Could a Computer Predict Emergency Hospitalizations?

BIRMINGHAM, Alabama (Feb. 27, 2020) — In a new study, the company Health at Scale found that its own artificial intelligence technology was able to partly predict whether people would visit the emergency room in the near future.

The study, “Precision Interception: Machine Intelligence for Actionable Prediction and Prevention of Emergency Department Visits,” tracked 12 months of claims data for a nationwide population of 2 million Medicare beneficiaries. The machine learning system generated a sub-cohort comprising 1% of that population estimated to have an increased likely of future emergency department (ED) visits over the next six months, as well as potential health issues for each member. The study was conducted and published by the company, not in a peer-reviewed journal.

The identified people and the predicted causes of their ED visits were compared to actual outcomes for these members over the next six months. About 81% of those people in the cohort flagged as at-risk did end up visit an ED at least once by the end of the six month evaluation period, and many had multiple emergency hospital visits.

“Emergency department visits are spiraling out of control with many of these visits, potentially preventable through appropriate early care,” Zeeshan Syed, CEO of Health at Scale, said in a news release. “If these trends of increasing emergency department utilization continue, more people will be showing up at hospitals than at their primary care physicians, placing a tremendous, unsustainable burden on the entire healthcare system and blocking patients experiencing true emergencies from getting the care they need in a timely manner. As our study shows, technology that can predict which patients are likely to experience emergency department visits and the causes underlying these visits holds the key to heading off this crisis.”

While the study focused on helping reduce the number and costs of emergency department visits, it could also be utilized to prevent hospitalizations by providing better preventative care, proactive care, and possibly care in the home.

According to the study, the average cost of ED visits and subsequent hospitalizations was $16,000 per member, and the total cost of ED visits and subsequent hospitalization in the sub-cohort totaled more than $266 million. And the authors argued that it has broader applications.

“The results achieved by Health at Scale’s machine intelligence interception are promising and can be scaled out across a variety of member populations for a broader set of health challenges besides ED visits, including hospital admissions and chronic disease onset, to fully realize the value of intelligent precision interception for complex managed populations,” the study says.