Machine learning models predict mortality risk in diabetic neuropathy patients using MIMIC-IV data.

Area of Science:

• Medical informatics
• Computational biology
• Clinical research

Background:

• Diabetic neuropathy (DN) is a common complication of diabetes, associated with increased mortality risk.
• Predicting mortality in DN patients is crucial for timely intervention and improved outcomes.
• Current risk stratification methods may benefit from advanced analytical approaches.

Purpose of the Study:

• To develop and validate interpretable machine learning (ML) models for predicting in-hospital mortality risk in patients with DN.
• To identify key risk factors associated with mortality in this patient population.
• To assess the clinical utility of ML models in supporting decision-making for DN patients.

Main Methods:

• Utilized data from the Medical Information Mart for Intensive Care (MIMIC-IV) database, including 1,313 patients with DN.
• Employed Least Absolute Shrinkage and Selection Operator (LASSO) for risk factor screening.
• Constructed and compared mortality prediction models using Random Forest (RF), XGBoost, SVM, and Logistic Regression (LR).
• Interpreted the best-performing model using SHapley Additive exPlanations (SHAP) analysis.

Main Results:

• The Random Forest (RF) model achieved the highest performance, with an Area Under the Curve (AUC) of 0.780 on the validation set.
• Red blood cell distribution width (RDW)_mean was identified as the most significant predictor of mortality.
• The interpretable ML model demonstrated strong predictive capability.

Conclusions:

• Machine learning offers a promising approach for predicting mortality risk in patients with diabetic neuropathy.
• Interpretable ML models can provide valuable insights into risk factors, potentially improving clinical decision-making and patient prognosis.
• Further validation and implementation of these models could enhance care for DN patients.