital treatment on survival after AMI.

                                                                      Design

                                                                      A prospective observational study collecting demographic and clinical data on cases of AMI admitted to hospitals in Yorkshire. The main outcome measures were survival status at discharge from hospital and two years later.

                                                                      Setting

                                                                      All district and university hospitals accepting emergency admissions in the former Yorkshire National Health Service (NHS) region of northern England.

                                                                      Participants

                                                                      3684 consecutive patients with a possible diagnosis of AMI admitted to hospitals in Yorkshire between 1 September and 30 November 1995.

                                                                      Main results

                                                                      AMI was confirmed by the attending consultant for 2196 admissions (2153 people, 850 women and 1303 men). Women were older and less likely than men to be smokers or have a history of ischaemic heart disease. Crude in hospital fatality rate was higher for women (30% versus 19% for men, crude odds ratio (OR) of death before discharge for women 1.8, 95% confidence intervals (95% CI) 1.5, 2.2). This difference persisted after adjustment for age, risk factors and comorbidities (adjusted OR 1.3, 95% CI 1.0, 1.6, p=0.02), but was not significant when treatment was taken into account. Women were less likely to be given thrombolysis (37% versus 46%, p<0.01) and aspirin (83% versus 90%, p<0.01), discharged on β blockers (33% versus 47%, p<0.01) and aspirin (82% versus 88%, p<0.01) or be scheduled for angiography, exercise testing or revascularisation. Adjustment for age removed much of the disparity in treatment. Crude fatality rate at two years was higher for women (OR 1.8, 95% CI 1.4, 2.3). Age, existing risk factors and acute treatment accounted for most of this difference, with treatment on discharge having little additional influence.

                                                                      Conclusions

                                                                      Patients admitted to hospital with AMI should be offered optimal treatment irrespective of age or sex. Women have a worse prognosis after AMI and under-treatment of older people with aspirin and thrombolysis may be contributing to this.

                                                                      Job satisfaction, self reported stress, cardiovascular risk factors and mortality

                                                                      Objectives

                                                                      To explore the association between job satisfaction, self perceived stress, cardiovascular risk factors and mortality.

                                                                      Design

                                                                      Prospective observational study.

                                                                      Setting

                                                                      27 workplaces in the West of Scotland.

                                                                      Participants

                                                                      1006 women and 6022 men in paid work.

                                                                      Outcomes

                                                                      Cardiovascular disease (CVD) and all cause mortality in a 21 year follow up period, according to job satisfaction at baseline, at second screening and comparison of job satisfaction between baseline and second screening.

                                                                      Results

                                                                      There was a strongly significant age adjusted association between job satisfaction and stress at baseline, at second screening and between the two periods The overall difference between occupational class groupings in the reporting of job satisfaction, cross sectionally and longitudinally, was significant for women only. After adjustment for age and occupational class, little job satisfaction was associated with high body mass index in men at baseline, and with a greater number of cigarettes smoked in men at second screening. Job satisfaction was not significantly associated with any of the CVD risk factors examined in women at baseline or second screening. For CVD mortality in a 21 year period, there was no significant difference in the reporting of job satisfaction at baseline, at second screening or between baseline and second screening for men or women. For all cause mortality in a 21 year period, there was no significant difference in the reporting of job satisfaction at baseline, at second screening or between baseline and second screening for men, but women who reported an increased level of job satisfaction between baseline and second screening were significantly less likely to die from any cause than women who were always satisfied with their jobs, after adjusting for age and occupational class (relative risk 0.44, 95% confidence intervals 0.20, 0.96).

                                                                      Conclusions

                                                                      While we found a strongly significant association between job satisfaction and perceived stress, there was little evidence to suggest that lack of job satisfaction was associated with cardiovascular risk factors in men or women, and no evidence to suggest that lack of job satisfaction was associated with cardiovascular or all cause mortality in men, once adjustment had been made for age and social class. There was some suggestion that women who reported increased satisfaction in their jobs between baseline and second screening were at less risk of dying from any cause than other women. We suggest that job (dis)satisfaction is an important form of stress for those in current employment and that there is a case for separating the experiences of men and women in future studies.

                                                                      Survival and outcome in a population-based cohort of pre-end stage chronic renal failure and their sociodemographic determinants

                                                                      Background and objectives

                                                                      Little is known of the outcomes of chronic renal failure (CRF) at a population-based level. There have been clinic-based studies of survival in advanced CRF, but survival in less severe CRF is not well understood. The objective of this study was to determine the outcomes and processes of health care in a population-based cohort of CRF.

                                                                      Design

                                                                      Newly diagnosed cases of CRF, in Southampton and SW Hampshire Health Authority (1992–1994), were determined by a persistently raised serum creatinine (SCr ⩾150 μmol/l for ⩾ six months; a sensitive and routinely used marker of renal failure) identified from chemical pathology records. Validation of diagnoses was by medical note search. A retrospective cohort was followed up for a mean of 5.5 years. Survival, including standardised mortality ratios (SMRs) and cause of death were determined by linkage to national mortality statistics and death certificates (99% complete). Patterns of referral and treatment were established by record linkage to routine data sets and a medical note search in a sample of 376 (35%).

                                                                      Results

                                                                      1076 cases of CRF were identified. Five year survival within the cohort was 34% with no gender difference. However, in comparison with the general population mortality was influenced by gender: SMRs for men and women were 2.10 and 2.85 respectively. Age was also a major determinant of mortality: SMRs for the age groups <50, 50–64, 65+ were 21.23, 8.62 and 2.02 respectively. Vascular causes comprised 46% of deaths and renal disease was only mentioned on 17% of death certificates. Referral to a nephrologist was low (24%) and significantly influenced by age, comorbidity and severity of disease. Referred cases had more extensive investigations and therapy. Eleven per cent of cases showed progression with a doubling of SCr. 27% (n=33) of those with an SCr ⩾300 μmol/l were accepted for renal replacement therapy, lower age was significantly related to acceptance.

                                                                      Discussion

                                                                      Survival in CRF is poor (about 50% of deaths were due to vascular causes). Only a minority were referred to a local nephrologist. The Renal Association recommends that all cases with an SCr ⩾150 μmol/l should be seen by a nephrologist, so there is scope for increased referral, but the cost effectiveness and the implications of an increased workload have not been assessed. There is a strong association between vascular disease and CRF, highlighting the importance of interventions to reduce vascular disease, which may be as important as those aimed at CRF.

                                                                      Appropriate coronary revascularisation: clinical outcomes in patients who do not receive it compared with those who do: the ACRE study

                                                                      Background

                                                                      The quantified judgements of expert panels might be a better guide to clinical practice than the pragmatic decisions of individual clinicians, yet there have been no prospective clinical outcome studies. We sought to determine whether patients rated appropriate for coronary revascularisation and who did not receive it had worse outcomes than those who did.

                                                                      Methods

                                                                      Prospective, population-based study of all patients undergoing coronary angiography (no exclusion criteria) during 1996/7 at three London, UK hospitals. Clinical indications for percutaneous transluminal coronary angioplasty (PTCA) and coronary artery bypass grafting (CABG) were rated as appropriate, uncertain or inappropriate by a nine member expert panel prior to patient recruitment; patients were managed without using these ratings. Altogether 4020 patients were followed up (median 30 months) for subsequent revascularisation, angina and functional status, non-fatal myocardial infarction and mortality (n=239 deaths).

                                                                      Results

                                                                      All effects were adjusted for age, number of diseased vessels and other prognostic factors. At the time of angiography 582 patients were rated appropriate for PTCA; the 385 patients (66%) who did not undergo PTCA had worse follow up angina status than those who did (odds ratio 1.56, 95% confidence intervals (95% CI) 1.1, 2.3). Some 943 patients were rated appropriate for CABG; the 419 (44%) who did not undergo CABG had worse angina status (odds ratio 1.88 (95% CI 1.4, 2.6)) and mortality (hazard ratio 3.34 (95% CI 2.3, 4.9)) than those who did.

                                                                      Conclusion

                                                                      Among patients judged appropriate for coronary revascularisation, those who did not receive it had worse clinical outcomes than those who did. Randomised trials are required to confirm whether decisions guided by expert panels are indeed superior to those of individual clinicians.

                                                                      To be anticoagulated or be given antiplatelet treatment for non-rheumatic atrial fibrillation: this is the question

                                                                      Background

                                                                      In the past decade, there has been widespread implementation of oral anticoagulation in favour of antiplatelet treatment in patients with non-rheumatic atrial fibrillation (NRAF) in the reduction of risk from stroke. This is based upon evidence from randomised controlled trials (RCT) comparing long term anticoagulation versus placebo and antiplatelet treatment versus placebo. However, these “indirect” comparisons of the effects of the two treatment options may (1) be biased by different selection criteria used in trials leading to differences in prognosis unrelated to treatment and (2) over-estimate the effects found. It is, therefore, necessary to undertake a direct, “head to head” comparison so that unbiased estimates of which is the better treatment option can be determined.

                                                                      Methods

                                                                      We conducted a systematic review of RCT comparing long term anticoagulation with antiplatelet treatment using the Cochrane library, Medline, Cinhal and Sigle for grey literature from 1966 to December 1999. Odds ratios (OR) and 95% confidence intervals (95% CI) were calculated to estimate treatment effects.

                                                                      Results

                                                                      Five RCT published between 1980–99 were identified. Using a fixed effects model on the pooled data, there were no differences between the two treatment options in stroke or cardiovascular death (stroke OR 0.91, 95% CI 0.47, 1.74; vascular OR 0.84, 95% CI 0.62, 1.14). There was a significant difference in non-fatal stroke in favour of anticoagulation (OR 0.68, 95% CI 0.47, 0.99). However, this difference was not seen when AFASAK 1 (in which there was selection bias, premature cessation of the study and unblinded observers) was excluded (OR 0.75, 95% CI 0.5, 1.13). Using a random effects model, there was also no difference in the combined fatal and non-fatal events (OR 0.74, 95% CI 0.53, 1.02). Major bleeding events among patients on anticoagulation tended to be higher than on antiplatelet treatment (OR 1.45, 95% CI 0.93, 2.27).

                                                                      Conclusion

                                                                      The heterogeneity between the trials and the limited data result in considerable uncertainty about the value of anticoagulation compared with antiplatelet treatment. The risks of bleeding and the higher cost of anticoagulation make it an even less convincing treatment option. The trials were small in number, two were stopped prematurely and the one that demonstrated a difference in effect was methodologically weak in design, skewing the true treatment effects. Further large scale RCT are needed to establish the value of long term anticoagulation in patients with NRAF.

                                                                      Reducing absolute risk of cardiovascular disease: randomised controlled trial of training interventions to improve information systems or knowledge of evidence-based medicine in primary care

                                                                      Objectives

                                                                      We sought to determine interventions designed with active participation from the practices to improve management of cardiovascular disease (CVD) patient information and improve the access to and interpretation of evidence of effectiveness for secondary CVD prevention, resulted in an improvement of risk management in patients with established CVD.

                                                                      Design

                                                                      A factorial cluster randomised controlled trial.

                                                                      Participants

                                                                      Participating practices: primary care teams from 17 general practices in West London. Participating patients: all patients with CVD, including angina, myocardial infarction, stroke, transient ischaemic attack or peripheral vascular disease patients and those who had undergone coronary artery bypass grafting, percutaneous transluminal coronary angioplasty or vascular surgery.

                                                                      Interventions

                                                                      Primary care teams were randomly allocated to complex training interventions to improve: (a) the management of patient information for CVD risk assessment; (b) access to and understanding of the evidence of effectiveness of secondary CVD prevention; (c) both; or (d) neither.

                                                                      Outcome measures

                                                                      Case note review was used to determine the impact of these interventions, between baseline (1998) and follow up (1999) on three outcomes: (i) complete recording of the six clinical risk factors necessary to calculate absolute five year risk of a further CVD event (age, sex, smoking status, blood pressure, cholesterol and diabetic status); (ii) appropriate treatment for modification of those risk factors—prescribing aspirin, control of cholesterol and blood pressure; and ultimately (iii) the modification of absolute risk.

                                                                      Results

                                                                      1261 CVD patients were identified at baseline, of which 959 (76.1%) were available for follow up. Completeness of recording the six clinical risk factors necessary to calculate absolute risk increased from 36.9% at baseline to 55.4% at follow up (p=0.02). Improvement from baseline was greatest in the group (c)—information management and access to evidence combined—from 34.0% to 59.8% (p=0.002) and group (b)—evidence only—from 41.4% to 58.3% (p=0.04). Improvements for group (a)—information management—of 14.5% and group (d)—control group—of 14.8% were not significantly greater than baseline.

                                                                      Conclusions

                                                                      In the first RCT of interventions designed by active participation from primary care teams to improve management of absolute risk in patients with CVD the results show that the combination of training in information management and improved access to and interpretation of evidence of effectiveness resulted in the largest increase in complete recording of absolute risk assessment.

                                                                      Gender differences in health and lifestyle after acute myocardial infarction: relevance for rehabilitation

                                                                      Background

                                                                      Studies of rehabilitation after acute myocardial infarction (MI) provide little information on gender differences. Most previous studies have focused on men and particularly men of working age, although a significant and increasing proportion of MI patients are women and half of all are aged 65 or more.

                                                                      Patients and methods

                                                                      The British Heart Foundation multicentre trial of rehabilitation after MI (n=2287) included 621 women. Data were collected on admission from clinical records, after discharge by structured interview using standard measures in patients own homes, at six months by repeated interview using the same standard measures and at 12 months clinical examination in outpatients.

                                                                      Results

                                                                      Comparisons showed many highly significant gender differences in health status, “natural history” of rehabilitation and lifestyle changes after MI. At discharge clinically significant anxiety was identified in 42% of women compared with 26% of men (p<0.0001), depression in 28% versus 16% (p<0.0001) and severe disability in 38% versus 15% (p<0.0001). At six month follow up these differences persisted and women experienced significantly more comorbidity (4.0 versus 3.5 prescribed medications per day p<0.001), fewer had changed their diet 55% versus 64% (p<0.01), returned to work (among those previously in paid employment) 31% versus 52% (p<0.001) or undertook any physical exercise 56% versus 83% (p<0.0001).

                                                                      Conclusions

                                                                      This study shows major differences between women and men in a number of standard measures both early after MI and six months later. These findings have implications for cardiac rehabilitation, which currently focuses on gym-based exercise. Programmes should be designed with sufficient flexibility to respond to different patient needs.

                                                                      Angina: the self care/primary care interface

                                                                      Objectives

                                                                      To describe and explain social class and gender variations in perceptions and primary care presentation of angina.

                                                                      Design

                                                                      Here, we present the qualitative stage of a three stage study that included a survey, using the Rose angina questionnaire, to identify people with chest pain (n=5033); a general practice casenote review of those thus identified (n=650) and a qualitative interview study (n=60).

                                                                      Setting

                                                                      Two socially contrasting areas of Glasgow.

                                                                      Participants

                                                                      Thirty men and thirty women, aged 45–64 with Rose angina.

                                                                      Results

                                                                      The casenote review showed (i) people from the more deprived area were more likely to present with chest pain than those from the affluent area and (ii) men were more likely to present than women. The qualitative study demonstrated that respondents' decisions of whether to present were influenced by their perceived vulnerability to heart disease and their perceived options for action. To assess their risk, participants took into account (i) their family history of heart disease, (ii) their cardiac risk behaviours and (iii) their perceived fit with a cardiac stereotype. They also considered competing explanations, such as other diseases or aging. Options for action included self care, lay consultation and primary care. The decision about action depended on respondents' capacity for self care, opportunities for lay consultation and the perceived quality of previous encounters with health service professionals. Compared with those from the affluent area, those from the socially deprived area generally felt more vulnerable to heart disease, yet were often reluctant to seek medical help for chest pain because of low expectations of their health and of health services. Women felt less vulnerable to heart disease than men and in many cases delayed seeking medical help despite having typical angina. This tendency to delay was magnified by women's greater propensity for self care and greater use of lay consultations.

                                                                      Conclusion

                                                                      Inequities in access to cardiology services have been described by others and it is increasingly recognised that research at the self care/primary care interface is important to understanding such inequities. This study outlines the factors that influence the progression from self care to primary care of people with angina. By understanding these processes, it should be possible, through patient education, to improve the appropriateness of primary care presentation for angina and to reduce inequities in access to cardiology services.

                                                                      Pattern of associations between social class, education and Townsend Score and coronary heart disease, diabetes/impaired glucose intolerance and risk factors: cross sectional comparison of South Asians with Europeans

                                                                      Background

                                                                      Studying health inequalities within ethnic groups in Britain is difficult because there are no widely applicable measures of socioeconomic position. Existing studies have given conflicting results. This paper applies three socioeconomic indicators widely used in European origin populations to three South Asian groups and to a wider range of health measures than studied hitherto.

                                                                      Objective

                                                                      To compare the pattern of associations between social class, education and Townsend Deprivation Score and coronary heart disease, glucose intolerance (impaired glucose intolerance and diabetes) and related risk factors in South Asians (Indians, Pakistanis, Bangladeshis) and Europeans, to test the hypothesis that associations would show worse risk profile or health in the socioeconomically disadvantaged in all ethnic groups (the predicted direction).

                                                                      Setting and population

                                                                      South Asians (n=684) comprising Indians (n=259), Pakistanis (n=305) and Bangladeshis (n=120), and Europeans (n=825), aged 25–74 years in Newcastle upon Tyne.

                                                                      Design, methods and main outcome measure

                                                                      Secondary analysis of cross sectional data from a questionnaire and clinical screening study. Eighty four associations were examined in each ethnic group relating to diseases (12 associations), lifestyles (18 associations), physical measures (24 associations), blood pressure (12 associations) and biochemistry (18 associations). Direction of association was the key outcome.

                                                                      Results

                                                                      Socioeconomic indicators were mostly associated with health measures as predicted in Europeans (71 of 84 (85%) associations) and less so in South Asians combined (56 of 84 (67%) associations). In Indians, only Townsend Deprivation Score was consistently associated as predicted (23 of 28 (82%)). In Pakistanis and Bangladeshis associations with all socioeconomic measures were inconsistent. There were some differences between Indians (52 of 84, 62% of associations as predicted), Pakistanis (39 of 84, 46%) and Bangladeshis (40 of 84, 48%). In South Asian men and women associations were as predicted with most anthropometric (18 of 24, 75%), biochemical (15 of 18, 83%), and lifestyle (13 of 18, 72%) measures, but those with blood pressure (4 of 12, 33%) and disease (6 of 12, 50%) were not. The pattern in Bangladeshis was often opposite to that predicted, even for physical measures (11 of 24, 46%) and biochemistry (44%).

                                                                      Conclusions

                                                                      Associations were mostly as predicted in Europeans, but were inconsistent in the South Asian groups. Associations were more consistently as predicted in Indians than Bangladeshis. South Asians' pattern of health inequalities differs from Europeans. Other explanations for the findings include artefacts from small sample size, differences in acculturation in Indians, Pakistanis and Bangladeshis, and that the chosen indicators were inappropriate for South Asian populations. Studies of inequalities in health should examine Indians, Pakistanis and Bangladeshis separately.

                                                                      Deprivation and death in Scotland: constructing a graphical cohort survival model

                                                                      Objective

                                                                      To present disease specific population mortality data in a clear format.

                                                                      Background

                                                                      Tables summarising the effects of deprivation on age specific mortality can be difficult to explain to lay audiences and non-specialists. Worse still, their initial confidence can be shattered with the arrival of the concept of competing mortality. One of the conceptually clearest forms of population mortality data is the survival of a birth cohort. We set out to construct cohorts of the Scottish population to examine and illustrate the effects of deprivation on disease specific and overall mortality.

                                                                      Setting

                                                                      Scottish population of 5.1 million.

                                                                      Methods

                                                                      For cohorts of older people, there will inevitably have been a number of deaths in early childhood and young adulthood (particularly war service). But from the point of view of chronic diseases, deaths are relatively rare until people reach their mid-50s. We therefore identified the population born in 1920, aged 55 in 1975 and divided them into quintiles of Carstairs deprivation scores. For each subsequent year we identified deaths by specific cause, and deprivation quintile. A disease specific survival chart would then be built up for each of the deprivation quintiles. The exercise was repeated for cohorts born in 1930 and 1940.

                                                                      Results

                                                                      The charts clearly demonstrated that there are no specific diseases of poverty. The smoking related conditions of lung cancer and heart disease are only slightly more common. Instead the dominant effect was clearly a general increase in risk of all the major diseases with age. Furthermore, people living in poorer circumstances had broadly the same mortality experience as affluent people of five to six years older. This has been described by Watt as the “more miles on the clock” phenomenon. It was possible to discern improvement in survival between the 1920 birth cohort and the 1930 cohort, however, the socioeconomic differentials persist.

                                                                      Conclusions

                                                                      This technique potentially offers a graphical method for illustrating the effects of factors such as deprivation on disease specific survival model, while accommodating the problem of competing mortality.

                                                                      Deprivation and poor health in rural areas: inequalities hidden by averages

                                                                      In the UK, poor health is strongly associated with social deprivation and financial resources for health services are allocated to areas in proportion to their aggregate levels of social deprivation. This seems to work reasonably well in urban areas, but previous research has shown that the link between population measures of ill health and poverty is weaker in rural settings. This research considered 570 census wards in East Anglia, UK, and found that poor health and social deprivation scores were much less associated in rural than in urban areas. Why is this? Two possible reasons are that the measures of social deprivation used are biased towards urban conditions and that wards are too large to capture social variations in rural areas. In the least accessible rural wards in East Anglia the deprivation measure most closely related to poor health was male unemployment, but use of this measure did not remove the urban-rural gradient of association strength over the region. Neither did replacing wards by smaller enumeration districts as the units of analysis. The differences between urban and rural correlations were removed by restricting the comparison to wards with the same unemployment range and combining pairs of rural wards with similar deprivation values. Apparent differences between rural and urban associations are therefore not due to the choice of deprivation indices or census areas but are artefacts of the greater internal variability, smaller average deprivation range and smaller population size of rural small areas. Deprived people with poor health in rural areas are hidden by favourable averages of health and deprivation measures and do not benefit from resource allocations based on area values.

                                                                      Spatial variation in mortality and morbidity and the relations with social deprivation and accessibility to health services

                                                                      Objectives

                                                                      To study the spatial variation in mortality and self perceived morbidity in a region with a significant rural population. To examine whether the generic deprivation indices adequately reflect deprivation in rural areas, and whether these can be improved upon by the use of customised measures. To explore the association of access to primary and secondary health care with such variation.

                                                                      Design

                                                                      A geographically based cross sectional study examining urban-rural and intra-rural variations in two health outcome measures, using several definitions of rurality. Correlation and regression analyses explore how well these are explained by generic deprivation indices. Multilevel Poisson modelling investigates whether Customised Deprivation Profiles (CDPs), area characteristics and access to GP surgeries and acute district general hospitals improve upon the explanatory power of the generic indices.

                                                                      Setting

                                                                      Nine counties in the south west of England, comprising of just over six million.

                                                                      Main outcome measures

                                                                      1991 Census limiting long term illness (LLTI) (0–64), and 1991–1996 all cause mortality (0–74).

                                                                      Results

                                                                      Intra-rural variation is apparent, with higher rates of premature LLTI, but not premature mortality, in remoter areas. The generic deprivation indices have strong positive relations with the health outcome measures in urban areas, but these are much weaker in semi-rural and rural locations. CDPs improve upon the generic indices, especially in the rural settings. A substantial reduction in unexplained variation in LLTI in rural areas is seen after controlling for the level of local isolation, with higher isolation, at the wider geographical scale, being related to higher levels of LLTI. The results for SMR will be presented as well as the associations between access to health care and both health outcomes.

                                                                      Conclusions

                                                                      This study highlights the need to treat rural areas as heterogeneous, although this has not been the tendency in health research. Generic deprivation indices are unlikely to be a true reflection of levels of deprivation in rural environments. The importance of CDPs that are specific to the area type and the health outcome measure is emphasised. The significance of physical isolation suggests that accessibility to public and health services may be an important issue. This analysis is in progress at present and findings will be presented at the conference.

                                                                      Variations in English hospital admissions for anaphylaxis by deprivation and geography

                                                                      Background

                                                                      Though the most severe of the allergic disorders, the epidemiology of anaphylaxis is poorly described. This is of particular concern as recent time trend studies of hospital admissions for acute anaphylaxis have shown admission rates to have almost doubled between 1991–1995.

                                                                      Objective

                                                                      To describe the epidemiology of hospital admissions for acute anaphylaxis by deprivation, residence in rural areas and geography.

                                                                      Design

                                                                      Descriptive study using routinely collected individual hospital discharge data.

                                                                      Setting

                                                                      Secondary care in England.

                                                                      Participants

                                                                      2323 emergency admissions for anaphylaxis to NHS hospitals between 1 April 1991 to 31 March 1995.

                                                                      Main outcome measures

                                                                      Admission adjusted anaphylaxis discharge rates (emergency anaphylaxis discharges per 100 000 emergency discharges). We defined Poisson regression models to examine the strength of the association between the admission adjusted rates and four potential risk factors: deprivation (UPA score of greater than 18); rural residence; residence in north/south; residence in east/west.

                                                                      Results

                                                                      Of the 13.5 million emergency discharges from NHS hospitals during the study period, 2323 patients had a primary diagnosis of anaphylaxis. Overall admission adjusted anaphylaxis rate was 17 per 100 000 emergency admissions. Three of the four factors studied were independent risk factors for anaphylaxis admission: south relative risk (RR) 1.35 (95% confidence intervals (95% CI) 1.24, 1.47), rural RR 1.33 (95% CI 1.16, 1.53), and non-deprived RR 1.24 (95% CI 1.12, 1.37).

                                                                      Conclusions

                                                                      Study of four years national hospital discharge data reveals that rates of emergency anaphylaxis hospital discharges vary considerably by deprivation, residence in rural areas and geography. The highest rates of anaphylaxis discharges were seen in patients resident in rural, affluent areas of the south west of England.

                                                                      Why surgeons don't follow guidelines

                                                                      Introduction

                                                                      Widespread and persistent medical and surgical practice variation is held to be a consequence of the failure of clinicians to incorporate research evidence in their everyday work. Practice guidelines are one of the key tactics advocated by the proponents of evidence-based medicine (EBM) to ensure that clinicians apply research evidence to their practice. The critics of EBM contend that there are fundamental differences between the formulised rules encapsulated in guidelines and the type of knowledge required to practise medicine. Much of the opposition to guidelines has been rooted in philosophical arguments about the tensions between “art” and “science”, and surgeons, in particular have been quick to argue that the exercise of clinical judgement makes the use of guidelines problematic. The aim of this paper is to suggest why surgeons ignore guidelines.

                                                                      Methods

                                                                      A qualitative study of 34 surgeons practising urology, gynaecology or pelvic surgery in the UK and USA, involved in treating women with stress urinary incontinence. Qualitative interviews and observational methods were used to explore surgeons' views of surgical practice, to examine the nature of everyday surgical work and consider the applicability of guidelines to this area of medical work. Analysis used techniques of constant comparison to generate themes and categories.

                                                                      Results

                                                                      Surgeons view surgical work as contingent: they describe it as both dependent on conditional factors and subject to chance. They respond to contingency by drawing on tacit knowledge and instinctive responses. Surgical judgement provides a strong justification for resisting the imposition of guidelines.

                                                                      Conclusion

                                                                      The technical knowledge embodied in guidelines may be difficult to reconcile with the individual and practical nature of everyday surgical work.

                                                                      Does “practice make perfect”? Volume of work and outcome in intensive care

                                                                      Background

                                                                      Before a policy to regionalise intensive care in the UK was implemented, it is important to have evidence that larger centres seeing greater numbers of patients achieve better outcomes for patients. There is little evidence, however, that larger centres provide better outcomes for patients. Previous analyses on 26 intensive care units (ICUs) using data from 1988–90 in the UK showed a statistically significant association between higher volumes of work and lower crude hospital death rates that could not be demonstrated after case mix adjustment. Given the small sample size of 26 ICUs in the earlier work, we repeated these analyses on a larger number of ICUs.

                                                                      Methods

                                                                      The relation between average daily ICU volume and hospital outcome was investigated for 46 587 admissions to 91 adult ICUs in the Case Mix Programme Database covering the period 1995–99. The average daily volume for each unit was calculated as the number of admissions divided by the number of days in the data collection period for each unit. The same analysis was repeated for solely surgical and non-surgical admissions.

                                                                      Results

                                                                      Unit volume varied from 2.3 to 26.6 admissions per week across ICUs while ultimate hospital mortality rates varied from 17.7% to 48.7%. The proportion of surgical admissions varied from 8.6% to 71.2% across ICUs. For all admissions to ICU, there was a statistically significant negative association between unit volume and ultimate hospital death rate (Pearson correlation coefficient ρ = −0.215, p value = 0.041). After adjustment for case mix, however, this negative association was no longer statistically significant (ρ = −0.190, p value = 0.071). Similar results were found for solely surgical admissions (crude: ρ = −0.275, p value 0.008; case mix adjusted: ρ = −0.127, p value = 0.229). For solely non-surgical admissions, the negative association remained statistically significant after adjustment for case mix admissions (crude: ρ = −0.220, p value = 0.036; case mix adjusted: ρ = −0.208, p value = 0.048).

                                                                      Conclusions

                                                                      Results showed no overall benefit of higher unit volume for case mix adjusted ultimate hospital mortality, although there was some marginal evidence of benefit for non-surgical admissions. To better understand the relation between volume and outcome, further analyses for subgroups of patients, with diseases that are rare in some ICUs, are warranted.

                                                                      Do hospitals influence their own death rates?

                                                                      Background

                                                                      From the standpoint of understanding both the underlying mechanisms of inequalities in health outcomes and the pressure for increased accountability of health care services, the question of the extent to which variability in the outcomes of hospital care is attributable to hospitals themselves is of growing importance. The International Study of Hospital Outcomes (based at the University of Pennsylvania) includes the objective of using routinely assembled hospital discharge data in order to address methodological aspects of this broad question. This report is based on the Scottish component of the study and uses linked data from the Scottish NHS together with organisational information about Scottish hospitals and data from a more detailed survey of nurses in acute hospitals. The analysis uses multilevel modelling with a three level model describing area, patient and hospital effects.

                                                                      Data

                                                                      The data cover all patients discharged from any one of the 29 acute hospitals in Scotland in 1997/98 in any of five diagnostic groups: 8440 patients with acute myocardial infarction (AMI), 4634 with bacterial pneumonia, 6716 with chronic obstructive pulmonary disease (COPD), 6328 with stroke and 5825 with chronic heart failure.

                                                                      Results

                                                                      Crude death rates for these conditions varied by factors of between two and four: as examples, unadjusted death rates for AMI ranged from 12.1% to 24.7% and those for stroke from 17.1% to 55.5%. Explanations for these differences, of course, require adjustment for several possible influences: within broad diagnostic categories it is necessary to take account of the influence of comorbidities; the demographic characteristics of a hospital's patients and its admission and discharge policies are clearly relevant, as is the population served by a particular hospital in terms of its patterns of health and social characteristics. As a further example, crude hospital death rates for COPD ranged between 2.7% and 11.5%; overall, a high proportion of these differences were explained by variables that were not within the control of individual hospitals.

                                                                      Conclusions

                                                                      Answers to the question of whether hospitals are able to influence death rates require more detailed exploration. More specific questions addressed in this paper include whether adjusted outcomes for specific diagnostic groups are correlated (implying links to other hospital attributes); whether hospital effects are focused on particular patient groups (such as those who have longer lengths of stay); and whether organisational features of different hospitals influence outcomes. Answers to these questions are relevant to “good practice” in the evaluation of hospital care, but are also germane to such wider questions as resource allocation and service planning.

                                                                      Changing the medical record—improving quality

                                                                      Objectives

                                                                      Medical records are often incomplete. This study aims to determine if completeness is improved by replacing the traditional medical record with structured forms in one specialty (urology).

                                                                      Methods

                                                                      All new patients attending urology clinics in two hospitals were randomised to a consultation recorded using traditional notes versus structured forms. The completeness of the notes was compared using a 15 point scale of essential data items. The time taken to complete the forms and traditional notes was recorded. All the clinicians were given a questionnaire to assess acceptability of the forms.

                                                                      Results

                                                                      Over six months, 200 patients were randomised between 15 clinicians. There were significant differences (p<0.05) in completeness between traditional notes (TN) and the structured forms for the following items; clinicians name (TN 32%, risk difference (RD) +53%, confidence intervals (CI) 41, 65), medical history (TN 61%, RD +38%, CI 28, 48), drug history (TN 50%, RD +40%, CI 29, 51), allergies (TN 32%, RD +61%, CI 51, 71), social history (TN 30%, RD +64%, CI 54, 74), examination (TN 68%, RD +23%, CI 12, 34), investigation (TN 80%, RD +16%, CI 7, 24), diagnosis (TN 31%, RD +48%, CI 36, 60), outcome (TN 44%, RD +51%, CI 40, 62) and signature (TN 60%, RD +37%, CI 27, 47). All but two of the clinicians said that they preferred the forms and would opt to use them if given the choice. There was no significant difference in the time taken to use the forms compared with TN.

                                                                      Conclusions

                                                                      Using structured forms significantly improved the completeness of urological documentation. The differences seen were most dramatic (risk difference >45%) in the following; clinician name, allergies, social history, diagnosis and consultation outcome. This simple and economical method of improving the quality of documentation in the medical record has obvious benefits for patient care, audit, research and medico-legal claims.

                                                                      The United Kingdom Neonatal Staffing Study

                                                                      Objectives

                                                                      To assess whether risk adjusted outcomes of neonatal intensive care are related to differences in patient volume, levels of nursing and medical staffing and workload.

                                                                      Design

                                                                      A prospective, study of outcomes for a cohort of infants consecutively admitted to a random sample of UK neonatal intensive care units, stratified in a 3×2×2 factorial matrix by high, medium, or low volume of patients; higher versus lower provision of nursing staff; and higher versus lower neonatal consultant availability.

                                                                      Setting

                                                                      54 UK neonatal intensive care units within the National Health Service.

                                                                      Participants

                                                                      14 436 infants consecutively admitted for neonatal intensive care between January 1998 and April 1999.

                                                                      Main outcome measures

                                                                      Death before hospital discharge; major brain damage of probable postnatal origin; bacteraemia or septicaemia of probable nosocomial origin, adjusted for risk using the Clinical Risk Index for Babies (CRIB) score and other case mix variables obtained in the first 12 hours after birth.

                                                                      Results

                                                                      High volume units were found to care for sicker infants than medium and low volume units. The percentage of infants with a CRIB Score greater than 0 was 32% in high volume units, 26% in medium volume units and 23% in low volume units. No differences were found in the risk adjusted odds of mortality, mortality or brain damage and nosocomial bacteraemia with patient volume, consultant availability and nursing provision. Infants admitted during periods when occupancy of cots was above 70% were found to have increased risk adjusted odds of mortality (1.64, 95% confidence intervals 1.10, 2.45) compared with periods of lower occupancy.

                                                                      Conclusions

                                                                      The current system of neonatal intensive care in the UK reveals no differences in risk adjusted outcomes in simple relation to these major organisational characteristics. There is an increased risk of mortality for all infants admitted when units approach maximum occupancy of cots. Neonatal intensive care may need to be re-configured to reduce the occasions when units approach maximum occupancy of cots.

                                                                      The Emperor's New Clothes: general practitioner views on evidence-based medicine and the role of clinical effectiveness evidence—a qualitative study

                                                                      Background

                                                                      Both development work to promote evidence-based medicine (EBM) and studies, which evaluate the impact of these activities, are mostly based on the assumption that GPs hold the same views and beliefs about what EBM is, to those of the EBM protagonists. There is growing evidence in the literature that this is not the case and this assumption may contribute to the low impact of attempts to promote EBM among GPs.

                                                                      Research aim

                                                                      How can EBM be best defined, understood and supported in general practice and what are the implications for Primary Care Groups/Trusts approaches to Clinical Governance?

                                                                      Method

                                                                      Qualitative methods were used in one case study, a Health Authority area. Semi-structured face to face interviews and group discussions were carried out. Both interview and discussion schedule were piloted. A purposeful sample of 58 GPs from a possible 148 were involved in interviews (12), group discussion (40) and pilot (6).

                                                                      Findings

                                                                      General practice was defined by GPs as a combination of art and science dealing with both the clinical and social aspects of patients' health and illness. Decision making in general practice was seen to be a patient and context specific involving a range of factors to be taken into account in a clinical judgement. Patient expectation was an important factor. Established definitions of EBM were not felt to fully recognise the complex dynamics of decision making and therefore did not fit well into general practice. Additionally, although GPs defined acceptable “evidence” as that produced through methodological approaches associated with the biomedical model, they identified major practical limitations to the application of clinical effectiveness evidence (CEE) produced in this way. These were lack of relevance and sensitivity to general practice patient populations and the individual patient context.

                                                                      Main conclusions

                                                                      EBM in general practice requires redefining. Measures of it should focus on the decision making process not the decision. Other scientific paradigms and associated methodologies need to be understood and accepted if evidence produced is to be relevant to the clinical reality of general practice.

                                                                      Policy implications

                                                                      Development of clinical governance indicators in primary care need to consider these findings, as do local strategies for promoting greater use of CEE in general practice. Particularly, attention should be given within such strategies to the role of the patient in the decision making process, patient information and patient expectation.

                                                                      The effect of evidence-based leaflets on informed choice in maternity care

                                                                      Background

                                                                      The Midwives Information and Resource Service and NHS Centre for Reviews and Dissemination produced a set of 10 evidence-based Informed Choice leaflets to help women make decisions in pregnancy and childbirth.

                                                                      Objective

                                                                      To assess the effect of the service wide use of these leaflets on promoting informed choice among users of maternity services.

                                                                      Design

                                                                      Pragmatic cluster trial, with randomisation of five maternity units to intervention and five to act as controls. Postal questionnaire to women to self report on aspects of informed choice and its possible consequences, before and after the intervention commenced. Antenatally: 1386 women reaching 28 weeks gestation before, and 1778 after, the intervention. Postnatally: 1741 women at eight weeks post delivery before, and 1547 after, the intervention.

                                                                      Intervention

                                                                      Informed Choice leaflets for all women delivering in an eight month period, leaflets for all midwives, and a two hour training session for staff in each intervention unit.

                                                                      Main outcome measures

                                                                      The primary outcome was the proportion of women reporting that they had exercised informed choice overall in their maternity care. Secondary outcomes were the components of informed choice, such as women's knowledge levels of topics covered by the leaflets and their satisfaction with information; and the possible consequences of informed choice, such as women's emotional health, satisfaction with services, and change in service usage.

                                                                      Results

                                                                      The overall response rate to the questionnaires was 64% (6452 of 10 070). The change in the proportion of women exercising informed choice was 1% (95% confidence intervals (95% CI) −13%, 15%) for women antenatally and −4% (95% CI −13%, 5%) for women postnatally. There was an increase in knowledge levels postnatally of 0.24 (95% CI 0.03, 0.45) on a 10 point scale and the uptake of screening tests for Down's syndrome and spina bifida decreased by 10% (95% CI −20, −1), although these were no longer statistically significant when adjusted for covariates.

                                                                      Conclusions

                                                                      The use of Informed Choice leaflets on a service wide basis was not effective in delivering informed choice. Qualitative research undertaken alongside the trial identified problems with the implementation of informed choice rather than the intervention itself.

                                                                      Understanding the role of opinion leaders in putting evidence into practice: the methodological challenge of reviewing qualitative research

                                                                      Objectives

                                                                      To explore the methodological problems involved in taking an overview of seven recent qualitative studies by two groups of organisational behaviour researchers, which examined the process of getting evidence into practice. To assess the benefits of this exercise in terms of its contribution to understanding the change process.

                                                                      Methods

                                                                      The seven studies used similar case study methods, using in depth semi-structured interviewing (both face to face and telephone interviews) and documentary analysis, and in some cases written questionnaires. All the authors undertook an initial analysis of the final reports of the studies; from this analysis, a framework of common themes and points of divergence was developed through a process of iterative reflection and debate. This framework is now being used to carry out a comparative analysis of the final reports; methodological issues and problems arising are being recorded and discussed as the exercise progresses.

                                                                      Results

                                                                      We have so far analysed one of the identified themes: the role of clinical opinion leaders. This emerged consistently as crucial for the flow of evidence into practice. However, subjective understandings of the role differ widely among respondents, and there is a spread of types of opinion leadership (notably expert and peer opinion leaders, and the impact of hostile opinion leaders). Researchers also understand the role differently. They attach subtly different meanings to the same term, or use different terms to mean similar things. One team preferred the term “opinion leaders” (encompassing hostile reactions); the other preferred the term “product champion” and treated hostile stakeholders separately. This perhaps raises a question about how far one can “systematically” review material—whether from case studies or randomised controlled trial studies—when such different terms and definitions are in use, and the reviewer is in part acting as translator. Different understandings of these terms will have affected how questions about opinion leaders were constructed and asked, and how responses were interpreted and categorised.

                                                                      Conclusions

                                                                      Qualitative research often produces small scale, non-generalisable results. Developing more cumulative findings by taking an overview across such studies could help validate and reinforce evidence from individual studies; our experience so far suggests consistent themes can usefully be identified, but a number of methodological difficulties remain to be worked through.

                                                                      Systematic sense—clinical nonsense

                                                                      Introduction

                                                                      Systematic reviews may be internally consistent and may satisfy the criteria for an adequately conducted study but unless they make clinical sense they will fail to convince clinicians. Reviews of the effectiveness of epidural corticosteroid injections for low back pain and sciatica illustrate this well. Practitioners want an answer to the question, “If I give an epidural corticosteroid injection to a patient with low back pain or sciatica in primary care or outpatients, will the patient get pain relief more quickly than if I do not?”

                                                                      Method

                                                                      Three systematic reviews on the subject, including a Cochrane review, have been published. The trials included in the reviews were scrutinised for the relevance of the study population, clinical appropriateness of the intervention, and the adequacy of the outcome measures.

                                                                      Results

                                                                      The conduct of the reviews did not make sense from a practitioner's viewpoint. The following faults were identified: (1) One review said more about the inadequacies of preceding research than about the condition under study by demonstrating that methodological quality was not related to statistical significance. (2) Two reviews included populations, such as postoperative patients, who are clinically distinct from the general populations treated by the majority of clinicians who would consult the reviews. (3) Two reviews included a study with an inappropriate intervention: patients in the placebo group were crossed over to active treatment so early as to invalidate the outcome assessment and to invalidate the eventual pooling. (4) Studies with poor outcome assessments were included. Their inclusion arose from the reviewers' reliance on methodological scoring systems which give lower weighting to outcome assessments (maximum of 5 points out of a potential total of 100) than to other aspects of study design. These older studies might have been acceptable at the time of the primary research but the assessments are unacceptable to practitioners now, who form the audience for this secondary research. (5) Two of the reviews combined heterogeneous treatments and conditions into a clinically meaningless mass and proceeded to analyse the pooled data.

                                                                      Conclusion

                                                                      These systematic reviews lost sight of the question asked by clinicians. Greater attention should be paid to understanding the condition and the interventions so that studies are selected on the basis of their relevance and appropriateness. Greater attention should also be paid to the adequacy of the outcome measures. Clinicians may be best suited for this role in systematic reviews.

                                                                      A systematic review of the definition and measurement of surgical wound infection

                                                                      Introduction

                                                                      The measurement and monitoring of surgical adverse events is of growing importance given the increase in innovative surgical techniques. Furthermore, the general decline in hospital length of stay has contributed to the need to review whether postoperative events are being accurately and comprehensively monitored.

                                                                      Objectives

                                                                      To review the definition and measurement of surgical wound infection.

                                                                      Study design

                                                                      A systematic review of prospective, follow up, cohort and longitudinal studies of surgical wound infection published in English between 1993 and 1999. This study was undertaken as part of an NHS R & D Health Technology Assessment methodological review.

                                                                      Main outcome measures

                                                                      (1) The definition of surgical wound infection, (2) identification of measurement, scoring or grading systems and (3) the assessment of the validity, reliability, accuracy and practicality of identified definitions and grading systems.

                                                                      Results

                                                                      Over 1400 abstracts were retrieved and read; 240 articles fulfilled eligibility criteria and were assessed by two independent reviewers. Preliminary findings suggest that there is no clear consensus on the definition of surgical wound infection. There is large variation in the measurement of wound infection and little evidence on the validity and reliability of wound scoring and grading systems currently in practice.

                                                                      Conclusion

                                                                      Despite international efforts at surveillance over at least 20 years, there is wide variation in the definition of surgical wound infection in the published literature. The measurement of postoperative wound infection is hampered by the lack of a standardised, repeatable and validated definition. Given the trend towards decreased hospital length of stay it is crucial that a single repeatable definition be used in the measurement of surgical wound infection in both hospital and post-discharge settings.

                                                                      Cardiovascular disease in women in Scotland: long term implications of hypertension in pregnancy

                                                                      Background

                                                                      Scotland has one of the highest rates of cardiovascular disease mortality in the world. Women are exposed to the same cardiovascular risk factors as men and to a number of gender specific risk factors such as pregnancy, menopause, hysterectomy and the use of exogenous hormones. Vital statistics data suggest that, compared with nulliparous women, parous women have higher mortality from hypertension, ischaemic and degenerative heart disease and cerebrovascular disease. Recent case-control studies have suggested an association between pregnancy related hypertensive diseases and later cardiovascular morbidity. However, these studies are potentially confounded by recall bias. We have conducted a cohort study to test this hypothesis, starting with reliable data on exposure to raised blood pressure during pregnancy.

                                                                      Methods

                                                                      Women who delivered their first baby in Aberdeen maternity hospitals between 1951 and 1970 were identified from the Aberdeen Maternity and Neonatal Databank. The women with pre-eclampsia/eclampsia (defined using internationally agreed criteria) were age matched with those with gestational hypertension and those with normotensive pregnancies. The women were traced through their medical records in Grampian and contacted with the knowledge of their GP. A total of 1876 were invited to complete a questionnaire and attend for an examination. Details of hospital discharges and mortality were also obtained for the entire cohort.

                                                                      Results

                                                                      A questionnaire response rate of 71 per cent was obtained and we were able to conduct a physical examination of 76 per cent of the questionnaire respondents. Differences between the three study groups were observed for body mass index and smoking history. Overall analysis indicates statistically significant excess risks of hypertension at follow up clinical examination, and of hospital discharge diagnoses of hypertension, ischaemic heart disease and circulatory disorders in women with previous hypertensive disease of pregnancy.

                                                                      Conclusions

                                                                      These data contribute to a growing understanding of the pathogenesis of cardiovascular disease in women. If the associations described here are real, we suggest that long term follow up of women who have raised blood pressure during pregnancy might be warranted in order to minimise morbidity due to cardiovascular and circulatory diseases in later life.

                                                                      The association between HbA1C levels in early pregnancy and poor pregnancy outcomes

                                                                      Objectives

                                                                      To assess the relation between HbA1c levels in the first trimester of pregnancy in women with pre-existing insulin dependent diabetes and the prevalence of congenital anomalies in their offspring.

                                                                      Method

                                                                      Data were collected on women with pre-existing insulin dependent diabetes mellitus in whom an HbA1c was recorded at 10±2 weeks gestation. Pregnancies resulting in a live or stillborn infant were included. HbA1c measurements were normalised using the criteria of the Standardisation Initiative for Glycated Haemoglobin.

                                                                      Design

                                                                      Prospective cohort study.

                                                                      Setting

                                                                      Geographically defined cohort, attending 10 centres for maternity care in north west England.

                                                                      Main outcome measures

                                                                      Adjusted HbA1c measurement in the first trimester of pregnancy. The prevalence of congenital anomalies.

                                                                      Results

                                                                      202 pregnancies were included in the analysis. The HbA1cvalues were divided into tertiles. The prevalence of congenital anomalies in the lowest tertile was not statistically different from that observed in the general population, (14/1000 total births; 95% confidence intervals (95% CI) 0, 43/1000). In contrast, the prevalence of congenital anomalies in women in the uppermost tertile was nine times higher (134/1000; 95% CI 53, 216/1000).

                                                                      Conclusions

                                                                      The findings suggest that there maybe a threshold level for HbA1c in pregnant women with insulin diabetes mellitus, below which the risk of congenital anomalies in their offspring approaches that seen in the general population.

                                                                      An analysis of the variation in mode of delivery, by ethnic group. East London 1988–1997

                                                                      Objectives

                                                                      To analyse the variation in mode of delivery in primiparous women, by ethnic group.

                                                                      Design

                                                                      Retrospective analysis of a computerised obstetric dataset.

                                                                      Setting

                                                                      East London 1988–1997.

                                                                      Population

                                                                      22 776 primiparous women with singleton pregnancies and no antenatal complications recorded.

                                                                      Main outcome measures

                                                                      The mode of delivery of primiparous women after spontaneous labour onset, by ethnic group. The mode of delivery of primiparous women after induction of labour, by ethnic group.

                                                                      Results

                                                                      There was no difference between ethnic groups in the proportion of women who delivered normally. The proportion of white women who entered labour spontaneously, and who delivered by emergency caesarean section was 3.4% (95% confidence intervals (95% CI) 3.04%, 3.76%). African and West Indian women were at increased risk of delivering by emergency caesarean section compared with white women, after spontaneous labour onset (African relative risk (RR) 2.46, 95% CI 2.07, 2.93; West Indian RR 1.56, 95% CI 1.30, 1.95). The proportion of white women who delivered by forceps or ventouse extraction was 13.7% (95% CI 13.01%, 14.39%), after spontaneous labour onset. Non-white ethnic groups were less likely to deliver by forceps or ventouse delivery after spontaneous labour onset, compared with white women: African RR 0.54, 95% CI 0.46, 0.64; West Indian RR 0.45, 95% CI 0.37, 0.56; Bangladeshi RR 0.72, 95% CI 0.64, 0.81; Indian RR 0.82, 95% CI 0.71, 0.95; and Pakistani women RR 0.78, 95% CI 0.63, 0.97. After induction of labour, the proportion of white women who delivered by emergency caesarean section was 14.00% (95% CI 12.50%, 15.46%), and by forceps or ventouse extraction was 24.8% (95% CI 21.3%, 24.8%). After induction of labour, African (RR 2.02, 95% CI 1.72, 2.38), and West Indian (RR 1.32, 95% CI 1.02, 1.70) women were more likely to deliver by emergency caesarean section, compared with white women. The risk of forceps or ventouse delivery was lower for African (RR 0.57, 95% CI 0.45, 0.73), West Indian (RR 0.65, 95% CI 0.5, 0.9) and Bangladeshi (RR 0.79, 95% CI 0.64, 0.98) women, after induction of labour, compared with white women.

                                                                      Conclusions

                                                                      Factors known to increase the rate of caesarean section such as maternal age, parity, maternal height, and birth weight do not explain the found variation in mode of delivery by ethnic group, although maternal weight and choice of analgesia may contribute. Prospective observation, measurement and recording of the process and progress of labour, and confounding variables, are required to elucidate the reasons for the found variation in mode of delivery by ethnic group.

                                                                      Changes with age in the socioeconomic gradient in hysterectomy: findings from a national cohort

                                                                      Objective

                                                                      To examine whether the striking social gradient in hysterectomy observed in a cohort of women at 43 years attenuated in the fifth decade of life.

                                                                      Methods

                                                                      1755 (69%) of 2547 women born in England, Scotland and Wales followed up since birth until the age of 52 years and who provided information on gynaecological surgery. Socioeconomic status was measured using highest educational qualifications, partners' and own social class and childhood social class. Cox's regression models were used to examine the risk of hysterectomy by education, fitted as a categorical factor with no qualifications as baseline, and social class, fitted as a linear trend with social class V as baseline. Time dependent covariates were used to assess whether the effects remained constant at all ages.

                                                                      Results

                                                                      The cumulative hysterectomy risk doubled (from 10 to 21%) between 43 and 52 years. By 52 years, women with less education were still more likely to have had a hysterectomy (p=0.005) but the hazard ratio for women in the highest three educational categories, compared with women with no qualifications, attenuated with increasing age. For ages up until 43 years the inverse gradient with partner's social class (hazard ratio (HR)=0.85; 95% confidence intervals (95% CI) 0.75, 0.95) found previously was confirmed, while an equivalent analysis with follow up to age 52 years indicated a weaker effect (HR=0.93; 95% CI 0.86, 1.01). A piecewise Cox's regression model with two time intervals (up to age 43 years and 44–52 years) provided evidence of a difference in effect for the two age groups (p=0.08). A lack of a gradient between 44–52 years (HR=1.01; 95% CI 0.89, 1.15) was observed. A gradient with childhood social class was observed (p=0.003), which remained constant across the whole time period (HR=0.88; 95% CI 0.81, 0.96). Adjustment for parity, obesity and prior sterilisation did not account for the differences by education or by father's and partner's social class. The effect of own social class was of marginal significance during both time periods.

                                                                      Conclusion

                                                                      This is the first study to report changing effects in the social gradient in the risk of hysterectomy. The greater social differentials at younger rather than older ages may be due to the increased proportion of hysterectomies for benign diagnoses among younger women. Alternatively, the diseases and conditions for which hysterectomy provides a treatment may occur later in women from higher social groups. The constant gradient by childhood social class may reflect differences in gynaecological health throughout life.

                                                                      What is meant by “heavy periods”?

                                                                      Introduction

                                                                      Although the formal definition of menorrhagia (excessive periods) is blood loss over 80 ml per period, objective assessment of volume of loss is rarely undertaken in routine clinical practice. The fact that management and treatment depends on subjective complaint of “heavy periods” occasions considerable unease, as it has been found that if blood loss is measured fewer than half of menorrhagia referrals have excessive loss by defined criteria.

                                                                      Design

                                                                      Cross sectional questionnaire survey.

                                                                      Setting

                                                                      Gynaecology outpatient clinics at three hospitals in Edinburgh and Glasgow.

                                                                      Participants

                                                                      Women aged 25–49 years, newly referred for menstrual problems (n=952).

                                                                      Main outcome measures

                                                                      Report of menstrual experience, deprivation code, subjective judgement of periods, referral for menorrhagia.

                                                                      Results

                                                                      Among 343 (36%) women reporting periods as very heavy, less than half (47%) state volume of loss as a severe problem, and only a third (35%) say it contributed to healthcare seeking. The majority of referrals were for excessive periods (719, 76%) but even within this subgroup only 43% reported their periods as “very heavy”. Logistic regression analyses were undertaken to construct a model explaining subjective judgement of periods as “very heavy”. This shows that “very heavy” periods are associated with a range of menstrual aspects being experienced as problematic, particularly “accidents”, change from normal, volume of loss, and pain, but also, very strongly, with large clots, and with having to get up to change protection at night and needing to use double protection. Reporting of “very heavy” periods was not associated with deprivation category but there was a strong deprivation gradient for reporting the various aspects of menstruation as “severe problem”, including extra washing caused and cost of sanitary protection, with up to a fourfold increase in prevalence with deprivation.

                                                                      Conclusions

                                                                      Less than half of menorrhagia referrals judge their periods as “very heavy”, which may partly explain why relatively few of similar women have excessive blood loss, if objectively measured. Furthermore, belief that periods are “very heavy” is based on more than solely subjective volume of loss. However, despite the fact that women from relatively deprived areas are more likely to report severe problems with periods, they are not more likely to judge their periods “very heavy”, nor to have been referred for excessive bleeding. Rather than focusing on volume of loss, healthcare need across socioeconomic groups would be better revealed by an integrated assessment of menstrual health, encompassing physical symptoms as well as psychosocial effects.

                                                                      Specialisation, survival and clinical practice in patients with pancreatic, oesophageal and gastric cancer

                                                                      Objectives

                                                                      To examine the relations between specialisation of cancer care (indicated by volumes of patients managed annually by doctors and hospitals) and clinical practice and patient mortality.

                                                                      Design

                                                                      Cohort study. Multiple logistic regression and Cox's proportional hazards models adjusted for relevant clinical and prognostic variables. Each cancer was examined separately.

                                                                      Setting

                                                                      Hospitals in south and west England, and (for pancreas) south Wales.

                                                                      Participants

                                                                      2294 patients newly diagnosed as having gastric, oesophageal or pancreatic cancer, between June 1996 and May 1997.

                                                                      Main outcome measures

                                                                      Test and treatments provided. Operative (30 day) mortality. Survival time.

                                                                      Results

                                                                      Patients of higher volume hospitals and doctors tended to have better prognostic factors. Several investigations were more likely with increasing doctor volume. Patients of higher volume doctors were more likely to have resections. “No active treatment” was more likely with lower doctor volumes for all three cancers and with lower hospital volumes for pancreatic cancer. Survival time was longer with higher doctor volumes for oesophageal cancer and with higher hospital volumes for gastric and pancreatic cancers (adjusted hazard ratios attributable to managing 40 more patients per year: 0.69 (95% confidence intervals (95% CI) 0.49, 0.98), 0.78 (95% CI 0.62, 0.97) and 0.64 (95% CI 0.49, 0.83) respectively). Operative mortality was less likely with increasing doctor volume for oesophageal and gastric cancers (adjusted odds ratios attributable to managing 10 more patients per year: 0.68 (95% CI 0.52, 0.96) and 0.60 (95% CI 0.39, 1.0) respectively), but for pancreatic cancers was not associated with doctor or hospital volumes.

                                                                      Conclusions

                                                                      Specialist cancer care, as indicated by patient volumes, was significantly and substantially associated with lower mortality. Clinical practice was influenced more by doctor specialisation than by hospital specialisation. The study supports the specialisation of cancer care. Specialisation of doctors is at least as important as specialisation of hospitals, especially for oesophageal and gastric cancers.

                                                                      Inequalities in survival from colorectal cancer: data from the Wessex Colorectal Cancer audit

                                                                      Introduction

                                                                      The relation between deprivation and survival from a wide range of cancers has been well documented, but the underlying reasons are not well understood. Delay in presentation, comorbidity and inequity of treatment have been cited as possible factors. We investigated the effect of deprivation on survival from colorectal cancer while controlling for prognostic factors of age, sex, site of cancer, Dukes's stage at diagnosis, comorbidity at presentation, emergency versus elective initial surgery and specialist versus non-specialist surgeon.

                                                                      Methods

                                                                      All incident cases of colorectal cancer in residents of Wessex, SW England, over a three year period between 1991 and 1994 were included in the study (n=5176). Patients with complete data on all prognostic factors were included in the survival analysis (n=4169). Deprivation was measured using the Townsend score of the patient's postcode of residence. Z tests were used to identify differences between included and excluded cases. Kaplan-Meier analysis was used to confirm the effect of the above variables on survival in this population, and survival of the most and least deprived quintiles of the study population was compared using Cox regression analysis. For patients presenting with Dukes's stage C cancer, numbers receiving chemotherapy in the most and least deprived population quintiles were compared.

                                                                      Results

                                                                      Patients excluded from the analysis had a similar level of deprivation to those included: (Z=0.18, p=0.5), but had shorter median survival times (45 versus 1096 days). The unadjusted hazard ratio for dying from colorectal cancer (most deprived versus most affluent 20%) was 1.22 (95% confidence intervals (95% CI) 1.07, 1.39). After adjustment for these prognostic factors, the hazard ratio was 1.26 (95% CI 1.11, 1.44). For patients presenting with Dukes's stage C cancer, receiving postoperative chemotherapy was significantly related to deprivation: 6.8% (n=13) of the most deprived quintile versus 18.3% (n=35) of the most affluent quintile (χ2=11.5, p=0.0007).

                                                                      Conclusions

                                                                      In this population-based cohort, survival from colorectal cancer is associated with material deprivation. The differences in survival cannot be explained in terms of known prognostic factors such as the stage of disease at diagnosis, initial health status, or surgical treatment. We found no evidence of differential exclusion of more deprived patients, but patients with short survival times were differentially excluded, raising our overall estimate of survival. There is some evidence that, in this population, chemotherapy treatment varied by deprivation, though this was not sufficient to explain survival differences. Further investigation at the level of individual patients is necessary to identify the underlying causes of such survival differences.

                                                                      Exploring some QALY assumptions: the views of bereaved relatives about the value of survival in malignant cerebral glioma

                                                                      Background

                                                                      Malignant cerebral glioma provides one example of a potential trade off between quality and length of survival. The median survival after surgery is increased from only three months to around 10 by radiotherapy and only 5–15% of patients survive two years. Radiotherapy takes six weeks and a range of adverse effects including fatigue, deterioration and cognitive problems may follow. An economic appraisal shows that the cost of achieving one QALY is over £100 000.

                                                                      Objective

                                                                      To explore the views of bereaved relatives of patients with malignant cerebral glioma about survival following radiotherapy.

                                                                      Design

                                                                      Semi-structured interviews with relatives of patients previously studied between diagnosis and death.

                                                                      Setting

                                                                      Patients treated with radiotherapy at five London hospitals between 1990–2.

                                                                      Subjects

                                                                      56 relatives (44 spouses; 12 others) seen 4–6 months after bereavement and 20 relatives re-interviewed at 13 months.

                                                                      Main outcome measures

                                                                      Relatives' views about quality of life and the value of radiotherapy.

                                                                      Results

                                                                      Relatives viewed quality of life as good or acceptable when they saw patients as having been fit, able to carry on some normal activities or to enjoy social relationships. They described restricted and dependent states, constant deterioration or loss of normal social interaction as providing poor quality of life. Most relatives (34 of 56) described poor quality of life and most were satisfied with radiotherapy (36 of 56). (Inter-rater reliability for interview ratings of quality of life and satisfaction with radiotherapy assessed by κ 0.73 and 0.83). Relatives' views were stable over time. Patients highly disabled at diagnosis were less often felt to have had a good or acceptable quality of life (3 of 19) than patients with no disability (9 of 12). Patients initially rated as highly distressed were also less often viewed as having an acceptable quality of life (2 of 12 versus 23 of 42). Relatives more often expressed dissatisfaction with radiotherapy when patients lived no months free from disability (13 of 21) than when they lived one to six months or more free from disability (7 of 35). Relatives of patients surviving less than six months also expressed more dissatisfaction with radiotherapy (12 of 19) than those living six months or more (7 of 36).

                                                                      Conclusions

                                                                      Bereaved relatives' views about quality of life relate to patient disability and distress at diagnosis. Relatives judge short periods free from disability or 6 to12 months of survival as worthwhile. These findings support the use of disability and distress in quality of life indices but suggest shorter periods than a full year free from disability are judged worthwhile.

                                                                      General practice approaches to increase attendance for breast screening: results from two randomised controlled trials targeting either recent non-attenders or all eligible women

                                                                      Introduction

                                                                      Breast cancer screening in the UK is offered three yearly to women aged 50 to 64 years by specialist screening units. It has been estimated that a 25% reduction in mortality from breast cancer could be realised if 70% of eligible women invited for screening subsequently attended. These studies examined the effectiveness of two primary care interventions aimed at improving uptake in areas with previously low uptake rates when targeting either recent non-attenders or all eligible women.

                                                                      Methods

                                                                      Two interventions aimed at improving screening attendance were investigated in two parallel factorial trials. The interventions were a systematic intervention (a letter of endorsement from all general practitioners in the practice) and an opportunistic intervention (a coloured flag in the notes to prompt the health professional to mention breast screening). General practices were eligible to participate in these trials if they had obtained a practice uptake of less than 60% in the previous screening round. The two trials differed as they targeted: all women prior to being invited for the third screening round in Trial 1; recent non-attenders in the third screening round in Trial 2. In addition, Trial 1 was cluster randomised by practice, while Trial 2 individually randomised women.

                                                                      Results

                                                                      In Trial 1, 6133 women from 24 GP practices were cluster randomised into the four intervention groups: 1721 to control, 1818 to letter, 1232 to flag, and 1362 to letter and flag. Attendance data were obtained for 5732 women (94%). In Trial 2, 1158 non-attenders were individually randomised: 289 to control, 291 to letter, 290 to flag and 288 to letter and flag. Subsequent attendance status was obtained for 1148 women (99%). The letter independently increased attendance rates in both trials (Trial 1: odds ratio (OR) = 1.3; 95% confidence intervals (95% CI) 1.1, 1.6 and Trial 2: OR = 1.5; 95% CI 1.0, 2.2), whereas the flag only significantly improved attendance for all eligible women (Trial 1:OR = 1.4; 95% CI 1.1, 1.8 and Trial 2: OR = 1.4; 95% CI 0.9, 2.1). Costs per extra attendance were £26 (Trial 1) and £35 (Trial 2) for the letter and £39 (Trial 1) and £63 (Trial 2) for the flag.

                                                                      Conclusion

                                                                      The letter was the most cost effective intervention when targeting recent non-attenders or all eligible women prior to screening. The flag intervention was effective for all eligible women, although slightly less cost effective than the letter.

                                                                      Transgenerational influences on inequalities in size at birth

                                                                      Background

                                                                      There is evidence for inter-generational continuities in birth outcome, however the mechanisms underlying these are not fully understood. This area is difficult to research in part because of the very few settings in which adequate information is available for two or more generations for both biological and social parameters. Developing an adequate understanding of these intergenerational effects, however, should throw new light upon the mechanisms underlying the perpetuation of socioeconomic inequalities in size at birth across generations.

                                                                      Study aim

                                                                      To identify the pathways through which a woman's social class at the time of her own birth can affect the size at birth of her own offspring.

                                                                      Methods

                                                                      In 1962 all Aberdeen primary school children were included in a study designed to investigate the antecedents of an IQ <60. Perinatal records including parental socioeconomic information and school health records detailing height and weight measurements at school entry were obtained concurrently. Using record linkage we have been able to identify the offspring of approximately 2500 of these women who subsequently gave birth to over 4500 infants between 1967–98.

                                                                      Results

                                                                      Social class of the mother in her own childhood was predictive of the birth weight of her own offspring. The mean birth weight of the offspring of women whose own fathers were in social class I and II was 3405 g, while for social class IV and V it was 3285 g, a difference of 120 g (p=0.02). Adjusting for each of maternal age, height, parity and smoking in pregnancy failed to reduce the difference to less than 82 g. However, the difference was reduced to 57 g on adjustment for the mother's own size at birth (p=0.10) and to 62 g adjusting for height at age 4–6 years (p=0.09). Simultaneous adjustment for both early factors reduced the difference further to 25 g (p = 0.44). Adjusting for all contemporary factors reduced the difference to 43 g (p=0.40).

                                                                      Conclusion

                                                                      Inequalities according to current social class exist in each generation's size at birth. These results demonstrate that the birth weight of an infant is also related to the mother's childhood socioeconomic environment. Contemporary socially patterned reproductive behaviours account for only a small proportion of this gradient. Instead, much of this effect is mediated through the effect of social class on the mother's own fetal and postnatal growth. These results demonstrate the way in which poor socioeconomic circumstances of one generation can have an adverse affect upon the fetal growth of the next.

                                                                      Association between breast feeding and growth in childhood through to adulthood: the Boyd Orr cohort study

                                                                      Background

                                                                      The long term influences of breast feeding on childhood and adult nutritional status are unclear. Based on a long term follow up study of the Carnegie (Boyd Orr) Survey of Diet and Health in Pre-War Britain (1937–39), we investigated the effects of breast feeding on later childhood and adult height and body mass index.

                                                                      Methods

                                                                      4999 children from 1352 families were included in the study. Information on infant feeding method and later childhood nutritional status was available for 2995 children. The main outcomes were mean differences between ever and never breast fed subjects for childhood and adult anthropometry.

                                                                      Findings

                                                                      Breast feeding was weakly associated with greater per capita income and increased weekly food expenditure, but was not associated with the number of children in the household, birth order or social class. In childhood, breast fed subjects were significantly taller than never breast fed subjects. The mean difference in Z score for childhood height was 0.15 standard deviations (SD) (95% confidence intervals (95% CI) 0.06, 0.24; p=0.001) and the mean difference in Z score for childhood leg length was 0.15 SD (95% CI 0.07, 0.24; p=0.001). The association between breast feeding and childhood height and leg length persisted when the analysis was restricted to within family height differences in relation to within family differences in breast feeding. Breast feeding was also associated with greater adult height and leg length (mean differences: 0.21 SD; 95% CI 0.06, 0.36; p=0.005; and 0.21 SD; 95% CI 0.04, 0.38; p=0.018; respectively). There was no association between breast feeding and childhood or adulthood body mass index.

                                                                      Interpretation

                                                                      Infants who were breast fed in the 1930s were taller in later childhood and became taller, but not more obese, adults. As stature is associated with health and life expectancy, the possible role of infant feeding in explaining adult mortality patterns is discussed.

                                                                      Birth weight, childhood growth and central obesity in adult life

                                                                      Objective

                                                                      Growth in early life may raise the risk of diabetes through its effect on insulin resistance. Central obesity is a major risk factor for both and thus its relation to fetal growth has received attention. Previous findings have been inconsistent and have not examined whether any effects are independent of childhood growth and socioeconomic circumstances. We examined the relation between birth weight, childhood growth, lifetime socioeconomic status and adult central obesity in a large national cohort.

                                                                      Methods

                                                                      3200 men and women born in England, Scotland and Wales, followed up since birth until age 43 years and with measurements on waist and hip circumference. Information on birth weight, weight relative to height at 4, 7, 11 and 15 years, body mass index (BMI) at 43 years and social class in childhood and adult life was collected prospectively. Waist hip ratios and waist circumference were examined by birth weight, relative weight in childhood and adult BMI grouped into equal fifths of their respective distributions. The relations were tested using linear regression.

                                                                      Results

                                                                      There was a small inverse effect of birth weight on waist hip ratio (p=0.01) but not waist circumference in women, after adjustment for current body size. In men, relative weight at age seven was inversely related to waist hip ratio and waist circumference after adjustment for current body size (p<0.001 for both). These relations were attenuated in men of large BMI (p <0.01 for interactions between relative weight at 7 years and BMI in both cases). Relative weights at 4, 11 and 15 years showed similar patterns to those observed at 7 years. These findings were independent of lifetime socioeconomic circumstances.

                                                                      Conclusions

                                                                      Women of low birth weight may have higher waist hip ratios because of a small pelvic size rather than abdominal obesity. Our findings for men may simply mean that lightness in childhood is a more sensitive marker than birth weight of impaired fetal growth. Alternatively, the relation may be independent of fetal growth, supporting studies that link poor childhood growth to insulin resistance. The stronger association in men with lower BMI in adult life (counter to the findings of other studies), may occur if being heavy in childhood or reaching puberty earlier is associated with a greater concentration of weight around the hips.

                                                                      An investigation of birth size and arterial compliance in a cohort of young adults

                                                                      Objectives

                                                                      It has been hypothesised that synthesis of elastin in the aorta and large arteries may be reduced in fetuses whose growth is impaired, leading to permanent stiffening of these vessels and raised blood pressure in later life. The aim of this study was to investigate the relation between birth size and arterial pulse wave velocity in a cohort of young adults.

                                                                      Design

                                                                      Follow up study of men and women who, along with their mothers, had been participants in the MRC Infant Growth and Nutrition Study (a randomised controlled trial of milk supplementation) between 1972 and 1979.

                                                                      Subjects

                                                                      603 men and women from the towns of Barry and Caerphilly in South Wales, who as part of the original study, had detailed anthropometric measurements from birth until age 5.

                                                                      Exposure

                                                                      Primary: body size, as measured by weight and length at birth. Secondary: changes in weight and length (or height) between birth and adulthood, and birth and 6 months.

                                                                      Outcome

                                                                      Pulse wave velocity, which is inversely related to arterial compliance and an indicator of alterations in elastin composition, in three arterial segments measured at the wrist, groin and foot.

                                                                      Results

                                                                      Pulse wave velocity was significantly faster in men than women at all three sites (p<0.001) and there was a non-significant inverse relation between pulse wave velocity and gestation period. Sex specific pulse wave velocity was not associated with either birth weight or birth length at any of the measurement sites, either before or after adjustment for gestation period. Pulse wave velocity was not associated with maternal weight, height, smoking or blood pressure, social class at 18 months of age or group in the original randomised trial. Pulse wave velocity increased with current blood pressure, smoking, alcohol and physical activity. There was no interaction effect on pulse wave velocity between either birth weight and adult size or birth weight and size at 6 months.

                                                                      Conclusion

                                                                      These data do not support the hypothesis that reduced arterial compliance in adulthood is more likely in individuals who were either light or short at birth. Neither was there any evidence that individuals who are light at birth but heavy or tall in adulthood are more likely to have reduced arterial compliance.

                                                                      How misleading can subgroup analyses be?

                                                                      Introduction

                                                                      Subgroups in randomised controlled trials arise in many different settings, for example, centres in a multi-centre trial, or groups of patients defined by age, gender or baseline risk. While the presentation of subgroup analyses in such trials is common, inappropriate analyses such as separate subgroup-specific tests of treatment effect are often presented and may lead to the incorrect conclusion of differential treatment effects across different subgroups. The aim of this paper is to explore the extent of this problem in different scenarios within the context of randomised controlled trials.

                                                                      Methods

                                                                      Data were computer simulated as if from two treatments across different subgroups and included the types of outcome variables commonly encountered in clinical trials. Factors thought to affect the outcome of subgroup analyses, such as the magnitude of the overall treatment effect and the sizes of the treatment arms and subgroups, were varied in a controlled manner to assess their impact on false positive and false negative rates. Analysis of simulated data considered the overall treatment effect, subgroup specific treatment effects and formal interaction tests.

                                                                      Results

                                                                      Many scenarios were examined. For example, data were randomly generated from the normal distribution for two treatments and two subgroups of equal size, to represent the case where the null hypothesis of no overall treatment and no subgroup effects is true. As expected, 5% of the simulated datasets produced a statistically significant overall treatment effect. Of those significant overall, subgroup specific tests found just one treatment effect to be significant in 55% to 64% of cases, for various sample sizes. Altering the subgroup ratio led to a marginal increase in this percentage (58% to 71%). Data were also simulated for the case where an overall treatment effect exists but no differential effect across subgroups. At 80% nominal power for the overall test, just one subgroup reached significance in 57% to 58% of cases, for various sample sizes. Altering the subgroup ratio led to an increase in this percentage (61% to 79%). No patterns were seen with a change in treatment ratio. For all these scenarios the percentage of significant interaction tests fluctuated around 5%.

                                                                      Conclusions

                                                                      The degree of error when performing subgroup specific treatment tests is high and may result in the incorrect conclusion of differential treatment effects across different subgroups. It is important that researchers and clinicians are aware of this potential problem when assessing the validity and interpreting the results of subgroup analyses.

                                                                      Estimating the degree of undercount in Congenital Anomalies Registers: How well does the capture-recapture technique perform?

                                                                      Congenital anomalies are a significant cause of ill health and mortality in childhood, with approximately 2.5% of all newborn babies having at least one major structural abnormality. The Trent Register for Congenital Anomalies started data collection in 1997, at which time Trent was the only region in mainland Britain not to be at least partly covered by a register. The benefits are not restricted to Trent, as the data (along with that from the new register in Wales) are now routinely transferred to the National database managed by the Office for National Statistics. This has led to substantial increases in the recorded rates of the great majority of anomalies.

                                                                      How many cases could the register still be overlooking? One method used for estimating the extent of undercount in registers in many areas of medicine—including congenital anomalies—is capture-recapture. This is based on a straightforward and attractive concept, using the pattern in which different notifiers to a register may notify some of the same children.

                                                                      Difficulties can occur when using capture-recapture in practice. Estimates can be very imprecise, as often only relatively small amounts of data are available. In particular, assumptions need to be made about the extent of correlation between the different sources of information. For example, an infant notified with a gastroschisis by an antenatal clinic has an increased probability of also being notified by a paediatric surgeon and a neonatal intensive care unit. This situation can be simplified by combining the notifiers into a number of groups, but the way in which this is done can have a large effect on the estimate.

                                                                      Depending upon the assumptions made, experimental estimates from Trent of the number of unrecorded cardiovascular defects include 179 as a relatively low estimate (yielding a rate of 27 per 10 000 registrable births; 95% confidence intervals (95% CI) 18, 134) and 1018 as a relatively high estimate (76 per 10 000 births; 95% CI 56, 108). Indeed, some sets of assumptions result in no useful estimate at all being obtained.

                                                                      This paper will present work currently in progress on the Trent data and introduce some of the potential solutions to the concerns expressed above. It is argued that these need to be addressed before embarking on a capture-recapture exercise with congenital anomaly register data.

                                                                      Recall bias in a study of cardiovascular disease and hypertension in pregnancy

                                                                      Background

                                                                      There is great interest in whether events during pregnancy and childbirth are associated with long term health outcomes for mothers and their children. The large cost of long term prospective studies has led to an increase in the use of retrospective data. Subjects are asked to remember events taking place months, years or even decades before. If the accuracy of these memories is generally poor then associations between recalled experiences and health outcomes are diminished. If there is differential accuracy of recall between sick and healthy subjects then there is great potential for these associations to be exaggerated in studies based entirely on recall.

                                                                      Methods

                                                                      In 1997, as part of a study of cardiovascular disease, a questionnaire was sent to women who delivered their first baby in Aberdeen between 1951 and 1970. A total of 1312 women returned completed questionnaires. Women's recall of hypertension in first pregnancy was compared with information collected during that pregnancy that was stored on the Aberdeen Maternity and Neonatal Databank. This databank contains comprehensive details of all pregnancies in Aberdeen from the 1920s to the present day. Odds ratios (OR) for the associations between adverse cardiovascular health outcomes and hypertensive disease in first pregnancy were calculated firstly using only recall data and secondly using only maternity records.

                                                                      Results

                                                                      Discrepancies between recall and records of hypertensive disease in pregnancy were found for 20% of those with pre-eclampsia or eclampsia in their first pregnancy and 10% of those with no hypertensive disease. When based solely on recall associations between hypertensive disease in first pregnancy and cardiovascular outcomes were generally exaggerated in comparison to similar associations based on maternity records. p Values tended to be smaller when based on recall data. A significant positive association was found between angina and recalled hypertensive disease in first pregnancy (OR=2.01, 95% confidence intervals (95% CI) 1.23, 3.29, p=0.005). This was not significant when maternity record data were used instead (OR=1.27, 95% CI 0.98, 1.64, p=0.07).

                                                                      Conclusions

                                                                      Many papers on retrospective studies of reproductive health include a passing reference to the possibility that recall bias may have influenced the results. If our study had been based only on recall we would be reporting some exaggerated effects and one spurious association between recalled events and long term cardiovascular outcomes. Some assessment of the accuracy of recall is required if we are to be informed rather than misled by retrospective studies.

                                                                      The comparative performance of the Rose angina questionnaire in South Asian populations

                                                                      Objective

                                                                      To assess the performance, in South Asian populations compared with Europeans, of two versions of the Rose angina questionnaire, as translated and implemented in the Newcastle Heart Project.

                                                                      Design

                                                                      Cross sectional study.

                                                                      Subjects

                                                                      A stratified random sample of 1509 Newcastle residents aged 25–74 years from European (n=825), Indian (n=259), Pakistani (n=305) and Bangladeshi (n=120) ethnic groups.

                                                                      Main measures

                                                                      Major abnormalities on a resting 12 lead ECG; prevalence of possible (RQP) or definite (RQD) Rose questionnaire angina; self reported doctor's diagnosis of angina; and associations between these measures.

                                                                      Results

                                                                      Major ECG abnormalities were commoner in South Asians than Europeans (5% versus 2% in men). The prevalences in South Asians and Europeans of RQP and a doctor's diagnosis of angina were similar (18% versus 19% for RQP and 7% versus 8% for a doctor's diagnosis in men, respectively), but RQD was less common (3% versus 7% in men). Among Indian men the prevalence of RQD (4%) was similar to that of a doctor's diagnosis (5%) and major ECG abnormalities (4%) but among Bangladeshi men RQD was about half as common (4%) as a doctor's diagnosis (9%) and major ECG abnormalities (8%). RQD showed a pattern of lower sensitivity and lower agreement with other measures in South Asians compared with Europeans. For example, sensitivity for a doctor's diagnosis was 25% in South Asian and 38% in European men. By contrast, RQP showed similar levels of sensitivity and agreement in the South Asians and Europeans. For RQP, the corresponding figures were 83% and 84%. Similar patterns were seen in women.

                                                                      Conclusions

                                                                      The performance of the Rose angina questionnaire, particularly the RQD form, was sufficiently inconsistent across different ethnic groups to warrant further work to achieve greater cross cultural validity. The RQP form of the Rose questionnaire performed more consistently across ethnic groups than the RQD.

                                                                      Beyond the burden of illness? What factors predict general practitioner consultations

                                                                      Objectives

                                                                      (a) To compare the effectiveness of different indicators of the burden of illness in predicting 12 month consultations with general practitioners; (b) to evaluate whether social support and indicators of social location have any additional explanatory value.

                                                                      Design

                                                                      Community health survey of two age cohorts of adults taking part in the West of Scotland Twenty-07 study.

                                                                      Respondents

                                                                      331 men and 423 women in their early 40s and 323 men and 400 women in their early 60s.

                                                                      Main outcome measures

                                                                      Number of general practitioner surgery contacts in previous 12 months reported by respondents.

                                                                      Methods

                                                                      Face to face interviews conducted by nurse interviewers included the collection of detailed data on morbidity. A series of regression models compared the amount of variation in consultation rates explained by (i) detailed indicators of burden of illness attributed to chronic illness (including number of conditions, type of condition, severity, frequency of pain); (ii) detailed data on current symptoms; (iii) a global self assessment of health. These dimensions of health were then added sequentially to regression models. Indicators of social support and social location were included to assess whether they had any predictive power after the various indicators of burden of illness had been taken into account. Finally these models were assessed against more parsimonious models.

                                                                      Results

                                                                      In isolation, the various indicators of burden of illness explained similar levels of variation in consultation (chronic illness 14.4%, current symptoms 16.5%, self assessed health 15.7%). When all three indicators were included 22.4% of the variation was explained. Taking account of social support and social location further enhanced the predictive power of the model (to 25.6% and 28.1% respectively). These models were assessed against simpler, more parsimonious models. One such model estimated that 23.6% of variation is explained by a small subset of just eight predictors.

                                                                      Conclusions

                                                                      The three general approaches to measurement of burden of illness were equally successful in explaining variation in 12 month consulting rates. However, combining the three approaches was even more successful. Social support and indicators of social location contributed to explaining 12 month consultation rates even in parsimonious models, confirming that non-illness factors influence the use of services.

                                                                      General practitioner's involvement in commissioning maternity care. Will it make a difference? Evidence from the total purchasing and extended fundholding experiments

                                                                      Objectives

                                                                      To evaluate the impact of general practitioners' commissioning of maternity services on women's experiences of care, and on resource use, and to consider the implications for primary care commissioning more generally.

                                                                      Design

                                                                      Comparison of women's experiences and resource use between 21 matched commissioning and non-commissioning general practices.

                                                                      Subjects

                                                                      Staff in general practices, NHS Trusts, and health authorities described organisation of care. Women registered with general practices responding to a postnatal postal questionnaire about their experience and resource use.

                                                                      Main outcome measure

                                                                      Women's self reported experience of information, choice in and control over care, and of resource use.

                                                                      Results

                                                                      After two reminders, 1957 women responded to the questionnaire (overall response rate 62% range (52%–81%)). Multilevel models adjusted for case mix showed no difference in women's experience of care, or their resource use, between commissioning and non-commissioning practices. Commissioning practices were more likely to be associated with more vertically integrated models of service organisation, but responses to only 3 of 21 questions about experience of information, choice and control over care, or about resource use, varied between models of service organisation.

                                                                      Conclusions

                                                                      The expectation that giving primary care organisations responsibility for commissioning care will result in improved patient experience of care or better use of resources, may be misguided. The presence of strong national policy (such as Changing Childbirth) may be equally important. Models of service organisation are not proxies for quality of care. The most powerful force shaping patients experiences of care may be professionals themselves.

                                                                      Lay expertise? The difficult role of lay members on primary care groups

                                                                      Objective

                                                                      To describe the diversity of understandings of the lay role in primary care groups (PCGs) and to assess the characteristics of the policy and practice context that foster or obstruct the development of this role.

                                                                      Background

                                                                      PCGs were established in 1999 as the heart of the Labour government's plans for a primary care lead health service, close to local communities and responsive to their needs. All PCG boards were required to appoint a single lay member, recruited from the local community, but a lack of detailed guidance has left individual lay members to work out for themselves what their role in practice should be.

                                                                      Methods

                                                                      A self complete survey was disseminated to chief executives and lay members of all 66 London PCGs. Respondents' views of the role of the lay member were subject to a content analysis. Subsequently, six PCGs were selected as case studies. In depth interviews with key local stakeholders in public involvement were undertaken, including the lay members. Thematic analysis of these interviews focused on the perceived role of the lay member and its relation to the local policy and practice environment.

                                                                      Results

                                                                      Questionnaires were received from 89% of chief executives and 74% of lay members in London PCGs. Their descriptions of the lay member's role revealed a tension between acting as any other board member, albeit with a non-professional view, and the adoption of a unique role dependent on the assumption of a special relationship with the local community. There was also no consensus over the specific part that the lay member should play in promoting public involvement in general. The experience in the case studies revealed that these tensions were at the heart of the difficulties that lay members faced. Even where PCGs took very corporate approaches to public involvement, lay members were expected to provide a lead or insight for which they were not always equipped. Although defined by their lack of medical expertise, lay members were none the less expected to bring much more to the board than a simply a lay perspective.

                                                                      Conclusion

                                                                      Public involvement in healthcare will require considerable investment if it is to move beyond the levels of informing and consultation (on Arnstein's ladder of participation). Although lay members have a role in this development, it is complicated by the tension between their lay status and expectations of special skills or community relationships. Greater clarity is needed about the future role of lay members within the development of public involvement in primary care.

                                                                      Patient determinants of mental health interventions in primary care

                                                                      Background

                                                                      A large proportion of a general practitioner's caseload comprises patients with mental health problems. It is important to ensure that care is provided appropriately, on the basis of clinical need. It is therefore necessary to investigate the determinants of the use of mental health care in the primary care sector and, in particular, to identify any non-clinical characteristics of patients that affect the likelihood of their receiving appropriate care.

                                                                      Aim

                                                                      To identify and compare the influence of non-clinical patient factors on general practitioners' acknowledgement of mental problems and on their provision of mental health care.

                                                                      Method

                                                                      Cross sectional study of adults aged 16–65 years (n=802) attending one of eight practices (20 general practitioners) in inner west London.

                                                                      Results

                                                                      Multivariable analysis showed that the combination of factors that best predict general practitioners' acknowledgement of the presence of mental problems are GHQ scores (odds ratio (OR) 1.10 per unit increase in score, 95% confidence intervals (95% CI) 1.07, 1.13), previous mental symptoms (OR 7.5, 95% CI 4.3, 12.9), increasing age (OR 1.03 per 1 year increase, 95% CI 1.01, 1.04) and physical health status (OR 0.98 per unit increase in SF36 score, 95% CI 0.96, 1.00). Multivariable analysis showed that the combination of factors that best predict intervention (prescription for psychotropic medication; return visit to general practitioner; referral to psychiatric inpatients/outpatients; referral to other (specified) health professionals, or social services) are previous symptoms (OR 7.4, 95% CI 3.8, 14.4), white ethnic group (OR 2.2, 95% CI 0.9, 5.5); and not owning a property (OR 2.1, 95% CI 1.1, 4.0). Life events influenced intervention only in the presence of low GHQ scores (OR 8.1, 95% CI 2.7, 24.0).

                                                                      Conclusions

                                                                      Mental problems are common in primary care and their acknowledgement is a necessary but not a sufficient condition for intervention. Our results show that general practitioners' decisions about mental health interventions can be influenced by non-clinical patient factors, regardless of patients' clinical needs. The results suggest that current practice may not always be equitable, and point to the need for better understanding of the basis of these potential inequalities and for focused training.

                                                                      Lp(a) lipoprotein and risk of coronary and peripheral arterial disease: Edinburgh Artery Study

                                                                      Background

                                                                      Lp(a) lipoprotein consists of a large glycoprotein, apolipoprotein(a), linked to a molecule of low density lipoprotein cholesterol and may be an important risk factor for the development of atherosclerosis. It is widely accepted that Lp(a) lipoprotein levels are raised in patients with pre-existing coronary artery disease, but there is some doubt about the causality of the relation. Little is known about the relation between Lp(a) lipoprotein and either stroke or peripheral arterial disease, nor about the role of Lp(a) lipoprotein in women.

                                                                      Methods

                                                                      1592 men and women aged 55–74 years were selected at random from 11 general practices in Edinburgh, Scotland and followed up for five years. Three diseases were defined: myocardial infarction (fatal and non-fatal), peripheral arterial disease (WHO intermittent claudication) and stroke (fatal and non-fatal).

                                                                      Results

                                                                      The incidences of myocardial infarction, intermittent claudication and stroke were 13.4%, 9.4% and 3.7% respectively. Raised Lp(a) lipoprotein levels at baseline were associated with an increased risk (95% confidence intervals (95% CI)) of myocardial infarction relative risk (RR) 1.15 (95% CI 1.00, 1.32), intermittent claudication RR 1.32 (95% CI 1.10, 1.57), but not significantly for stroke RR 1.24 (95% CI 0.93, 1.64). This increased risk persisted for intermittent claudication after adjustment for baseline cardiovascular disease and other risk factors RR 1.20 (95% CI 1.00, 1.44), but for myocardial infarction became non-significant RR 1.07 (95% CI 0.93, 1.24). The risk of disease associated with raised Lp(a) lipoprotein was slightly higher in women than in men, especially for intermittent claudication (men RR 1.10 (95% CI 0.88, 1.37) compared with women RR 1.38 (95% CI 1.01, 1.89)).

                                                                      Conclusion

                                                                      We found that Lp(a) lipoprotein was an independent predictor of cardiovascular events in both sexes. The association between Lp(a) lipoprotein and cardiovascular events may have been stronger in women than in men, and for peripheral arterial disease than myocardial infarction or stroke.

                                                                      A population-based analysis of the relation between socioeconomic deprivation and death without hospitalisation from a first acute myocardial infarction in Scotland

                                                                      Objective

                                                                      To examine population mortality rates according to socioeconomic deprivation for those individuals who experienced a first acute myocardial infarction (AMI) but did not survive to reach hospital.

                                                                      Design

                                                                      Population-based study.

                                                                      Setting

                                                                      Scotland.

                                                                      Subjects

                                                                      All Scottish residents dying between 1986 and 1995 for whom a first AMI was the principal cause of death. (First AMI was defined as no prior hospitalisation for AMI since 1981).

                                                                      Main outcome measures

                                                                      Death from first AMI (ICD code 410) between 1986 and 1995 according to age, sex and deprivation category.

                                                                      Results

                                                                      Between 1986 and 1995, 48 481 men (mean (SD) age 71.2 (11) years) and 42 297 women (mean (SD) age 77.8 (10) years) died without hospitalisation following a first AMI in Scotland. There was a marked socioeconomic gradient, which was greater in men and in younger age groups (p < 0.001 for all age groups). The population-based mortality rate in deprivation category five was more than twice that of category one in men and women aged < 65 years. In this age group alone, there were 665 and 388 more male and female deaths than expected. Overall, there were 2007 excess deaths in the lowest socioeconomic category.

                                                                      Conclusion

                                                                      Socioeconomic deprivation profoundly affects population-based mortality rate for first AMIs not surviving to reach hospital. This effect was greatest in the young and in men.

                                                                      The authors are grateful to the British Heart Foundation for funding for this work.

                                                                      Geographical variation of incidence of coronary heart disease in Britain: the contribution of established risk factors

                                                                      Objective

                                                                      To quantify the degree of geographical variation in incidence of coronary heart disease (CHD), and to estimate how much may be explained by conventional risk factors.

                                                                      Design

                                                                      Prospective study.

                                                                      Setting

                                                                      24 British towns.

                                                                      Subjects

                                                                      7735 men followed up from screening in 1978–80 for 15 years.

                                                                      Main outcomes

                                                                      Percentage of variance between the towns in CHD incidence that can be explained by attributes of men in the towns.

                                                                      Results

                                                                      Incidence rates over 15 years varied from 7.7% in Lowestoft to 16.0% in Dewsbury, and tended to follow the well known pattern of being higher in Scottish and northern English towns, and lower in southern English towns (“north-south gradient”). Town incidence rates were strongly related to average systolic blood pressure, prevalence of current cigarette smoking, of leisure time physical activity, and social class distribution. Allowing for sampling error only, we estimated that true age adjusted CHD incidence (over 15 years) would vary from 8.8% to 15.2% among British towns (95% range). After adjusting for baseline blood pressure, cholesterol, body mass index, smoking status, and physical activity, this variation would reduce by 59%, and by 65% if adjustment for social class was also included. A model based on these six variables accounts partially but not completely for the north-south gradient.

                                                                      Conclusion

                                                                      Almost two thirds of the variation in CHD incidence between British towns was accounted for by conventional risk variables; remaining unexplained variation could be related to environmental factors such as climate, or simply measurement error in the known risk variables.

                                                                      Are dietary fibre and bowel habit risk factors for varicose veins in the general population?: Edinburgh Vein Study

                                                                      Background

                                                                      Venous disease is common, resulting in considerable morbidity and a heavy burden on national healthcare resources. For many years, it has been postulated that diets deficient in fibre-rich plant foods are a fundamental cause of varicose veins in the Western world. Such a refined diet results in larger, harder stools that are more difficult to pass leading to constipation and regular straining. The objective of this study was to determine if a relation exists between fibre intake, constipation and clinical venous disease within a Westernised population.

                                                                      Methods

                                                                      The Edinburgh Vein Study is the first study in the United Kingdom to investigate venous disease in the general population. Men and women aged 18–64 years were selected at random from the age-sex registers of 12 general practices and invited to screening. A total of 1566 subjects completed a validated questionnaire enquiring about dietary fibre intake and bowel habit. The presence and severity of varicose veins were assessed during a standardised clinical examination.

                                                                      Results

                                                                      Fibre intake, intestinal transit time, defecation frequency and the prevalence of straining at stool were all found to be significantly different between the sexes. Men who reported that they strained in order to start passing a motion showed a higher prevalence of mild (46%) and severe (12%) trunk varices compared with men who did not strain (32.85 and 6.1% respectively). After adjustment for age, social class, body mass index and mobility at work, this group of men showed a significantly increased risk of having severe trunk varices (odds ratio (OR) 2.76, 95% confidence intervals (95% CI) 1.16, 6.58). No other consistent relations between dietary fibre, bowel habit and varicose veins were seen among men or women.

                                                                      Conclusion

                                                                      No strong and consistent evidence was found to support a role of dietary fibre and constipation in the aetiology of varicose veins within a Westernised population, although straining at stool in men may be mildly protective.

                                                                      Birth weight and ethnicity as predictors of coronary artery calcification

                                                                      Background

                                                                      High rates of coronary heart disease (CHD) are seen in first generation South Asians in the UK, and there is increasing evidence of an inverse association between size at birth and CHD among men born before the second world war. However, it is unknown whether these observations are relevant to those born in Britain to South Asian parents and to younger post-war cohorts. Electron beam computed tomography (EBCT) is a new non-invasive measure of coronary artery calcification and thus indirectly of coronary atheroma. We have used EBCT in a study of young men to address these uncertainties.

                                                                      Study aim

                                                                      To determine whether birth weight and ethnicity are associated with calcified coronary atheroma in men born in west London 1964–68.

                                                                      Methods

                                                                      Obstetric records of 19 000 men born in west London hospitals 1964–68 were abstracted. Men in the top and bottom 15% of the birth weight for gestational age distribution and all men born at term to mothers with South Asian names were traced. All those currently registered with a London GP were invited to participate. To date, 313 men (53 South Asian) have been examined. In the non-South Asian group, 91 are low birth weight.

                                                                      Results

                                                                      The overall prevalence of coronary calcification was 21% (calcification defined as calcium score >3). Body mass index (p<0.001), weight (p<0.001), height (p=0.49) and waist hip ratio (p<0.001) were all independently and positively associated with coronary calcification. South Asian ethnicity was associated with an increased risk of coronary artery calcification (age adjusted odds ratio (OR): 1.94: 95% confidence intervals (95% CI) 0.99, 3.82). Adjusting for concurrent body size increased the strength of association (OR: 2.53: 95% CI 1.01, 6.42). Low birth weight was associated with a lower prevalence of coronary artery calcification, but this association was not statistically significant, even when adjusted for anthropometric measures (OR 0.78: 95% CI 0.31, 1.95).

                                                                      Conclusions

                                                                      These results indicate that second generation South Asians in the UK have a higher prevalence of coronary atheroma than their non-South Asian counterparts, suggesting that the higher rates of CHD evident in first generation South Asians may be exhibited in their children. The lack of a significant association between low birth weight and coronary calcification raises interesting questions regarding the mechanisms underlying the size at birth—CHD association and its relevance to those born more recently.

                                                                      Folate, vitamin B6, vitamin B12 and coronary heart disease in the Caerphilly study

                                                                      Objective

                                                                      To assess the risk of coronary heart disease (CHD) associated with dietary folate, vitamin B6 and vitamin B12.

                                                                      Design

                                                                      Nested case-control study.

                                                                      Setting

                                                                      Caerphilly and surrounding villages in South Wales, UK.

                                                                      Participants

                                                                      2512 men recruited in 1979 to phase I of the study. After 15 years of follow up, 337 men developed CHD and were compared with 1348 randomly selected age frequency matched controls.

                                                                      Main outcome measure

                                                                      Acute myocardial infarction or death due to CHD.

                                                                      Results

                                                                      The adjusted odds ratio of CHD per standard deviation change in nutrient as measured by a food frequency questionnaire was 0.85 (95% confidence intervals (95% CI) 0.7, 1.0) p=0.02 for folate, 0.81 (95% CI 0.7, 0.9) p=0.003 for vitamin B6 and 0.95 (95% CI 0.8, 1.1) p=0.4 for vitamin B12. Sixty two men with three separate measures (phase I to III) of dietary folate, vitamin B6 and vitamin B12 developed CHD between phase III and IV of the study. Comparing these 62 cases with 248 age frequency matched controls show an odds ratio per standard deviation change in mean folate of 0.81 (0.6, 1.1) p=0.2 using the phase I measurement, 0.85 (0.6, 1.1) p=0.3 using the mean of phase I and II, and 0.83 (0.62, 1.1) p=0.2 using the mean of phase I, II and III. When more than one measure of vitamin B6 is used, the odds ratio of CHD is 0.77 (95% CI 0.6, 1.1) p=0.1 for one measurement, 0.84 (95% CI 0.6, 1.1) p=0.2 for the mean of two measurements and 0.85 (95% CI 0.6, 1.2) p=0.2 using the mean of three measurements.

                                                                      Conclusions

                                                                      These findings support the hypothesis that folate and B6 are protective against CHD. Using the mean of more than one measurement does not alter the odds ratio greatly. Randomised controlled trials of folic acid and the B vitamins are needed to determine if this is a true association.

                                                                      Psychosocial risk factors for coronary disease in white, South Asian and Afro-Caribbean civil servants: the Whitehall II study

                                                                      Background

                                                                      Psychosocial factors are associated with the aetiology and prognosis of coronary heart disease (CHD) in white populations, but previous studies have not examined the distribution of psychosocial factors in ethnic groups with coronary rates higher (South Asian) and lower (Afro-Caribbean) than a white population.

                                                                      Study objective

                                                                      To determine whether ethnic differences in psychosocial risk factors parallel those in CHD mortality.

                                                                      Design

                                                                      Cross sectional survey.

                                                                      Setting

                                                                      20 civil service departments in London.

                                                                      Participants

                                                                      8973 white, 577 South Asian (62% Indian) and 360 Afro-Caribbean office based civil servants aged 35–55 years. Ethnicity was observer and self assigned (agreement beyond chance κ 0.85 (95% confidence intervals (95%CI) 0.83, 0.87).

                                                                      Outcome measures

                                                                      Minor psychiatric morbidity (General Health Questionnaire), social supports (marital status, social networks, negative aspects of support, confiding/emotional support, social support at work), psychosocial work characteristics (job control, effort-reward imbalance), hostility and Type A personality.

                                                                      Results

                                                                      South Asians and Afro-Caribbeans were considerably more likely than white subjects to be in lower employment grades; within grades, South Asians were more likely than white subjects or Afro-Caribbeans to have a car, own their own home or be highly educated. South Asians, compared with the white population, had more depression, higher negative supports, less social support at work, less job control, more effort-reward imbalance and higher hostility, when adjusting for age and sex. Afro-Caribbeans, compared with white subjects, had lower minor psychiatric morbidity and lower Type A scores. Thus, the odds of being in the adverse tertile of the depression sub-scale of the GHQ was higher among South Asians (odds ratio 1.42 (95%CI 1.2, 1.7)) and lower among Afro-Caribbeans (0.65 (95%CI 0.5, 0.8) than among white office workers. The remaining psychosocial factors showed either no ethnic differences in distribution, or effects opposite in direction to those predicted from coronary event rates. Further adjustment for employment grade made little difference to these associations.

                                                                      Conclusion

                                                                      Among South Asians, the majority of whom were Indian, the distribution of psychosocial factors was consistent with ethnic differences in coronary rates; the pattern for Afro-Caribbeans was less consistent. Further work is required to test the ability of psychosocial factors to predict events within ethnic groups and to characterise better psychosocial measures.

                                                                      Personality and social predictors of atherosclerotic progression: Edinburgh Artery Study

                                                                      Background

                                                                      If personality and social factors relate not just to the occurrence of acute cardiovascular events, but also to the progression of atherosclerosis, then research can be better targeted to examine the putative mechanisms driving the disease process. An objective, non-invasive and reliable measure of atherosclerosis is the ratio of ankle systolic pressure to arm systolic pressure (ABPI). The ABPI is related inversely to the degree of atherosclerosis in the legs (and throughout the vascular system). Studying personality and social factors in relation to ABPI change may indicate whether these factors are part of the mechanism that accelerates atherosclerosis.

                                                                      Methods

                                                                      In the Edinburgh Artery Study, 1592 men and women were sampled randomly from the general population and had their ABPI measured at baseline and at the end of a five year follow up. Trait submissiveness and hostility were assessed at baseline using the Bedford-Foulds personality deviance scales. Data on other baseline risk factors, including social and physiological factors, were also collected.

                                                                      Results

                                                                      The worsening of atherosclerosis over five years was correlated with increased baseline age and smoking in both men and women (age; men:r=−0.10; women:r= −0.25; smoking; men:r=−0.09; women:r=−0.11; p⩽0.05). Other significant (p⩽ 0.05) correlations with atherosclerotic progression in men were decreased baseline alcohol consumption (r=0.10) and higher submissiveness (r=−0.09). In women, baseline cholesterol levels (r=−0.11, p⩽0.01) and alcohol consumption (r=0.09, p⩽0.05) were also correlated with atherosclerotic progression. In multiple linear regression models, in men, smoking, alcohol consumption and submissiveness accounted for 2% of the variance in ABPI change. In women, only age related to ABPI change, accounting for 6% of the variance. Well fitting structural equation models in both sexes revealed complex associations: age directly influenced both baseline ABPI and change in ABPI; smoking and social deprivation directly affected baseline ABPI; but the effect of hostility, and some of the effect of social deprivation, was mediated by smoking.

                                                                      Conclusions

                                                                      In addition to biological factors, social and psychological variables have an impact on cardiovascular disease. This is important for a clearer understanding of the complex interaction of risks and for more effective disease prevention.

                                                                      The influence of socioeconomic circumstances in early and later life on stroke risk among men in a Scottish cohort study

                                                                      Objectives

                                                                      The purpose of the study was to investigate stroke risk in relation to socioeconomic position at different stages of the life course.

                                                                      Design

                                                                      Prospective cohort study.

                                                                      Setting

                                                                      27 workplaces in Scotland.

                                                                      Subjects

                                                                      5765 working men aged 35–64 at the time of screening in 1970–1973, who provided information on their occupation, their first occupation, their father's occupation, their age leaving full time education and their home address.

                                                                      Main outcome measures

                                                                      Hospital admission for, or death from stroke (ICD8 or ICD9 codes 430—438 and ICD10 codes I60—I69 and G45) in a 25 year follow up period.

                                                                      Results

                                                                      There were 416 men who had a stroke. Men with manual occupations when screened, men whose first occupation was manual and men whose fathers had manual occupations had significantly higher rates of stroke than men in the non-manual categories. Men who left full time education at 16 years or under also had significantly higher rates of stroke. Men living in more deprived areas had non-significantly higher rates of stroke. The most marked difference was in relation to father's social class (age adjusted relative rate was 1.70 (95% confidence intervals (95% CI) (1.31, 2.20)) for manual father's social class compared with non-manual). Father's social class was divided into three categories (non-manual, III manual and IV or V) and even after adjusting for risk factors for stroke, men whose fathers were in manual social classes had higher relative rates of stroke than men whose fathers were in non-manual classes (adjusted relative rate for father's social class III manual was 1.37 (95% CI 1.03, 1.81) and for father's social class IV or V was 1.46 (95% CI 1.09, 1.96)). Men who were upwardly mobile (father's social class manual, own social class non-manual) had a similar rate of stroke to stable manual men.

                                                                      Conclusions

                                                                      Poorer socioeconomic circumstance was associated with greater stroke risk. Adverse early life circumstances were of particular importance and men who improved their social class in adulthood did not affect their stroke risk. Improved early life conditions may reduce socioeconomic inequalities in stroke.

                                                                      Activity and impact of first wave NHS Direct sites

                                                                      Objectives

                                                                      NHS Direct, the 24 hour nurse led telephone advice service, will cover the whole of England by the end of 2000. Although the public aim of the service is to provide “easier and faster access” to care there have also been hopes, and concerns, about the potential impact on demand for other services. We report here findings on the activity and impact of the three first wave NHS Direct sites.

                                                                      Methods

                                                                      Call data from 16 months of site logs and anonymised transcripts of a random sample of 267 calls have been analysed to describe the activity and casemix of NHS Direct. Local and national data from immediate care services, together with data from a population survey of health care use before and after the start of NHS Direct, have been examined to identify changes that may be attributable to NHS Direct.

                                                                      Results

                                                                      Population call rates to NHS Direct have risen steadily and it now accounts for a substantial proportion of all contacts with immediate care services. Over two thirds of calls are out of hours, with a patient and casemix profile resembling that of “acute primary care”. Analysis of other service activity and population survey data suggest there has been no discernible impact on overall demand for ambulance or A&E services. However, the data suggest that the introduction of NHS Direct was associated with an interruption in the pre-existing upward trend in demand for out of hours general practice, so that GP co-op workload is no longer increasing.

                                                                      Conclusions

                                                                      The available data on activity, casemix and caller intention suggest that NHS Direct is being used particularly as an alternative to out of hours contact with a general practitioner. The finding that the service may have restrained growth in out of hours demand for general practice, but has had no impact on A&E or ambulance services, is consistent with this pattern of use. The emerging impact of NHS Direct on the future organisation of primary care, both in and out of hours, will be explored in the light of local projects to integrate NHS Direct with GP cooperatives, the national review of out of hours primary care services and the development of the new walk in centres.

                                                                      Evaluation of the NHS Direct Hampshire 2nd Wave Pilot Site

                                                                      Objectives

                                                                      To determine: (1) Callers' perceptions of ease of access and satisfaction with the service offered by NHS Direct, Hampshire. (2) Safety of the service in terms of adverse incidents following a call.

                                                                      Study design

                                                                      (1) Postal survey of NHS Direct callers using a structured questionnaire, with space for free text comments. (2) Comparison of caller data with coroners' records over a six month period to identify deaths within seven days of consulting the service.

                                                                      Setting

                                                                      NHS Direct pilot site, Hampshire.

                                                                      Participants

                                                                      1000 consecutive callers to NHS Direct over one week. Anonymous questionnaires, based on a previously tested model, were withheld in extreme emergencies or if a caller declined.

                                                                      Main outcome measures

                                                                      (1) Caller satisfaction and dissatisfaction with the service. (2) Number of deaths within seven days of calling.

                                                                      Results

                                                                      (1) 700 people (70%) responded after repeat mailing. Ages of callers ranged from 16–84 years, peaking in the 25–34 years age group. Female callers outnumbered male by more than 3:1. Forty four per cent called for themselves and 45% on behalf of others. Ninety five per cent indicated satisfaction with the service. Of 176 who commented, 33 (19%) cited potential benefits to the NHS of fewer visits to GPs and A&E departments. Main benefits included reassurance, and helpful, appropriate advice. Negative comments (13%) concerned length of calls, excessive questioning, and waiting time. (2) Eighteen patients from 19 335 callers (0.09%) died within seven days of consulting NHS Direct.

                                                                      Discussion

                                                                      (1) NHS Direct is being used predominantly by younger women. They appear to consider it an efficient, professional service run by friendly, sympathetic staff who give appropriate advice. Some comments reveal poor understanding of its purpose and indicate a need for clearer advertising. The view expressed by some of excessive questioning reflects the use of clinical decision support software in the consultation. (2) No adverse incidents resulting from calls to NHS Direct were detected. The rates of deaths of patients within seven days of consulting the service are similar to those observed in the SWOOP study.

                                                                      Conclusions

                                                                      NHS Direct Hampshire is a safe service that is popular with members of the public. It seems likely that these results will support the government's determination to roll out the service to the whole UK by October 2000. Callers' perception that the service will ease demand on busy doctors and A&E departments, while yet unproven, supports calls for closer integration with primary care.

                                                                      Health Improvement Programmes: what do health authorities mean by health improvement, and how do they plan to achieve it?

                                                                      Background

                                                                      “The new NHS. Modern. Dependable.” emphasised that the NHS should strive to improve the general health of the population. Health authorities (HAs) are now required to draw up Health Improvement Programmes (HImPs), and to revise and extend them each year. HImPs must take account of national priorities for health (for example, the four Our Healthier Nation priority areas), managerial/organisational issues (for example, the NHS Modernisation Fund), and include Service and Financial Frameworks.

                                                                      Objective

                                                                      To discover how HAs interpret “health improvement”, and how they have used HImPs to focus activity on improving health.

                                                                      Design

                                                                      Structured analysis of the first and revised HImPs of a sample of 36 HAs.

                                                                      Results

                                                                      In the first HImPs, priorities relating to population health receive considerably more emphasis than do bureaucratic priorities. The health priorities chosen are predominantly those set out in Our Healthier Nation, particularly coronary heart disease, cancer and mental health. Population health priorities are more likely to be targeted at specific diseases than at the socioeconomic determinants of health, although the need for health promotion to people with harmful lifestyles is often mentioned. In most cases, intended processes for achieving priorities lack specific measurable targets and do not name the people or agencies responsible for key tasks. Most HImPs make some reference to consultation and partnership working with NHS and non-NHS organisations, and with the public. However, a few make very little mention of primary care groups (PCGs) despite their central importance in delivering the HImP agenda. In general, therefore, the first HImPs provide broad frameworks for NHS activity, rather than detailed strategies for health improvement at a local population level. The analysis of the revised HImPs (April 2000) will examine evidence of whether the emphasis on population health is maintained, whether strategic planning for health focus has become any sharper, and whether PCGs are pursuing HImP priorities in locally specific ways.

                                                                      Conclusion

                                                                      The challenge for HAs is to devise a deliverable strategy for health gain at population level that can be implemented locally by PCGs in partnership with other agencies. Future HImPs need to include more detailed strategies for improving health that have local relevance and ownership, so that stakeholders can work together effectively and accountably to improve the health of their populations.

                                                                      How have primary care groups managed their core functions in their first year? Findings from the National Evaluation

                                                                      Introduction

                                                                      Primary Care Groups/Trusts (PCGs/Ts) were introduced in 1997 as part of Labour's policies for modernising the NHS. They comprise groups of general practices, covering populations of approximately 100 000. PCGs have three core functions: to develop primary and community care; to commission hospital and community health services; and to improve the health of the local population. The King's Fund and National Primary Care Research Development Centre were commissioned by the Department of Health to undertake a national evaluation of PCG/T development. This paper will present findings from the first year of the study.

                                                                      Objectives

                                                                      (1) To describe how PCGs have tackled their core functions and the obstacles and enabling factors encountered. (2) To identify features associated with the successful delivery of their core functions drawing on lessons from previous forms of primary care commissioning.

                                                                      Design

                                                                      A random sample of 72 PCGs (15% of PCGs in England) were selected to form a three year longitudinal cohort. Structured face to face interviews were held with each PCG Chair, chief executive and nominated health authority lead during September–October 1999. Postal questionnaires were used to collect further data from other key stakeholders on the PCG boards. Relevant documents were collected from each PCG, including Health Improvement Programmes, Primary Care Investment Plans and Annual Accountability Agreements.

                                                                      Results

                                                                      PCGs spent much of their first year developing the organisation. Management costs varied considerably across PCGs; 17% had fewer than two staff with smaller PCGs generally receiving lower management costs and fewer staff. PCGs had begun to take stock of their existing practice infrastructure and provision and had developed specific initiatives on prescribing and clinical governance. PCGs were also beginning to tackle their commissioning role, with 50% having fully delegated responsibility for commissioning hospital and community health services. Health improvement was usually defined as tackling inequalities in access, rather than health inequalities. Forty per cent are considering applying for PCT status for April 2001 but were not clearly more advanced in term of their organisation or core functions.

                                                                      Conclusion

                                                                      PCGs/Ts have made significant steps in the first year, although much variation was observed among them. Not surprisingly they have tended to concentrate on areas that are more familiar to them, such as prescribing and general practice provision. However, the majority have plans to develop their commissioning role. PCGs need scope for local innovations if they are to consolidate their early achievements.

                                                                      Clinical governance in the UK: a principal-agent analysis

                                                                      Background

                                                                      A key feature of the latest round of UK health reform is the placing on senior health service managers of a statutory duty for clinical quality. Accountability for clinical performance is to be achieved by new structural, procedural and cultural changes collected under the umbrella term of “clinical governance”.

                                                                      Objective

                                                                      This paper analyses the latest proposals for clinical governance in the UK. The objective is to elucidate shifts in the emphasis of government policy regarding physician control. Key aims include: description of the predicates of current policy; analysis of the implications of new policy mandates; and a SWOT analysis (strengths, weaknesses, opportunities, threats) of the proposals as currently constituted.

                                                                      Methods

                                                                      Analysis of the key government policy documents, and assessment of the proposals contained therein, in the light of the established literature on performance management and control strategies. The dominant perspective of the analysis is a principal-agent framework.

                                                                      Principal findings

                                                                      The UK NHS can be envisaged as a diverse collection of overlapping principal-agent relationships. Yet asymmetries of information and a lack of congruence of objective functions hamper principals' ability to control clinical behaviour. This policy analysis suggests much greater and more prescriptive attention to the former, and a comparative neglect of the latter. Thus recent policy shifts in the UK regarding physician control emphasise national standard setting, detailed measurement of local clinical practice, and external scrutiny. The incentive framework to reward excellence and punish ineptitude, however, remains under-articulated. Internal modes of control (for example, individual ethics, professional norms, and satisfaction in self efficacy) receive little concrete attention. Although frequent mention is made of inculcating an appropriate quality culture and the importance of leadership it remains unclear how this is to be accomplished.

                                                                      Conclusions

                                                                      An over-reliance on bureaucratic control facilitated by performance measurement and objective standards tackles only one of the major asymmetries inherent in the principal-agent relationship. The comparative neglect of possible incongruities in objective functions and the potential for the deleterious impact of top-down control mechanisms need further consideration. Bureaucratic control brings one level of reassurance but may impact adversely on the development of an appropriate quality focused culture and reflective practice. Thus knitting together these diverse objectives of public policy remains a key challenge.

                                                                      The acute effects of air pollution reduction in Westminster on disease outcomes

                                                                      Objectives

                                                                      To model the effects of achieving the UK National Air Quality Strategy objectives on mortality and morbidity.

                                                                      Design

                                                                      Modelling that combines epidemiological evidence with policy options to influence determinants of disease.

                                                                      Setting

                                                                      City of Westminster.

                                                                      Main outcome measures

                                                                      Estimated lives no longer shortened and emergency cardiac and respiratory hospital admissions no longer caused or brought forward.

                                                                      Background

                                                                      Westminster City Council has declared an air quality management area: levels of PM10 and NO2 are predicted to exceed the UK National Air Quality Strategy targets unless additional action is taken. I have developed spreadsheets to enable local or health authorities to calculate the health effects of reducing particulate and NO2 pollution in their area.

                                                                      Method

                                                                      I derived three models, representing minimum, maximum and an intermediate fall in daily pollution to achieve the UK objectives. Using 1996–98 ONS and HES data, I modelled the effects of achieving various reductions in ambient concentrations of PM10 and NO2 from 1996–1998 ambient levels. Results are based on effect estimates from the most recent time series studies in London and on WHO meta-analyses.

                                                                      Results

                                                                      The minimum estimate for lives no longer shortened when the 24 hour PM10 objective for 31 December 2004 is met was 1 per year (London or WHO) when particulate levels are “capped” at that concentration. This increased to 4 (London) to 11 (WHO) when the percentage fall in concentration each day is the same as the reduction needed for the highest non-permitted level. Based on London studies, reaching the 31 December 2009 24 hour PM10 objective results in 1 (minimum) to 7 (maximum) lives no longer shortened. Using the WHO meta-analysis, the figures are 3 (minimum) to 18 (maximum) lives. Half are cardiovascular and half respiratory deaths. Achieving the PM10 annual objectives of 40 μg/m3and 20 μg/m3 (gravimetric) would delay between 2 and 8 deaths respectively (London) or 4 and 21 (WHO). Reducing the NO2 annual mean to 21 ppb would delay 9 (London) to 44 (WHO) deaths. Based on London studies, reducing PM10 to the higher annual mean objective (40 μg/m3) delays or prevents 4–6 respiratory and 4 cardiovascular emergency hospital admissions. Reaching 20 μg/m3 (gravimetric) affects 20–32 respiratory and 17 cardiovascular admissions.

                                                                      Conclusion

                                                                      It is possible to use modelling to estimate the health impacts of achieving environmental targets and to compare different strategies. With an assumption of no threshold, greater reductions are obtained when air pollution is reduced each day than with a model that removes the exceedences alone.

                                                                      The resurgence of tuberculosis in an industrialised city in northern England: the relative effects of poverty, ethnicity and crowding

                                                                      Since the mid-1980s, several industrialised countries have seen a resurgence in tuberculosis (TB). In England, the city of Liverpool has experienced one of the fastest growing rates for the disease. This study overcomes some of the limitation of routinely published TB statistics by using individual level data from the city's notification register. It presents a ward level ecological analysis of standardised annual TB notification rates between 1974 and 1995; categorised by age, sex, type of TB and ethnic group. Mapped distributions of relative risks illustrate the cross sectional variation in disease and how this has changed over time. The relation between TB and various socioeconomic, demographic and ethnic measures is analysed using Poisson regression modelling techniques. The relative contributions to disease levels made by exogenous infection and endogenous re-activation are discussed.

                                                                      For all TB categories, a cluster of high relative risk was evident in inner city wards. In contrast, mapped distributions of trends over time indicated that most of the increase in TB has occurred elsewhere in the city. Cross sectional Poisson models confirmed the known strong relation between TB and poverty. For indigenous white and ethnic minority groups alike, the Jarman index constituted the best predictor of variation in disease. For men under 65 years, an interaction effect was found between poverty and the proportion of ward level population from the Indian subcontinent. In more affluent wards TB was negatively associated with Indian subcontinent population but, as poverty increased, the relation became increasingly positive. Models of trends indicated that the rise in TB has occurred, not in the poorest parts of the city, but in those wards with relatively high population densities and low levels of household overcrowding. For those over 65 years, increased relative risk was characterised by those areas that have seen least improvement in the level of household overcrowding. Explanations for the observed relations include poverty related immunosuppresence, population mixing and migration because of slum clearance and re-housing programmes.

                                                                      A population-based study of the impacts of exposure to environmental ultraviolet B radiation on blood pressure

                                                                      Objectives

                                                                      To determine the geographical relation between exposure to environmental ultraviolet B radiation, blood pressure, and mortality from hypertension related conditions in populations.

                                                                      Design

                                                                      A geographical study using aggregate data.

                                                                      Setting

                                                                      Participating centres in the international INTERSALT study, and populations of local authority districts in Great Britain.

                                                                      Main outcome measures

                                                                      Measures of diastolic and systolic blood pressure, and mortality from ischaemic heart disease (ICD410–414) and cerebrovascular disease (ICD430–438).

                                                                      Results

                                                                      For the INTERSALT centres, mean population blood pressure was positively associated with estimated population exposure to ultraviolet B radiation (reduction in mean diastolic pressure for a 200 000 J/m2 increase in annual environmental ultraviolet B radiation = 1.9 mm Hg, p<0.001). Within Great Britain a similar positive association was found between exposure to ultraviolet B radiation and mortality from ischaemic heart disease (reduction in adjusted relative risk for a 50 000 J/m2 increase in annual environmental ultraviolet B dose = 0.13, p<0.001) and cerebrovascular disease (reduction in adjusted relative risk for a 50000 J/m2 increase in annual environmental ultraviolet B dose = 0.08, p=0.005).

                                                                      Conclusions

                                                                      There is evidence that exposure to environmental ultraviolet B radiation may be protective against hypertension and related conditions within populations. The mechanism by which this protection is afforded is not fully understood, but seems to be associated with the process by which levels of 1,25-dihydroxyvitamin D are regulated within the body. The generally held view of the negative health impacts of exposure to environmental ultraviolet B radiation could require a fundamental reassessment.

                                                                      The health consequences of frequent job changes

                                                                      Background

                                                                      The health consequences of frequent job changes have received little attention. Such employment histories are becoming more common in an increasingly flexible labour market. It is hypothesised that an employment history characterised by frequent job changes will be associated with poorer health. Psychosocial stress, and its influence on health linked behaviours and physiological factors, will be considered as a possible mediator in such a relation.

                                                                      Design

                                                                      Cross sectional study.

                                                                      Setting

                                                                      27 workplaces in the west of Scotland.

                                                                      Participants

                                                                      5399 men and 945 women in paid work, interviewed between 1970 and 1973.

                                                                      Main measures

                                                                      Number of job changes since leaving school (categorised as 0, 1–2, 3, 4–5, 6–8, 9+ changes), “Reeder” stress inventory, health behaviours (cigarette smoking, alcohol consumption, physical exercise), physiology (diastolic blood pressure, body mass index, forced expiratory volume, plasma cholesterol concentration), and current health (angina and myocardial ischaemia).

                                                                      Results

                                                                      The statistics presented here are odds ratios (OR), or differences in means where units are given, for each increase in job change category, adjusted for age, lifetime socioeconomic position and time since most recent job change. The 95% confidence intervals (95% CI) are also provided. Frequent job changes were associated with greater psychosocial stress for women (OR 1.25, 95% CI 1.11, 1.40) but not men (OR 1.03, 95% CI 0.99, 1.09). There was no association between frequent job changes and physiological measures for females, while for men there was evidence of lower diastolic blood pressure (−0.32 mm Hg, 95% CI −0.56, −0.09), cholesterol (−1.74 mmol/l, 95% CI −2.62, −0.85) and perhaps lung function (−0.42%, 95% CI −0.87, 0.02). There was strong evidence for a link between frequent job changes and unhealthy behaviour. For men there were greater odds of being a smoker (OR 1.05, 95% CI 1.00, 1.10), and of drinking more than 22 units of alcohol per week (OR 1.11, 95% CI 1.05, 1.18). Similar relations were observed for women. There was weak evidence of an association between frequent job changes and a lower risk of ischaemia in males (OR 0.91, 95% CI 0.83, 1.00) but perhaps a higher risk in women (OR 1.12, 95% CI 0.90, 1.41). There was no effect on angina for men or women.

                                                                      Conclusions

                                                                      There was little evidence for the expected relation between frequent job changes and poor health despite an association of more frequent changes with high stress in women and greater cigarette and alcohol consumption in both sexes. An intensified healthy worker effect among frequent job changers will be proposed as a possible explanation of these findings.

                                                                      Do parental occupations involving social mixing and infectious contacts affect the risk of childhood type I diabetes mellitus?

                                                                      Objective

                                                                      To investigate the hypothesis that increased exposure to infections, through parental jobs involving high levels of social mixing, reduces the risk of childhood type I diabetes.

                                                                      Design

                                                                      Two population-based case-control studies of children diagnosed with type I diabetes.

                                                                      Setting

                                                                      Yorkshire and Northern Ireland.

                                                                      Subjects

                                                                      220 cases and 433 controls from Yorkshire (aged 0–15 years); 189 cases and 465 controls from Northern Ireland (aged 0–14 years).

                                                                      Main outcome measures

                                                                      Associations between parental occupational social mixing and childhood type I diabetes were assessed using odds ratios (OR) adjusted for age and sex. For each OR, 95% confidence intervals (95% CI) and two sided tests of statistical significance were obtained. Analyses were performed by level of parental occupational social mixing and age at diagnosis.

                                                                      Results

                                                                      Parental occupations were coded using a standard occupational classification and each job allocated to high, medium or low levels of social mixing according to a predefined categorisation. One hundred and six (29%) occupations out of 371 were identified as having potentially increased levels of social mixing: 75 classified as “high” and 31 as “medium”. The remaining 265 (71%) occupations, whose likely social mixing was not judged to be unusual, classified as low. Because of small numbers within the medium exposure group (Yorkshire— mothers: 12 cases and 22 controls; fathers: 13 cases and 21 controls: Northern Ireland—heads of household, usually the father: 8 cases and 27 controls), low and medium exposure levels were combined. Childhood type I diabetes was not associated with high levels of parental occupational social mixing (Yorkshire—mothers: OR 1.07, 95% CI 0.76, 1.50, based on 88 exposed cases; fathers: OR 1.15, 95% CI 0.75, 1.76, based on 41 exposed cases; Northern Ireland—heads of household, usually the father: OR 0.78, 95% CI 0.49, 1.25, based on 27 exposed cases). A larger proportion of mothers (39%) compared with fathers (18% Yorkshire, 17% Northern Ireland) had jobs involving high levels of social mixing. Mothers with high social mixing jobs conferred a non-significant reduced risk of diabetes among children diagnosed under 5 years (OR 0.58, 95% CI 0.24, 1.38) compared with those diagnosed at age 5 and over (OR 1.14, 95% CI 0.77, 1.69).

                                                                      Conclusion

                                                                      No association between parental occupational social mixing and childhood type I diabetes was detected for all ages combined. Mothers were more likely to have jobs involving high levels of social mixing than fathers. The possible protective effect of maternal high occupational social mixing on children diagnosed under 5 merits investigation.

                                                                      Comparing the length of NHS waiting times: can we trust official statistics?

                                                                      Objectives

                                                                      To compare two measures of the promptness of elective admission, namely the proportion of valid elective episodes “admitted” within three months and the likelihood of elective admission within three months. To assess the extent to which ranked performance on one measure predicts ranked performance on the other.

                                                                      Methods

                                                                      We obtained information on each elective episode with a date of enrolment other than “15 Oct 1582” and a date of admission for trauma and orthopaedic surgery at each of 34 NHS Trusts in South Thames Region between 1 July and 31 December 1994 inclusive. We calculated the proportion of valid elective episodes “admitted”within three months. We also obtained the KH06, KH07 and KH07A counts submitted for these waiting lists for the quarters ending 30 September and 31 December 1994. We calculated the proportion eventually admitted—that is, the proportion of waiting times no longer eligible for inclusion in the waiting list census as a result of elective admission. Finally, we calculated the likelihood of elective admission within three months, among all those at risk, as the product of these two proportions.

                                                                      Results

                                                                      The proportion of elective admissions occurring within three months of enrolment ranged from 0.62 to 0.27. The proportion eventually admitted ranged from 0.93 to 0.31. As a result, the likelihood of elective admission within three months, among all those at risk, ranged from 0.55 to 0.12. This measure confirms that elective admission may be very much less prompt than suggested by the Government Statistical Service estimate and that ranked performance may be very different from that suggested by the official statistics.

                                                                      Conclusion

                                                                      The published statistics allow those already admitted to look back over their shoulder and assess how typical their experience was. And the published statistics allow those destined to be admitted to assess their chance of admission within any given “time since enrolment”. But patients, clinicians, managers and politicians all want to know how long new recruits might expect to wait and cannot predict whose “time since enrolment” will end in admission and whose “time since enrolment” will end in some competing event. As a result, the published caveats fail to protect users from misinterpreting official statistics: the proportion of valid elective episodes admitted is of no interest unless it estimates how long new recruits might expect to wait so users assume that this is what it does!

                                                                      Health variations: a qualitative study of the reasons for differences in health service use and access to services

                                                                      Background

                                                                      This qualitative study is part of a research programme investigating inequalities as they relate to the use of, and access to, health services.

                                                                      Aims and objectives

                                                                      To develop understanding of reasons for differences in the use of, and access to, health services; to generate questions for a population survey in order to assess the relative importance of the reasons.

                                                                      Method

                                                                      Fourteen focus groups, involving purposive recruitment of 103 participants from the general public and health care occupations, were conducted to explore people's various experiences and knowledge of the health service and discuss key issues and factors they consider pertinent to differences in use and access.

                                                                      Analysis

                                                                      Verbatim transcripts from each session were coded independently and then together. Data were indexed manually, as well as by computer, using the qualitative data analysis software Atlas.ti. Similar themes and issues within and across sessions were grouped together, then associations and relations between the categories were identified. The detail, sequence, emphasis and consensus/divergence within the discussions was emphasised rather than just frequency, as befits the methodology.

                                                                      Results

                                                                      Prior health related experiences; attitudes and beliefs (general and health related); knowledge, skills and abilities (general and health related); and social/economic circumstances; operating within a service context (organisation and quantity), combine together to influence use of services (by the public) or provide access to them (by the professional). Each category involves many elements, with participants/groups emphasising different issues with varying levels and directions of influence. Some elements relate to each other across categories: for example, some circumstantial factors directly influence use of services—such as access to/ownership of a car—whereas others, such as poverty, influence use indirectly, via relations to attitudes to life. Most importantly it is experience of health care encounters that influences use and access: the participants' approach, either as user or provider, strongly influences the response of the other; interpersonal aspects of the professional's approach are identified as especially crucial, and more so than technical ones. For example, experiences of an encouraging/discouraging health professional strongly impact on decisions relating to whether, how and when a person uses the services subsequently; similarly, experiences of assertive patients and/or perceptions of motivation influence professionals' decisions to provide treatment and/or referral. Diagrammatic representations of the key issues and processes in use of and access to services have been developed, which will also be presented.

                                                                      Open SESAME—the impact of socioeconomic status on health care seeking behaviour

                                                                      Objectives

                                                                      The study aims to explore how variations in health care seeking behaviour can hinder the goal of equitable health care access, specifically testing the hypothesis that sociodemographic factors such as socioeconomic status, gender and age influence an individual's perception of the need and urgency for seeking health care.

                                                                      Design

                                                                      1500 people aged 30 to 80 years, from a general practice in the south of England were sent a questionnaire asking about sociodemographic characteristics, attitudes to health and health care, and intended response to a clinical vignette about a lump under the armpit. Initially it is non-tender (part I) but then a few weeks later it is associated with night sweats and a throbbing pain (part II). The practice was selected because it contained an even social mix according to census data.

                                                                      Main outcome variables

                                                                      The main outcome measure was the proportion indicating that they would seek immediate care (hospital emergency department or general practitioner) for the symptoms described in each case scenario.

                                                                      Results

                                                                      The final sample was 1287 and the questionnaire response rate was 911 (70.8%). The social class distribution was: SCI and II 31%, IIINM 25%, IIIM 19% and IV and V 24% and mean age of respondents 53 years (range 30–84). A five group socioeconomic ordinal scale was created based on a composite of several measures. For part I of the vignette, the age and sex adjusted odds ratio (OR) for trend across the scale was 1.29 (95% confidence intervals (95% CI) 1.13, 1.46; p<0.001), so that lower socioeconomic status was associated with greater reporting of seeking immediate care. However, for part II, this had now disappeared (OR for trend across groups 1.02, 95% CI 0.93, 1.14). Other variables that were strong predictors of immediate care seeking were being older, high degree of anxiety, agreeing that good health is very important and agreeing that one should always do what doctors say. Multivariable analysis, however showed that adjustment for these variables attenuated, but did not abolish the association (OR 1.19, 95% CI 1.03, 1.38, p=0.02).

                                                                      Conclusion

                                                                      These data suggest that lower socioeconomic status is associated with increased care seeking behaviour. It is important not to over-generalise these results as different symptom clusters appear to alter social patterns of health seeking behaviours. However, social class in itself was weaker than other psychosocial variables or the degree of anxiety generated by the vignette. These factors only partially explained the social gradient.

                                                                      Age-based decision making: Is it rationing?

                                                                      Background

                                                                      Age Concern claim that older patients are not receiving their fair share yet NHS policy has never explicitly sanctioned rationing on the basis of age. The distinction between “implicit” age-based rationing and age discrimination is not clear cut. Empirical evidence to fuel the debate is lacking.

                                                                      Objectives

                                                                      To use routine data sources to explore the possibility of age related decision making in the hospital management of colorectal cancer and to ascertain whether there are differences that cannot be explained on clinical grounds.

                                                                      Design

                                                                      Retrospective analysis of linked Scottish cancer registry and hospital discharge data for colorectal cancer (1992–1996).

                                                                      Main outcome measures

                                                                      Histological verification was used to indicate the “gold standard” of investigation. Definitive surgery and chemotherapy were used as indicators of treatment received.

                                                                      Results

                                                                      After adjustments for demographic factors, tumour sub-site, comorbidity and type of admission, increasing age was associated with decreased histological verification, surgery and chemotherapy.

                                                                      Conclusion

                                                                      Although all factors could not be accounted for, it was possible to conclude, drawing on published literature, that clinical factors were not the sole determinants of care. It is suggested that decisions on the grounds of chronological age may be being made by clinicians because of beliefs and expectations about “quality of life” in older people that are out of date with the present evidence base. Furthermore, it is thought that, where the basis of the decision is belief and expectation rather than economic pressure, withholding treatment without clinical justification should be seen ethically as discrimination rather than as rationing.

                                                                      An introduction to DIPEx—a database of individual patient experience

                                                                      Objectives

                                                                      DIPEx is a multimedia web site and CD ROM that links patients' experiences with evidence-based information about treatments and the illness itself, with a range of other resources that may be useful, including support groups and links to other web sites. DIPEx aims to identify the questions that matter to people when they are ill and may be used for informing patients, educating health care professionals and providing a patient centred perspective to researchers and those who manage the health service. An important feature of the database is that it addresses the needs of professionals and the general public. The same database will be available for patients, carers, professionals and students, policy makers and researchers. In this paper the first complete DIPEx site, for hypertension, will be demonstrated.

                                                                      Methods

                                                                      A purposive sample was chosen to represent the widest practical range of experiences of hypertension. Volunteers for the project were sought through GPs, support groups, radio broadcasts and newspaper articles. Interviews took place in the respondents' home. Experienced qualitative researchers conducted semi-structured, narrative interviews with 40 respondents, recorded on digital video or audio tape. The information about treatments and additional resources was compiled by a steering group including lay and professional experts.

                                                                      Results

                                                                      The analysis used grounded theory to identify the issues that are important to the respondents. Themes are illustrated with video, and audio clips and written excerpts from the interviews. The information about hypertension and the answers to “frequently asked questions” on the site have been developed in response to the information needs of the interviewees. These include questions about prevention, causes and effects of hypertension and explanation of the meaning of blood pressure readings.

                                                                      Conclusion

                                                                      We believe that DIPEx is unique in the field of patient and healthcare communication. The demonstration of the hypertension site includes a short film about the project and video clips from the interviews.

                                                                      Family history of diabetes in UK South Asians

                                                                      Objective

                                                                      To describe the associations between diabetes mellitus (DM) and impaired glucose tolerance (IGT) in study subjects and reported diabetes in their parents and siblings.

                                                                      Design

                                                                      Cross sectional study.

                                                                      Subjects

                                                                      A stratified random sample of 1509 Newcastle residents aged 25–74 years from European (n=825), Indian (n=259), Pakistani (n=305) and Bangladeshi (n=120) ethnic groups.

                                                                      Main measures

                                                                      Self reported history of diabetes in parents and siblings of subjects.

                                                                      Results

                                                                      Among Europeans 8.3% of those with normal glucose tolerance, 13.3% of those with IGT and 18.9% of those with type II DM reported parental diabetes. Among South Asians the prevalence of diabetes in parents was similar regardless of the glucose tolerance status of the respondent (26.2%, 25.8% and 25.4% of those with normal and impaired glucose tolerance and type II DM respectively). In both Europeans and South Asians the likelihood of reporting a sibling with diabetes was least in those with normal glucose tolerance, greater in those with IGT and greatest in those with type II DM. Among Europeans diabetic subjects were almost five times as likely as those with normal glucose tolerance to report a sibling with diabetes (17.9% versus 3.8%), while among South Asians diabetics were less than twice as likely to report a sibling with diabetes (17.8% versus 9.3%). The same overall patterns were observed in Indians, Pakistanis and Bangladeshis, and when male and female respondents were analysed separately.

                                                                      Conclusions

                                                                      An association between the glucose tolerance status of the respondent and the frequency of diabetes in parents and siblings is consistent with a genetic basis for diabetes. An association with sibling but not parental diabetes is consistent with changing environmental risk factors, as the exposure status of respondents would be expected to resemble their siblings more than their parents. Our data suggest that risk factors for diabetes in UK South Asians are more likely to be shared with their siblings than with their parents. In explaining the well recognised familial pattern of type II DM, environmental influences among UK South Asians may be more, and genetic factors less, important than is usually acknowledged.

                                                                      Smoking and risk of type II diabetes

                                                                      Introduction

                                                                      It is known that smoking causes insulin resistance. However, the relation of smoking to risk of type II diabetes has not been well defined.

                                                                      Aim

                                                                      To estimate the relative risk of type II diabetes in smokers relative to non-smokers and to assess the effect of smoking cessation on diabetes risk.

                                                                      Design

                                                                      Prospective cohort study involving a group of 7735 middle aged men followed up for an average period of 16.8 years.

                                                                      Results

                                                                      Baseline data on smoking status were available from 7124 men. Upon exclusion of known diabetics at screening, and those with recall of ischaemic heart disease and stroke, there were 290 incident cases of diabetes in this group of 7128 men during follow up. All current smokers combined had a significantly increased risk of diabetes compared with never smokers, relative risk (RR) (95% confidence intervals (95% CI)) 1.7 (1.2, 2.4), adjusted for body mass index and other potential confounders including physical activity, social class, alcohol intake, pre-existing coronary heart disease and antihypertensive treatment. Primary pipe/cigar smokers showed similar risk to never smokers but secondary pipe/cigar smokers showed significantly higher risk than never smokers, adjusted RR 1.9 (95% CI 1.1, 2.8). Ex-smokers showed lower risk than current smokers and diabetes risk tended to decrease with increasing years since quitting. After adjustment for age, body mass index and other confounders the benefit of giving up smoking was apparent in those who had given up at least five years before to screening. Diabetes risk reverted to that of never smokers in those who had given up at least 20 years before screening.

                                                                      Conclusion

                                                                      The findings suggest that smoking is an independent and reversible risk factor for type II diabetes.

                                                                      Work disability in patients with rheumatoid arthritis

                                                                      Background

                                                                      There are strong links between the development of rheumatoid arthritis (RA) and the ability to remain in work. Work loss often occurs early in the disease process. We previously reported work disability rates of 14% at one year and 28.6% at three years from RA onset. The move to earlier, more aggressive treatment in RA may have resulted in patients being able to remain economically active.

                                                                      Objectives

                                                                      (1) To re-examine the magnitude of work disability in a new community based cohort of patients with recent onset RA. (2) To look at the further loss of employment in the original cohort, followed up for up to 10 years.

                                                                      Subjects

                                                                      The Norfolk Arthritis Register (NOAR)—a primary care based inception cohort of patients with inflammatory polyarthritis (IP) with longitudinal follow up.

                                                                      Methods

                                                                      (1) 134 consecutive patients (52 male, 82 female) with symptom onset in 1994–1997, who met the 1987 ACR criteria for RA and were economically active at onset, were followed up by NOAR for at least two years from onset. Occupational and clinical data were collected annually. Changes in employment status were analysed in 1999 —a mean of 48 months from onset. (2) The original cohort of 160 working RA patients with an onset in 1989–1992 were followed up until 1999.

                                                                      Results

                                                                      (1) 81 of the 134 cases (60.4%) were still in employment, three had stopped working for non-health reasons and 50 (37.3%) were work disabled because of RA. Some 26.6% had stopped within 12 months of onset and 32% within two years. The mean interval between onset and stopping work was 11.4 months. (2) A further 10 (6.2%) of the original cohort had stopped work prematurely because of RA.

                                                                      Conclusions

                                                                      Work disability is still a major outcome in RA, occurring soon after onset and frequently before second line treatment can become effective. People continue to lose their jobs because of RA after the first few years of disease. There is a need for early advice and education for patients regarding employment.

                                                                      Associations between self reported physical activity and a measure of heel bone ultrasound in EPIC—Norfolk

                                                                      Objectives

                                                                      To study associations between self reported physical activity and ultrasound measurement of heel bone density in men and women.

                                                                      Design

                                                                      A cross sectional population-based study.

                                                                      Subjects

                                                                      2296 men and 2914 women, aged 45–74 years who were registered with general practices in Norfolk, UK and surveyed as part of the EPIC—Norfolk study between 1995 and 1998.

                                                                      Methods/main outcome measure

                                                                      Time spent participating in recreational physical activity was calculated for four groups that were defined according to the level of impact from questionnaire data. The questionnaire also quantified stair climbing and frequency of inactivity (time spent viewing television/video). At least two measurements of bone density, per foot, at the heel were recorded by broadband ultrasound attenuation (BUA). BUA has previously been shown to predict hip fracture risk.

                                                                      Results

                                                                      Self reported time spent in high impact physical activity was strongly and positively associated with heel bone density, independently of age, weight and other confounding factors. Men who reported participating in ⩾2 hours per week of high impact activity compared with men who report no such participation, had 8.40 dB/MHz (95% confidence intervals (95% CI) 4.49, 12.32) higher heel bone density. In women, the difference in heel bone density between reporting any versus no time spent in high impact activity was 2.36 dB/MHz (95% CI 0.42, 4.31). The size of this effect was equivalent to that of four years in age. There was no significant association between time spent in moderate impact activity on heel bone density in either men or women. This null association persisted when the analysis was restricted only to people who did not participate in high impact activity. For each additional five flights of stairs climbed per day the increase in heel bone density was 0.68 dB/MHz (95% CI 0.22, 1.13), p<0.005 for women. There was a significant negative association between increasing televison/video viewing hours per week and heel bone density in women. The effect of each additional hour of television/video viewing per week was −0.08 dB/MHz (95% CI −0.14, −0.02), p<0.009. These associations were independent of possible confounding factors; age, weight, height, cigarette smoking habit and hormone replacement therapy in women.

                                                                      Conclusion

                                                                      This cross sectional study demonstrates an independent relation between high impact physical activity and a measure of bone density (by BUA) in men and women. Interventions to increase participation in these activities may have important public health consequences.

                                                                      Barriers to utilisation of total joint replacements

                                                                      Introduction

                                                                      While total joint replacement (TJR) is an effective treatment for severe joint disease, research evidence consistently shows variation in provision and thus potential for unmet need. We investigated barriers to appropriate utilisation of TJR.

                                                                      Methods

                                                                      Published literature from standard electronic databases concerned with treatments for osteoarthritis was reviewed. Consensus panels including primary care physicians, rheumatologists and orthopaedic surgeons were formed to explore expert views about barriers to utilisation. In depth interviews were undertaken with people with severe hip/knee disease and low levels of service utilisation selected from a community prevalence survey (Somerset and Avon Survey of Health) to explore why they had not sought help. Interviews were audiotaped and transcribed. Analysis was according to the method of constant comparison.

                                                                      Results

                                                                      Evidence of inequality in the provision of TJR was found in the literature, with wide variations in surgical rates nationally and internationally. Expert discussions and literature identified three potential barriers to appropriate utilisation: people with joint disease not presenting to primary care physicians, primary care physicians not referring people to specialists, and surgeons refusing to operate on particular groups (for example, young, obese). Recurring themes in interviews to explain why people did not seek health care included perceptions that they were too old or unwell for TJR, a view that joint problems were a normal part of aging, not wishing to undergo surgery, poor TJR outcomes in friends/relatives, and unwillingness to initiate reconsideration for referral/surgery after previous refusal.

                                                                      Conclusions

                                                                      There is the potential to develop strategies that might reduce inequality and unmet need, including referral and review guidelines for primary care physicians. It is also clear that there is a need for further research to establish appropriate indications for the timing of TJR that incorporate patient preferences and clinical factors.

                                                                      Trends in mortality and hospitalisation following a first acute myocardial infarction: a population-based study

                                                                      Objectives

                                                                      To examine the trends in: (a) population mortality rates following a first acute myocardial infarction (AMI) in those people who did not survive to reach hospital and (b) the proportion of people experiencing a first AMI who were admitted to hospital in Scotland between 1986 and 1995.

                                                                      Study design and setting

                                                                      Population-based study in Scotland (population 5.1 million).

                                                                      Subjects

                                                                      All 208 527 men and women experiencing a first AMI in Scotland between 1986 and 1995, including the 117 749 people who were admitted to hospital, plus the 90 778 people who did not survive to reach hospital. (A first AMI was defined as ICD9 code 410 with no prior hospitalisation for AMI since 1981).

                                                                      Results

                                                                      Between 1986 and 1995, a total of 48 481 men and 42 297 women had a first AMI and did not survive to reach hospital. Overall, population-based death rates increased with age. Thus, in 1986 the death rates in men and women aged <65 years were 73 and 24/100 000 population respectively, rising with age to 1930 and 1210/100 000 respectively in those aged >74 years. Significant declines were observed in all age groups between 1986 and 1995, and were greatest in men and the young. Thus, death rates halved in men aged <65 years; (from 73 to 36/100 000). The proportion of people who survived to reach hospital increased over this period. Although this proportional increase occurred in men and women in all age groups, it was most evident in younger cohorts. According to multivariate analysis performed separately for men and women, year of admission, age and extent of social deprivation were all significant predictors of probability of admission to hospital.

                                                                      Conclusion

                                                                      Population-based death rates following a first AMI, but without hospitalisation, declined significantly between 1986 and 1995. This trend was most evident in men and younger age groups. These data are consistent with the impact of primary prevention.

                                                                      The association between socioeconomic deprivation and the management of acute myocardial infarction and survival at two years: a prospective observational study

                                                                      Background

                                                                      Coronary heart disease is a major factor in the widening social divide in health. The decline in death rates since the 1970s has been least marked in lower socioeconomic groups; variations in lifestyle and environmental factors are likely to account for most of this difference. With the advent of effective treatments such as thrombolysis, ACE inhibitors and aspirin, the management of acute myocardial infarction (AMI) may be an additional factor contributing to socioeconomic variation in outcome.

                                                                      Objective

                                                                      To investigate the association between socioeconomic deprivation and the management of, and survival after, AMI.

                                                                      Design

                                                                      Prospective observational study collecting demographic and clinical data on all cases of AMI admitted to hospitals in Yorkshire.

                                                                      Setting

                                                                      Acute admitting district and university hospitals in the Yorkshire region of northern England.

                                                                      Participants

                                                                      3684 consecutive patients with a possible diagnosis of AMI admitted to hospitals in Yorkshire between 1 September and 30 November 1995.

                                                                      Main outcome measures

                                                                      Types of ward on admission, treatments received in hospital, survival status at discharge from hospital and two years, by quintiles of Townsend score.

                                                                      Results

                                                                      2153 people had a confirmed AMI. Patients from all Townsend quintiles waited a similar time before seeking medical help. After adjusting for clinical variables and hospital of admission, deprived patients were less likely to be admitted directly to a coronary care unit (adjusted odds ratio (OR) 0.96, 95% confidence intervals (95% CI) 0.93, 0.99, p=0.005 per unit of Townsend score) and more likely to attend accident and emergency. Once in hospital, deprivation did not influence treatment with aspirin or thrombolysis, treatment on discharge or investigations planned after discharge. No relation was seen between deprivation and death before discharge (adjusted OR 0.99, 95% CI 0.95, 1.03, p=0.60) or two years (adjusted OR 1.01, 95% CI 0.98, 1.06, p=0.41).

                                                                      Conclusions

                                                                      If social inequities in the management of AMI in secondary care do exist, they are not an important contribution to the social divide in coronary heart disease outcomes. However, there may be important differences in the primary care of AMI that merit further attention.

                                                                      Identifying the risk factors and indicators of ischaemic heart disease in primary and secondary care: is there a consensual model?

                                                                      Background

                                                                      Medical research has identified numerous cues that are associated with ischaemic heart disease (IHD), but no one clear pathogenic predictor. Given time and resource constraints in clinical practice, diagnosis is likely to be based upon a subset of these: those a physician considers to be the most predictive of IHD. It is not known which factors are most influential in reaching a diagnosis, nor whether there are systematic differences between primary and secondary care physicians.

                                                                      Objectives

                                                                      Our overall aim was to examine the degree of consensus between and within primary and secondary care physicians' subjective models of diagnosis of IHD. Our first objective was to elicit those cues considered by physicians in the diagnosis of IHD. Our second objective was to ascertain the strength of association between these cues and IHD. Our third objective was to examine the interrelation between these cues.

                                                                      Method

                                                                      The study was divided into two experiments. The first experiment used an open-ended cue identification task. Sixty primary and secondary care specialists were asked to identify the cues, which in their view were predictive of a diagnosis of IHD. Each cue was also subjectively weighted—that is, the strength of association between the cue and IHD was rated. The second experiment used a network diagram technique to elicit each person's causal model of IHD. Physicians formulated a causal (belief) diagram identifying the relations between cues and the strength of the identified associations.

                                                                      Results

                                                                      Seventy two different cues were identified by at least one physician as being associated with IHD. Both primary and secondary care physicians consistently identified six risk factors (angina, diabetes mellitus, hyperlipidaemia, hypertension, smoking and a positive family history of IHD) and four positive test results (ECG, ETT, angiography and thallium scan) as predictive of IHD. Although these six cues were both rated highly and identified by most people, and cues with low ratings tended to be identified by few people, there were some cues that only a few people identified but which were rated very highly.

                                                                      Conclusion

                                                                      This is the first study to use causal modelling to examine the role of information used by primary and secondary care physicians in the diagnosis of IHD. Although our model showed agreement on the core cues involved in a diagnosis, we have also identified cues that could lead to a difference in opinion with any one patient.

                                                                      Is “Saving Lives” feasible? Estimating the potential for further reductions in UK coronary heart disease deaths

                                                                      Objective

                                                                      To assess the potential to reduce coronary heart disease (CHD) death rates in the UK.

                                                                      Background

                                                                      The recent public health white paper “Saving Lives: Our Healthier Nation”, sets a target of a 40% (200 000) reduction in CHD deaths in people under 75 between 1997 and 2010. Is this feasible through further risk factor reductions?

                                                                      Settings

                                                                      (a) Scottish population of 5.1 million, (b) UK population of 59 million.

                                                                      Methods

                                                                      A previously validated cell-based mortality model combining effectiveness data from published meta-analyses with available information on: (a) uptake of all CHD treatments in all patient categories, (b) risk factor trends (smoking, blood pressure, cholesterol, deprivation) by sex/age group. Applying data from the WHO MONICA (Monitoring of Trends and Determinants in Cardiovascular Disease) project and elsewhere, the model was used to estimate the additional deaths that might have been prevented by a variety of plausible scenarios, such as if: (1) reductions in smoking prevalence in Scotland had been as great among women as among men; (2) reductions in population mean cholesterol level in Scotland had been as great as in Sweden; (3) reductions in population mean diastolic blood pressure had been 50% higher than those observed (an additional 4 mm Hg over all age/sex groups in the model). The robustness of the model results to uncertainties surrounding all key parameters was examined by extensive sensitivity analyses.

                                                                      Results

                                                                      Between 1975 and 1994, smoking prevalence in Scotland declined by 60% in men and 34% in women. In contrast, population mean cholesterol levels and blood pressure declined by only approximately 5% and 9% respectively. These observed risk factor changes explained approximately 50% of the deaths prevented in 1994 compared with 1975. If the reduction in smoking prevalence among women had equalled that in men, several hundred further deaths might have been prevented in 1994, mostly in older age groups (over 75). Significant additional deaths might have been prevented by further blood pressure reduction, and death rates could have been reduced substantially if the decline in population mean cholesterol in Scotland had mirrored that in Gothenberg, Sweden. Comparable reductions in CHD deaths were seen when the model was extended from 1994 to 2010.

                                                                      Conclusions

                                                                      Cautious extrapolation to the UK population of 59 million suggests that if comparable risk factor reductions were achieved, approximately 20 000 further CHD deaths per annum might be prevented. Given that cardiological treatments should also have a substantial impact, the “Saving Lives” target seems entirely feasible.

                                                                      Is exercise an effective intervention for the management of depression? A systematic review

                                                                      Background and Objective

                                                                      Depression is common and an important cause of morbidity and mortality worldwide. Despite the availability of effective pharmacological interventions much depression remains inadequately treated and compliance with antidepressant medication is poor. An effective alternative treatment that might be more acceptable to patients would therefore be beneficial. Research into the effect of exercise on depression goes back several decades and there is a growing body of literature on this subject. This review examines the effectiveness of exercise as an intervention for the management of depression.

                                                                      Design

                                                                      Systematic review of all randomised controlled trials, in any language, obtained using five electronic databases, (Medline, Embase, PsycLit, Sport discus, Cochrane Library) contact with experts in the field, bibliographic searches and hand searches of recent copies of relevant journals. Meta-analysis and meta-regression of trials for which complete data were obtained.

                                                                      Main outcome measures

                                                                      Standardised mean difference in effect size and weighted mean difference in Beck Depression Inventory (BDI).

                                                                      Results

                                                                      16 articles referring to 14 studies met our inclusion criteria. The key results are: (a) There is a paucity of good evidence. All studies had important methodological weaknesses with randomisation being adequately concealed in only three and intention to treat analysis undertaken in only two. (b) The majority (nine) of the studies were of non-clinical community volunteers sometimes with financial or other incentives to participate and complete. In most of these an adequate diagnosis of clinical depression was not made. (c) Most studies did not present data to enable statistical pooling or confirmation of author's conclusions. After contact with authors adequate data were available for 12 of the 14 studies. (d) Exercise is efficacious, when compared with no exercise, in reducing depressive symptoms—standardised mean difference −1.1 (95% confidence intervals (95% CI) −1.5, −0.6), weighted mean difference in outcome BDI = −7.3 (95% CI −10.0, −4.6). (e) There were systematic differences (heterogeneity) between studies that were not explained by study quality, setting or depression severity but were explained by publication type and length of follow up. (f) Exercise has similar efficacy to cognitive therapy in reducing depressive symptoms−standardised mean difference=−0.3 (95% CI −0.7, +0.1).

                                                                      Conclusions

                                                                      Exercise is efficacious in the management of depressive symptoms but these results may be exaggerated by the inclusion of two conference abstracts in the analysis and the short-term follow up of studies. The clinical effectiveness of exercise in the treatment of depression cannot be determined with currently available evidence.

                                                                      Explanations for the rise in youth suicide: a European perspective

                                                                      Background

                                                                      Suicide rates have doubled in men aged <30 in England and Wales (E&W) since 1970, while in young women rates have declined. With the exception of the former West Germany many other European countries have experienced similar increases in young male suicide and elsewhere in Europe trends in female suicide have generally followed those in men.

                                                                      Methods

                                                                      Using age and gender specific social and economic data from four countries with different trends in youth suicide—E&W, West Germany, France and Norway —we have investigated whether changes either in social and economic conditions or in the lethality of suicide methods underlie these differing trends.

                                                                      Results

                                                                      In young men in E&W, France and Norway suicide rates increased by over 70% between 1970–1990, whereas in West Germany, over the same period, rates declined. In young women, rates decreased in E&W and West Germany but increased in both France and Norway. In all four countries there have been reductions in female overdose suicide mortality but in France and Norway these have been offset by increased use of other methods, particularly hanging. Trends in markers of social and economic conditions are broadly similar in the four countries. Between 1970–1990 levels of unemployment rose steeply in each country. In E&W and France the timing of the increase in unemployment coincided with the rise in suicide. While divorce rates have also increased markedly in all four countries, the timing of these rises differs from that for the increases in suicide in all countries except France. Marriage rates declined in all four countries from around 1970. Changes in all these risk factors have been greatest in people aged <30. There are no clear differences between the countries in trends in alcohol consumption or GDP, both of which have increased. Trends in income inequality show no consistent association with suicide trends.

                                                                      Summary

                                                                      Changes in the social and economic risk factors examined do not seem to explain differing trends in youth suicide. Changes in the lethality of methods used for suicide may have influenced trends in women. Further research is required into reasons for the discordance in suicide trends in Germany compared with other European countries, explanations are relevant to understanding the aetiology of suicide and in developing preventive strategies. Particular features of Germany in the past 50 years are postwar reconstruction, changes in its national borders and reunification in 1989. It is notable that similar reductions in youth suicide occurred in Japan 1970–1990.

                                                                      Urban-rural differences in the rise in youth suicide in England and Wales

                                                                      Background

                                                                      Suicide rates have doubled in young men over the past 30 years in most industrialised countries. Explanations for these rises are unclear, but research from Australia indicates that the steepest rises have occurred in rural areas. It is speculated that these changes reflect a decline in the rural economy and its effects on rural communities, including the out-migration of healthier people. We have investigated whether similar geographical differences in trends in suicide exist in England and Wales.

                                                                      Setting

                                                                      England and Wales, 1981–92.

                                                                      Methods

                                                                      The wards of England and Wales were categorised into four quartiles of rurality using an index of population potential derived from the 1991 census. The index is based on the populations of all wards in Britain, each weighted by its distance from the centre of the index ward. Differences in suicide rates (ICD9 codes E950–959 and E980–989 excluding E988.8) between 1981–85 and 1986–92 were calculated in each quartile for 15–24 and 25–44 year old men and women separately using routine mortality data.

                                                                      Results

                                                                      In men, between 1981–85 and 1986–92, suicide rates increased by 4.9 (95% confidence intervals (95% CI) 4.3, 5.6) per 100 000 person years in 15–24 year olds and by 1.0 (95% CI 0.4, 1.7) in 25–44 year olds. In women, the changes were 0.6 (95% CI 0.3, 1.0) and −1.4 (95% CI −1.8, −1.1) in the same age bands. In men aged 15–24, rises in suicide rates were similar in both rural and urban areas. In men aged 25–44, however, there were striking differences across the four categories of rurality—most of the increase occurred in the most rural wards: 3.0 (95% CI 1.3, 4.7) while just 0.02 (95% CI −1.0, 1.0) in the most urban wards. In women aged 15–24, there were similar marked differences depending on rurality: 1.4 (95% CI 0.5, 2.3) in the most rural and 0.2 (95% CI −0.4, 0.8) in the most urban quartile. In women aged 25–44, suicide rates decreased over the years studied across all quartiles: −2.0 (95% CI −2.6, −1.4) in urban and −1.5 (95% CI −2.5, −0.6) in rural areas. It is unlikely that changes in unemployment explain these trends, as between 1981 and 1991 the greatest increases in unemployment were experienced in urban areas (correlation between change in unemployment and population potential 0.20; p<0.001 in men and 0.33; p<0.001 in women).

                                                                      Summary

                                                                      These findings suggest that the greatest increases in youth suicide have occurred in rural areas, in particular in 15–24 year old women and 25–44 year old men. The influence of changes in preferred methods of suicide as well as other social and economic changes in rural wards will be presented at the conference.

                                                                      The impact of legislation and changing vehicle propulsion methods on suicides using motor vehicle exhaust gases across the UK

                                                                      Background

                                                                      Catalytic converters (CATs) have been compulsory on all new petrol cars sold in the European Union since 31 December 1992. As CATS reduce the level of carbon monoxide (CO) in motor vehicle emissions by approximately 86% it has been postulated to have an unintended benefit of reducing suicides using motor vehicle emissions gas. In addition, there has been a major change in the UK car fleet in the past 15 years with an increasing proportion of diesel engine cars, which generate lower levels of CO than petrol engines fitted with CATs. A coincidental decline in suicides and the introduction of CATs has been reported in US and in Scotland. However, the US studies are limited in their generalisability because of the preponderance of petrol vehicles and the access to firearms as an alternative method of suicide, and the Scottish study failed to examine the changes in the car fleet.

                                                                      Objectives

                                                                      To consider: what impact did the uptake of diesel vehicles have on suicide rates; and has any substitution of methods occurred.

                                                                      Design

                                                                      Retrospective analysis of mortality reports from the Registrar Generals (1980–1998).

                                                                      Methods

                                                                      Deaths as a result of suicide (ICD9 codes E950–E959), including undetermined nature (E980–E989), were compared with the number of vehicles powered by petrol or diesel engines. The question of substitution was examined by producing annual rates by suicide method. Linear regression was used to test the effect of increasing diesel ownership and the decline in cars without CATs on suicide trends.

                                                                      Results

                                                                      The suicide rate using motor vehicle gases (E952+E982) increased steadily to peak in 1992 at 2.51 per 100 000 before dropping to 1.30 per 100 000 by 1997, while suicides from all other methods remained relatively constant around 9.3 per 100 000. However, rates for hanging (E953+E953) have increased from 2.54 per 100 000 to 3.30 per 100 000. Initial findings indicate that the increase in diesel vehicles did not have the expected effect on the level of suicides before the introduction of CAT fitted petrol vehicles.

                                                                      Conclusion

                                                                      This study shows for the first time that substitution is occurring with a marked shift towards the use of hanging. It is interesting that the uptake of diesel vehicles did have the same impact on suicides as the introduction of CATs. This raises the question whether those who own diesel vehicles are somehow different to those with petrol ones.

                                                                      Slipping through the net—risk factors for unmet treatment need in children born with cleft lip and palate

                                                                      Background

                                                                      Children born with cleft lip and palate require long term follow up with multidisciplinary specialist treatment, including surgery, speech therapy and dentistry, from birth until adulthood. After a national audit of the outcome of cleft care the number of cleft teams in the UK is being reduced to create regional specialist centres. The risk in this strategy is that there will be reduced access to specialist services in vulnerable population groups.

                                                                      Aims

                                                                      To examine the characteristics of children born with unilateral cleft lip and palate (UCLP) in the UK who have unmet needs for speech therapy and/or dental treatment.

                                                                      Design and setting

                                                                      Cross sectional study of subjects under the care of UK cleft teams.

                                                                      Subjects

                                                                      Children born with complete UCLP between 1 April 1982 and 31 March 1984 (“twelve year olds”) and 1 April 1989 and 31 March 1991 (“five year olds”). An independent research team collected outcomes for 238 five year olds and 218 twelve year olds (71% of cases identified). Socioeconomic status was determined from postcodes.

                                                                      Main outcome measures

                                                                      Children with one or more decayed teeth needed dental treatment. Children with one or more “serious” consonant errors of speech needed speech therapy.

                                                                      Results

                                                                      Forty per cent of five year olds and 20% of twelve year olds needed dental treatment; 38% of five year olds and 20% of twelve year olds needed speech therapy. Low socioeconomic status was a risk factor for need for dental care in five year olds (odds ratio (OR) 1.73, 95% confidence intervals (95% CI) 1.01, 2.96 p=0.046) and need for speech therapy in both age cohorts (OR 2.15, 95% CI 1.40, 3.30 p= 0.000). Being under the care of a “high volume” cleft team (more than 25 cleft referrals per year) was associated with unmet need for dental treatment in five year olds (OR 2.45, 95% CI 1.44, 4.18 p=0.001). Adjusting for socioeconomic status and being registered with a dentist had little effect on this association. In both age cohorts, children under the exclusive care of the cleft team therapist were less likely to need speech therapy than children who were managed by a combination of team and local therapists (OR 2.51, 95% CI 1.45, 4.36 p= 0.001). This association was independent of socioeconomic status.

                                                                      Conclusions

                                                                      Cleft services in the UK are failing to meet the treatment needs of a significant number of children born with UCLP, especially those from low socioeconomic backgrounds. It is important that cleft teams, particularly those treating larger numbers of patients, have adequate arrangements in place to ensure that children receive appropriate treatment and follow up.

                                                                      Social determinants of children with behavioural problems

                                                                      Objective

                                                                      To identify the key social determinants of children with behavioural problems.

                                                                      Design

                                                                      Population based cross sectional study; information collected by health visitors.

                                                                      Setting

                                                                      Torbay, Devon.

                                                                      Subjects

                                                                      10 015 households with children under the age of 5 years.

                                                                      Main outcome measure

                                                                      Household having a child with behavioural problems where a planned programme of intervention is in place.

                                                                      Results

                                                                      Overall there were 866 children (8.5%) with behavioural problems who were also having a planned programme of intervention. Of the various household and family factors studied (n=21), nine remained significant after adjustment for possible confounders. However, the three most important determinants (adjusted odds ratios (OR) and 95% confidence intervals (95% CI)) as defined by the best model (goodness of fitr 2 = 0.068) were: (a) either parent being depressed/mentally ill (OR 3.45, 95% CI 2.90, 4.11), (b) low income families (OR 1.88, 95% CI 1.59, 2.21) and (c) one parent families (OR 1.99, 95% CI 1.66, 2.38). The low income group (who were working but dependent on benefits) seemed to be worse off, as compared with the unemployed (OR 1.14, 95% CI 0.93, 1.40). The one parent family (excluding those with extended family support) effect seemed to be independent of the effect of recent divorce (OR 1.31, 95% CI 1.00, 1.72).

                                                                      Conclusions

                                                                      Behavioural problems in children are on the rise and to some extent reflect on the changing social environment of early childhood. This study identifies some of the key social determinants that may prove useful to policy makers involved in developing programmes such as “Sure start”.

                                                                      Age and gender differences in utilisation of asthma medication in children and young adults in Tayside, Scotland

                                                                      Background

                                                                      There are few studies that examine the age and gender differences in children and young adults aged 0–25 in the utilisation of asthma medications.

                                                                      Methods

                                                                      The Medicines Monitoring Unit (MEMO) captures all dispensed medication for the Tayside area of Scotland (population 400 000). These data have a unique patient identifier attached that allows demographic and drug safety studies to be carried out. This study examined the dispensing of asthma medications for a population resident and registered with a GP in Tayside between 1993 and 1995.

                                                                      Results

                                                                      A total of 130 372 subjects aged 25 or under were identified. Of this group 18.1% of males and 16.3% of females had received at least one prescription for asthma medication. Stratifiying by age in four year bands the age group 1–4 years had the highest proportion with an asthma prescription (29.4% male and 23.5% female, p=0.001 χ2 test). The proportion of male subjects with an asthma prescription was consistently higher than female subjects until the age of 19 after which there was a higher proportion of female subjects. Logistic regression analysis showed that overall male subjects were more likely to be dispensed a prescription for asthma medication than female subjects (odds ratio (OR) 1.14, 95% confidence intervals (95% CI) 1.11, 1.18, p<0.001). Those aged under 20 were twice as likely to have a prescription as those aged 20–25 (OR 2.02, 95% CI 1.95, 2.10, p<0.001). There was a significant age by gender interaction (p<0.001) whereby male subjects were more likely to have a prescription for asthma than female subjects before the age of 20 (OR 1.26, 95% CI 1.22, 1.30, p<0.001) and less likely after the age of 20 (OR 0.73, 95% CI 0.69, 0.79, p<0.001).

                                                                      Discussion

                                                                      The difference in results for male subjects may reflect a higher incidence of asthma under age 20 or a possible prescribing bias. Gender should be taken into account when studying asthma in children and young adults.

                                                                      Trends in teenage pregnancies in Scotland and the associations with deprivation and rurality

                                                                      Background

                                                                      The recent Scottish white paper “Towards a Healthier Scotland” set a target to reduce pregnancies among 13–15 year olds by 20% by the year 2010.

                                                                      Objectives

                                                                      To describe trends in teenage pregnancies at the national and local level and associations with deprivation and rurality.

                                                                      Subjects

                                                                      Teenage conceptions treated in NHS hospitals in 1981–96, abstracted from hospital discharge records (SMR1, SMR2).

                                                                      Main outcome measures

                                                                      Annual age specific conception rates per 1000 population and the proportions resulting in a maternity for 13–15, 16–17 and 18–19 year olds.

                                                                      Methods

                                                                      Multivariate multilevel Poisson and logistic regression.

                                                                      Results

                                                                      Conception rates increased for 13–15 (from 5 to 9 per 1000) and 16–17 year olds (from 45 to 60 per 1000) in 1981–96, while rates remained constant at around 85 per 1000 for 18–19 year olds. The per cent resulting in a maternity decreased from 80% to 65% for 18–19 year olds, from 70% to 65% for 16–17 year olds and stayed constant at 50% for 13–15 year olds. These trends varied across districts but differences between localities were largely maintained so that districts above the average at the start of the 1980s were still above average in the mid-90s. At the small area level, deprivation, measured by 1991 Carstairs scores, was positively related to both conceptions and maternities. Relative to 13–15 year olds resident in affluent areas, those living in deprived areas were three times as likely to conceive (95% confidence intervals (95% CI) 2.6, 3.5) and were nearly three times as likely to give birth following the conception (odds ratio (OR) 2.9, 95% CI 2.3, 3.7). When levels of deprivation were taken into account, there was no difference in conception rates between urban and rural areas but 16–17 and 18–19 year olds in urban areas were less likely to have a maternity (OR 0.88, 95% CI 0.80, 0.97 and OR 0.65, 95% CI 0.59, 0.72) than those resident in rural areas.

                                                                      Conclusion

                                                                      Pregnancies in young teenagers have increased in the previous two decades while the percent leading to a maternity has changed very little. The consistency of district differentials implies that local strategies to reduce unwanted pregnancies are desirable. While increased maternities in rural areas might reflect cultural differences, the impact of access to services must also be considered. To achieve government targets, population-based interventions aimed at preventing unwanted teenage pregnancies and supporting teenage mothers must take account of different social, economic and cultural circumstances and the consequential health service needs of teenagers living in both deprived and rural communities.

                                                                      The relation of body mass index in early and mid-adulthood to mortality risk

                                                                      Objective

                                                                      To examine the relation between body mass index (BMI) in early and mid-adulthood and subsequent mortality.

                                                                      Design

                                                                      Cohort study.

                                                                      Setting

                                                                      University of Glasgow student health service. Weight and height were measured by a physician while the students were at university, mean age 23 years. Self reported weight was obtained from a survey 16 years later.

                                                                      Participants

                                                                      The study included 487 men, born between 1902 and 1932, who attended the student health service in 1948–49, who replied to a postal survey (1963–66) and who were traced through the NHS central register and followed up to August 1999.

                                                                      Main outcome measures

                                                                      All cause, cardiovascular disease (CVD) and cancer mortality.

                                                                      Results

                                                                      From age 23 to 39, BMI increased by an average 2.7 kg/m2. The two BMI measures, based on the original height measure and weight readings 16 years apart were correlated (r=0.67). Ninety two men (19%) died during the mean follow up period (from the second weight measurement) of 33 years. Fifty two deaths were attributable to CVD and 23 to cancer. BMI at ages 23 and 39 was positively associated with subsequent mortality. Having adjusted for smoking, alcohol consumption, blood pressure, pulse rate and father's social class (each measured at 23 years), the hazard ratios (HR) (95% confidence intervals (95% CI)) per 1 kg/m2 increase in BMI were HR 1.16 (95% CI 1.06, 1.27) and HR 1.12 (95% CI 1.05, 1.20) respectively. Cause specific analyses showed BMI at both ages was strongly predictive of CVD. The later but not the earlier BMI measure was related to cancer mortality. There were insufficient deaths to determine whether BMI was related to site specific cancers. Change in BMI was positively associated with all cause and CVD mortality; per 1 kg/m2 increase in BMI change, the adjusted HR were 1.07 (95% CI 0.97, 1.08) and 1.11 (95% CI 0.97, 1.18) respectively. Controlling for mean BMI substantially attenuated these relations. However, having controlled for mean BMI, a 1 kg/m2 increase in BMI change was associated with increased risk of cancer mortality HR 1.18 (95% CI 0.96, 1.45).

                                                                      Conclusions

                                                                      The association between early life BMI and mortality is only partially explained by the tracking of BMI from early to later adulthood. Efforts to decrease BMI in early life and to minimise weight gain from early to mid adulthood may have substantial impact on subsequent health.

                                                                      Stress and mortality—confounding or causation?

                                                                      Background

                                                                      Perceived psychological stress has been proposed as an important determinant of physical health. Stress may influence health directly (through neuroendocrine mechanisms that increase physiological risk) or indirectly (through the promotion of unhealthy behaviour). Perception of stress is socially patterned therefore other correlates of social position may confound the apparent association between stress and health.

                                                                      Objective

                                                                      To examine the association between self reported stress, disease risk, social position and mortality.

                                                                      Design

                                                                      Workplace based prospective observational study.

                                                                      Setting

                                                                      27 workplaces in west and central Scotland.

                                                                      Subjects

                                                                      5718 men aged 35–64 at recruitment and followed up for 21 years.

                                                                      Main outcome measures

                                                                      Hazard ratios for death from all causes, from cardiovascular disease (ICD9 codes 390–459), from smoking related cancers (ICD9 codes 140–41,143–50, 157, 160–63, 188–89) and from other cancers (remainder of ICD9 codes 140–208).

                                                                      Results

                                                                      Higher stress (measured by the “Reeder” stress inventory) was significantly associated with both higher occupational class and unhealthy behaviour (less exercise, more cigarettes smoked, greater alcohol consumption) but not with increased physiological risk. High and medium (compared with low) perceived stress appeared protective to health in terms of all cause mortality, cardiovascular mortality, and cancers. These associations were generally attenuated on adjustment for occupational class. (Hazard ratios (HR) for mortality associated with “high” and “medium” stress adjusted for age only and adjusted for age and occupational class. All cause: high 0.89 (95% confidence intervals (95% CI) 0.75, 1.03), medium 0.85 (95% CI 0.76, 0.95) and high 0.93 (95% CI 0.83, 1.03) medium 0.94 (95% CI 0.81, 1.10). Cardiovascular: high 0.84 (95% CI 0.67, 1.04), medium 0.82 (95%CI 0.71, 0.96) and high 0.90 (95% CI 0.72, 1.12), medium 0.90 (95% CI 0.77, 1.05). Smoking related cancers: high 0.66 (95% CI 0.44, 0.98), medium 0.80 (95% CI 0.62, 1.02) and high 0.73 (95% CI 0.49, 1.09), medium 0.91 (95% CI 0.71, 1.18). Other cancers: high 0.86 (95% CI 0.57, 1.31), medium 0.95 (95% CI 0.72, 1.26) and high 0.86 (95% CI 0.56, 0.32), medium 0.95 (95% CI 0.71, 1.26)). Adjustment for other risk factors and reporting bias had little effect on these estimates.

                                                                      Conclusions

                                                                      The apparent protective effect of stress on health was the result of confounding by social position. This “protective” effect was seen despite the expected association between stress and unhealthy behaviour, suggesting that material circumstances may be more important than individual behaviour in determining health.

                                                                      Personality in young adults and later mortality: prospective observational study

                                                                      Objectives

                                                                      There is evidence that anxiety and depression in middle age are associated with increased cardiovascular disease. We examined whether psychological health measured much earlier in the lifecourse, in a cohort of young adults born 1920–50, is also associated with later mortality patterns.

                                                                      Design

                                                                      Observational study of students attending the University of Glasgow from 1948–68 who participated in a health survey. A box labelled “personality” prompted a free text assessment by the examining physician. Before analyses, these were coded into eight categories: anxious, schizoid, depressed, immature, hypochondriacal, unstable, hypomanic, paranoid, inadequate, obsessive and odd.

                                                                      Participants

                                                                      8394 former males Glasgow University students aged 16–30 years, (mean 20.5) at the time of first examination.

                                                                      Main outcome measures

                                                                      All cause and cause specific mortality after a median follow up time of 41.3 years. Cox proportional hazards models were used to estimate the association between personality and mortality.

                                                                      Results

                                                                      There were 830 deaths. The number of participants with at least one personality category was 820 (9.8%). The most common personality category was anxiety with 481 (5.7%) men given this label. Only 12 (0.1%) men were labelled depressed. The presence of “any” coding was associated with an increased risk of stroke, hazard ratio (HR) (95% confidence intervals (95% CI)) 1.83 (1.00, 3.56). Both anxiety and depression codings were positively associated with all cause mortality and cancer. For anxiety the HR were 1.33 (95% CI 1.06, 1.68) and 1.48 (95% CI 1.02, 2.14) for all cause and cancer mortality respectively; for depression the these were 4.02 (95% CI 1.50, 10.73) and 5.54 (95% CI 1.38, 22.27) for the same causes of death respectively. Depression was also positively associated with cardiovascular disease and coronary heart disease and stroke and with both cancers related to smoking and cancers not related to smoking. There were no significant associations between other categories of personality and mortality. These results were unchanged after controlling for the potential confounding variables of cigarette smoking, father's social class, body mass index, systolic blood pressure and vital status of parents.

                                                                      Conclusions

                                                                      The results indicate that certain aspects of “personality” in early adulthood may be associated with later mortality. Although state and trait aspects of personality were conflated and assessed subjectively (and only a small number of students were depressed) these findings are suggestive of the importance of psychological well being in early adulthood in determining future mortality risk. Along with other accumulating evidence the results point to the need to optimise mental health in young adults.

                                                                      Current trends in mortality among adults in Scotland

                                                                      Objective

                                                                      To describe recent trends in death rates in Scotland with particular reference to mortality among young adults.

                                                                      Data and methods

                                                                      Trends in age specific death rates for all cause and specific causes since 1981 from mortality records gathered by the General Register Office (Scotland). Comparison of the percentage change in rates, for all Scotland and each local government district, between the grouped years 1981–83, 1988–91 and 1996–98.

                                                                      Results

                                                                      Between 1981–83 and 1988–91 death rates in Scotland began to rise among young men aged 20–24 (6% increase) while for those aged 25 and over mortality continued to decline across all age groups, with the greatest fall being experienced at ages 40 to 59 (20% fall). When death rates during 1996–98 were compared with rates in 1988–91 this pattern had changed. During the 1990s death rates among 20 to 34 year olds increased by 13%, with a slight rise at ages 35–39 (3%). At older ages overall mortality continued to decline. However, for those aged between 40 and 59 the relative decline experienced in death rates during the 1980s were not sustained and the greatest fall was now among men aged 60–69 (15% decline). Trends among women were more complicated although they did share similarities with men. For both men and women the overall pattern was suggestive of an age cohort process, but falls in mortality from chronic conditions such as heart disease, stroke and cancers were also being offset by increases in other causes of death across all age groups. The causes of death that have contributed to the increased mortality among young adults include to various degrees suicides, drug related deaths, psychiatric illness, alcohol and AIDS related deaths. When changes in death rates were examined by local government district, those districts with the highest concentrations of deprived populations were found to show the greatest increase in mortality among young adults. This picture, however, was somewhat complicated by the number of deaths from AIDS in Dundee and Edinburgh.

                                                                      Conclusion

                                                                      In Scotland the rate of decline in mortality among men aged 59 and below is slowing down, and death rates among young men aged 20–39 are increasing. If these trends continue there is a suggestion that future death rates will also see a rise at older ages. The failure to maintain earlier gains in mortality has important implications for the ways in which public health policies are prioritised.

                                                                      Body image and weight change in middle age: a qualitative study

                                                                      Objectives

                                                                      To explore experiences of weight change in adulthood and views of the medical, social and practical problems associated with different body shapes.

                                                                      Design

                                                                      Qualitative study using semi-structured, tape recorded interviews. Views about weight change in adulthood, experiences of changes and motivations to change were explored. Pictures of a set of eight body shapes designed to represent a range from underweight (1) to borderline obese (8) were used to encourage discussion about body image preferences for men and women. Respondents were asked to describe identify any health, social or practical problems they would associate with the different body shapes.

                                                                      Setting

                                                                      A purposive sample was identified through two health centres. Interviews were conducted in the respondents' home by researchers trained in qualitative interviewing.

                                                                      Participants

                                                                      Seventy two men and women aged between 35 and 55 with body mass indices (BMI) between 22 and 29. (That is, of recommended weight or moderate overweight).

                                                                      Results

                                                                      Experience of weight gain in adulthood was widespread in this group: 97% of men and 95% of women had noticed gain since they were in their 20s. However, less than half thought that weight gain was inevitable in middle age. Over a third of the men, but few women, said that they had never tried to lose weight. 42% of the men, but only 13% of the women with BMI below 26.9, had ever tried to gain weight. Responses to the pictures of body shapes were consistent with this finding: more men (41%) than women (26%) expressed a preference for a body shape numbered “4” or above for themselves. Forty seven per cent of men and 60% of women with BMI under 26.9 reported that they had successfully lost weight at least once. Problems associated with overweight included heart disease, joint problems, breathlessness, low self esteem, negative reactions from others, mobility, fitting into seats, and difficulty finding clothes, jobs and partners. Weight gain was attributed to slower metabolism, ageing, quitting smoking and more sedentary lifestyle, although childbearing, comfort eating and “letting oneself go” were thought to be reasons for women's gain while being less sporty and drinking beer were cited for men.

                                                                      Discussion

                                                                      Although none of the respondents' weight would be seen as a medical problem the interviews provide rich accounts of struggles with weight gain in middle age. Respondents were aware of holding contradictory views about ideal weight. Body image preferences bear little relation to recognised medical risk.

                                                                      Incidence of HIV infection among gay men seeking a repeat HIV test

                                                                      Objective

                                                                      To estimate the incidence of HIV infection among gay men who have previously tested negative for HIV.

                                                                      Methods

                                                                      Of 2100 people attending the same day HIV testing clinic at this hospital between September 1997 to July 1998, 1580 (75%) completed an anonymous questionnaire concerning sexual risk behaviour, number of previous HIV tests and date of last test. Repeat testers were those clinic attenders who had previously tested negative for HIV and were returning for another test. Only those whose previous test was at least three months before the present test were included in this analysis. HIV incidence was estimated by dividing the number of newly diagnosed cases of HIV among repeat testers by the person years of exposure since their last negative test.

                                                                      Results

                                                                      470 clinic attenders were gay men of whom 337 (72%) were repeat testers; 275 provided information on date of last test (median time since last test 24 months), age (median 31 years) and number of previous tests; 151 (55%) had had one or two previous negative tests while 124 (45%) reported three or more. Of these 275 men, 12 tested HIV positive. Overall HIV incidence was estimated to be 1.8 per 100 person years (12/655.2) (95% confidence intervals (95% CI) 0.8, 2.9). HIV incidence was increased for men whose previous test was within the last 12 months (4.7%) compared with those who had tested more than 12 months before (1.4%) (p=0.06). HIV incidence was also higher among men reporting three or more previous HIV tests (3.6%) than men with one or two previous tests (1.1%) (p=0.05). The highest HIV incidence (8.0%) was seen among gay men with a history of three or more previous tests who had tested negative within the past 12 months. Forty two per cent of gay men with three or more previous HIV tests reported “high risk” unprotected anal intercourse in the past three months (that is, with a partner whose HIV status was either positive or unknown) compared with 25% of those who had had one or two previous tests (p=0.002).

                                                                      Conclusion

                                                                      HIV incidence among gay men who had tested negative for HIV in the previous 12 months was high (4.7%) and was even higher for those who reported three or more previous negative tests (8.0%). For some gay men a negative HIV test result may produce a disinhibiting effect and reinforce risky behaviour. This needs to be tackled by health promotion programmes.

                                                                      The impact of bicycle helmets on impacts involving bicycles

                                                                      Background

                                                                      The protective effect of bicycle helmets remains unclear, with case-control studies accused of inadequate adjustment for confounding factors and existing time trend studies possibly affected by changing levels of bicycle use.

                                                                      Objective

                                                                      To examine patterns of serious head injuries among cyclists during a period of increased helmet usage. Associations with age, sex and deprivation are explored.

                                                                      Design

                                                                      Analysis of hospital admissions for the years 1991/92 to 1994/95, using the hospital episode statistics database.

                                                                      Setting

                                                                      All emergency admissions to NHS hospitals in England.

                                                                      Participants

                                                                      35 056 bicycle related admissions (ICD-9 E8261, E810–E825, 4th digit=6).

                                                                      Main outcome measures

                                                                      Monthly counts of admissions. Head injuries defined as fracture of vault or base of skull, or intracranial injury (ICD-9 800, 801, 850–854).

                                                                      Results–Head injuries were the primary diagnosis in 34% of cases (n=11 985), over half of these being in children. Head injury admissions declined significantly over the four years, by 9% annually (95% confidence intervals (95% CI) 7%, 10%). The decline was significant in three age subgroups: 6–10 years, p<0.001; 11–15 years, p<0.001; and >15 years, p<0.001. The decline did not differ between males and females but was significantly lower among those from deprived areas, 4% annually compared with 10% elsewhere (p<0.001). For all cyclist emergency admissions, 75% were male while 21% were resident in deprived areas (Jarman under privileged area score > 20). Numbers of admissions varied strongly by season, from an average of 1173 in July to 352 in January. Over the study period total admissions of cyclists increased, non-significantly, by 0.2% (95% CI, −1%, 1.1%). The increase did not differ between the sexes but was significantly higher in those from deprived areas, 6.3% (95% CI 2.6%, 10%) compared with −1.2% (95% CI −2%, 0%).

                                                                      Conclusion

                                                                      The level of cycling changed little during the study period, taking total admissions as a marker of cycle use. The observed year on year decline in head injuries occurred during a period of steadily increasing helmet use and provides strong evidence of their having a protective effect at all ages. Residents of deprived areas did however appear to cycle more, again taking total admissions as a marker for cycle use. This may explain the smaller decline they experienced in numbers of head injuries, rather than a difference in helmet wearing patterns.

                                                                      Print media response to a major malpractice inquiry in the UK

                                                                      Background

                                                                      In 1997–8, the body with the statutory duty to uphold professional standards, the General Medical Council (GMC), held its longest ever disciplinary hearing centred on events at the Bristol Royal Infirmary. Three doctors (two surgeons and a senior health service manager) were eventually found guilty of serious professional misconduct and were severely sanctioned. The case received unprecedented attention in both the professional and the lay media, and provoked intense government scrutiny culminating in a public inquiry.

                                                                      Objective

                                                                      To analyse lay print media reaction to The Bristol Case as a way of gaining insight into the public debates over health care quality assurance in the UK NHS.

                                                                      Methods

                                                                      We reviewed all major daily and Sunday newspapers in the UK for the five week period surrounding the announcement of the GMC verdict and sentencing (14 newspaper titles in all; total of 230 issues). A total of 184 separate media items were retrieved. A qualitative content analysis of the material was conducted in an attempt to draw out the major ideas and viewpoints. A small number of very prominent themes appeared repeatedly; these themes were confirmed by independent review of the material by each of the authors.

                                                                      Results

                                                                      The print media reporting of The Bristol Case was intense, emotive and hostile. Almost all the papers devoted editorial space to the case, some repeatedly. The Bristol Case was seen less as an unusual aberration by individual doctors and more as a symptom of systematic failings in the health system. Many articles referred to diminished public trust in health services and expressed disdain for professional self regulation. The key reforms demanded were greater publication of performance data (for example, individual surgeons' mortality rates) and increased external scrutiny. These findings will be illustrated with direct quotes from newspaper articles.

                                                                      Conclusions

                                                                      The media (and indeed government and professional) interest in The Bristol Case was unprecedented. The print media was characterised by hostility towards doctors, scepticism about self regulation and demands for greater accountability. UK government policy on health care quality will have to pay much greater heed to public sensibilities than hitherto. Any attempts at reforming physician regulation must now take place in the face of a highly sensitised and sceptical print media. Fine tuned arguments on the relative merits of professional self regulation and external scrutiny may thus be hard to sustain. Demands for greater access to data on clinical performance are unlikely to be deflected.

                                                                      View Abstract

                                                                      Request permissions

                                                                      If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.