OTHER PROJECT INFORMATION – Project Summary/Abstract Project 3 – Inside the Black Box of Clinical Decision-Making: Provider Practice Habits, Provider Caseloads, and Implications for Patients with ADRD The process of arriving at a clinical decision depends on a complex and hard to observe set of factors beyond the biomedical decision itself, including the clinical environment, patient conditions and characteristics, belief formation, team interactions, and information flow. These subtle factors may be especially influential for patients with ADRD, who present unique challenges to medical providers. Project 3 aims to fill a gap in knowledge about the economics of clinical decision-making for patients with ADRD. We will leverage unique, high frequency audit log data on providers’ use of the Electronic Health Records (EHR) system for all UCSF Emergency Department (ED) patients from 2017-2019. We use these data to analyze the process of diagnosis and treatment decisions and their effects on the quality of clinical care and patient health outcomes. The project will generate information needed for the design of technologies and organizations to improve quality of clinical decision making, care and outcomes for patients with ADRD. First, we will describe the health profiles and needs of patients with ADRD in the UCSF ED and relate those profiles and needs to the broader nationwide context for ADRD patient care. Second, we will document the impact of provider characteristics on clinical documentation, practice patterns and quality of health care for patients with ADRD. We will assess differences within physicians in how they care for patients with and without ADRD and differences across physicians in how they care for patients with ADRD. We will relate these differences to key observable characteristics (e.g., type of provider, level of provider, gender, race, age) as well as unobservable factors that control for these observable characteristics. Third, we will leverage the quasi-random assignment of patients to physicians in the ED to assess how treatment decisions and health outcomes for ADRD patients are impacted by the cognitive load of a physician. Finally, we will develop machine learning methods to identify patient diagnoses based on patient characteristics, clinical orders, and the results of those orders, with a focus on modeling physician information acquisition. We will apply these methods to assess how providers arrive at diagnoses for patients with symptoms of ADRD as compared to those without them. While Project 3 shares the ED setting with other projects, the intent focus on decision-making processes in the ED, including how pressured environments affect providers and their care of patients with ADRD, closely complements the study of other influences on healthcare, outcomes, and disparities in the other P01 projects.