Comparison of self-report data and medical records data: results from a case-control study on prostate cancer.

BACKGROUND: Self-report and review of medical records are the most common methods for the assessment of past exposures. However, information obtained from self-reports and medical records may not be consistent. This study compared information provided in a self-administered questionnaire with medical records data.
METHODS: Self-report and medical records data came from a case-control study on prostate cancer. Cases were 181 patients with primary prostate cancer and controls were 297 men without the disease, enrolled in Group Health Cooperative (GHC) in Seattle. The consistencies between the two data sources were examined.
RESULTS: In general, agreement between the two data sources was almost perfect for demographic and anthropometric variables, substantial for the history of inguinal hernia and kidney stones, and moderate for vasectomy, family history of prostate cancer, smoking and alcohol consumption. However, the two data sources generally were poorly concordant for prior genitourinary diseases that have less explicit diagnostic criteria such as benign prostatic hyperplasia and prostatitis. Analyses of discordant data showed that men were more likely to report an exposure or medical condition that could not be verified from medical records. No discernible patterns in the difference of agreement were found according to age, GHC membership length or case-control status.
CONCLUSIONS: This study suggests that agreement between self-reported data and medical records data varies depending upon the study variables. While both data sources are subject to some problems, self-report may provide more complete and comparable information, at least for variables unrelated to diagnosis.