Exploring the role of pain-related fear and lifting biomechanics in predicting low back pain incidence using supervised machine learning.

Greater lifting-specific pain-related fear has been associated with reduced lumbar spine motion during lifting, suggesting fear-driven protective movement strategies with potential negative consequences. However, the role of task-specific pain-related fear and lifting kinematics in the development of low back pain (LBP) remains unclear. This study aimed to develop and evaluate a supervised machine learning model to predict one-year LBP incidence and to identify the most important predictors. Baseline data from 156 healthy participants included pain-related fear, demographic, health, and lifestyle factors as well as lumbar spine range of motion (ROM) and whole-body lifting strategy during 15-kg lifting. LBP incidence was assessed using biweekly follow-up questionnaires. We trained and evaluated an Explainable Boosting Machine to predict one-year LBP incidence from baseline variables. The final model achieved an accuracy of 0.815 and an ROC AUC of 0.839 for out-of-sample predictions. The most important predictors were higher BMI, lower sleep quality, greater lifting-specific pain-related fear, and reduced lumbar spine ROM. This predictive performance suggests that a machine learning approach may help identify individuals at higher risk of developing LBP according to the applied criteria. These findings emphasize the role of lifting-specific pain-related fear and lifting kinematics in the development of LBP and highlight multifactorial contributors to LBP incidence.