Since 1995, the Medical Evidence Report for end-stage renal disease (Form 2728) has been used nationally to collect information on comorbid conditions. To date, these data have not been validated. A national cross-sectional study of 1005 incident dialysis patients (734 hemodialysis and 271 peritoneal dialysis) enrolled between October 1995 and June 1998 was conducted using clinical data to validate 17 comorbid conditions on Form 2728. Sensitivity and specificity were calculated for each condition. The relationship between patient characteristics and sensitivity was assessed in multivariate analysis. Sensitivity was fairly high (0.67 to 0.83) for HIV disease, diabetes, and hypertension; intermediate (0.40 to 0.52) for peripheral vascular disease, neoplasm, myocardial infarction, cerebrovascular disease, coronary artery disease, cardiac arrest, and congestive heart failure; and poor (<0.36) for dysrhythmia, ambulation status, pericarditis, chronic obstructive pulmonary disease, and smoking. Sensitivity did not change significantly over calendar time. The sensitivity of Form 2728 averaged across all 17 conditions was 0.59 (95% confidence interval, 0.43 to 0.75). The average sensitivity was 0.10 greater in peritoneal dialysis than hemodialysis patients. 0.11 greater in diabetic patients than nondiabetic patients, and 0.04 less with each added comorbid condition. The specificity was very good for hypertension (0.91) and excellent (>0.95) for the other 16 conditions. Comorbid conditions are significantly underreported on Form 2728, but diagnoses are not falsely attributed to patients. Scientific research, quality of care comparisons, and payment policies that use Form 2728 data should take into account these limitations. Considerable effort should be expended to improve Form 2728 coding if it is to provide accurate estimates of total disease burden in end-stage renal disease patients.