RESUMO
OBJECTIVE: Pain often accompanies carpal tunnel syndrome and affects patients' health-related quality of life. The aim was to develop and validate a predictive model for the pain intensity of carpal tunnel syndrome using demographic, clinical, electrophysiological, and ultrasound findings. METHODS: We conducted a secondary analysis of data from a large sample of patients (May 2017 to December 2022) with carpal tunnel syndrome. A total of 520 (53.0â¯%) mild, 276 (28.1â¯%) moderate, and 186 (18.9â¯%) severe syndromes were included in the complete data set of 982 hands (61.1â¯% female). The mean age was 57.8 (10.7) years and the median duration [interquartile range] of the symptoms was 4 [2,10] months. A regression model was developed and validated to predict pain intensity on a numerical rating scale using a tree-based machine learning algorithm. RESULTS: The validation of the regression model showed good performance with a root mean squared error, R-squared, and mean absolute error of 1.35, 0.42, and 1.05, respectively. Overall, the top significant predictors of pain intensity were compound motor nerve action potential latency, nocturnal pain, and thenar weakness. These were followed by the cross-sectional area of the median nerve, sensory nerve action potential, bowing of the flexor retinaculum, disease duration, and body mass index. We did not find strong associations between the median nerve transcarpal latency, age, sex, and diabetes with the pain intensity of carpal tunnel syndrome. CONCLUSION: Our model showed good performance in predicting the subjective pain intensity of carpal tunnel syndrome, even in the context of non-linear relations.