AI Models Can Predict Individual Risk of Hospitalization for COVID Using HIE Data

In a new study, researchers at Regenstrief Institute (Indianapolis, IN, USA) and Indiana University (Bloomington, IN, USA) demonstrated that machine learning models trained using clinical data from a statewide health information exchange can predict, on a patient level, the likelihood of hospitalization of individuals with the virus. The researchers used clinical data from 96,026 individuals from all 957 zip codes in Indiana to train decision models that predicted healthcare resource utilization.

“It has been quite challenging to bring the bread-and-butter data generated by healthcare systems together with public health decision-making - entities which have long been separate and distinct,” said study senior author Shaun Grannis, M.D., M.S., Regenstrief Institute vice president for data and analytics and professor of family medicine at Indiana University School of Medicine. “Our work shows how you can build and employ AI (artificial intelligence) models to securely utilize the clinical information in a health information exchange to support public health needs such as predicting hospital utilization within one week and within six weeks of onset of COVID infection.

“When new circumstances requiring rapid response arise, such as emergence of omicron or other new variants, once there are sufficient cases to train models, one can confidently access and plug clinical data into these readily available models to make accurate public health predictions and provide valuable insights into patient-level need for healthcare resource utilization,” said Dr. Grannis.

“Since the onset of COVID-19, researchers, healthcare systems, public health departments and others have leveraged existing data repositories and health information infrastructure for rapid analytics,” said study first author Suranga Kasturi, PhD, a Regenstrief Institute research scientist and an assistant professor of pediatrics at IU School of Medicine. “Machine learning has been invaluable in these efforts.”

“But any model is only as good as the data that goes into it,” he added. “The broad, robust data from the Indiana Network for Patient Care is representative of the US population. What we have done could be characterized as a precursor of how AI tools can be deployed across the entire country with the important caveat that whatever models are used should be evaluated for fairness across all subpopulations.”

Related Links:
Regenstrief Institute 
Indiana University