Abstract Body (Do not enter title and authors here): Background
Despite well-established benefits on morbidity and mortality, implementation of guideline-directed medical therapy (GDMT) in heart failure (HF) population remains suboptimal. Quality improvement interventions to improve GDMT uptake often necessitate labor-intensive manual chart reviews to identify eligible patients. To automate this process, we implemented and evaluated a Rule-Based System (RBS) designed to identify HF patients eligible for remote medication optimization program.
Methods
To identify symptomatic HF patients for enrollment in our Cooperative Program for ImpLementation of Optimal Therapy in Heart Failure (COPILOT-HF), an RBS was hierarchically developed to first identify HF patients meeting program criteria (Figure 1A), then exclude those with contraindications or at increased risk of adverse drug reactions. The RBS processed structured data from electronic health records (EHR) from Mass General Brigham’s enterprise data warehouse to compile a list of potentially eligible patients as per program criteria. This list was then reviewed by program staff, who manually verified patient eligibility.
Results
A total of 5,460 patients were identified by the RBS as potentially eligible, of whom 1,754 (32.1%) were confirmed as eligible on manual chart review with positive predictive value of 47.3%. Among the false positives, 25% did not meet the program criteria for symptomatic HF, 12.4% with LVEF >50% were already on SGLT2i, 11.0% were too medically complex for our program, 9.8% did not have a recorded echocardiogram within 2 years, 6.6% were receiving anticancer therapies, 4.0% had outside cardiologist, 3.9% with LVEF <50% were either on GDMT or intolerant and 23.3% had other reasons for exclusion (Figure 1B).
Discussion
These findings suggest that RBS may be a useful tool for modeling and applying guideline logic in comparison to the traditional method of querying institutional health record repositories. While the RBS demonstrated acceptable performance in mass screening for remote GDMT optimization program, future efforts should focus on improving coding and integrating the natural language processing tool for processing of clinical notes to enhance screening accuracy and efficiency.