Health AI

The Health AI team joins deep clinical and data science expertise with broad knowledge of health informatics to address UCSF Health’s top priorities.

Leveraging our extensive knowledge of EHR data infrastructure and clinical workflows, we use data science, visualization, and machine learning to improve care delivery and outcomes. To do this, we:

Collaboratively identify opportunities for improvement with clinical and operational leaders
Take a data-driven approach to rapidly prototype tools
Implement validated and effective solutions both within and outside of the electronic heatlh record

Our group's expertise spans data science, data visualization, machine learning, process improvement, nursing informatics, and medical informatics. Our tools include Microsoft SQL Server, PostgreSQL, Python, R, and Tableau for data analysis and modeling and GitHub, Trello, Jenkins, and Slack for DevOps and project management.

Read about our team: "How UCSF’s data science team took on COVID"

Project Highlights:

Data Science, Visualization & Machine Learning

COVID-19 data mart, analytics, and visualizations to support UCSF COVID-19 response and recovery
Analytics and predictive modeling for enterprise-wide delirium prevention and treatment program
Predictive modeling to enable targeted outreach and interventions that reduce clinic no-shows
Standardization and evaluation of health equity metrics
Analysis of high-frequency decision-support alerts, with improvements to increase their effectiveness
Complex process flow analysis to identify barriers to discharge and provide analytics solutions to improve patient throughput and length-of-stay

Process Improvement

Standardize opioid metrics for the enterprise to facilitate measurement of UCSF Health's impact on the opioid epidemic
Use data and analytics to automate manual compliance audits of high-risk medication administration
Leverage EHR data and processes to display real-time status of impending discharges and enable day-of-discharge order prioritization

Clinical Project Consultation & Development

Translate clinical data requirements into analytic solutions that enable clinicians to improve patient care
Translate and align visions of clinical content improvement from diverse stakeholders into usable tools and effective processes in the electronic health record
Gather and prioritize requirements for real-time dashboards in the outpatient setting

See more about our recent work:

Artificial Intelligence in the Health System: Governance and Infrastructure for Safe Deployment	Equitable, Targeted Interventions to Decrease No-Shows: An Analytic Approach	Preventing Residency Work Hour Violations: A Novel Use of EHR Data	Extracting Actionable Insights from External Rankings	dashboards.ucsf: The Enterprise Dashboard Destination
We have developed infrastructure (HIPAC) and a governance process to enable safe deployment of artificial intelligence (AI) in the health system.	We analyzed disparities in existing Apex tools for reducing no-shows in outpatient appointments. We built analytic tools to address those disparities and to support equitable and targeted interventions for reducing no-shows.	We built a tool to automatically measure and help prevent work hour violations. The Internal Medicine residency program reduced their inpatient work-hour violation rate by 50% by implementing interventions centered around daily use of this tool.	We designed and deployed analytic tools to identify where strategic improvements will have the most meaningful impact.	We created a one-stop shop for unrestricted, secure enterprise-level dashboards for the entire UCSF community.

People

Sara Murray, MD, MAS, Vice President, Chief Health AI Officer for UCSF Health; Associate Chief Medical Information Officer, Inpatient Care and Data Science; Associate Professor, Hospital Medicine