Multicenter Clinical Trial Examines Predictive Analytics to Support Decision-Making in Pediatric Cardiac Critical Care
February 16, 2026
Avihu Z. Gazit, MD, inaugural Patrick Dick Memorial Chair in Pediatric Cardiology, chief of the Pediatric Cardiac Critical Care Division, Department of Critical Care Medicine, and co-director of the Heart Institute at UPMC Children’s Hospital of Pittsburgh, was the lead author on a multicenter study evaluating the use of real-time predictive analytics to support clinical decision-making in pediatric cardiac intensive care. The research examined if physiologic risk analytics integrated into bedside care dynamics can help to standardize decisions around weaning patients use of vasoactive medications after congenital heart surgery.
“This aspect of patient care after heart surgery for a congenital condition is an area where practice variation remains common despite its clinical importance,” Dr. Gazit says.
The study found that incorporating analytics-informed decision support into clinical rounds was associated with a reduction in the duration of vasoactive infusion use without an increase in adverse events or weaning failure. The findings suggest that predictive tools can help clinicians identify when a patient is ready for de-escalation of therapy while maintaining safety. The study’s findings support a more data-informed approach to postoperative management in this patient population.
Interview in HealthLeaders Media
In a recent interview with HealthLeaders Media, Dr. Gazit discusses the broader implications of this research, including practical considerations related to the adoption of artificial intelligence–enabled tools in pediatric critical care settings, clinician trust in decision support systems, and the operational factors that can influence successful implementation.
Viewpoint Article in JAMA Pediatrics on AI in Critical Care Medicine
In January, Dr. Gazit and colleagues published a viewpoint article in JAMA Pediatrics on ”AI in Critical Care – Use for De-escalation Rather than Escalation of Care.”
This opinion piece emphasizes the paucity of implementation of predictive analytics models in pediatric intensive care, commonly referred to as the AI chasm, and proposes a novel approach to bridge that gap, namely: utilizing AI-driven clinical decision support tools during the de-escalation process rather than the acute phase of critical illness.
Dr. Gazit and colleagues rationale is that during the acute phase, patients exhibit extreme biological variability that requires intense team focus and rapid responses to clinical changes. Hence, rendering near-real-time predictive algorithms redundant. Moreover, during this phase of care, the likelihood that critical care teams would be equipoised in their use of predictive AI tools is minimal.
As for the de-escalation phase, Dr. Gazit and colleagues hypothesize that, because of decreased biological variability and overall improvement in patients’ clinical status, fewer resources are allocated to them. As a result, the response to changes is slower, and greater variability in care is observed. Utilizing AI-driven clinical decision support tools during this phase, as shown in the multicenter study Gazit et al conducted, has the potential to decrease variability in care and possibly decrease the exposure of these vulnerable patients to the intensive care unit. Equipoise should, therefore, be easier to obtain.
Read the full viewpoint in JAMA Pediatrics.
Study Reference
Gazit AZ, Futterman C, Baronov D, Tomczak A, Goldsmith MP, Talisa VB, Nadkarni VM, Laussen PC, Salvin JW. Risk Analytics Clinical Decision Support Decreases Duration of Vasoactive Infusions Following Pediatric Cardiac Surgery: A Multicenter Before and After Clinical Trial. Crit Care Med. 2025 Jul 1; 53(7): e1355-e1364.
