Measuring the validity of survey questions on breast, cervical, colorectal, and lung cancer screening

The National Health Interview Survey (NHIS) is used to measure progress on cancer screening. We assessed the validity of questions on cervical, colorectal, breast, and lung cancer screening from the 2020, 2021, and 2022 NHIS, using electronic medical records (EMRs) as our standard for accuracy. We surveyed 1770 adults aged 21+ years for breast, cervical, colorectal, and lung cancer from 4 US health systems. We made slight changes in question order and wording to improve the understanding of questions. We compared survey responses for screening adherence with EMR data as a gold standard, calculating sensitivity, specificity, positive predictive value, negative predictive value, Cohen's Kappa, and reports-to-records ratio. Self-reported screening adherence had high sensitivity for most cancer types (range 0.79-0.96). We found good agreement for breast cancer screening using Cohen's Kappa (0.81) and more modest agreement for the other 3 cancer sites (0.59-0.65). The reports-to-records ratio showed overreporting of cancer screening, ranging from 12% to ⁓50% more screening reported for the United States Preventive Services Task Force (USPSTF)-recommended screening periodicity compared to medical record data. NHIS questions that assess cancer screening provide reasonably accurate estimates. However, some misclassification with expected bias toward overreporting of screening suggests that improvements in measuring screening adherence are needed.