A constraint satisfaction approach to data-driven implementation of clinical practice guidelines.

Despite significant research efforts, the implementation of computerized clinical practice guidelines (CPG) in practice remains problematic for a number of reasons. In particular most guideline representation models do not deal adequately with incomplete or inconsistent clinical data. We present a constraint satisfaction approach to address such shortcomings by focusing on CPG data rather than CPG representation. We model a CPG as a set of data-driven constraints which are used to generate complete solutions for describing a patient state from incomplete clinical data, where the patient state is confirmed by the user. Inconsistent input data can be temporarily eliminated and final feasible solutions (permitted complete solutions from a CPG) can pinpoint inconsistencies in original input data alongside allowable guideline data. We demonstrate a sample implementation of the approach for a pediatric asthma CPG.