Predicting social isolation in maintenance hemodialysis patients using machine learning methods: a cross-sectional...

Area of Science:

• Nephrology
• Artificial Intelligence
• Public Health

Background:

• Social isolation is a significant concern for patients undergoing maintenance hemodialysis (MHD).
• Identifying at-risk populations is crucial for timely intervention.
• Understanding key predictors can inform targeted support strategies.

Purpose of the Study:

• To develop and validate a machine learning (ML)-based risk prediction model for social isolation in MHD patients.
• To identify significant risk factors associated with social isolation in this cohort.

Main Methods:

• Seven ML algorithms were implemented and compared: Random Forest (RF), Decision Tree (DT), K-Nearest Neighbor (KNN), Logistic Regression (LR), Elastic Net (EN), Extreme Gradient Boosting (XGB), and Support Vector Machine (SVM).
• A cohort of 362 MHD patients was recruited and divided into training and detection groups.
• Feature importance analysis was conducted to identify key predictors.

Main Results:

• The incidence of social isolation in the MHD cohort was 45.856%.
• The Random Forest (RF) model demonstrated the best predictive performance with an AUC of 0.95.
• Significant predictors identified include place of residence, heart failure (HF), anxiety, monthly household income, age, and sleep condition.

Conclusions:

• The RF-based prediction model is effective in identifying MHD patients at risk of social isolation.
• These findings support clinicians in stratifying high-risk populations for targeted interventions.
• Further validation in multicenter studies with larger cohorts is recommended.