Abstract
OBJECTIVES: The International Association for the Study of Pain defines 3 pain types presumed to involve different mechanisms-nociceptive, neuropathic, and nociplastic. Based on the hypothesis that pain types should guide the matching of patients with treatments, work has been undertaken to identify features to discriminate between them for clinical use. This study aimed to evaluate the validity of features to discriminate between pain types. MATERIALS AND METHODS: Subjective and physical features were evaluated in a cohort of 350 individuals with chronic musculoskeletal pain attending a chronic pain management program. The analysis tested the hypothesis that, if features nominated for each pain type represent 3 different groups, then (1) cluster analysis should identify 3 main clusters of patients, (2) these clusters should align with the pain type allocated by an experienced clinician, (3) patients within a cluster should have high expression of the candidate features proposed to assist identification of that pain type. Supervised machine learning interrogated features with the greatest and least importance for discrimination, and probabilistic analysis probed the potential for the coexistence of multiple pain types. RESULTS: Results confirmed that data could be best explained by 3 clusters. Clusters were characterized by a priori specified features and agreed with the designation of the experienced clinician with 82% accuracy. Supervised analysis highlighted features that contributed most and least to the classification of pain type, and probabilistic analysis reinforced the presence of mixed pain types. DISCUSSION: These findings support the foundation for further refinement of a clinical tool to discriminate between pain types.