Abstract
Background: Elderly patients with Type 2 diabetes mellitus (T2DM) often experience prolonged length of stay (LOS) and 30-day readmission. This study was aimed at identifying factors influencing these outcomes and develop predictive models for them. Methods: The least absolute shrinkage and selection operator (LASSO) combined with logistic regression was utilized to construct the prediction models, which were subsequently visualized through nomograms. The performance of these models was comprehensively evaluated in terms of discrimination, calibration, and clinical utility. Specifically, the discrimination capacity was assessed using the area under the receiver operating characteristic curve (AUROC), while calibration was evaluated via calibration curves and the Brier score. Clinical utility was examined through decision curve analysis (DCA) and clinical impact curve (CIC). Additionally, to verify the robustness and generalizability of the developed prediction models, subgroup analyses were conducted across various strata of the study population. Results:A total of 24 variables for 8800 patients were included for predicting prolonged LOS, and 38 variables were used for 30-day readmission prediction. In the training set, 28.42% of patients had prolonged LOS and 13.68% were readmitted within 30 days. The prolonged LOS model had an AUROC of 0.720 (95% CI: 0.703-0.737), while the 30-day readmission model achieved 0.766 (95% CI: 0.745-0.787). The Brier scores were 0.174 (95% CI: 0.168-0.180) and 0.102 (95% CI: 0.096-0.108), respectively. Both models showed good clinical utility in DCA and CIC analyses. Subgroup validation across different age groups showed consistent performance, with all AUROCs above 0.60. Albumin was identified as the most significant predictor in both models. Conclusion: The predictive models developed in this study demonstrated robust performance in forecasting common outcomes in elderly patients with T2DM. Moreover, albumin level was strongly associated with both prolonged LOS and 30-day readmission, making it a key factor in patient management.