Construction of predictive model for osteoporosis related factors among postmenopausal women on the basis of logistic regression and Bayesian network

•A cross-sectional survey with data spanning ten years and credible results.•Our study included a broader range of variables that encompassed lifestyle and chronic diseases.•Logistic regression and Bayesian network were used to build the prediction model for postmenopausal women with osteoporosis.•The relationship between influencing factors has been shown in the form of visual graphics. Osteoporosis is a prevalent chronic disease that often goes unnoticed in postmenopausal women. Early identification of risk factors for osteoporosis in postmenopausal women is essential. This study aimed to develop predictive models for osteoporosis-related factors among postmenopausal women in the U.S. and explore the influencing factors. In this cross-sectional study, we included 4417 postmenopausal women from the NHANES (2009–2010, 2013–2014, and 2017–2020). Through multiple regression analysis, we found that age, minutes of sedentary activity, prednisone or cortisone usage, arthritis, bone loss around teeth, and trouble sleeping were risk factors for osteoporosis after menopause. Conversely, height, BMI, and age at the last menstrual period were identified as protective factors. The findings from the Bayesian network analysis indicated that several factors influenced osteoporosis, including age, BMI, bone loss around teeth, prednisone or cortisone usage, arthritis, and age at the last menstrual period. On the other hand, minutes of sedentary activity and height might have indirect effects, while trouble sleeping may not have a significant impact. Both logistic regression and Bayesian network models demonstrated good predictive capabilities in predicting osteoporosis among postmenopausal women. In addition, Bayesian networks offer a more intuitive depiction of the intricate network risk mechanism between diseases and factors.