Concurrent comorbidities substantially alter long-term health behaviours and outcomes of headache patients

ObjectivesTo investigate factors affecting the long-term health behaviours and outcomes of headache patients and to design a risk-stratification tool. MethodsA population based observational-longitudinal cohort study using primary care electronic healthcare records from the UK Clinical Practice Research Datalink. 550,103 participants with a headache or migraine diagnosis of which 170,336 patients were identified as having recurrent headache-related events. Univariate and multivariable survival analysis was performed to determine factors influencing remission and a neural network classifier was developed. ResultsThere were almost twice as many female patients (352,330) as males (197,802). The median age at which a patient first saw their GP in males was 38 years, and in females was 37 years. Whilst age, gender and social deprivation significantly influenced the likelihood of seeing a GP, these factors had little effect on likelihood of remission from a period of recurrent headache-related events. In contrast, a comorbidity during the recurrent headache period reduced the risk of remission and increased median survival time from approximately 400-days up 1600. Increasing numbers of comorbidities progressively reduced the hazard risks for remission. The prediction models on remission within two and five years, demonstrated high precision and recall, with an F1 score of 0.795 and 0.887, respectively. InterpretationHeadache patients who suffer comorbidities have a substantively reduced likelihood of remission, highlighting an opportunity to considerably improve health outcomes in recurrent headache patients through addressing multi-morbidity more effectively. The prediction model could be used to help stratify patients most at risk of poor long-term outcomes.