New research uses AI to potentially reduce hospital readmissions for children

Mukul Sehgal, M.D., assistant professor of pediatrics, cares for a patient at USA Health Children's & Women's Hospital. His research is using artificial intelligence to potentially reduce readmission rates in hospitalized pediatric patients with pulmonary hypertension.

For many children living with hypertension and other chronic conditions, their young lives can become a series of endless doctor visits, needle jabs for lab work and other dreaded treatments, punctuated by extended hospital stays when their conditions worsen. So if physicians could more accurately predict when a child with a chronic illness was going to become critically ill, they could take steps to intervene before their health declined and hospitalization was required.

That’s the idea behind new research from Mukul Sehgal, M.D., assistant professor of pediatrics at the USA College of Medicine and a pediatric critical care physician at USA Health Children’s & Women’s Hospital.

“Working in a pediatric intensive care unit, we see quite a few patients who get admitted for similar medical problems multiple times a year,” Sehgal said. “This causes increased morbidity among children who spend prime years of their life recovering from these diseases. We decided to explore ways in which we can identify these high-risk patients and came across the use of artificial intelligence (AI).”

An evolving tool increasingly used in healthcare, AI can help researchers construct prediction models to identify the patients most at risk with much higher accuracy than traditional methods, Sehgal said.

Sehgal and his colleagues used AI to examine the National Readmission Database from 2017 for a review of 5.52 million previous instances where children age 18 and younger with hypertension were readmitted to hospitals. What he discovered is the topic of a January 2022 research article published in the journal Critical Care Medicine. 

Amod Amritphale, M.D., assistant professor of internal medicine at the USA College of Medicine and an interventional cardiologist with USA Health, is a contributing author on the study. 

The research, which took about 10 months to complete, also will be presented at an international conference of the Society of Critical Care Medicine.

Specifically, the research and article highlight the importance of AI in reducing the unplanned 30-day readmission rates in hospitalized pediatric patients with pulmonary hypertension, Sehgal said. By identifying certain characteristics of patients who are discharged from the hospital, clinicians can better predict their probability of being readmitted in 30 days after discharge.

Among the findings was that respiratory infections and mechanical ventilation were associated with increased chances of readmissions among patients suffering from pulmonary hypertension.

“This can help us recognize those patients,” he said, “and then schedule an early follow-up with a cardiologist for them, and thus help prevent their readmission” and the escalation of their illness.

Pediatric hypertension has been on the rise for more than 30 years and often is associated with other health conditions. Research from the American Heart Association found that more than three million children in the United States have the condition. Recent AHA heart disease and stroke statistics also show the number impacted may be even higher, estimating some 15% of adolescents have abnormal blood pressure.

Sehgal is excited to present the findings to an international audience: “Sharing our scientific research with such a community opens the door for future collaboration and should help many children across the world. While we did our research on pulmonary hypertension, this can be easily expanded to other diseases and in other age groups.”

Read the article in Critical Care Medicine.