Characterizing Patterns of Care Using Administrative Claims Data

Comparative effectiveness research (CER) is concerned with determining which treatment, among known effective treatments, may provide the most benefit to an individual patient. CER stresses treatment in the "real-world," taking into account multiple sources of patient, provider, and treatment heterogeneity. This is opposed to randomized clinical trials which may enforce strict treatment guidelines and may exclude patient populations from participation. One way to access treatment outcomes in these general settings is through administrative databases such as Medicaid claims. While they usually have variables linking records to individual patients over time, we believe a strategy that may aid in CER is to use administrative databases longitudinally. We describe two approaches that can be used with administrative data to characterize longitudinal patterns of treatment, trajectory analysis and multi-state Markov models. We apply these models to Medicaid retail pharmacy claims and behavioral service claims to describe the attention deficit hyperactivity disorder treatment patterns in youths in general clinical settings.