Completeness and reliability of USRDS data: comparisons with the Michigan Kidney Registry.

A census of 12,730 Medicare-eligible ESRD patients identified in the USRDS database as living in Michigan or as having received treatment in Michigan between 1977 and 1987 was compared to the 13,424 ESRD patients listed in the Michigan Kidney Registry (MKR) who were alive between 1977 and 1987. Of 7,600 USRDS patients with both residence and ESRD treatment in Michigan, 97.8 percent were also found in the MKR, thus providing evidence that the USRDS is not overstating the size of the ESRD population. Overall, 11.4 percent of the cases in the MKR could not be matched to the USRDS; however, the rate of matching showed a slight upward trend over time. Subsequent analysis shows that after excluding those cases expected not to be captured by the Medicare system, only 5 percent of MKR cases were left unmatched to USRDS cases. Some of these unmatched cases from the MKR were found to have short follow-ups, while others belong to groups not covered by Medicare without being noted as such by the MKR database. A high level of exact agreement on data elements for cases found in both databases was documented for demographic variables, above 90 percent. However, differences were found between the USRDS and the MKR databases on elements related to treatment history. It is likely that a lower precision of the dates associated with treatment episodes and difficulty in recording short term treatment episodes in the USRDS could account for the differences found.(ABSTRACT TRUNCATED AT 250 WORDS)