Background: Electronic health records are frequently used for cancer epidemiology. We report on their quality for ascertaining colorectal cancer (CRC) in UK women.
Methods: Population-based, retrospective cohort study nested within the UK Collaborative Trial of Ovarian Cancer Screening (UKCTOCS). Postmenopausal women aged 50-74 who were diagnosed with CRC during 2001-11 following randomisation to the UKCTOCS were identified and their diagnosis confirmed with their treating clinician. The sensitivity and positive predictive value (PPV) of cancer and death registries, hospital episode statistics, and self-reporting were calculated by pairwise comparisons to the treating clinician's confirmation, while specificity and negative predictive value were estimated relative to expected cases.
Results: Notification of CRC events were received for 1,085 women as of 24 May 2011. Responses were received from 61% (660/1,085) of clinicians contacted. Nineteen women were excluded (18 no diagnosis date, one diagnosed after cut-off). Of the 641 eligible, 514 had CRC, 24 had a benign polyp, and 103 had neither diagnosis. The sensitivity of cancer registrations at one- and six-years post-diagnosis was 92 (95% CI 90-94) and 99% (97-100), respectively, with a PPV of 95% (95% CI 92/93-97). The sensitivity & PPV of cancer registrations (at one-year post-diagnosis) & hospital episode statistics combined were 98 (96-99) and 92% (89-94), respectively.
Conclusions: Cancer and death registrations in the UK are a reliable resource for CRC ascertainment in women. Hospital episode statistics can supplement delays in cancer registration. Self-reporting seems less reliable.
Keywords: Colorectal neoplasms; Death certificates; Electronic health records; Hospital episode statistics; Registries; Self report.
Copyright © 2018 The Authors. Published by Elsevier Ltd.. All rights reserved.