Establishment and Evaluation of a Nomogram Prediction Model for the Risks of Nontraumatic Fracture in Older Adults with Type 2 Diabetes Mellitus

  Objective  To analyze the risk factors for nontraumatic fractures in older adults with type 2 diabetes mellitus, to establish a nomogram prediction model, and to evaluate the model.

  Methods  The clinical data of 278 older adults with type 2 diabetes mellitus were collected as the modeling group, and the clinical data of 109 older adults with type 2 diabetes mellitus were collected as the validation group. In both groups, patients were divided into a fracture subgroup and a non-fracture subgroup according to whether there were nontraumatic fractures after patients developed type 2 diabetes mellitus. Multivariate logistic regression was done to identify factors influencing the risks of non-traumatic fracture in older patients with type 2 diabetes mellitus. R software was used to construct a nomogram prediction model, and then the accuracy and clinical validity of the nomogram (area under the ROC curve, H-L fit curve, and calibration curve) were evaluated.

  Results  In the modeling group, the incidence of nontraumatic fractures in older adults with type 2 diabetes mellitus was 24.46% (68/278). The two subgroups showed significant differences in age, diabetic peripheral neuropathy, smoking history, drinking history, serum triglyceride (TG), high-density lipoprotein cholesterol (HDL-C), low-density lipoprotein cholesterol (LDL-C), glycated hemoglobin (HbA1c), and hypertension history (P<0.05). Age, diabetic peripheral neuropathy, HbA1c and history of hypertension were independent risk factors for nontraumatic fractures in older patients with type 2 diabetes mellitus (P<0.05). A nomogram prediction model was constructed accordingly and the internal verification results of the prediction model were as follows: the area under the ROC curve was 0.774 (0.680-0.869), the slope of the calibration curve was close to 1, and the H-L fit curve was χ²=12.643, P=0.125. External validation was conducted with the patients in the validation group. The results showed that the area under the ROC curve was 0.780 (0.670-0.890). The prediction probability of the calibration curve was close to the actual probability, suggesting that the model had good discrimination and accuracy.

  Conclusion  Age, diabetic peripheral neuropathy, HbA1c, and hypertension history are independent risk factors for nontraumatic fractures in older adults with type 2 diabetes mellitus, and the prediction model established consequently has high accuracy and discrimination. Medical workers can take preventive measures based on individual patient factors to reduce the possibility of nontraumatic fractures in older adults with type 2 diabetes mellitus.