Association between cluster analysis for multiple measures and International Classification of Diseases 11th revision as classification of chronic pain patients

Cluster analysis can classify patients with chronic pain using multiple scales, and classification of chronic pain will be adopted in the International Classification of Diseases 11 th revision (ICD– 11 ) in 2022 . In the present study, we aimed to investi-gate whether cluster analysis was practical for classifying chronic pain and to determine the association between these two classifications for chronic pain. This study included 229 patients with chronic pain who completed a self–reported questionnaire at the first visit to a pain clinic in a university hospital. Patients were clustered using a two–step cluster analysis (TSCA), a machine learning method, for the scores of nine questionnaires. Thereafter, the proportions of clusters among major and several minor classifications were tested using the analysis of covariance adjusted for age and doctor. The following three clusters were calculated using TSCA: mild, moderate, and severe symptoms. Among the major classifications of chronic pain in ICD– 11 , the distribution of clusters significantly differed, but the proportions of these three clusters in each chronic pain classification did not differ. Our findings suggested that TSCA for multiple measures may be a better approach for the classification of chronic pain, but its classification is not associated with the classification of chronic pain in ICD– 11 . The P–values of chronic widespread primary pain and others were calculated for comparison with chronic localized primary pain by the analysis of covariance using Dunnett’s test. The P–values of chronic centralized and other neuropathic pain were calculated for comparison with chronic peripheral neuropathic pain by the analysis of covariance using Dunnett’s test. The P–values of chronic non–specific and other pain were calculat ed for comparison with chronic structurally changed musculoskeletal pain by the analysis of covariance using Dunnett’s test. The analysis of covariance was adjusted for age and doctor.