What is the effect of delays in radiology procedures on length of stay (LOS) in the hospital? Did the integration of patient records from outside organizations into Apex reduce number diagnostic tests ordered in the emergency department? What is the impact of delaying antibiotics on mortality in patients with sepsis?
Data from electronic health records (EHRs) are increasingly used to answer such questions that seek to quantify the causal effect of an intervention or exposure (e.g., procedural delays, antibiotics) on an outcome of interest (e.g., LOS, mortality). In medicine, this has historically been accomplished with prospective randomized controlled trials, but these are often cost-prohibitive or ethically unacceptable. Alternatively, under certain conditions, causal effects can be also estimated from retrospectively collected, observational data, guiding future policies without requiring prospective experiments.
In this session, we will discuss basic statistical principles that impact our ability to use retrospective EHR data to understand important clinical questions, including confounding and selection bias. We will use real-world examples from UCSF to illustrate these principles, and we will present strategies for optimizing analytic design.
Slides: https://ucsf.box.com/s/yxr17l8m1gv6tjqbfsqmr8uttpzay5gb (MyAccess login required)
This session is designed to be accessible to a general audience of researchers and data analysts.
Hossein Soleimani, PhD, is a Data Scientist in the Department of Health Informatics at the University of California, San Francisco (UCSF). He received his Ph.D. degree from Pennsylvania State University, PA, in 2016 in Electrical Engineering with Minor in Statistics. He specializes in developing and deploying machine learning-based clinical decision support tools, causal inference, and data analytics.
Sara Murray, MD, MAS, is an Assistant Clinical Professor of Medicine in the Division of Hospital Medicine at the University of California, San Francisco (UCSF), and serves as the Medical Director of Clinical Informatics for UCSF Health. She works with the clinical systems teams at UCSF to optimize the current electronic health record (EHR) infrastructure and design novel informatics solutions for the medical center. In her role leading the Advanced Analytics and Innovation team, she is involved in large analytic projects using EHR data to inform quality and value improvement efforts at the medical center. She is interested in predictive analytics and has done research in EHR phenotyping. Her clinical time is spent as a hospitalist, attending on the teaching and non-teaching medicine services.