Method: HF experts identified high priority evidence-based measures from American College of Cardiology/American Heart Association that serve as the basis for National Quality Forum performance measures. We defined complex numerator, denominator, and exclusion criteria, and mapped each criterion to structured patient data (ICD9 codes, CPT codes, labs, vital signs, and prescriptions). For each criterion we defined the time period for the data, and decision rules for missing data. We encoded this clinical knowledge into computable formats in a Protégé knowledge base (KB). We built a HF performance measurement pipeline: first, SQL queries extract filtered patient data; next, a Java program formats and presents data to EON, the execution engine; EON processes the patient data with Protégé KB to generate the automated performance measurement results; finally, these results are stored in a SQL Server database. The system output specifies (1) a patient’s eligibility for a recommendation during a defined performance measurement period (denominator) and (2) whether the patient’s medical regimen during the measurement period adheres to the recommendation (numerator). We created SQL queries to present aggregated views for the end users, HF health professionals.
Results: We encoded 6 inclusion criteria and 34 exclusion criteria for performance measures that address inpatient/outpatient ACE-inhibitors/ARB and beta-blocker therapy into the KB. We conducted preliminary evaluation of the accuracy of this prototype system using a convenience sample of 20 outpatient cases and 17 inpatient hospitalizations. We compared conclusions from human review of the input data with system output; the system made the correct conclusions regarding applicability of guidelines and adherence to guidelines in all cases.
Conclusions: It is possible to automate performance measurement of HF management taking account of complex criteria and multiple types of structured patient data. Potential uses include clinical decision support for health professionals and quality assessment across health care systems. Future research will include integration of data extracted from free-text portions of the electronic health record such as echocardiography reports.
Views expressed are those of the authors and not necessarily of the Department of Veterans Affairs.