Documented pain phenotypes and interventional pain-care exposure in patients with dystonia: a retrospective text-mining cohort study

Background Pain is a non-motor manifestation of dystonia, but its anatomical heterogeneity and treatment exposure are incompletely characterized. We described pain phenotypes in patients with dystonia referred to a tertiary pain service. Methods In this retrospective cohort study, electronic health record text from 2010 to 2026 was collapsed from 108,797 raw rows to 1860 encounters among 129 patients. A Hebrew-English rule-based pipeline identified pain phenotypes and documented treatment exposures. A category-stratified random sample of 129 records was manually audited by a senior neurosurgeon blinded to pipeline output. The primary objective was phenotype distribution; treatment exposures and phenotype-treatment associations were secondary or exploratory. Results Mean age was 67.0 years (SD 6.0), 61 patients (47.3%) were men, mean age at dystonia onset was 60 years, and all patients had at least four years of documented disease. Phenotype-codable pain was identified in 125 patients (96.9%). The most frequent phenotypes were limb/extremity (72.9%), spine/radicular (72.1%), cervico-occipital (53.5%), neuropathic-descriptor (43.4%), and myofascial/muscle pain (37.2%). At least one prespecified interventional procedure category was documented in 86 patients (66.7%). Cervical dystonia was associated with cervico-occipital pain (69.4% vs 43.8%; OR 2.91, 95% CI 1.38–6.18; P = .006). Treatment associations were interpreted as co-documentation patterns affected by indication bias. Conclusions Pain in this referral-enriched cohort was anatomically and mechanistically heterogeneous and frequently accompanied by documented procedural and medication exposure. The findings support prospective assessment using validated pain instruments, but do not establish diagnoses, treatment efficacy, or population prevalence.