A prospective comparison of methods for determining if cardiovascular disease is a predictor of mortality in dialysis patients.

A random sample of 464 dialysis patients was surveyed between December 1990 and June 1991 to compare methods for determining the relationship between cardiovascular disease (CVD) and mortality. The following three methods were used to identify the prevalence of CVD: standard epidemiologic questionnaires, recall by the patient, and a review of the medical record. The 1-year mortality rate during this prospective study (average follow-up, 17.5 months) was 19%. The measure of prevalent CVD found to be the best predictor of the risk of mortality was the review of the medical record. Specifically, after controlling for the effects on mortality of age, sex, race, cause of renal failure, serum albumin level, and performance status (determined by the Karnofsky score), a patient with a history of angina pectoris documented in the medical record had a relative risk (95% confidence interval) of mortality of 1.8 (1.1 to 2.8), and a patient with peripheral vascular disease recorded in the medical record had a relative risk of 1.6 (1.0 to 2.4). Estimates of CVD obtained from either the questionnaires or patient recall resulted in associations between CVD and mortality that were substantially weaker than those for the medical record. We conclude that at present the medical record is the best source of information for estimating the presence of CVD as a mortality risk factor in dialysis patients. We recommend inclusion of a medical record history of CVD as a mortality case-mix factor when comparing dialysis populations.