Fragility fractures and their consequences are the most common signs of osteoporosis, the most prevalent disease related to the adult skeleton. Identifying patients at high risk of fracture prior to a fracture occurring is a critical component of osteoporosis care. This problem continues to be a major concern for researchers and physicians worldwide. Despite the fact that several algorithms have been created to either identify persons with osteoporosis or forecast their risk of fracture, concerns remain about their accuracy and usefulness. Scientific breakthroughs, such as machine learning technologies, are rapidly gaining acceptance as alternative approaches for improving risk assessment and existing practice.
Osteoporosis is a severe condition that primarily affects postmenopausal women. The standard-of-care test for osteoporosis includes estimating bone mineral density (BMD) in the proximal femur, the lumbar spine, and, in certain cases, the forearm using dual-energy X-ray absorptiometry (DXA). The BMD is then compared to that of a reference group, including a sex-matched and ethnicity-matched healthy, premenopausal adult population (e.g., how much lower it is regarding standard deviations, or the T-score) for diagnosis1,2. Professional organizations, such as the International Society for Clinical Densitometry, the United States Preventive Services Task Force, and the International Osteoporosis Foundation, all promote screening strategies for older women, but determining when and how to conduct the screenings is more contentious.
The Osteoporosis Self-Assessment Tool (OST) is one of the oldest and simplest ways to identify people at risk of osteoporosis. This tool uses weight and age to identify men and women in various populations likely to have osteoporosis3,4,5,6,7,8,9,10,11,12,13. For the Asian female population, prediction tools were developed by integrating the magnitude of the correlation between age and weight with BMD to estimate the likelihood of osteoporosis5,14,15.
By extending the number of factors used to determine osteoporosis, 12 input parameters, such as demography, lifestyle, and medical history were included in the Fracture Risk Assessment Tool (FRAX)16. Similarly, other complex tools, including ORAI17, SCORE18, ORISIS19, ABONE20 and MOST21, have incorporated additional characteristics in order to improve the performance of OST detection. Although combining numerous recognized risk factors was expected to improve the utility of screening tools, studies have revealed that basic tools, such as OST, might work as well as those with more complicated algorithms, while recent systematic reviews have emphasized the potential and limitations of these approaches.
Machine learning models to predict risk of osteoporosis
Scientists worldwide recognize osteoporosis as a significant public health issue. While therapy can reduce fracture risk by 33% to 50%22, only a small percentage of patients, including those who have previously experienced osteoporotic fractures, receive the proper diagnosis and treatment. Furthermore, accurately identifying high-risk and/or high-cost patients in a fast and accurate way is expected to enhance effective healthcare management, as well as enhance clinical decision-making and to improve service planning and policy23,24. In recent decades, machine learning models have been increasingly integrated into osteoporosis prediction, along with the effective uses of healthcare big data, leading to improvements in the quality and efficiency of healthcare planning and delivery. Additionally, early illness identification (through simpler intervention and treatment, for example), customized health management, and the efficient detection of fraudulent behavior in healthcare are some of potential advantages25. Artificial intelligence (AI) technology has been developed based on mathematical modelling over years. AI software has been applied to different majors, including epidemiological survey26, drug discovery27, and diagnostic radiology28. At this point, the computer-assisted devices have been integrated to clinical routine practice to detect the abnormalities relating to respiratory diseases on chest X-ray images29,30,31. The on-site implementation of AI softwares proves its advantages to minimize the diagnostic bias, overcome the burnout issues and enhance the active case finding in community. Codlin et al. included 12 AI softwares to predict tuberculosis on chest X-ray images into the independent performance evaluation, which stated that a half of the AI softwares had the higher specificity values than ones of an intermediate radiologist28.
In a study by Erjiang et al.32, seven machine learning models—CatBoost, eXtreme Gradient Boosting, Neural Networks (NN), Bagged Flexible Discriminant Analysis, Random Forest (RF), Logistic Regression (LoR) and Support Vector Machines (SVM)—were implemented to derive the best fit models to differentiate between patients with and without osteoporosis using DXA T-scores. Ho-Pham et al. applied four machine learning models—artificial neural networks (ANN), LoR, SVMs and k-nearest neighborhood—to the BMD hip data of Australian women to identify hip fractures33. Ou Yang et al. implemented five ML models—ANN, SVM, RF, K-nearest neighbors (KNN), LoR—with many features, which were categorized into different areas related to bone health34. This study examined 16 input features for men and 19 input features for women in order to identify the relationship between the presence of certain features and risk of osteoporosis in a Taiwanese population. Other machine learning methods using OST to predict osteoporosis were reviewed by Ferizi et al.1.
Osteoporosis and the resulting fragility fractures are recognized as major public health issues throughout many developing countries, especially Vietnam. The lack of DXA equipment to diagnose osteoporosis in these countries requires a prediction model for individualized assessment. Ho-Pham et al. proposed a prediction model for individualized assessments of osteoporosis based on age and weight for men and women14. In this study, an LoR model using data from a population in Ho Chi Minh City was applied to develop the tool for each gender, with good accuracy. The researchers developed and validated a prediction model based on age and weight to estimate the absolute risk of osteoporosis in the Vietnamese population14. However, few studies have focused on OST-based prediction of osteoporosis in other areas in Vietnam. Our study’s primary objective was to build tools to assess the risk of osteoporosis from OST data in women over 50 years old in the Northern Vietnam. On the other hand, little is known about the performance of the model proposed by Ho-Pham et al.14 when applied on a new population. The Ho-Pham’s model proved its good accuracy during the internal validation14 whereas its performance to predict osteoporosis during an external validation was not noticeably reported. Therefore, our secondary objective was to independently validate the Osteoporosis Self-assessment Tool for Asians (OSTA) model and the model developed by Ho-Pham et al.14 on a new population. In addition to age and weight, our data now include height, geographic location (urban/rural area) and blood test results of uric acid, cholesterol, creatinine, FT4, glucose, HbA1c, Ure, AST, TSH, calcium and GGT. In their study, Ou Yang et al. concluded that specific blood test parameter is relevant to the OST (e.g., creatinine) to predict osteoporosis, while others (e.g., TSH) were of insignificant value in predicting osteoporosis in the Taiwanese population34. However, in contrast to Ou Yang et al. regarding the influence of TSH in North American patients, Jamal et al. recommended that patients with suspected osteoporosis based on their OST score undergoing the TSH test34,35. Therefore, the third objective of our research was to validate the conclusion of Ou Yang34 and Jamal35 using the dataset collected at Hanoi Medical University Hospital as well as to discover more novel factors linked to OST results.
Significance of the study
The findings of this study would provide valuable evidence to strengthen the potentials of machine learning algorithms for use as decision making support tools in the context of the widely osteoporosis screening. The study would also present the supportive findings to promote the digital transformation of medical diagnosis in Vietnam. Additionally, the covariates which showed the significant contribution to the risk of osteoporosis would be pointed out and might be a valuable consideration for governmental policy makers in Vietnam.
#Predicting #risk #osteoporosis #older #Vietnamese #women #machine #learning #approaches