Development and validation of a model predicting post-traumatic headache six months after a motor vehicle collision in adults

Abstract

ImportanceThe prognosis of post-traumatic headache is poorly understood. ObjectiveTo develop and validate a prognostic model to predict the presence of post-traumatic headache six months after a traffic collision in adults with incident post-traumatic headache. DesignSecondary analyses of adults with incident post-traumatic headache injured in traffic collisions between November 1997 and December 1999 in Saskatchewan, Canada (development cohort); and between January 2004 and January 2005 in Sweden (validation cohort). SettingThe Saskatchewan cohort (development) was population-based (N = 4162). The Swedish cohort (validation) (N = 379) were claimants from two insurance companies covering 20 % of cars driven in Sweden in 2004. ParticipantsAll adults injured in traffic collisions who completed a baseline questionnaire within 30 days of collision. Excluded were those hospitalized >2 days, lost consciousness >30 min, or reported headache <3/10 on the numerical rating scale. Follow-up rates for both cohorts were approximately 80 %. PredictorsBaseline sociodemographic, pre-injury, and injury factors. OutcomeSelf-reported headache pain intensity ≥3 (numerical rating scale) six months after injury. ResultsBoth cohorts were predominantly female (69.7 % in Saskatchewan, 65.2 % in Sweden), with median ages 35.9 years (Saskatchewan), and 38.0 years (Sweden). Predictors were age, work status, headache pain intensity, symptoms in arms or hands, dizziness or unsteadiness, stiffness in neck, pre-existing headache, and lower recovery expectations. With a positive score (i.e., ≥0.75 probability), the model can rule in the presence of post-traumatic headache at six months (development: specificity = 99.8 %, 95 % CI 99.5 %–99.9 %; sensitivity = 1.6 %, 95 % CI 1.0 %–2.6 %; positive likelihood ratio (LR+) = 8.0, 95 % CI 2.7−24.1; negative likelihood ratio (LR-) = 1.0, 95 % CI 1.0−1.0; validation: specificity = 95.5 %, 95 % CI 91.1 %–97.8 %; sensitivity = 27.2 %, 95 % CI 20.4 %–35.2 %); LR+ = 6.0, 95 % CI 2.8−13.2; LR- = 0.8, 95 % CI 0.7−0.8). Conclusions and relevanceClinicians can collect patient information on the eight predictors of our model to identify patients that will report ongoing post-traumatic headache six months after a traffic collision. Future research should focus on selecting patients at high risk of poor outcomes (using our model) for inclusion in intervention studies, and determining effective interventions for these patients.