From Paper to AI: Houston Methodist's Journey to Smarter Patient Handoffs

Few things burden hospital employees as much as distilling and relaying essential information from one caregiver to another during patient handoffs.

Busy emergency rooms, high surgical volume and capacity-filled medical and post-surgical units generate an outpouring of documentation. At Houston Methodist, a 10-day stay results in some 3,000 pages of records — from clinical notes to medication history to test and procedure outcomes — all constantly being updated.

Given that U.S. hospitals average about 4,000 patient handoffs per day, the question becomes: Is there a more efficient way to transfer information besides relying on employees to comb through the records themselves?

Houston Methodist's answer: generative artificial intelligence.

In the past couple years, Houston Methodist kicked off a pilot program using generative AI to produce real-time patient summaries and predict discharge dates within patients' electronic health records. The synthesized summaries guide caregiver handoffs at shift change, length of stay management, team communication, discharge barrier resolution and the need for post-discharge care.

Early results show reduced lengths of stay and lowered readmission rates.

"We used to have to sit in a conference room for an hour to get input from multidisciplinary teams on where the patient is going, what the barriers for discharge were and when they could leave safely," recalls Jennifer Jaromahum, director of Nursing at Houston Methodist Willowbrook Hospital and a leader of the pilot program. "We no longer have to do this because AI has already summarized it for us in a structured, easy to read format."

Instead of scouring patient charts, nurses are now able to spend more time inside patient rooms talking to them about their plan of care.

"It's great not just for hospital efficiency, but in keeping our patients and families informed," Jaromahum said. "Nursing has truly come a long way from paper and pen," she adds.

Since the implementation of the program, Houston Methodist's preliminary HCAHPS data has already shown improvement in care coordination and doctor/nursing communication domains, Jaromahum noted.

The hospital uses software produced by Dallas-based Pieces Technologies, which uses natural language processing and a "SafeRead" system to pull insights from clinical notes and records.

Jaromahum says the program doesn't just produce patient chart summaries that highlight key information about hospitalization and changes in condition, it improves multiple interactions, including physician to physician, nurse to nurse, physician to patient family and emergency department to inpatient floor.

It also provides doctors a working summary with real-time documentation suggestions about the use of specific medical terminology based on best practice; captures diagnoses; and highlights prior medical history notes that were previously buried in the medical record. And because it is constantly updating, it helps rapid response teams get a quick snapshot of a patient on the way to responding to a code or other emergency.

"This program illustrates Houston Methodist's commitment to an innovative culture, with hands-on involvement from leadership and consistent feedback from staff regarding the program's development," said Jaromahum. "Such vital input, rooted in real-world clinical expertise, directly fuels the refinement and accuracy of our generative AI. It's incredible how much the software has already learned and improved on over a short period of time."

The early returns are impressive. An analysis of the program found that AI-generated patient summaries required edits less than 5% of the time. It also found the program detected 34,000 barriers to discharge within just one month and identified an important subset of patients at more than a five-fold higher risk for ICU transfer.

With a clear focus on enhancing patient care, Jaromahum says the next focus of generative AI at Houston Methodist will be a continued reduction in physicians' administrative clinical documentation burdens, allowing for more time to direct quality patient care.

This article originally appeared in Innovation Spotlight, a LinkedIn newsletter on how Houston Methodist is leveraging technology to shape a smarter health system.