A four-predictor clinical model demonstrated strong accuracy in forecasting SDM among hospitalized dermatology patients receiving systemic glucocorticoid therapy. Published in the International Journal of Diabetes in Developing Countries, the study examined whether routine clinical and metabolic variables could help identify patients at higher SDM risk.

The analysis included 293 dermatology inpatients, of whom 42 developed SDM between 2019 and 2024. After selection using the least absolute shrinkage and selection operator (LASSO) and multivariable logistic regression, the final model retained four key predictors: family history of diabetes, immunosuppressant usage, average daily glucocorticoid dose, and triglyceride level. The model showed excellent discrimination with an area under the curve (AUC) of 0.860 and good calibration, supported by a Brier score of 0.075. Cross-validation yielded a mean AUC of 0.839, and decision curve analysis showed net clinical benefit across a broad range of thresholds.

In an age-matched sensitivity subset of 168 patients, performance remained strong (AUC 0.847), with calibration and error metrics indicating robust internal reliability.
These results show that a simple four-factor model can support early SDM risk stratification in dermatology inpatients. External validation in prospective, multicenter settings is now required to establish generalizability before broader adoption.