Barrett MA, Allen A, Vuong VT, Zhu D, Rainey AJ, McConnell WM, Kho AN, Salisbury A, French DD
Objectives
This study evaluated the impact of an electronic health record (EHR) system enhanced with artificial intelligence and machine learning (EHR+AI) on quality measures in nursing homes in the United States.
Design
A difference-in-differences (DiD) design was used to estimate the effect of the EHR+AI intervention on quality measures among nursing homes with and without the AI intervention. The intervention included a feature that analyzed 150 daily clinical data elements per patient, alerting staff to changes in conditions, acuity, fall risk, and medication monitoring.
Setting and Participants
The analysis included 218 nursing homes, with 94 using EHR+AI and 124 using EHR only. Baseline differences in organizational characteristics, acuity index, neighborhood affluence, and racial or ethnic composition were evaluated.
Methods
Eighteen quality measures from the Centers for Medicare and Medicaid Services (CMS) were analyzed over 6 quarters before and 5 quarters after EHR+AI implementation. A DiD approach with linear mixed effects models was used, adjusting for significantly different baseline characteristics.
Results
Statistically greater improvements were observed in 16 of 18 quality measures (89%) in EHR+AI sites, with 11 measures (61%) also meeting the parallel trends assumption. Notably, EHR+AI sites demonstrated larger improvements in functional status, including greater reductions in major falls (−9%, 95% CI –17, −1; P = .034) and residents needing help with daily activities (−22%, 95% CI –29, −15; P < .001), and a 5% larger increase in residents who made improvements in function (95% CI 2, 7; P = .001). Higher decline in depressive symptoms and the use of antipsychotic, antianxiety, or hypnotic medications were also noted. These results were observed among sites with higher patient acuity and neighborhood diversity.
Conclusions and Implications
These findings suggest that an EHR enhanced with AI can improve the quality and efficiency of care in nursing homes through real-time monitoring and response of resident assessment protocol triggers for clinical modification, but further research is needed.