Children's National, KenSci studying the use of AI to improve pediatric critical care

Exterior of Children's National
Children's National Health System and KenSci plan to use artificial intelligence to better predict pediatric patients at risk of further deterioration in the intensive care unit. (Children's National)

Children’s National Health System in Washington, D.C., is collaborating with technology company KenSci to study the use of artificial intelligence to improve pediatric risk scores in the intensive care unit.

Through the research collaborative, Children's National and KenSci, a company that develops machine learning software, will work together to develop new models to understand the factors that impact criticality in pediatric patients. The organizations aim to enhance pediatric risk scores to improve clinical decision-making in critical care.   

Risk scores have been used since the mid-'80s to predict mortality risks in pediatric ICUs, according to Murray Pollack, M.D., director of outcomes research at Children’s National and a professor of pediatrics at George Washington University School of Medicine. In most cases, these scores are used for quality assessment, he said.

Whitepaper

Key Realities Pushing Healthcare Into a Digital Future

Paper forms, contracts, and documents are the quicksand that bogs down both patient care and provider business. However, that does not have to be the case. Download this whitepaper to learn the three key realities that are pushing healthcare past paper-based processes and into a digital, more streamlined future.

“Our collaborative goals are to study the temporal variation in data, taking the first step towards dynamic risk scoring for pediatric ICUs," Pollack said.

RELATED: Verily’s India project reveals the work of applying AI to real-world clinical care

Hiroki Morizono, Ph.D., director of biomedical informatics at the Children’s National Center for Genetic Medicine Research, said the AI modeling could predict an individual patient’s likelihood of deterioration or improvement.

“We see tremendous possibilities for how this research can be used safely and securely to supplement the clinician’s judgment,” Morizono said.

The joint team will use KenSci's AI platform to test different models and compare their predictive effectiveness to prior baselines developed by Children’s National and George Washington University. 

Ankur Teredesai, KenSci’s co-founder and chief technology officer, said in a statement: “Time is our best ally if integrated appropriately with other variables in healthcare machine learning and AI. Adding dynamism holds tremendous promise to be assistive for critical care."

RELATED: Kaiser Permanente develops machine learning tool to predict HIV risk

Hospitals and health systems are increasingly using AI and machine learning to analyze data to support diagnosis and treatment. AI is being applied to oncology to help physicians manage volumes of data and to rapidly annotate the results of tumor genome sequencing, identifying potential therapeutic options.

Eye doctors at Aravind Eye Hospital in Madurai, India, are using a machine learning algorithm developed by technology company Verily to screen patients for diabetic retinopathy, intending to use technology to fill gaps in patient access to eye doctors. 

HCA Healthcare developed an AI algorithm to more quickly identify patients with sepsis. The health system said its sepsis prediction and optimization of therapy technology so far has been used with 2.5 million patients and, together with the use of clinical interventions, has helped save an estimated 8,000 lives in the last five years.

Suggested Articles

Specialty drugs made up 1% of prescriptions for employers but accounted for 40% of total drug spending last year, an analysis found.

A collaboration between California payers and providers yielded millions in savings and prevented thousands of unneeded ER visits and admissions. 

Physicians certified by the American Board of Internal Medicine will soon have a new option that takes some of the pain out of MOC.