Alternative population sampling frames produced important differences in estimates of association: a case-control study of vasculitis

OBJECTIVE: A common population sampling frame in countries with universal health care is health service registers. We have evaluated the use of such a register, in the United Kingdom, against a commercially available database claiming large population coverage, an alternative that offers ease of access and flexibility of use. STUDY DESIGN AND SETTING: A case-control study of vasculitis, which recruited cases from secondary care clinics in Scotland, compared two alternative sampling frames for population controls, namely the registers of National Health Service (NHS) primary care practices and a commercially available database. The characteristics of controls recruited from both sources were compared in addition to separate case-control comparison using logistic regression. RESULTS: A total of 166 of 189 cases participated (88% participation rate), while both the commercial database and NHS Central Register (NHSCR) controls achieved a participation rate of 24% among persons assumed to have received the invitation. On several measures, the NHSCR patients reported poorer health than the commercial database controls: low scores on the physical component score of the Short Form 36 (odds ratio [OR]: 2.3; 95% confidence interval [CI]: 1.3-4.1), chronic widespread pain (OR: 2.3; CI: 1.1-4.7), and high levels of fatigue (OR: 2.0; CI: 1.3-3.1). These had an important influence on the estimates of association with case status with one association (pain) showing a strong and significant association using commercial database controls, which was absent with NHSCR controls. CONCLUSION: There are important differences in self-reported measures of health and quality of life using controls from two alternative population sampling frames. It emphasizes the importance of methodological rigor and prior assessment in choosing sampling frames for case-control studies.