We studied the accuracy of Medicare part B coding for cataract extraction to provide validation for research involving Medicare data. Hospital and physician office records associated with a sample of 802 paid claims for cataract surgery were reviewed. The sample was randomly selected from 118,420 Medicare part B claims for cataract surgery submitted by physicians in an 11-state sample during the first quarter of 1988. Medical records were successfully obtained for 796 cataract surgery episodes (99.2%), of which 794 (99.7%) indicated that cataract extraction had been performed. In the remaining two cases, cataract surgery was attempted but aborted. In 24 (3%) of the 794 cases, the surgical approach (intracapsular or extracapsular) indicated in the operative note differed from the coded on the physician's bill. In all cases in which the operative note indicated a secondary procedure performed at the time of surgery, the billing information was in agreement. We conclude that, at least in the case of cataract surgery, the Medicare part B database is 99% accurate (95% confidence interval, +/- 0.6%) for cataract surgery having occurred and 96% accurate (95% confidence interval, +/- 1.4%) in terms of surgical approach.