Using AI to Forecast Patient Deterioration

Using AI to Forecast Patient Deterioration

A predictive tool to identify trends and stay ahead of deterioration reduced crisis response calls by nearly 40% in an acute care unit.
Medical staff examines patient and records notes in EMR.

For years, clinicians at Akron Children's relied on the Pediatric Early Warning Score (PEWS) system to evaluate the condition of its admitted patients. Though it was effective, hospital leaders wanted to improve the tool.

“We wanted a more comprehensive, big-picture look at the patient’s clinical status,” said Erin Pitchure, BSN, RN, quality initiative specialist at Akron Children’s.

That quest led the hospital to launch a pilot program in 2019 based on the Rothman Index, a predictive analytics program that helps clinicians assess a patient’s condition and predict the possibility of deterioration over time.

During the pilot stage, the tool reduced Medical Response Team (MRT) calls by nearly 40%.

Akron Children's assembled a multi-disciplinary team to build and deploy the tool referred to internally as the “deterioration index.” The initial team included intensive care physicians, infection control, information technology, electronic medical record (EMR) vendor representatives, and nursing informatics and leadership. Over time, the need for such a large team has shifted to a smaller group, but one aspect remains constant: empowering the voices of frontline staff members.

“We’ve always been very open to suggestions. It’s really important to ensure that anyone involved with the tool has a say,” Pitchure said. Provider input in the deterioration index ruleset provides consistency in practice and language around patient care, according to Pitchure.

Deterioration index: How it works

The tool integrates seamlessly into standard clinical processes. Patient data from nursing assessments, vitals, and lab results is automatically pulled from the EMR into the model as it is documented. Based on the data, the algorithm produces a deterioration index score indicating a patient’s risk factor and populates a graph illustrating the patient’s trendline over time. Care providers can see trends developing at a glance and use the index scores to validate their own judgments when they feel something is not right with a patient.

The tool also issues warnings when a patient’s score decreases over a period of time (lower scores are more concerning), prompting a bedside huddle of the care team. Additionally, kiosks on the floor enable nursing leaders to instantly assess the index metrics for all patients in the unit.

Though the tool was originally designed for the adult population, Akron Children’s worked closely with its developers to tailor it to pediatrics. For example, the algorithm accounts for metrics with significant age-based variances, including heart and respiration rates, blood pressure, and serum creatinine levels.

Because the tool leverages already-documented information and runs continuously in the background, it has added no strain to clinician workloads.

“With PEWS, you had one more item you needed to document,” said Ericka Moore, MSN, RN, CPN, clinical informatics specialist. “But this tool generates a score based on everything you're already doing, so there’s no manual legwork involved.”

As the program has rolled out from the pilot to deployment across the hospital, Pitchure, Moore, and their team have standardized implementation to ensure consistency. They provided education modules, user guides, bedside huddle scripting and documentation, and decision trees with unit-specific guidance for escalation and communication.

Improving outcomes

The deterioration index has proven valuable. After the nearly 40% reduction in MRT calls during the pilot period, the hospital has sustained a 14% decrease in MRT calls across the hospital following the tool’s rollout. Additionally, the hospital’s clinicians report improved situational awareness, better communication among team members, and reduced cognitive workload with the tool’s implementation.

Pitchure and Moore’s key piece of advice: be patient.

“Take your time. This is not something that happened overnight,” Pitchure said. “Investigate it, work out the kinks, and make it benefit your staff and their workflows. Had we not done that, we wouldn’t be where we are today.”

Read the Latest Issue of Children's Hospitals Today

Children's Hospitals Today Summer 2024

Don't miss the latest industry news, insights and ideas.

View