Sensitivity and specificity of self-reported cancer history compared to cancer registry.

Abstract

234 Background: The Ontario Health Study (OHS) is a large prospective epidemiologic cohort study in which any Ontario resident eighteen years of older may enroll regardless of prior medical history. Baseline data are collected using web-based tools. As part of the consent process, participants are asked for consent to link study data with administrative and health care claims databases, including the Ontario Cancer Registry (OCR). There is an option to enter their Health Insurance Number (HIN) for this purpose. The purpose of this study was to link these data and evaluate the accuracy of self-reported cancer history compared to the cancer registry. Methods: Consenting participants that provided HINs were deterministically linked to the administrative data. Those that did not were probabilistically linked using name, sex and date of birth. Cancer registry records indicating a cancer diagnosed before the date of completion of the OHS baseline survey were considered the gold standard. Concordance, sensitivity, and specificity were assessed. Results: OHS records were successfully linked to administrative claims data and the Ontario Cancer Registry (OCR) with an 85.13% match rate. The final cohort consisted of 139,798 participants. A personal history of cancer was reported by 13,171 of these subjects, out of which 10,066 were found with a record in the OCR. The sensitivity of self-report was 77% and the specificity 95%. Excluding cancers diagnosed after the completion of the baseline survey increased sensitivity of self-report to 93%. The main area of discrepancy causing low sensitivity was the self-reporting of non-melanomatous skin cancers in the OHS questionnaire. Conclusions: While some self-over-reporting of cancer history may occur, cancers with lower metastatic potential tend to be under-captured in our provincial cancer registry. These findings have implications for cohort creation for research and quality improvement.