Article Text

Download PDFPDF
Trends and socioeconomic disparities in preadolescent's health in the UK: evidence from two birth cohorts 32 years apart
  1. Nichola Shackleton,
  2. Daniel Hale,
  3. Russell M Viner
  1. Department of General and Adolescent Paediatrics, Population, Policy & Practice Programme, Institute Of Child Health, University College London, London, UK
  1. Correspondence to Dr Nichola Shackleton, UCL Institute of Child Health, 30 Guilford St. London WC1N 1EH, UK; n.shackleton{at}ucl.ac.uk

Abstract

Background Compared to children and adults, little is known about changes in adolescent health over time. This study profiles the health of preadolescents in two distinct time periods, 1980 and 2012.

Methods Secondary analysis of the British Cohort Study and the Millennium Cohort Study for preadolescents mostly aged between 10 and 11.5 years (range 9.75–13 years). The prevalence/average of, and socioeconomic gradients in, the following were compared between 1980 and 2012; general health problems, anthropometrics, allergic conditions, infectious diseases, health service use, smoking and parental smoking behaviour and maternal adiposity. Socioeconomic status was measured by familial income (3 groups).

Results There were decreases in the proportion of hospital admissions (Δ=−0.10(−0.09:−0.11)) smoking (Δ=−0.12(−0.11:−0.13)), parental smoking (mother: (Δ=−0.19(−0.18:−0.21)) father: (Δ=−0.35(−0.34:−0.37), infectious diseases (measles: (Δ=−0.46(−0.45:−0.47)) whooping cough: (Δ=−0.07(−0.06:−0.07)) and hearing problems (Δ=−0.04(−0.03:−0.05)). There were no changes in limiting long-standing illness (Δ=0.00 (−0.00:0.00)), or the proportion of children having two or more accidents requiring medical attention (Δ=−0.02(−0.00:−0.02)). There were increases in the proportion overweight (Δ=0.18(0.17:0.19)), maternal overweight(Δ=0.22(0.21:0.23)) and obesity(Δ=0.12(0.11:0.13)), height for age(Δ=0.47(0.44:0.49)), weight for age (Δ=0.68(0.65:0.71)), proportion reporting chicken pox (Δ=0.28(0.27:0.29)), allergic conditions (eczema Δ=0.19(0.18:0.20)), asthma Δ=0.12(0.11:0.13), hay fever Δ=0.15(0.14:0.16)) and wearing glasses (Δ=0.08(0.07:0.09)). There were increases in socioeconomic gradients for limiting long-standing illness, smoking, overweight, weight for age, height for age, wearing glasses, asthma and the onset of puberty.

Conclusions There have been reductions in infectious diseases and tobacco exposure among British preadolescents, but overweight and allergic conditions have risen dramatically. Children from deprived families have benefitted least from improvements in health status, and have experienced the largest increases in health risks.

  • PAEDIATRIC
  • PUBLIC HEALTH
  • CHILD HEALTH
  • INEQUALITIES

Statistics from Altmetric.com

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.

Background

In the UK, the burden of disease is now largely related to non-communicable conditions and injuries, as the burden of infectious diseases continues to decrease.1 We know that this health transition has resulted in dramatic improvements in the health of infants and younger children, with, for example, UK infant mortality now approximately one-third of what it was in 1980.2

However, improvements in the health of adolescents may have been much smaller in magnitude. Mortality data suggests that young people benefitted much less than children from the health transition, with rises in injuries and some non-communicable diseases outweighing benefits from the reduction in more traditional causes of mortality.3 ,4 Adolescents may be especially sensitive to the health effects of social and economic change. Early adolescence is increasingly recognised as a second critical developmental period in growth and development.5 Not only is there dramatic and far-reaching maturation of the brain and all body systems,6 but it is the time when many of the key risk factors for later disease are initiated (eg, smoking, alcohol consumption, drug use) or exacerbated (eg, poor diet, inactivity, obesity).7 Yet, we know little about progress in the health of UK young people over the past 30 years outside of mortality, despite some evidence that mental health problems in particular have increased among adolescents over the past 20 years.8

We used data from two nationally representative UK birth cohorts assessed in very early adolescence 32 years apart. Our aims were to (1) compare the health of contemporary young adolescents with their historical peers on a range of health outcomes including non-communicable diseases, infectious diseases and health service use and (2) to examine changes in the social patterning of health between the two cohorts.

Method

Sample

We used data from longitudinal birth cohorts born in 1970 (British Cohort Study (BCS)) and 2000/2001 (Millennium Cohort study (MCS)). The BCS sample consists of all births in Great Britain in 1 week in April 1970 (n=17 287). The MCS sample includes children born in England and Wales between September 2000 and August 2001 and those born in Scotland and Northern Ireland between November 2000 and January 2002 (n=18 819).9

We used data from sweep 2 (1980) of the BCS, with 89% of respondents aged 10 to 10.5 years (range 9.75–11.25 years). Response rates for this BCS sweep were very high (88.9% of the eligible sample).10 Data from the MCS came from sweep 5 (2012), when the majority (84.8%) of children were aged 10.5–11.5 years (range 10.25–13 years). MCS response rate for this sweep was 81% of the eligible sample.11 Further details regarding the cohort studies is available from the Centre for Longitudinal Studies (http://www.cls.ioe.ac.uk).

Health measures

We examined the two cohorts to identify all health indicators with highly comparable items available in both data sets. Those indicators which we judged to have high consistency across the two data sets are shown in table 1. Details of the wording of the specific questions in each study and the sources of the responses are available in online supplementary appendix A.

Table 1

Comparing health outcomes in the British Cohort Study (1980) and the Millennium Cohort Study (2012)

General health

Parental report was available in both cohorts on whether the child had a limiting long-standing illness, whether the child ever had hearing problems and whether the child wore glasses. The medical examiner reported whether the child had a stammer or stutter in the BCS, this was determined by parental report in the MCS.

Anthropometrics

Height and weight were measured by medical examiners in the BCS and by trained interviewers in the MCS. In both, we used the International Obesity Task Force's (IOTF) criteria to categorise preadolescents as overweight, or not, based on their body mass index (BMI), age and sex. Compared to more direct measures of body fat, the IOTF criteria for overweight are very specific, meaning they are very unlikely to classify somebody as overweight if they have healthy levels of body fat.12–14 Linear interpolation of the half year overweight cut-off criteria were used to create cut-off values relative to the child's age in months.15 Overweight includes the obese group. Cole et al's16 measure of child underweight (cut-off value of BMI of 17 at 18 years) was used to identify underweight children based on age and sex. Height for age and weight for age z scores were calculated using the 1990 British growth charts.17

Pubertal onset

Signs of pubertal onset were measured by the medical examiners in the BCS and reported by parents in the MCS. For both we defined the category of interest as any signs of puberty compared to no signs of puberty.

Allergic conditions

The presence of hay fever and eczema were reported by parents in both cohorts. The presence of asthma was reported by the medical examiner in the BCS and by parental report in the MCS.

Childhood infectious diseases

In both cohorts parents reported whether the child had ever had whooping cough, measles and chicken pox.

Health service use

All health service use variables were from parental report.

  1. Hospital admissions in the BCS referred to admissions since the child's fifth birthday, while in the MCS hospital admissions were reported since the previous MCS survey (age 7).

  2. The number of accidents requiring medical attention was reported in both cohorts as being since the last cohort survey, that is, age 5 years in the BCS and age 7 in the MCS. This was dichotomised into ≥2 accidents and <2 accidents.

  3. Visits to the dentist: parents reported in both cohorts whether the child had visited a dentist in the past 12 months.

Smoking and exposure to smoking

In both cohorts, preadolescents reported whether they had ever tried a cigarette and how many of their friends smoke cigarettes. In the BCS cohort, parental smoking behaviour was reported mostly by mothers on the father's behalf. In the MCS both parents were asked to provide this information. In both cases they were asked to report whether they were current smokers.

Maternal overweight and obesity

Maternal height and weight were self-reported in both cohorts and body mass index (BMI) calculated. Maternal overweight and obesity status were calculated using the WHO cut-off criteria of ≥25 kg/m2 for overweight and ≥30 kg/m2 for obese based on BMI.

Socioeconomic status

We elected to use household income as the most consistent proxy measure for socioeconomic status available in both cohorts. As original income bandings differed in the cohorts, we constructed a three category measure of income so that the BCS and MCS income groups had similar proportions of preadolescents within each income group. In the BCS, parents were asked to report band membership for total gross weekly income including earned and unearned income, excluding child benefit. In the MCS, parents were asked to report band membership based on joint net income (weekly, monthly or yearly) including total take home income from earnings, benefits and other sources of regular income. In the BCS income was originally banded into seven categories, whereas in the MCS reported familial income was banded into 19 categories of varying increments.

We divided income in the BCS into three categories, attempting to make those categories as similar in size as possible. Those earning between <£35 and £99 per week were classified as the lowest group (37%),those earning between £100 and £149 per week were classified as the middle group (34%), and those earning £150+ per week were classified as the highest group (28%). Income in the MCS was coded so that three income groups had similar proportions to the BCS. Income bands 1–9 (<£0–£299 per week) were classified as the lowest income group (37%). Income bands 10–13 (£300–£599 per week) were classified as the middle income group (35%) and income band 14 and above (£600+ per week) were classified as the highest income group (28%)

Analysis

All analyses were conducted using Stata V.13.18 The proportions of preadolescents reporting the health outcomes described above, or the mean value for the outcome, are presented for each cohort separately together with the 95% CIs.

Cohort data sets were reorganised so that all variables were coded in the same way and had the same name. The data sets were appended using the ‘append’ command in stata. To obtain the change in values over time with CIs, regression analyses were run with a cohort dummy variable included. For the majority of outcomes these regressions were logistic regressions. The exceptions were weight for age and height for age, where we used standard OLS regressions. The change in the outcomes over time, which is signified by the Δ, is also provided with 95% CIs.

To test the socioeconomic inequalities in health outcomes regression analyses were run in each cohort data set separately with income group included as the only covariate. We present the ORs for the lowest income group compared to the highest income group. To test whether the socioeconomic inequalities had changed over time, we ran regression analyses in the appended data set which included an interaction between income group and cohort.

Where the cohort data was analysed separately, the complex sampling design of the MCS has been accounted for using the ‘svy’ commands in Stata. For all the results presented here, the MCS is weighted using the ‘weight2’, which adjusts observations to be nationally representative based on the probability of selection into the survey. Where the data were appended, the weighting variable was set to ‘1’ for all responses in the BCS and given the values of ‘weight2’ for the MCS. Listwise deletion was applied.

As a means of robustness checking, the analyses presented here have also been conducted using the ‘Eovwt2’ weight, which further adjusts for non-response based on observable characteristics. The analyses were also repeated using the three category version of the National Statistics Socioeconomic Classification (NS-SEC), a measure of social class based on occupation instead of income.

Results

Prevalence of health conditions

The prevalence of the health outcomes in both cohorts are shown in table 1. Between the BCS and MCS surveys, there were large decreases in the proportion reporting ever having measles (Δ=−0.46 (−0.45:−0.47), preadolescents reporting ever having tried a cigarette (Δ=−0.12(−0.11:−0.13) or having friends that smoked (Δ=−0.11(−0.10:−0.12) and having parents that smoke (Δmother=−0.19(−0.18:−0.21)|Δfather=−0.35 (−0.34:−0.37)). There were moderate decreases in the proportion of preadolescents being admitted to hospital (Δ=−0.10(−0.09:−0.11), or ever having whooping cough (Δ=−0.07(0.06:0.07)). There were small decreases in the proportion reporting ever having had hearing problems (Δ=−0.04(−0.03:−0.05)), and very small decreases for ever having a stutter or a stammer (Δ=−0.01(0.00:0.01)) and being classified as underweight (Δ=−0.01(0.01:0.01)).

There were large increases in the proportion of preadolescents classified as overweight or obese (Δ=0.18(0.17:0.019)), ever having hay fever (Δ=0.15(0.14:0.016)), eczema (Δ=0.19(0.18:0.020)) and asthma (Δ=0.12(0.12:0.013)), ever having chicken pox (Δ=0.28(0.27:0.29)), showing signs of puberty(Δ=0.41(0.40:0.42)), having overweight and obese mothers (overweight Δ=0.22(0.21:0.24)| obese Δ=0.12(0.11:0.13)) and large increases in preadolescents height for age (Δ=0.47(0.44:0.49)) and weight for age (Δ=0.68(0.65:0.71)). There were moderate increases in the proportion of preadolescents wearing glasses (Δ=0.08(0.07:0.09)), and small increases in the proportion visiting the dentist in the past 12 months (Δ=0.03(0.02:0.04)). There were no changes in the proportion of preadolescents with limiting long-standing illness (Δ=−0.00(−0.01:0.01)) or having two or more accidents requiring medical attention (Δ=−0.01(−0.02:−0.00)).

Socioeconomic inequalities

The socioeconomic inequalities in preadolescent health are shown in table 2.

Table 2

Comparing socioeconomic inequalities in health outcomes in 1980 and 2012

In 1980 preadolescents in lower income families had significantly higher odds of having a limiting long-standing illness (OR=1.30 (1.10:1.53)); this gradient is significantly larger in 2012 (OR=2.01 (1.62:2.50)). Yet the overall prevalence did not change, suggesting a rise in limiting long-standing illness for low-income groups and a decrease for high-income groups.

Differences in the odds between the lowest and highest income groups also increased significantly for whether the child had ever tried a cigarette (1980: OR=1.01(0.88:1.16)|2012: OR=3.30(2.23:4.87)), and whether their friends smoked (1980: OR=1.43(1.23:1.63)|2012: OR=3.36 (2.49:4.54)). Given the change in prevalence, this suggests that the majority of preadolescents who have ever tried a cigarette, or who have friends that smoke, are from low income families, but the actual prevalence is very small.

Increased socioeconomic inequalities in child overweight accompanied an increase in the prevalence of overweight (1980: OR=1.06(0.89:1.25)| 2012: OR=1.94(1.69:2.24)), suggesting larger increases in prevalence for low-income groups. The same is true for whether the child wears glasses (1980: OR=1.05(0.92:1.21)| 2012: OR=1.50(1.29:1.74)), and evidence of pubertal status (1980: OR=0.97(0.85:1.10)| 2012: OR=1.38(1.23:1.55)).

Allergic conditions showed a reversal or disappearance of social gradients across the two cohorts. In 1980 preadolescents in lower income families had decreased odds of having hay fever (OR=0.55 (0.47:0.63)), eczema (OR=0.71(0.63:0.81)) and asthma (OR=0.78(0.63:0.96)). By 2012, preadolescents in low-income families had increased odds of having asthma (OR=1.48 (1.28:1.71)) but there were no income gradients for eczema and hay fever. Given the increases in prevalence of allergic conditions over time, this suggests larger increases in allergic conditions for preadolescents from low-income families.

In both 1980 (OR=0.56(0.51:0.61)) and 2012 (OR=0.48(0.39:0.60)) preadolescents in low-income families had lower odds of ever having had chicken pox. The differences between the low and high-income groups became more pronounced over time with increasing prevalence, suggesting that increases were smaller for preadolescents from low-income families.

Weight for age also showed a reversal in the social gradient across the two cohorts. Preadolescents in low-income families in the BCS cohort were lighter on average (−0.16(0.12:0.20)) than preadolescents in high-income families, whereas in the MCS cohort preadolescents from low-income families were heavier on average (0.18(0.13:0.24)). In both cohorts preadolescents from low-income families were shorter for their age on average (1980: (−0.29(−0.33:−0.24))| 2012: (−0.09(−0.14:−0.04))), but the difference is smaller in 2012, compared to 1980.

In both 1980 (OR=2.18(1.95:2.42)) and 2012 (OR=1.62 (1.41:1.86)) mothers in lower income families had higher odds of being overweight, but this gradient was significantly smaller in 2012. Given that the proportion of mothers being classified as overweight almost doubled between 1980 and 2012, this suggests larger increases in overweight prevalence for mothers in high-income families. The same pattern is shown for mother's obesity, with mothers from low-income families more likely to be obese in 1980 (OR=3.09(2.50:3.82)) and 2012 (OR=2.22(1.89:2.59)), but the difference between low and high-income groups being significantly smaller in 2012, despite the prevalence increasing threefold.

Mothers in low-income families had higher odds of being smokers in 1980 (OR=1.87(1.70:2.04)) and 2012 (OR=4.87(4.10:5.78)), as did fathers (1980: OR=1.78(1.62:1.95)|2012: OR=3.99(3.39:4.71)), but the differences in the odds were significantly larger in 2012 than in 1980. This suggests that reductions in smoking have been much larger for high-income groups.

Tests of the interaction between income and cohort suggested that social patterning did not change over time for hospital admissions, having two or more accidents requiring medical attention, ever having measles, whooping cough or visiting the dentist.

Robustness

The analysis was repeated using the ‘Eovwt2’ weight, which further adjusts for non-response based on observable characteristics. This made no substantive difference to the results (available on request). Further, the results from the sensitivity analysis using the NS-SEC in place of income as a measure of socioeconomic position are largely consistent with those presented here and are available in online supplementary appendix B.

Discussion

The health of young people at the start of adolescence in Britain has changed over the past 30 years. Once common childhood infections such as measles are now rare, smoking and exposure to tobacco smoke has diminished greatly, while rates of allergic conditions and obesity have risen markedly. Other problems such as long-term conditions and accidents did not change significantly, we identified evidence for growing health inequalities in limiting long-standing illness, smoking and exposure to smoking, overweight, weight for age, height for age, glasses wearing, asthma and the onset of puberty.

Many of these changes reflect elements of the epidemiological transition away from infectious diseases towards a disease burden composed primarily of injuries and non-communicable diseases.1 This is likely to be one of the main causes of the reduction in the number of preadolescents having ever been admitted to hospital, although it may reflect changing hospital admission policies to manage more childhood illness at home or in the community.19 ,20 Immunisation21 and reduction in infectious diseases may also explain the decreasing prevalence of hearing problems in preadolescents, as infectious diseases are the leading causes of acquired hearing loss.22 However, chicken pox increased rather than decreased across the study period, consistent with other reports which attributed this increase to lack of widespread use of varicella vaccines in the UK and increased use of formal child care and the opportunity this provides to mix with infected children.23 ,24 Improved management and or earlier detection of problems likely explains reductions in having a stutter or stammer.25

The most notable improvements in the health of Britain's early adolescents relate to reduction in exposure to tobacco smoke, both through active smoking by young people and by passive exposure to smoking in the home. This corresponds to other national statistics which show similar decreases over time among similar age groups.26 ,27 Reports on smoking drinking and drug use in England and Scotland suggest that the number of adolescents smoking has reduced over time, along with adolescents’ use of substances more generally.26 ,28

We did not find evidence for an increased burden of long-term conditions and injuries, but did identify major increases in overweight and obesity, a key risk factor for future non-communicable diseases. The epidemic of obesity across the Western world in children and adults over the past 30 years is well documented.29 We found a threefold increase in childhood overweight and halving of underweight across the study period, consistent with previous national trend data30 ,31 and coinciding with a threefold increase in maternal overweight. As higher body fat is an trigger for the onset of puberty,32 increases in overweight may explain higher proportions of preadolescents in puberty in the MCS compared with the BCS, together with the slightly older age in the MCS cohort.

The increase in allergic conditions we identified is also well-described in Western countries over the past three decades,33–35 however the reasons for this remain unclear. One explanation involves improved living standards and how this influences preadolescent's exposure to germs and infectious agents, and consequently how their immune system reacts to irritants.36 ,37 Other explanations involve changes in the environment and lifestyles such as increased pollution, changes in diets or indoor allergen exposure. Given that disadvantaged children were less likely to have allergic conditions in 1980, it may be that differences in terms of living standards or household environment had a protective effect on disadvantaged children in 1980, but that socioeconomic differences in living standards or household environments were not as pronounced in 2012.

The rise in the proportion of, and social patterning of, preadolescents wearing glasses likely reflects changes in preadolescent's lifestyles and activity levels and the social patterning of these. Throughout many developed countries the prevalence of myopia (short sightedness) in children and young people has increased.38 ,39 Lifestyle changes such as increases in the amount of time children spend on ‘near-work’, and a decrease in the amount of time children spend in outdoor activities have been forward as potential explanations.39 ,40

Positive changes in young people's health, such as reduced tobacco exposure, were marred by evidence of growing inequalities in health across many of the outcomes studied, consistent with evidence for a general widening of socioeconomic inequality in many health outcomes across the UK population.41 In tobacco, despite very major falls in average exposure, we found twofold to fourfold increases in risk of tobacco exposure in poorer compared with richer families between the BCS and MCS. This indicates that it is preadolescents from the most affluent families that have benefitted most from reduction in tobacco exposure. Social gradients in childhood overweight also increased alongside marked increases in overweight prevalence, indicating that preadolescents from the most deprived families experienced a greater rise in overweight prevalence.

The socioeconomic inequalities in smoking, overweight and glasses wearing all suggest a pattern of behaviour or lifestyle that can be altered to improve health outcomes for disadvantaged preadolescents. Small lifestyle changes that involve being more active and reducing time spent in sedentary activities that involve focusing on close objects, such as looking at mobile phones or computer screens would be beneficial for improving the health of preadolescents across these health outcomes.

Strengths and limitations

We used two large nationally representative birth cohorts containing a wide range of comparable health and healthcare indicators. One limitation is that neither sweep level non-response nor item non-response were taken into account in the analysis. It is likely that the non-respondents are from lower socioeconomic groups and are more likely to have poor health; therefore our results may underestimate the relationship between SES and the health outcomes. However, response rates in the BCS were very high, and the analysis presented here has been repeated using survey weights with the MCS which adjust for non-response, with no difference in the overall story.

A potential limitation is the difference in ages between the two cohorts. It may be that differences in age explain some of the differences we report in prevalence, or socioeconomic inequalities. Age is accounted for in the measurement of overweight, underweight, weight and height for age. Furthermore, given that the MCS cohort is older on average it is likely that we are underestimating reductions in ever preadolescents smoking. Nevertheless it is plausible that allergic conditions and eyesight problems are more likely to be detected at secondary school.

Another limitation is that for some outcomes, responses were obtained from the medical examiner in the BCS, where they were obtained from the parents in the MCS. It may be that some of the differences observed in these variables are due to differences in data collection methods, especially for more subjective judgements of health such as questions referring to changes in the child body (pubertal status). Furthermore, because self-report and parental reports are used to measure many of the health outcomes, some of the differences we observe between the two cohorts may result from changes in what is deemed a socially acceptable response at the time.

Finally, both hospital admissions and accidents requiring medical attention refer to the previous sweep of the survey as a reference point. For the BCS this is age 5 and for the MCS this is age 7. Hence there are 5 years on average between the previous survey and the current survey in the BCS and 4 years on average in the MCS. This additional year may have resulted in an increased number of events in the BCS.

Conclusion

The past 30 years has seen reductions in infectious diseases and tobacco exposure among British preadolescents at the cusp of adolescence. Yet, overweight and atopic conditions have risen dramatically. Preadolescents from deprived families benefitted least from improvements in health status (eg, reduction in tobacco exposure) and suffered most from increases in health risk (eg, increasing prevalence of overweight). This study supports gains made in the reduction of serious illness among British preadolescents, but with increasing inequalities, the focus must now shift to targeting lifestyle behaviours to reduce ill-health.

What is already known on the subject?

  • Early adolescence is increasingly recognised as a second critical developmental period in growth and development.

  • Mortality data suggests that for young people rises in injuries and some non-communicable diseases are outweighing benefits from the reduction in more traditional causes of mortality.

What this study adds?

  • This paper provides an overview of the health of preadolescents from two nationally representative birth cohorts born 32 years apart.

  • There have been reductions in infectious diseases and tobacco exposure among British preadolescents, but overweight and allergic conditions have risen dramatically.

  • Preadolescents from deprived families benefitted least from improvements in health status and suffered most from increases in health risk.

Acknowledgments

The authors are grateful to The Centre for Longitudinal Studies, Institute of Education for the use of these data and to the UK Data Archive and Economic and Social Data Service for making them available. The analysis and interpretation of these data is our own. Data are available from the ESDS website (http://www.esds.ac.uk) and are used within the remits of the End User License. The authors would also like to thank members of the Policy Research Unit for the health of children, young people and families: Terence Stephenson, Ruth Gilbert, Catherine Law, Miranda Wolpert, Amanda Edwards, Steve Morris, Helen Roberts and Cathy Street.

References

Footnotes

  • Twitter Follow Russell Viner at @russellviner

  • Contributors NS drafted all sections to the paper and performed the analysis. RMV conceived the research questions, contributed to the design of the analysis plan and revised the draft paper. DH contributed to the drafting of the paper and the analysis plan.

  • Funding The Policy Research Unit is funded by the Department of Health Policy Research Programme. This is an independent report commissioned and funded by the Department of Health. The views expressed are not necessarily those of the Department.

  • Competing interests None declared.

  • Provenance and peer review Not commissioned; externally peer reviewed.