Data Science Core

点击翻译按钮获取中文摘要

PROJECT SUMMARY The broad goal of the Data Science Core (DSC) is to advance and apply modern data science methods to support the practice of precision medicine in rheumatology. We define precision medicine as discovering mechanistically anchored and clinically relevant disease subgroups for which optimal strategies can be followed. The precision medicine statistical methods we develop and apply will improve the clinical assessment of an individual's disease status, trajectory, and likely benefits of competing interventions and thereby benefit people living with rheumatic diseases. During the past funding period, a team led by P30 PI Antony Rosen and DSC Director Scott Zeger, designed and implemented the JH Precision Medicine Analytics Platform (PMAP), a cloud-based data system: (1) receives a nightly download of all JHM clinical data and projects the data into a secure clinical cohort database; (2) provides an environment for collaborative, modern statistical analyses; and (3) provides tools to visualize from within the Electronic Medical Record those analytic results relevant to a particular patient's decisions. PMAP has already been implemented in the Scleroderma and Myositis Center. During the coming period, PMAP will become active in each rheumatology Center of Excellence so that current clinical cohort databases will form the information foundation for the next phase of this P30 program. This novel infrastructure makes it possible for the DSC to develop and apply novel statistical tools that identify disease subgroups, for whom optimal treatments can be evaluated. In this way, JHM Rheumatology will discover the key data science tools required by a learning healthcare system. The Specific Aims of the Data Science Core are to: (1) provide study design and analysis support that enables investigators to generate, manage, analyze, and interpret complex data using modern statistical and computing methods; (2) develop and apply multivariate hierarchical models (MHMs) to longitudinal datasets to identify patient subsets predictive of disease trajectories and major clinical events; and (3) apply modern statistical analytic approaches to observational and experimental studies that rigorously estimate treatment efficacy and safely, acknowledging the likely heterogeneity among individuals in their responses to treatments and the potential biases inherent in using observational data to address causal questions.