Using administrative data for case-control studies: the case of the Papanicolaou smear

Administrative data have not been used extensively for case-control studies. As an example, data from the Manitoba Health Services Commission were used to assess the effectiveness of Papanicolaou (Pap) testing in screening for cancer of the cervix. Case patients were 415 women with invasive cancer (aged 25 to 64 years) from the Manitoba Cancer Registry and control subjects were women aged 25 to 64 years chosen from a random sample of the Manitoba female population (N = 29,926). For 82% of case patients there was information about Pap testing in a prior 5-year period, as compared to 91% of the controls. Since diagnostic tests could not be distinguished from screening tests, data for the year prior to diagnosis were not used. The odds ratios for Pap testing reducing the risk of cancer of the cervix was 0.82 (95% confidence interval [CI]: 0.53, 1.26) for women aged 25 to 34, 0.55 (95% CI: 0.35, 0.85) for women aged 35 to 44, 0.40 (95% CI: 0.23, 0.69) for women aged 45 to 54, and 0.45 (95% CI: 0.26, 0.78) for women aged 55 to 64. The overall 4-year relative odds adjusted for age and number of physician visits was 0.51 (95% CI: 0.43, 0.61). These results were similar to previously published values obtained from interview studies. Selection and recall bias are minimized and using claims avoids the expense of seeking neighborhood control subjects, carrying out interviews, or performing follow-up searches. As well, true population-based controls can be used.(ABSTRACT TRUNCATED AT 250 WORDS)