Analytics in Action News

Machine-Learning Model Predicts Risk of Pediatric Deterioration

Nationwide Children's Hospital researchers utilized a machine- learning tool with an EHR-integrated risk index algorithm to alert providers of early pediatric deterioration.

Source: Getty Images

By Sarai Rodriguez

- Nationwide Children's Hospital developed and deployed a machine-learning (ML) model that uses the deterioration risk index to promptly predict hospitalized children at risk for pediatric deterioration earlier than previously implemented programs, according to a study published in the Pediatric Critical Care Medicine journal.

Earlier identification of high-risk patients is crucial in preventing adverse events and code blue situations, as patient deterioration can rapidly escalate from seemingly ordinary to critical. For organizations that see large numbers of medically complex patients, risk-scoring methods are particularly helpful.

"Predictive algorithms focused on improving clinical care have been increasingly developed over the years, but the vast majority are not operationalized," Laura Rust, MD, emergency medicine physician and physician informaticist at Nationwide Children's and lead author of the paper, said in a press release. "Transitioning the algorithm from the computer to the bedside can be a long process and requires engagement and collaboration from clinicians, data scientists and clinical informaticists. This project has been a 5-plus year journey, and we are really proud of the successful integration within our safety culture and the impact on patient outcomes."

The Deterioration Risk Index (DRI), based on a Watchstander program already used at Nationwide Children's Hospital, leverages familiar alert responses to promote adoption, such as patient assessments, care team huddles within 30 minutes, risk mitigation, and escalation plans.

EHRs often contain excessive clinical data, particularly after patient care transitions, and the model is designed to alleviate clinician burden by automatically processing risk criteria, Rust explained. Further, its integration within the EHR enables access to data from all previous points in time across the care continuum.

Three predictive models were trained using data from three diagnostic groups: cardiac, malignancy, and general (neither cardiac nor malignancy). Researchers then used the models to develop the algorithms leveraged by the tool.

The patient population included all individuals admitted to inpatient units participating in the preexisting situational awareness program from Oct. 20, 2015, to Dec. 31, 2019. Exclusions were made for patients over 18 years old at admission and those with a neonatal intensive care unit stay during their hospitalization.

"One of the design features that helped build trust with clinical teams is that we didn't necessarily identify any new criteria. Our model simply identifies which existing situational awareness criteria are most important and weighs them accordingly," said Tyler Gorham, data scientist in IT research and innovation at Nationwide Children's and co-author of the publication, in the press release.

The study revealed that the DRI exhibited 2.4 times greater sensitivity than the existing situational awareness program, with a four-fold increase in sensitivity observed for the cardiac group and a three-fold increase for the malignancy group. Moreover, the model demonstrated more precise alerting, requiring 2.3 times fewer alarms per detected event.

Following implementation, the pilot study reported a 77 percent reduction in deterioration events during the first 18 months compared to the situational awareness program.

Transparency was the most significant aspect of the model, researchers stated.

"This is not a black box. We show clinicians what goes in and how the algorithm evaluates the data to trigger alarms," said Gorham. "The tool helps support clinical decision making because the clinical team is able to see why an alarm was triggered."

Lately, risk prediction ML models have gained popularity throughout the healthcare sector.

For example, researchers at Ohio State University (OSU) recently announced a new ML model to accurately estimate optimal timing for sepsis treatment.

Sepsis is a life-threatening condition that can rapidly lead to tissue damage, organ failure, and death without timely treatment. However, the symptoms of sepsis — such as breathing problems, high heart rate, low blood pressure, and fever — can resemble those of multiple other conditions,

The team developed an ML model to determine the best moment to administer antibiotics to patients with suspected sepsis cases, enhancing clinical decision-making support.