Using Large Electronic Health Records and Advanced Analytics to Develop Predictive Frailty Trajectories in Patients with Heart Failure

点击翻译按钮获取中文摘要

Candidate objective: My objective for this award is to become an independent quantitative scientist in analytical clinical research through structured training and mentored research experience. My goal is to become an academic leader and developer of advanced predictive models of health trajectories using electronic health records (EHR). Training objectives: I seek to sharpen my skill set as a clinical quantitative scientist using clinical informatics, EHR data warehouses, and advanced computational models. I will use the protected time provided by this award to gain proficiency in patient-clinician interactions, clinical informatics, natural language processing, and advanced survival analysis to accomplish my research aims. Background: Frailty is a complex clinical syndrome associated with aging and chronic illness. It decreases physiological reserves and increases vulnerability to stressors. The prevalence of frailty in patients with heart failure is 74%. The interplay of frailty and heart failure increases the risk for death, prolonged hospital stays, and functional dependence. One conceptual framework to operationalize frailty is accumulation of deficits: the frailty index (FI). The FI provides a risk score based on the assumption that the more ailments a patient has, the higher the risk of adverse outcomes, including mortality. Prior FI models have not been used in routine clinical practice due to the following limitations: insufficient number and range of clinical variables, lack of personalized deficit detection, use of data not commonly found in EHRs, insufficient use of longitudinal analytical models including survival analysis techniques, and the reduction of FI to a cross-sectional health status rather than a health trajectory. Research Aim: The overarching goal of this application is to develop a frailty trajectory (FT) for heart failure patients that provides information integrating prior functional impairment, current functional status, and future risk of mortality. In Aim 1, we will develop a novel cross-sectional FI that uses the full breadth of outpatient EHR data and innovative machine learning data science methods to predict mortality. In Aim 2, we will use serial cross-sectional FIs to build FTs and identify clusters of individuals following a similar progression of frailty over time. In Aim 3, we will compare the prognostic value of cross-sectional FI versus FT. The VA national EHR offers the ideal context for this study, as it provides longitudinal data since 1999 and can link to administrative data from non- VA sources, including linked Medicare databases. Mentoring & environment: A multidisciplinary mentoring team will supervise my training and will oversee my mentored research projects, formal coursework, directed reading, and career development. The proposed activities will provide a foundation for transitioning to an independent quantitative data scientist developing clinical decision aids to guide patient care. Baylor College of Medicine and the Center for Innovations in Quality, Effectiveness, and Safety have a national reputation of mentoring and supporting junior faculty members from diverse academic backgrounds to independent careers as clinical-investigators.