Construction and validation of a prediction model for osteoporosis in urban and rural residents in Changshu City

Objective:To construct a prediction model for osteoporosis (OP) in urban and rural residents in Changshu City and validate it.Methods:A total of 2, 270 cases from the imaging department of the First People’s Hospital of Changshu with dual-energy X-ray bone density testing database from April 2018 to March 2021 were selected as the study population and divided into OP group (T≤-2.5, 337 cases), bone loss group (-2.5< T< -1.0, 701 cases) and normal bone mass group (T≥-1.0, 1, 232 cases) according to T value, comparing the 3 groups’general demographic data, clinical characteristics, grip strength, routine blood indicators, liver function, renal function, blood calcium, blood phosphorus, 25 hydroxyvitamin D25(OH)D3and parathyroid hormone (PTH) levels.Binary logistic regression was used to analyze the influencing factors associated with OP, using the R language rms software package to draw a column line graph model for predicting OP.Bootstrap method was used for internal and external validation, and the receiver operating characteristic (ROC) curve was used to analyze the predictive ability of the OP column line graph prediction model.Results:The OP group had higher age, female, diabetes, cognitive impairment, sleep duration ≥9 h, alkaline phosphatase, and PTH than the bone loss group, and lower regular calcium supplementation, regular intake of calcium-containing dairy products, grip strength, blood calcium, and 25(OH)D3 than the bone loss group(P< 0.05); the bone loss group had higher age, female, diabetes, cognitive impairment, sleep duration ≥9 h, alkaline phosphatase, and PTH than the normal bone mass group, and lower regular calcium supplementation, regular intake of calcium-containing dairy products, grip strength, blood calcium, and 25(OH)D3 than the normal bone mass group(P< 0.05).Binary logistic regression analysis showed that age, female, diabetes, cognitive impairment, sleep duration ≥9 h, alkaline phosphatase, and PTH were risk factors associated with OP, and regular calcium supplementation, regular intake of calcium-containing dairy products, grip strength, blood calcium, and 25(OH)D3 were protective factors associated with OP (P< 0.05); column line graph model predicting OP drawn based on each of these influencing factors showed its predictive risk ability index (C-index) was 0.944, with good discrimination.The ROC analysis found that the area under the ROC (AUC) of the column line graph model for predicting OP was 0.944(95%CI:0.923-0.960), suggesting that the column line graph model for predicting OP had good differentiation and predictive ability; the internal calibration plot using Bootstrap method found that the calibration curve was close to the standard curve, suggesting that the column line graph model for predicting OP was in good agreement with the actual observation.The external validation showed that its AUC for predicting the risk of death was 0.950(95%CI:0.945-0.999), and the external calibration plot found that the calibration curve was still close to the standard curve, suggesting that it still had a high predictive value in the external data.Conclusion:Age, female, diabetes, cognitive impairment, sleep duration ≥9 h, alkaline phosphatase, PTH, regular calcium supplementation, regular intake of calcium-containing dairy products, grip strength, blood calcium, and 25(OH)D3 are all influencing factors of OP, and the column line graph model constructed based on the above factors shows high predictive value, which can provide reference for early screening of high-risk population and targeted prevention of OP in this region.