AI-Driven Interventions for Imminent Hospital Admissions in Patients with End-Stage Kidney Disease: A Medicare and EMR-Based Analysis

Patients with end-stage kidney disease (ESKD) have a high rate of hospitalizations related to fluid overload and infections. Artificial intelligence (AI)-driven models may improve patient care by predicting the risk of hospitalization. The authors conducted a retrospective, observational matched cohort study of adult patients with ESKD who were receiving value-based hemodialysis at integrated kidney care clinics across the United States in 2023. Two AI-powered machine learning models calculated risk scores (range: 0-1) and the models identified patients with a risk score of 0.64 or above who were at risk for hospitalization within 7 days in relation to infections or fluid status abnormalities. To prevent avoidable hospitalizations, case reviews and interventions were conducted for the patients identified by the models. The AI models generated scores for all patients, but only high-risk scores triggered case review and possible intervention. The authors linked electronic medical records and Medicare claims data and conducted multivariate logistic regression analyses to examine the impact of AI-driven interventions on the odds of all-cause hospitalization in patients with ESKD. A total of 10,294 patients representing 83,928 risk scores were included in the analysis. AI-driven intervention was associated with an 8% reduction in the odds of hospitalization within 7 days (odds ratio=0.92; P=0.025). These interventions were most effective for high-risk patients with scores between 0.64 and 0.85, but had no statistically significant effect for patients with scores above 0.85. Factors that were independently associated with higher rates of hospital admission included a higher risk score (>0.75), chronic high-risk scores, older age, and a higher number of hospital admissions in the year prior. AI-driven interventions were associated with a reduction in the odds of hospitalization among patients with ESKD receiving managed kidney care. These findings underscore AI's potential to assist health care providers with targeted risk interventions for patients with ESKD.