What type of study is this?

This is a Validation Study study.

What question did this study set out to answer?

The aim is to develop a reliable machine learning pipeline for predicting incident diabetes using hospital data.

November 13, 2025ElectronicsOpen Access

View Full Paper

Ask AI

Bookmark

Authors

This model predicts diabetes incidence in patients hospitalized within a year, indicating strong calibration and discrimination.

The aim is to develop a reliable machine learning pipeline for predicting incident diabetes using hospital data.
Utilized data from the MIMIC-IV hospital module for developing the prediction model.
Implemented a leakage-safe pipeline ensuring temporal and patient-level guards during predictions.
Evaluated models using performance metrics such as Brier score, AUROC, and AUPRC.
Two baselines (logistic regression and random forest) were compared with cross-validation and calibration.
Provided global explanations of model predictions using permutation importance and SHAP.
Calibrated logistic regression achieved AUROC of approximately 0.70 and AUPRC of about 0.31.
The random forest model exhibited lower performance compared to logistic regression.
The calibrated models yielded clinically relevant risk patterns based on age, length of stay, sex, and ethnicity.

Corte et al. (2025) studied this question.

View Full Paper

Ask AI

Bookmark