Disability Risk Prediction in Older Inpatients Using EMR Data and Machine Learning

523 patient records from two tertiary hospitals in Northeast China were used to develop machine learning models for predicting four levels of disability based on the Barthel Index. The study, authored by Guangpeng Chen and last updated in 2026, identified 17 critical predictors including age, muscle mass indicators, and inflammatory markers. The Logistic Regression and XGBoost models demonstrated the best performance with an AUC of 0.83 and Accuracy of 0.91.

Use Cases

• Predicting disability risk levels in older inpatients based on demographic and clinical features mentioned in the description.
• Training machine learning models for early clinical decision support using electronic medical record data.
• Interpreting model predictions for disability risk using SHAP analysis on identified predictors like neurological comorbidities and nutritional-inflammatory status.

Strengths

• Includes data from 523 patients aged 60 and over.
• Identified 17 critical predictors through LASSO regression.
• Models were validated with 10-fold cross-validation, achieving an AUC of 0.83 and Accuracy of 0.91.

Limitations

• Row count is unknown, which may limit suitability assessment.
• Column-level documentation is absent; field semantics must be inferred after download.
• Data may reflect geographic bias inherent to the two hospitals in Northeast China.

Provenance

Source

Guangpeng Chen via figshare.

Collection Method

Data were retrospectively collected from two tertiary hospitals.

Freshness

Last updated 2026-05-01 05:28:00; freshness should be verified.

Geography

Northeast China.