OBJECTIVES: The purpose of this study is to provide a source document in the creation of a dataset for analysis to develop a predictor algorithm for disease management. METHODS: A review of the literature was performed to determine best practices in the development of a dataset to design a predictor model.  Algorithms improve applicability across health settings to meet various disease management aims. Peer reviewed articles and reports were retrieved from the published literature from 2004 to 2014.  To summarize results a case study is also provided related to the need for a patient to enroll in a disease state management program based on predicting their chance to have a heart attack.

RESULTS: Results revealed a total of 33 articles and reports. A summary of this information is provided as a step-by-step guide on how to setup a dataset to predict disease state management patient outcomes. The purpose of data collection should be discussed in detail (e.g. collection of data to identify high risk members). Using predictive modeling tools, data can be synthesized such as diagnoses, hospitalizations, emergency room encounters, expenditures and demographics to develop individualized risk profiles. The assessment period and outcomes of interest should be clearly defined followed by an analysis of model sensitivity and specificity.  Members are assigned a chronic illness intensity index score based on these factors. Once scored, members are filtered through clinical criteria that prioritize individuals with clinically manageable conditions. The resulting list represents those members with the most acute and complex illness burden. CONCLUSIONS: Findings from this study have implications for clinical care, patient outcomes, research, and policy. Many patients are not receiving appropriate preventive care, lack recommended care and receive contraindicated care.  An increase in the creation of databases to be used effectively to develop predictor models is needed to intervene early in the disease cycle.