How Hartford HealthCare uses AI to cut length of stay and improve patient flow
Physician experience varies. That shows up in discharge decisions, and it adds days patients don't need to spend in beds - with higher exposure to infections, falls and deconditioning.
Hartford HealthCare partnered with MIT applied mathematician Dimitris Bertsimas to co-create H2O (Holistic Hospital Operations), an analytics platform that uses machine learning to improve operational efficiency - specifically length of stay and staffing efficiency.
The challenge
Less experienced clinicians often take a more cautious approach to discharge readiness. That caution is understandable, but across a system it translates into longer stays and more avoidable days.
The cost isn't just financial. Every extra day increases risk and slows throughput for the next patient waiting for a bed.
The approach
H2O analyzes de-identified, patient-centered data to map flow from admission through discharge. The target: actionable tools that help frontline teams make better, faster decisions about readiness for discharge.
Physician leaders were embedded from day one. The team piloted across multiple inpatient units, took feedback from the floor, and iterated until the model matched the reality of daily operations.
How it fits into the workflow
Hartford HealthCare runs unit-based progression rounds to review each patient, identify barriers, and set a physician-determined expected date of discharge (EDD). H2O adds a second signal - an AI prediction of discharge readiness.
When the EDD and the AI prediction don't match, the team digs in. That tension surfaces fixable barriers, aligns expectations and reduces unnecessary prolongation of stay.
The tool started as a standalone app, then was embedded directly into the Epic EHR based on physician feedback. That move boosted usability, adoption and day-to-day impact.
Results that matter to operations
Nearly every admitted medical patient now has a discharge readiness prediction within 24 hours of admission. Those predictions are used in daily care management workflows, not parked in a dashboard.
Combined with standardized progression rounds, clear EDD ownership, and unit-based physician alignment (geographic rounding), Hartford HealthCare reports about a 5% reduction in overall length of stay versus the pre-implementation period.
- Predictions available within 24 hours for almost all admitted medical patients
- ~5% reduction in length of stay systemwide
- Greater accountability to planned discharge dates through EDD vs. AI comparison
- Better integration and adoption after embedding into Epic
Why this works
It's not AI in a vacuum. It's AI paired with disciplined operational routines. The model provides an early, consistent signal; the team uses that signal to remove barriers and move patients when it's safe to do so.
Keeping physicians in the loop - both in design and in daily rounds - builds trust and ensures the final decision stays clinical, not purely algorithmic.
How to replicate in your hospital
- Build a cross-functional team with physician leaders, nursing, care management and IT.
- Use de-identified data and clear governance; socialize privacy protections early.
- Pilot on a few units, gather frontline feedback and iterate before scaling.
- Anchor the process in standardized progression rounds and a physician-owned EDD.
- Integrate the tool into your EHR to reduce click burden and improve adoption.
- Use EDD vs. AI discrepancies to drive targeted problem-solving on barriers.
- Track a core metric set: percent of patients with predictions < 24h, LOS, avoidable days, and throughput KPIs.
What's next
The same data and methods that shorten stays can inform staffing efficiency and bed capacity planning. Forecasting discharge volume by unit and day unlocks better staffing alignment and fewer last-minute scrambles.
Related resources: MIT on analytics in operations * Epic EHR
If you're building AI literacy across your operations team, see curated options here: AI courses by job.
Bottom line: Pair a reliable prediction with a disciplined rounding process and clear ownership of the discharge plan. That's how you turn analytics into shorter stays, safer care and smoother flow.
Your membership also unlocks:
