Development and validation of a new nomogram for OA based on machine learning

Abstract

Introduction: Osteoarthritis (OA) is a chronic joint disease with the global number of OA patients exceeds 300 million currently, posing a significant economic burden on patients and society. Currently, there is no cure for OA, making early identification and appropriate management of individuals at risk crucial. Thus, the development of a novel OA prediction model to screen for high-risk individuals, enabling early diagnosis and intervention, holds great importance in improving patient prognosis. Methods: Based on the National Health and Nutrition Examination Survey (NHANES) for the periods of 2011-2012, 2013-2014, and 2015-2016, the study was a retrospective cross-sectional study involving 11,366 participants. Least absolute shrinkage and selection operator (LASSO) regression, XGBoost algorithm, and random forest (RF) algorithm were used to identify significant indicators associated with OA, and a OA prediction nomogram was developed. The nomogram was evaluated by measuring the the area under receiver operating characteristic curve (AUC), calibration curve, and decision curve analysis (DCA) curve of training and validation sets. Results: In this study, we identified 5 predictors from 19 variables, including age, gender, hypertension, BMI and coffee intake, and developed an OA nomogram. In both the training and validation cohorts, the OA nomogram exhibited good predictive performance (with AUCs of 0.804 and 0.814, respectively), good consistency and stability in calibration curve and high net benefit in DCA. Conclusion: This nomogram based on 5 variables predicted the risk of OA with a high degree of accuracy, suggesting that it is a convenient tool for clinicians to identify high-risk populations of OA.