Real-world evidence on migraine treatment in routine clinical practice: insights from the Türkiye migraine registry study using artificial intelligence algorithms
TL;DR
Artificial intelligence-assisted EHR analysis revealed underdiagnosis, high imaging use, and limited adoption of mechanism-specific preventives in Türkiye, underscoring the need for improved diagnostic accuracy and broader access to modern therapies.
Abstract
Background Despite available diagnostic and therapeutic options, a major unmet need persists in migraine care. This study applied artificial intelligence (AI) to extract structured insights from routine care and assess migraine burden, diagnosis, and treatment in Türkiye. Methods The Türkiye Migraine Registry Study was a retrospective, observational study across 23 hospitals and clinics in eight cities. Anonymized electronic health records (EHRs) of 11,023 adults treated between January 2021 and October 2023 were analyzed. A validated AI framework with natural language processing and machine learning extracted demographics, comorbidities, imaging, and prescriptions. Descriptive statistics summarized outcomes. Results Of 90,381 neurology patients, 16,751 (18.5%) had migraine or suspected migraine; 11,023 met inclusion. Only 46.9% carried an ICD-10 migraine code, while 53.1% were coded under other headache syndromes yet frequently received migraine-specific therapies. The cohort was 73% female, mean age 41.5. Neuroimaging was frequent (58.7% MRI, 15.2% CT), and 18.4% visited emergency care. Preventive therapy was prescribed to 63.7%, mainly antidepressants (31.4%) and beta-blockers (19.4%). OnabotulinumtoxinA and CGRP monoclonal antibodies were used in 10.3 and 3.4%. Conclusion AI-assisted EHR analysis revealed underdiagnosis, high imaging use, and limited adoption of mechanism-specific preventives in Türkiye, underscoring the need for improved diagnostic accuracy and broader access to modern therapies.