Machine learning algorithm to predict fragility fractures and identification of important features – an explainable approach

Abstract In this study, we developed ML algorithms to predict fragility fractures, considering the occurrence of fractures at different skeletal sites. We investigated seven ML algorithms (LASSO, Elastic Net, Random Forest, Decision Tree, Neural Network, XGBoost and Logistic Regression) using the data from the Canadian Multicentre Osteoporosis Study (CaMos) with participants aged 50 years or older. We considered 73 baseline features, including age, sex, menopause status, and bone mineral density (BMD), and the outcome was the first incidence of fracture at any of the following sites: hip, spine, pelvis, ribs, shoulder, and forearm, over a 19-year follow-up period. Data were divided into training (70%) and testing (30%) datasets. The ML algorithms were trained on the training dataset and evaluated on the test dataset in terms of the ROC_AUC. SHapley Additive exPlanations (SHAP) analysis was performed to identify the important features that contribute to the prediction of fracture, and to investigate the interaction among these features. In total, 7,753 subjects were included in the study. Approximately 72% were female, and the average age was 67 years. We found that the XGBoost algorithm had a slightly better ROC_AUC (0.70; 95% CI: 0.67, 0.73). From the SHAP analysis, we found that BMD was the most important feature that contributed to the prediction. The other important features include age, previous fracture, osteoporosis and menopausal status. Total hip BMD interacted the most with femoral neck BMD, lumbar spine BMD interacted the most with weight, previous fracture status interacted the most with femoral neck BMD, and age interacted the most with lumbar spine BMD. This study demonstrated that XGBoost was the most effective algorithm for predicting fragility fractures. In addition, we identified important features that contribute to the prediction of fragility fractures. Intervention focusing on these features will help to prevent the incidence of these fractures. Lay summaries We developed machine learning (ML) algorithms to predict fragility fractures, considering the incidence of fractures at different skeletal sites, including the hip, spine, pelvis, ribs, shoulder, or forearm, using 19 years of follow-up data from the Canadian Multicentre Osteoporosis Study (CaMos). We investigated seven ML algorithms and found that XGBoost had slightly better performance compared to other algorithms. We identified important factors that increase the risk of fractures, including BMD, age, and previous fracture. We also demonstrated how the interaction between these factors increases the risk of fractures. The intervention focusing on these factors will help to prevent fragility fractures. Competing Interest Statement Acharya and Drs. Borhan, Thabane, Hanely, Berger, and Morin declared no conflict of interest. Dr. Papaioannou reported receiving honoria from Amgen and funding from Osteoporosis Canada. Dr. Goltzman reported receiving funding from the Canadian Institute of Health Research (CIHR), one-time royalties from UpToDate, one consulting fee from Biosyent, patents: 2457928(Canada), 60/384122(USA); 2343713(Canada) issued to the McGill University, and provided clinical expert assessment of Burosumab for treatment of X-linked Hypophosphatemia(XLH). Dr. Adachi reported receiving funding from CIHR, Eli Lily, Merck, Procter & Gamble, Sanofi, Amgen, consulting fees and honoraria from Amgen. Dr. Raina reported receiving funding from the Canadian Institute of Health Research (CIHR) and the Canada Foundation for Innovation and being involved with the WHO working group on life course. Funding Statement Dr. Borhan received partial funding through the OC-CaMos fellowship from Osteoporosis Canada to conduct this study. Author Declarations I confirm all relevant ethical guidelines have been followed, and any necessary IRB and/or ethics committee approvals have been obtained. Yes The details of the IRB/oversight body that provided approval or exemption for the research described are given below: Hamilton Integrated Research Ethics Board (HiREB) gave ethical approval of this work. I confirm that all necessary patient/participant consent has been obtained and the appropriate institutional forms have been archived, and that any patient/participant/sample identifiers included were not known to anyone (e.g., hospital staff, patients or participants themselves) outside the research group so cannot be used to identify individuals. Yes I understand that all clinical trials and any other prospective interventional studies must be registered with an ICMJE-approved registry, such as ClinicalTrials.gov. I confirm that any such study reported in the manuscript has been registered and the trial registration ID is provided (note: if posting a prospective study registered retrospectively, please provide a statement in the trial ID field explaining why the study was not registered in advance). Yes I have followed all appropriate research reporting guidelines, such as any relevant EQUATOR Network research reporting checklist(s) and other pertinent material, if applicable. Yes Footnotes Conflict of interest disclosure: Acharya and Drs. Borhan, Thabane, Hanely, Berger, and Morin declared no conflict of interest. Dr. Papaioannou reported receiving honoria from Amgen and funding from Osteoporosis Canada. Dr. Goltzman reported receiving funding from the Canadian Institute of Health Research (CIHR), one-time royalties from UpToDate, one consulting fee from Biosyent, patents: 2457928(Canada), 60/384122(USA); 2343713(Canada) issued to the McGill University, and provided clinical expert assessment of Burosumab for treatment of X-linked Hypophosphatemia(XLH). Dr. Adachi reported receiving funding from CIHR, Eli Lily, Merck, Procter & Gamble, Sanofi, Amgen, consulting fees and honoraria from Amgen. Dr. Raina reported receiving funding from the Canadian Institute of Health Research (CIHR) and the Canada Foundation for Innovation and being involved with the WHO working group on life course. Funding: Dr. Borhan received partial funding through the OC-CaMos fellowship from Osteoporosis Canada to conduct this study. Data availability statement: Data are not available to share. The manuscript has been formatted for another journal. Data Availability Data are not available to share