Validation of Self-Report of Chest X-Ray Exam at a Prostate, Lung, Colorectal and Ovarian Cancer Screening Trial Center

It is imperative to measure the degree of contamination throughout the course of randomized controlled trials, as contamination, the receipt of the intervention arm regimen by control arm participants, can affect trial power. In the Prostate, Lung, Colorectal and Ovarian (PLCO) Cancer Screening Trial, contamination was estimated through use of the self-administered Health Status Questionnaire (HSQ) annually to a randomly-selected subset of control arm participants. We examined agreement of self-report of chest x-ray on the HSQ with clinic records at one of the 10 PLCO screening centers (Henry Ford Health System, or HFHS). We focus on HFHS participants covered by the Health Alliance Plan (HAP), a managed health care insurance plan owned and operated by HFHS, because claims for care received both at HFHS and other facilities are available in HFHS databases for HAP enrollees. We examined agreement for the six years prior to HSQ completion, with HFHS clinic records considered to be the gold standard. For those who had complete HAP coverage during the six years, percent agreement was 0.69, sensitivity was 0.84, and positive predictive value was 0.76. Specificity and negative predicted value were low, however (0.28 and 0.38, respectively), and Cohen's kappa was 0.13. For groups with incomplete or no HAP coverage, and when timing of exam was considered, performance measures typically became lower, in some instances below 0.20. These data suggest that self-report of chest x-ray screening may not be accurate, although high prevalence of chest x-ray may make performance measures less interpretable.