Identifying Heart Failure from Electronic Health Records: A Systematic Evidence Review

BackgroundHeart failure (HF) is a complex syndrome associated with significant morbidity, mortality, and healthcare costs. Electronic health records (EHRs) are widely used to identify patients with HF. Despite the development of many HF phenotyping algorithms, it is unclear if the characteristics of identified populations reflect the known spectrum of disease in HF patients. MethodsWe performed a systematic evidence review to assess the methods used for HF algorithm development from US based data sources and the relevant information facilitating their application to research and clinical care. We queried PubMed for articles published up to November 2021. Out of 428 studies screened, 35 articles were included for primary analysis and 23 studies using only International Classification of Diseases (ICD) codes were evaluated for secondary analysis. Results are reported descriptively. ResultsHF algorithms were most often developed at academic medical centers, though these were not evenly distributed across the US. HF and congestive HF were the most frequent labels for observed phenotypes. Diagnoses were the most common data type used to identify HF patients and echocardiography was the second most frequent. Most studies used rule-based algorithms. Validation of algorithms varied considerably with 36% of HF subtype algorithms validated and 67% of acute HF algorithms validated. Demographics of any study population were reported in 57% of algorithm studies and 70% of ICD-only studies, however fewer than half (40% of algorithm and 46% of ICD-only) of studies reported demographics of the HF population identified by their computable phenotype. Of those reporting, most identified majority (>50%) male populations, including both algorithms for HF with preserved ejection fraction identified in this study. ConclusionThere is significant heterogeneity in phenotyping algorithms used to ascertain HF from EHRs. Validation of algorithms is inconsistent and largely relies on manual review of medical records. The concentration of algorithm development at a few sites limits their generalizability to identify HF patients in other settings. Differences between the reported demographics of algorithm-identified HF populations and those expected based on HF epidemiology suggest that current algorithms do not reflect the full spectrum of HF in the general population.