Analysis And Evaluation Of The Risk Factors And Novel Screening Models For Osteoporosis Based On Big Health Data

Background: Osteoporosis has become a public health issue worldwide due to the high prevalence,high economic burden,high disability rate,and high mortality rate of osteoporosis.Therefore,osteoporosis management increasingly emphasized the early prevention,early detection,and early treatment of osteoporosis.This current paper aimed to analyze the potential risk factors of osteoporosis from multiple perspectives using big health data based on the early prevention and early detection of osteoporosis.Moreover,this study aimed to develop novel osteoporosis screening models and tools based on the disease risk factors and building principles for the prediction model,which would provide new options and strategies for osteoporosis prevention and screening in the future.Objective: 1.This paper aimed to analyze and explore the potential risk factors for osteoporosis based on big health data from the perspectives of several aspects,including lifestyle,diet/nutrition,and laboratory examination.2.This paper aimed to develop novel osteoporosis screening models using the information on risk factors for osteoporosis and to develop the corresponding practical screening tools.Moreover,this paper would evaluate and optimize the osteoporosis screening models and tools based on a prospective single-center study in China to provide new theoretical and methodological support for future osteoporosis disease management.Methods: This study used linear regression,logistic regression,and nonlinear fitting to evaluate the relationship between bone mineral density(BMD)(and osteoporosis risk)and independent factors from lifestyle,diet/nutrition,and laboratory examination,including nighttime sleep duration,sleep patterns,dietary fiber intake(DFI),high-density lipoprotein cholesterol(HDL-C),and inflammatory markers.2.Based on big health data,this paper used information on osteoporosis risk factors to develop a Primary Osteoporosis Screening Tool(POST)and a Secondary Osteoporosis Screening Tool(SOST)for middle-aged and elderly populations aged 50 years and over.Moreover,this paper used R language and other software to develop corresponding practical screening tools based on POST and SOST,including scoring tools and online web programs.In addition,this paper evaluated and validated the performance of the POST and the SOST in osteoporosis screening(including model discrimination,model calibration,sensitivity,specificity,and others).Finally,we conducted a single-center,prospective clinical study based on populations in China to evaluate and optimize the POST and the SOST.Results: 1.Based on big health data,this paper found that:(1)Regarding lifestyle:long(> eight h/day)nighttime sleep duration and unhealthy sleep patterns(especially sleep patterns of long sleep duration combined with later bedtime)were significantly associated with reduced BMD or the increased risk of osteoporosis.(2)Regarding diet/nutrition: no linear correlation between DFI and BMD was observed after adjusting for all covariates.However,the results of nonlinear fitting showed that DFI displayed an inverted U-shaped relationship with BMD among men.(3)Regarding laboratory examination: HDL-C was negatively associated with BMD,and high levels of HDL-C were significantly associated with the increased risk of osteoporosis or osteopenia,which were mainly observed among women aged 50 years and over.Increased systemic immune-inflammation index or neutrophil-to-lymphocyte ratio levels were significantly associated with reduced BMD and the increased risk of osteoporosis among postmenopausal women.2.This study developed two osteoporosis screening models and tools based on big health data and risk factors for osteoporosis.(1)POST: POST,which was established based on the information on age,sex,and weight,showed good performance in identifying individuals with osteoporosis [area under the curve(AUC)of receiver operating characteristic(ROC)is 0.81(95% confidence interval: 0.79-0.83)].Moreover,the sensitivity,specificity,positive predictive value(PPV),and negative predictive value(NPV)of the POST(threshold: POST score ≥7 points)in osteoporosis screening(in validation cohort)were 91.57%,42.28%,22.18%,and 96.54%,respectively.(2)SOST: SOST,which was established based on clinical information such as age,sex,body mass index,nighttime sleep duration,and history of fractures,showed excellent discrimination(AUC: 0.849,95% confidence interval: 0.820-0.878)and useful calibration(Brier Score: 0.062,95% confidence interval: 0.054-0.070)in osteoporosis screening.Moreover,the sensitivity and specificity(SOST)under the optimal threshold determined by the maximum Youden index are 84.10% and 72.90%,respectively.In addition,this study has also developed a practical online web-based tool based on the SOST model for osteoporosis screening.Finally,this study demonstrated that either the POST or the SOST,alone or in combination,showed good performance and effect in osteoporosis screening among middle-aged and elderly populations aged 50 years and over in China.Conclusions: 1.Clinicians need to be alert to middle-aged and older adults with long nighttime sleep duration,unhealthy sleep patterns,high levels of HDL-C,or increased levels of inflammatory markers,which may indicate an increased risk of osteoporosis.Our findings would provide new theoretical and empirical support for the early prevention of osteoporosis in the future.2.This paper has demonstrated that our self-developed novel osteoporosis screening models and tools have shown good performance and effect in osteoporosis screening among middle-aged and older populations.The novel screening models and tools this study developed are expected to provide new options and ideas for the early detection of osteoporosis in the future.