Interpretable machine learning helps improve patient satisfaction

Researchers at Penn State and Geisinger Health System have used explainable AI to uncover factors that influence inpatients’ opinions when they’re completing satisfaction surveys following hospitalization.

Senior author Eric Reich of Geisinger’s Steele Institute for Health Innovation and colleagues combined serpentine EHR data with scores and accounts from completed satisfaction surveys. They trained an interpretable machine learning framework to wring actionable insights from the disparate inputs.

Presenting their findings in IEEE Journal of Biomedical and Health Informatics, Reich and co-authors report that promptness and attentiveness to patients’ concerns or complaints rose to the top of factors affecting satisfaction scores across the board.

Also important was pain management for patients to whom it applied. So too variables contributing to perceptions of courtesy, respect and communications among and between doctors, nurses and other care-team members.

In their study abstract, the authors note the established connection between satisfaction in the hospital and subsequent compliance with discharge instructions at home.

In turn, treatment adherence can affect clinical outcomes and rehospitalization rates for good or ill.

Commenting on the findings for Penn State News, lead author Ning Liu, PhD, notes that hospitalized patients interact with lots of healthcare workers and support staff in many different settings along the arc of their stay.

“It’s important for providers to understand the needs of each patient group, like those receiving surgery, cancer treatments or emergency visits,” Liu says.

He underscores the importance of the tested framework’s use of interpretable AI, the results of which can help tailor corrections to the particular care points in which they’re most needed.

Compared with black-box AI, the interpretable type can help hospitals “implement change to improve patient satisfaction across various levels, from the top down to the individual unit workers,” Liu says.

Reich adds that the model’s now-demonstrated utility should encourage new research combining patient satisfaction metrics with advanced analytics.

“Healthcare systems can use these findings to drive targeted improvements in patient satisfaction to the point where we know if patients with a certain set of characteristics are getting their knee replaced, then we believe these are the top three items that are going to ensure the patient has a very positive experience,” Reich says. “To discover the key drivers behind patient satisfaction is a critical initiator for improving the quality of patient-centered healthcare.”

Study link here (behind paywall), Penn State News item here.

Dave Pearson

Dave P. has worked in journalism, marketing and public relations for more than 30 years, frequently concentrating on hospitals, healthcare technology and Catholic communications. He has also specialized in fundraising communications, ghostwriting for CEOs of local, national and global charities, nonprofits and foundations.

Trimed Popup
Trimed Popup