Machine Learning Techniques for Identifying Lifestyle Factors Associated With Low Back Pain in Adults Aged 50 and Older Using Data From the Korean National Health and Nutrition Examination Survey

Abstract

Chronic low back pain (cLBP) is shaped by multiple lifestyle factors, yet models with practical behavioral cutoffs remain scarce. This study developed a machine learning model to identify key lifestyle factors and specific thresholds linked to cLBP in adults aged 50 and older, using data from 5607 participants in the Korean National Health and Nutrition Examination Survey. Machine learning algorithms were trained and validated, with performance assessed via AUROC and SHAP for interpretability. The logistic regression model performed best (AUROC = 0.721, 95% CI: 0.699-0.742). SHAP analysis revealed that higher cLBP risk was associated with older age, female gender, prolonged sitting (>= 6 h/day), low walking frequency (< 4-5 times/week), infrequent strength training (< 1 time/week), moderate-intensity work, elevated stress, and smoking over five packs lifetime. Diet also mattered: cLBP risk rose among those dining out less than similar to 2.2 times/week, consuming under 2.9 servings/day of protein, or with carbohydrate intake outside 55%-65% of total energy. These practical cutoffs can help clinicians identify high-risk individuals through simple assessments, guiding tailored interventions in physical activity, diet, smoking cessation, and stress management to prevent cLBP.