Machine Learning Approaches to Predict Chronic Lower Back Pain in People Aged over 50 Years

• Published in: Medicina (Kaunas, Lithuania) Vol. 57; no. 11; p. 1230
• Main Authors: Shim, Jae-Geum, Ryu, Kyoung-Ho, Cho, Eun-Ah, Ahn, Jin Hee, Kim, Hong Kyoon, Lee, Yoon-Ju, Lee, Sung Hyun
• Format: Journal Article
• Language: English
• Published: Basel MDPI AG 11.11.2021; MDPI
• Subjects: Age; Alcohol use; Algorithms; Arthritis; artificial neural network; Back pain; Body mass index; Cardiovascular disease; chronic lower back pain; Datasets; decision tree; Decision trees; Diabetes; Disease control; Disease prevention; Family income; Feature selection; logistic regression k-nearest neighbors; Machine learning; naïve Bayes; Osteoarthritis; Patients; Support vector machines
• ISSN: 1648-9144; 1010-660X; 1648-9144
• DOI: 10.3390/medicina57111230

Background and Objectives: Chronic lower back pain (LBP) is a common clinical disorder. The early identification of patients who will develop chronic LBP would help develop preventive measures and treatment. We aimed to develop machine learning models that can accurately predict the risk of chronic LBP. Materials and Methods: Data from the Sixth Korea National Health and Nutrition Examination Survey conducted in 2014 and 2015 (KNHANES VI-2, 3) were screened for selecting patients with chronic LBP. LBP lasting >30 days in the past 3 months was defined as chronic LBP in the survey. The following classification models with machine learning algorithms were developed and validated to predict chronic LBP: logistic regression (LR), k-nearest neighbors (KNN), naïve Bayes (NB), decision tree (DT), random forest (RF), gradient boosting machine (GBM), support vector machine (SVM), and artificial neural network (ANN). The performance of these models was compared with respect to the area under the receiver operating characteristic curve (AUROC). Results: A total of 6119 patients were analyzed in this study, of which 1394 had LBP. The feature selected data consisted of 13 variables. The LR, KNN, NB, DT, RF, GBM, SVM, and ANN models showed performances (in terms of AUROCs) of 0.656, 0.656, 0.712, 0.671, 0.699, 0.660, 0.707, and 0.716, respectively, with ten-fold cross-validation. Conclusions: In this study, the ANN model was identified as the best machine learning classification model for predicting the occurrence of chronic LBP. Therefore, machine learning could be effectively applied in the identification of populations at high risk of chronic LBP.