Development and evaluation of a system that executes an interventional cardiology risk model based on patient phenotypes automatically extracted from the EMR

Lack of electronic medical record (EMR) integration hampers effective clinical operationalization of clinical decision support. Accordingly, this study sought to develop and evaluate a system that automatically executes an interventional cardiology risk model based on EMR content accessed through a web service. Phenotyping rules were developed to automatically determine the models risk factors, advanced age, anemia, congestive heart failure, diabetes, reduced eGFR and hypotension. Performance of the rules was determined using a ground truth based on chart review of 76 patients. Except for one instance, PPV and NPV of the phenotyping rules exceeded 0.93. The predictions produced by the risk model were accurate in 82% of patients. Performance of the phenotyping rules was determined by the ability to model the rules after the risk model's definition and the availability and accuracy of the accessed data sources. Data retrieved by the rules can be presented for efficient physician review and confirmation.