Health system slashes LOS with Epic-linked AI predictive analytics

Healthcare IT News – Wednesday, January 28, 2026
By Bill Siwicki

A physician’s expertise and clinical experience vary widely, from seasoned clinicians to newly trained graduates.

THE CHALLENGE

This variability can lead to differences in how physicians assess a patient’s readiness for discharge. Less experienced physicians might be more cautious, which is one contributing factor to the longer length of stay observed among their patients when compared with those managed by more experienced colleagues, said Dr. Daniel Craig Kombert, associate vice president of medical affairs at Hartford HealthCare.

Prolonged hospitalization beyond what is clinically necessary increases a patient’s exposure to hospital-associated harms, including infections, falls and deconditioning related to decreased mobility, he added.

PROPOSAL

Holistic Hospital Operations, more commonly known as H2O, is a company created by Hartford HealthCare and MIT applied mathematician Dimitris Bertsimas. The health IT tool built by H2O uses artificial intelligence and machine learning to optimize hospital operational efficiency including improving length of stay and staffing efficiency.

“The H2O team analyzed patient-centered data in a de-identified manner, ensuring patient privacy while enabling robust system-level insights,” Kombert explained. “The objective was to leverage this extensive dataset to better understand patient flow across the continuum of hospitalization – from admission through discharge.

“By applying advanced analytics to these data, the goal was to develop actionable tools that support frontline care teams, enhance operational decision-making, and improve how hospitals manage patient flow and throughput,” he noted.

MEETING THE CHALLENGE

As development progressed, physician leaders were intentionally embedded within the design team to enhance model accuracy and ensure the solutions generated by the H2O tool directly addressed real-world patient throughput challenges. The tool initially was deployed in a pilot across multiple inpatient units, allowing for refinement based on frontline feedback.

“Output from the predictive model was integrated into existing unit-based progression rounds across Hartford HealthCare,” Kombert said. “These multidisciplinary rounds focus on reviewing each patient, identifying barriers to discharge, and establishing a physician-determined expected date of discharge following team discussion.”

The AI tool was incorporated into this process to support accountability to the planned discharge date.

“By comparing the physician-determined EDD with the AI-generated discharge readiness prediction, teams are able to identify discordance between anticipated and predicted discharge timelines,” he explained. “These discrepancies prompt focused discussion to uncover modifiable barriers, align expectations and reduce unnecessary prolongation of length of stay.

“Initially accessed through a standalone platform, the tool was subsequently embedded directly into the Hartford HealthCare Epic environment in response to physician feedback, significantly improving usability, adoption and integration into daily clinical workflows,” he continued.

RESULTS

Within Hartford HealthCare, nearly 100% of admitted medical patients have a predicted discharge readiness assessment available within 24 hours of hospital admission. These predictions are operationalized and integrated into daily care management workflows.

“By leveraging the H2O predictive analytics tool – alongside standardized progression rounds, the establishment of a clearly defined, physician-determined expected date of discharge, and enhanced unit-based leadership through geographic rounding with physicians primarily assigned to specific units whenever feasible – HHC has achieved around a 5% reduction in overall length of stay compared to the pre-implementation period,” Kombert reported.

“This improvement reflects the combined impact of predictive analytics, disciplined operational processes, and improved physician-unit alignment on systemwide patient throughput and discharge efficiency,” he concluded.

Access this article at its original source.