Optimizing the Care of Acute Heart and Lung Diseases through Precision Triage and Inpatient Bed Assignment (OPTIBED) - PROJECT SUMMARY Up to 16% of patients hospitalized with acute heart and lung diseases will die during their hospitalization. Mortality is up to three times higher when patients who need ICU-level care are initially triaged to a general care ward. Even after triage decisions are made, prolonged emergency department (ED) stays while awaiting an inpatient bed are increasingly common and associated with poorer outcomes for patients requiring time- sensitive or complex care. Yet, processes for ICU triage (“where should a patient be admitted”) and inpatient bed assignment (“when should they be transferred to an inpatient bed”) have failed to capitalize on modern advances in data science, causal modeling, and decision support. This results in triage and bed assignment decisions that are imprecise, impersonal, and highly variable, culminating in preventable hospital deaths. Our overall objective is to improve the care of patients with acute heart and lung diseases by developing data- driven models to support intensive care unit (ICU) triage and bed assignment. Our overall hypothesis is that personalized triage and bed assignment models can safely reduce rates of clinical deterioration and death among patients with acute heart and lung diseases. To test this hypothesis, we will complete three specific aims among patients with acute heart and lung diseases at five diverse hospitals: (1) we will identify multilevel determinants of ICU triage and bed assignment using a sequential, explanatory mixed methods study, (2) we will develop a clinical decision support model to estimate the benefit of ICU level of care for patients being admitted from the ED and evaluate it in an emulated pragmatic trial, and (3) we will develop an optimization model for inpatient bed assignment and evaluate it in an emulated pragmatic trial. Our multidisciplinary team is uniquely suited for the proposed research due to complementary expertise in data science, optimization modeling, causal inference, target trial emulation, health services research, mixed methods research, and clinical medicine. Integrating this expertise will allow us to develop innovative models guiding ICU triage and bed assignment decisions for patients with acute heart and lung diseases. We will also generate strong observational evidence of their safety and efficacy using target trial emulation, a novel way of evaluating clinical models; these results will expedite progress towards beside implementation and testing in a follow-up R01 proposal. Finally, this work will provide a framework for broader efforts to improve care delivery for patients with acute heart and lung diseases by applying innovative methods and causal inference for model development and evaluation.
