Development and validation of prediction models for the risks of diabetes-related hospitalization and in-hospital mortality in patients with type 2 diabetes

Abstract

ObjectivesDiabetes is a major cause of hospitalization and in-hospital mortality. However, a scoring system that can be used to identify diabetic patients at risk of diabetes-related hospitalization and in-hospital mortality is lacking. MethodsWe included 32,653 patients in this retrospective cohort study. All recruited patients had type 2 diabetes, were 30–84 years of age, and were enrolled in the National Diabetes Care Management Program over the period of 2001–2003. We used the Cox proportional hazard regression model to derive risk scores. The predictive accuracy of the models was evaluated using receiver operating characteristic curves. We conducted the Hosmer–Lemeshow test to assess the agreement between predicted and observed risks. ResultsOver a follow-up period of eight years, 6243 patients were hospitalized for diabetes-related events, and 2048 deaths were registered in hospital records. For the one-, three-, five-, and eight-year periods, the areas under the curve (AUC) for diabetes-related hospitalization in the validation set were 0.80, 077, 0.76, and 0.74, respectively. The corresponding values for in-hospital mortality in the validation set were 0.87, 080, 0.77, and 0.76. The goodness-of-fit test showed that the predicted and observed probabilities in the one-, three-, five-, and eight-year periods were similar for diabetes-related hospitalization and in-hospital mortality in the validation set (all p values > 0.05). ConclusionWe developed models for the estimation of the risks of diabetes-related hospitalization and in-hospital mortality in patients with type 2 diabetes. The models may be used to identify diabetic patients who are at high risk for hospital admission and in-hospital mortality.