Construction of a nomogram prediction model for individualized prediction of vertebral compression fracture risk in postmenopausal osteoporosis population

To construct a nomogram risk prediction model for osteoporotic vertebral compression fracture (OVCF) in postmenopausal osteoporosis (POP). A retrospective study was conducted on 326 POP patients from July 2023 to November 2024. The patients were randomly included into a training cohort of 228 cases and a validation cohort of 98 cases based on a 7:3 ratio; The training queue was assigned into a fracture group of 82 cases and a non fracture group of 146 cases. Risk factors were screened based on multiple logistic regression analysis. R software was used to construct nomogram prediction models. ROC curve, calibration curve, and clinical decision curve were performed to evaluate model discrimination, calibration, and practicality. For the non fracture group, the fracture group showed manifest increases in age, physical exercise time < 1 h/d, gastrointestinal diseases, uric acid, and type I collagen carboxy terminal peptide (β-CTX), and manifest decreases in bone mineral density (BMD) (T value), hemoglobin (Hb), and 5-hydroxyvitamin D [25-(OH)D] (P < 0.05). Increase in age, decrease in BMD (T value), physical exercise time < 1 h/d, presence of gastrointestinal diseases, increase in uric acid, increase in β-CTX, and decrease in 25-(OH)D were independent risk factors for OVCF in POP (P < 0.05). The calibration curves of the training queue and validation queue indicated a high degree of agreement between the predicted results of the model and the actual results, Hosmer-Lemeshow χ² = 5.155, 4.490, P = 0.631, 0.810. The AUC of ROC curve was 0.974 and 0.909. In the decision curve analysis, the model demonstrated a net clinical benefit across threshold probability ranges of 0.05-0.97 in the training cohort and 0.02-0.99 in the validation cohort. The established nomogram prediction model can be used to reasonably evaluate the risk of OVCF in POP patients.