Adrian Haimovich, MD, PhD

Yale School of Medicine

"Computational Phenotyping of Heart Failure Patients in the Emergency Departments"
SAEMF/RAMS Resident Research Grant




Heart failure (HF) is a common presentation in emergency departments (EDs) worldwide and is associated with significant morbidity and mortality. The vast majority of patients presenting with HF are admitted, but ED providers play a significant role in early management. While HF has now been widely recognized as a complex clinical syndrome, treatment pathways in the ED remain largely homogeneous. Furthermore, undifferentiated presentations of patients with histories of HF can pose significant diagnostic and prognostic challenges to practitioners. Few efforts to phenotype patients with HF have addressed the ED setting, partially due to difficulties identifying this patient population early in a hospital course. Here, we propose a retrospective study of patients presenting to one academic and two community EDs with documented histories of HF. We have previously shown that abstracted patient histories and triage data derived from electronic health record (EHRs) can be leveraged to create robust predictive algorithms for patient disposition. We have furthered this work, showing that patient phenotype discovery algorithms identify patient clusters with significant differences in risk of admission. We hypothesize that coupling HF patient EHR-data at initial ED presentation to outcomes data will enable the unbiased discovery of patient phenotypes with significant implications for treatment and prognostication.

Specific Aims: This proposal includes two specific aims. The first is to perform computational phenotype discovery on the ED visits of patients with histories of heart failure in order to identify groups of patients with similar presentations at time of triage. To do so, we will employ cutting-edge unsupervised machine learning techniques to analyze data extracted from our system EHR. The second is to expand our dataset using available ED outcomes metrics to enable phenotype-based patient risk stratification.

The overall aim of this proposal is to identify clinically-relevant phenotypes in a population of patients with a history of heart failure at the time of ED triage. In doing so, we seek to set the stage for subsequent phenotype-based interventions and to aid in prognostication.

Research Results

Dr. Rotoli is still completing the project.