Validation of self-reported colorectal cancer screening behavior from a mixed-mode survey of veterans

Objective: The aim of the study was to validate self-reported colorectal cancer (CRC) screening using the National Cancer Institute Colorectal Cancer Screening questionnaire. Materials and Methods: 890 patients, ages 50 to 75 years, from the Minneapolis Veterans Affairs (VA) Medical Center were surveyed by mail. Phone administration was attempted with mail nonresponders. VA and nonVA records were combined for the reference standard. Sensitivity, specificity, concordance, and report-to-records ratio (R2R) were estimated for overall and test-specific CRC adherence among respondents providing complete medical records. Secondary analyses examined variation in estimates by patient characteristics, treatment of missing and uncertain responses, and whether a strict or liberal time interval was used for assessing concordance. Results: Complete medical records were available for 345 of the 686 survey responders. For overall adherence, sensitivity was 0.98, specificity was 0.59, concordance was 0.88, and R2R was 1.14. Sensitivity was 0.82 for fecal occult blood test (FOBT), 0.75 for sigmoidoscopy, 0.97 for colonoscopy, and 0.63 for double-contrast barium enema (DCBE). Specificity was 0.89 for FOBT, 0.76 for sigmoidoscopy, 0.72 for colonoscopy, and 0.85 for DCBE. Concordance was >0.80 for all tests other than sigmoidoscopy (0.76). R2R was 1.31 for FOBT, 1.33 for sigmoidoscopy, 1.42 for colonoscopy, and 6.13 for DCBE. The R2R was lower for a combined sigmoidoscopy and colonoscopy measure. Overreporting was more pronounced for older, less-educated individuals with no family history of CRC. Sensitivity and R2R improved using a liberal interval and treating uncertain responses as nonadherent (versus missing), but differences were not statistically significant. Conclusions: Self-reported CRC screening validity is generally acceptable and robust across definitional decisions, but varies by screening test and patient characteristics.