Background: Preliminary data suggest a high positive predictive value of both prevalent (84%) and incident self-report of diabetes (85%) from WHI and a low false negative rate of about 1%. These results indicate that self-reports of diabetes may be sufficiently accurate for use in epidemiologic studies. The Women’s Health Initiative (WHI) clinical trials offer a wealth of data on patient outcomes including cancer, coronary heart disease, and hip fracture, which were confirmed via medical record review. Diabetes was self-reported, but was not validated by independent review. The literature shows wide variations on the validity of self-reported diabetes in diverse populations.
Aims: To compare the rates of positive and negative predictive values of self-reports of incident and prevalent diabetes during the WHI trials.
Methods: At four WHI field centers (Minneapolis MN, Winston-Salem NC, Birmingham AL, Portland OR), all WHI trial participants with a self-report of diabetes at baseline or follow-up were identified. A random sample of participants who did not self-report diabetes was also identified. Women were surveyed regarding details of their diagnosis and treatment of diabetes. Medical records were obtained and reviewed for consenting subjects using a standardized protocol based on documented treatment with anti-diabetic medications, or physician diagnosis of diabetes supported by laboratory measurements of glucose.
Results: A total of 1280 eligible subjects were available across the four sites. To date, subject information from the Minneapolis field center has been collected and analyzed. Of the 295 eligible subjects contacted at the Minneapolis site, 227 (77%) consented to participate and provided survey data. Medical records were obtained for 224 (57 prevalent, 81 incident, and 86 no self-reported diabetes). Medical records confirmed 84% (95% CI 74% - 94%) of self-reported prevalent diabetes and 85% (95% CI 74% - 94%) of self-reported incident diabetes. Among those who never self-reported diabetes there was no medical record evidence for diabetes in 99% (95% CI 92% to 100%). Results from all four sites will be presented.