As of 2023, over 1.2 million individuals in the US receive nursing home care, the majority of whom are covered by Medicare. Nursing homes provide clinical care to patients recovering from illness, those undergoing rehabilitation after injury, and persons with long-term physical, cognitive, or mental impairments. Over the past decade, however, an increasing number of facilities have closed. Within the past four years alone, approximately 28,421 residents were displaced.
Nursing homes face numerous challenges, including chronic underfunding, understaffing, and high staff turnover, which lead to poorer quality of care for residents. As ongoing financial and workforce pressures threaten the stability of facilities, identifying factors associated with nursing home closures and predicting future risks is imperative.
To address this need, Rahul Fernandez, senior research assistant, Dr. Robert Tyler Braun, assistant professor of population health sciences, and Dr. Yongkang Zhang, assistant professor of population health sciences, conducted a study published in npj Health Systems that identified factors that best predict nursing home closures, as well as means of improving closure predictions. Using this information, they developed machine learning and deep learning models applied to longitudinal data to shed light on key drivers of closures and to enhance the accuracy of closure prediction.
Results show that nursing home closures tend to be concentrated in socioeconomically disadvantaged areas and that rural facilities have higher odds of closure. This indicates that communities with limited resources also suffer disproportionately from facility loss. The study also demonstrated that machine learning and deep learning methods could predict US nursing home closures with reasonable accuracy, which carries several implications for state and federal regulators. Researchers emphasize that these methods can identify at-risk facilities in advance, thereby supporting more proactive oversight, targeted interventions, and equitable resource allocation.
- Highlights