Reliability of Medicare Claim Forms for Outcome Studies in Kidney Transplant Recipients: Epidemiology in Clinical Outcome Trials

Medical databases may offer substantial opportunities for outcomes research to investigators in the field of renal transplantation. An increasing number of analyses have attempted to merge Medicare billing claim forms with large, national registries. Large registries such as the US Renal Data System and the Scientific Registry of Transplant Recipients have afforded researchers the opportunity for large-scale, population-based analysis; however, these registries have come under increasing scrutiny in terms of their use for associative outcome studies. It has been postulated that by merging Medicare claims with other large registries, some of the limitations outlined by critics can be overcome. Despite the increasing number of publications that are taking this approach, this practice has not been fully validated in the renal transplantation population.

Four main elements that characterize a medical database are (1) population, (2) medical events, (3) coding systems, and (4) data management (1). Each of the registries mentioned in the previous paragraph, to some degree, do not capture all of the essential elements. This may introduce nonrandom and uncaptured bias, potentially leading to nonindependent associations. In addition, the large amount of variables available can lead to type 1 error if the association found is not robust across a number of situations and does not have some biologic plausibility.

In population-based registries, inadequate capture of certain data is unavoidable (2). Eligibility criteria may be difficult to obtain properly, with imprecise and inconsistent definitions limiting the accuracy of the data used (3). Also, only selected medical events are …