Deployed systems and outcomes

A standardized discharge checklist, a machine learning readmission risk model, and automated follow-up were built into the electronic health record for every patient admitted with heart failure, work that earned the ZSFG Performance/Quality Improvement Award for standardizing inpatient heart failure care.

Readmissions fell from 27.9% to 23.9%, heart failure mortality fell by 6%, and the racial gap in readmissions was eliminated. The program retained $7M for the safety net, a 7:1 return, and ZSFG earned U.S. News High Performing recognition for heart failure in 2026.

A decision-support pathway helps emergency clinicians evaluate chest pain consistently and discharge low-risk patients safely.

The pathway avoids more than 100 admissions each year, approximately two every week. Thirty-day heart attacks fell significantly after go-live, long-term mortality fell in survival analysis after the pathway launched, and patients are evaluated faster with fewer downstream tests.

AI identifies severe valve disease and routes patients to treatment.

Patients were four times more likely to receive timely specialist care.

The platform identifies social needs from clinical and social records and matches patients to resources. It summarizes complex medical and social information so care teams can spend more time with patients and less time searching through charts.

It agrees with expert human reviewers in 89% of cases and cuts chart review time by more than 60%. It is deploying with ZSFG social workers now, who helped shape the project.