Keywords: medical devices, observational studies, propensity scores, orthopedics, registries
Observational studies are pervasive in evaluating effectiveness of medical devices, in part because device failures are typically quite rare. The rarity of device failures makes conducting a randomized trial prior to the device’s introduction to the market prohibitive and thereby necessitates use of relatively large observational datasets to track performance only after the device has been introduced to the market. However, even large individual registries may be inadequate for surveillance of devices, therefore a collection or network of registries are desirable. In this case study we consider issues in the design of a multi-registry study that evaluates the comparative effectiveness of a particular hip replacement device. The design issues center on coordinating the inclusion/exclusion of measured variables, approaches to balancing on measured covariates, and evaluating the treatment-outcome relationship.