C2. Guidelines for the use of claims databases for outcomes research

Abstract

Outcomes research is used to identify, measure, and compare the economic, clinical, and humanistic outcomes of programs or services to determine what works best for which patients. It is believed that improved methods and use of outcomes research may reduce costs and improve quality of care. Although prospective outcomes research studies are desirable, resource constraints often limit researchers to a retrospective study design involving a claims database. Because claims databases were originally designed to process medical claims for the insured population, they are not perfectly suited for conducting outcomes studies. An extensive review of the literature identified the limitations associated with the use of claims databases. These limitations can be categorized into threats to internal, external, and construct validity and will be discussed within these contexts. Based on the aforementioned threats to validity, guidelines have been proposed that suggest ways of avoiding bias in claims database outcomes research: (1) use a within-patient design that allows for control of channeling bias; (2) provide a priori an explicit theory (ie, treatment theory) about the mechanism through which the treatment is expected to produce an outcome; (3) conduct a sensitivity analysis to assess the robustness of results to changes in any inclusion and exclusion criteria; and (4) replicate the study in a comparable population. By recognizing the threats to validity and adhering to the guidelines described, the validity and usefulness of claims database outcomes research results will be greatly enhanced.