ipopba/istock via Getty Images
Clinical deterioration AI contributes to reduced care escalation risk
An AI-driven clinical deterioration intervention was associated with a 10.4 percent reduction in the risk of escalations of care during hospitalization.
Researchers have demonstrated that an artificial intelligence (AI) model designed to detect clinical deterioration was associated with a significantly decreased risk of inpatient escalations in care, according to a study recently published in JAMA Internal Medicine.
Clinical deterioration is a major driver of morbidity and mortality, but identifying early warning signs of deterioration can be a challenge for care teams – contributing to an estimated 15 percent of avoidable deaths in the hospital.
Recognition of the signs of clinical deterioration can be improved through the use of clinically relevant interventions. The research team indicated that early warning scores have been developed to help flag high-risk patients, but evidence of their effectiveness is limited.
To address this, the researchers set out to assess an AI-driven clinical deterioration detection tool’s ability to reduce the risk of care escalations among hospitalized adults and gauge the causal inference between the two.
The analysis included data from 9,938 patients hospitalized across four internal medicine units within one academic medical center from January 17, 2021, through November 16, 2022. The AI-enabled intervention consisted of Epic Deterioration Index (EDI)-based alerts and an associated clinical workflow.
Escalations in care were characterized as rapid response team activation, transfer to the intensive care unit or cardiopulmonary arrest during hospitalization.
Overall, the intervention was associated with an absolute risk reduction of 10.4 percentage points in care escalations for patients at the EDI score threshold.
These findings led the researchers to conclude that the implementation of an AI-driven clinical deterioration model can significantly reduce the risk of inpatient escalations in care. Further, they indicated that their study provides evidence for the effectiveness of such an intervention and necessitates testing of the model in additional care settings.
Other studies have also highlighted the potential of deterioration models to improve care for hospitalized patients.
Last year, researchers from Nationwide Children's Hospital found that a machine learning model could accurately predict risk of pediatric deterioration.
The tool uses the Deterioration Risk Index (DRI) to forecast the risk of cardiac, malignancy and general (neither cardiac nor malignancy)-associated deterioration.
The model was both more accurate and precise than existing tools, exhibiting 2.4 times greater sensitivity overall, and a three- and four-fold increase in sensitivity in the malignancy and cardiac groups, respectively, while requiring over two times fewer alerts per each detected event.
Following a pilot project deploying the tool in the clinical setting, the researchers noted a 77 percent reduction in deterioration events for the 18 months following implementation compared to a previous situational awareness-based program.
