Hitachi and Partners Connected Heart
are utilizing AI to predict with higher
accuracy hospital readmission risks
among heart failure patients

Hitachi and Partners Connected Health assess readmission for heart failure with AI

December 15, 2017
by John R. Fischer, Senior Reporter
Hitachi Ltd. has teamed up with Partners Connected Health in a collaboration overseeing the use of artificial intelligence in predicting the risk of hospital readmission among heart failure patients, with explanations for its reasoning.

The Japanese-based company is utilizing its AI technology to predict with high accuracy which patients are most likely to be readmitted within 30 days of their initial release with those selected then enrolled into Partners Connected Cardiac Care Program, a remote monitoring and education program for patients with heart failure.

“The reason to limit the research target to a certain illness is to avoid complicating the problem. The reason to choose cardiac patients is that its number of patients is large, and it is a representative illness for higher medical cost from readmission,” Toru Hisamitsu, project manager of the collaboration for Hitachi, told HCB News. “It's not a challenge that is just limited to heart failure. We think that providing high quality and suitable medical care for every patient is required and the importance of value-based health care is receiving attention worldwide.”

The thirty-day readmission rate is an important aspect of hospital management and one that can incur penalties for hospitals from the U.S. Centers for Medicare and Medicaid (CMS) in accordance with the Affordable Care Act (ACA).

The Partners Connected Health innovation team simulated the readmission prediction program among heart failure patients participating in CCCP, while Hitachi used its AI technology to construct the prediction model.

The results of the project were compared to information on approximately 12,000 heart failure patients hospitalized and discharged by Partners HealthCare hospital network in 2014 and 2015. The prediction algorithm displayed a high accuracy rate of 0.71 in area under the curve (AUC) and proved it could significantly reduce the number of patient readmissions with an expectation of approximately $7,000 saved per patient per year among the cohort of CCCP patients.

The technology also extracts factors of information from each patient that contributed to the predictions, allowing it to explain its reasoning to enable physicians to make better medical decisions for clinical practice.

Both companies plan to conduct a joint study to evaluate the prediction program by clinicians and examine how to integrate it within clinical workflows.

“PCH will investigate the optimization of its care program based on the risk factors extracted by Hitachi's AI for the cardiac patients who have higher risk to readmit,” Himatsu said. “Also, Hitachi will investigate the brush-up of the AI algorithm for it. At the end, we will contribute to the improvement of medical treatment and the realization of cost reduction by providing [care] that [is] suitable to each patient.”