Article Text

Download PDFPDF

Linking questionnaires to primary care records: factors affecting consent in older people
  1. Tess Harris,
  2. Derek G Cook,
  3. Christina Victor,
  4. Carole Beighton,
  5. Stephen DeWilde,
  6. Iain Carey
  1. Department of Community Health Sciences, St George’s Hospital Medical School, London, UK
  1. Correspondence to:
 Dr T Harris
 Department of Community Health Sciences, St George’s Hospital Medical School, Cranmer Terrace, Tooting, London SW17 0RE, UK;

Statistics from

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.

Primary care databases provide rich consultation and prescribing information,1 but lack individual background detail. Linking to information from questionnaires can overcome this, but requires individual informed consent.2,3 The generalisability of linked data will be constrained by low consent rates and any differences between those giving and those withholding consent, in addition to any non-response bias. Most studies examining the consent issue, have either focused on patients with specific illnesses,3 or have had limited comparison measures.4 A recent large study found that men, younger people, and subjects with the symptom under investigation, were more likely to give consent, but lacked information on service use, or other measures of physical and psychological health.5 We conducted a wide ranging health and social survey and compared older people who gave and withheld consent to link questionnaire data to their primary care records.


A detailed account of the survey methods has already been published.6 A postal survey was conducted, with an offer of assistance in completion, with patients aged ⩾65 years from two London practices. The questionnaire is available on request from the authors and is also on the journal web site ( and included details on: demographic factors; physical health; psychological measures (including depression, anxiety, and health beliefs); social support; socioeconomic measures; and a request for consent to link questionnaire data with subjects’ computerised primary care records. Logistic regression was used to examine associations between other factors and consent to link, with results expressed as odds ratios (OR) and 95% confidence intervals (CI). Three groups (questionnaire non-responders; responders withholding consent to link; and responders giving their consent to link) were created on the practice computer system. The number of primary care consultations for the 12 month period after the questionnaire, was downloaded for each group, with no identifying individual data. Crude mean consultation rates for each group were estimated from total number of consultations and number of subjects per group. The 95% confidence limits were based on assuming these were generated by a Poisson process with over-dispersion parameter 2.1 (estimated from earlier data on individual level consultation rates from responders who gave consent to linkage).

The survey response rate was 75% (1704 of 2276) of whom 92% (1565 of 1704) gave their consent to link questionnaire and primary care data. Older subjects had higher response rates (65–74 year olds 785 of 1090 (72.0%) ⩾75 years 919 of1186 (77.5%) χ2 = 8.7, p = 0.00) but there was no sex difference in response (men 629 of 842 (74.7%) women 1075 of 1434 (75.0%) χ2 = 0.01, p = 0.93). Subjects with poorer health were more likely to give their consent, both for physical health factors and for anxiety and depression (table 1). Demographic factors (age, sex, practice, ethnic group) health locus of control, smoking, alcohol use, and social support measures were not associated with consent. For several socioeconomic measures, there was a trend for the more affluent to be more likely to give consent (income, occupational pension, and private health insurance) but this was only statistically significant for owning a car OR 1.8 (1.2 to 2.6). These findings were similar when analysed for each practice separately.

Table 1

 Factors associated with giving consent to link questionnaire with primary care records

Overall mean consultation rates (number of consultations/number of subjects) (95% CI) for the 12 months after the questionnaire were as follows: non-responders 7.8 (4436/572) (7.3 to 8.2); responders withholding consent 8.3 (1160/139) (7.8 to 8.8); responders giving consent 9.6 (14946/1565) (9.2 to 9.9).


A high percentage of older people gave consent for their primary care records and questionnaire results to be linked, consistent with other studies.3,5 No significant age or gender differences in consent were seen in our study, although the trend was for men and younger subjects to be more likely to consent, as reported by Dunn in their larger study with greater power.5 We found that those giving consent were more likely to have poorer physical and psychological health, and to use primary care services more. These findings were consistent across both practices and with findings among adults that those with the symptom under investigation were more likely to give consent.5 These differences between those giving and withholding consent (consent bias) add to reported differences in service use between survey responders and non-responders7 (response bias).

Our study findings therefore have important implications for researchers collecting non-anonymised data from primary care. Firstly, although 92% of responders gave consent to linkage, this provides further data attrition on top of that attributable to non-response, and needs consideration in sample size calculations. Secondly, the finding that, in older subjects at least, those giving consent are likely to have poorer physical and psychological health and to use primary care services more, implies that studies based on such subjects could overestimate disease prevalence rates. This provides more ammunition for those arguing against the necessity of informed consent for all studies using identifiable data, on the basis that the incomplete data obtained may be unrepresentative.2,8 Maximising consent rates will reduce, but not eliminate this consent bias.


We thank the two practices involved (Benhill and Belmont Practice, Sutton, Surrey and Bridge Lane Health Centre, Battersea, London) and all of the participants who gave their time to respond to the survey. Thanks also to Pit Rink and Sunil Shah for their involve-ment early on in the study, particularly with questionnaire development.

 TH, DC, and CV were all involved in designing the study and with data analysis (TH and DC) and interpretation (TH, DC, CV). CB was involved in the collection, processing, and analysis of data. SDW advised and supervised downloading of data from the practice computer systems and was involved in interpretation of data. IC was involved in processing data from the computer system and with data analysis and interpretation. All authors were involved in drafting the paper and in approving the final version of the paper. TH will act as guarantor for the paper.


Supplementary materials

  • The appendix is available as a downloadable PDF (printer friendly file).

    If you do not have Adobe Reader installed on your computer,
    you can download this free-of-charge, please Click here


    Files in this Data Supplement:

    • [view PDF] - The South London Health Study questionnaire


  • Funding: the BUPA Foundation funded the study. All researchers are independent of them.

  • Conflicts of interest: none declared.

  • Ethical approval: this was given by the Wandsworth Local Research Ethics Committee.

Linked Articles

  • In this issue
    Carlos Alvarez-Dardet John R Ashton