The relationship between headache-attributed disability and lost productivity:3 Attack frequency is the dominating variable

Background: In an earlier paper, we examined the relationship between headache-attributed disability, measured as proportion of time in ictal state, and lost productivity. In a linear model, we found positive and significant associations with lost paid worktime, lost household worktime and total lost productivity (paid + household), but with high variance, which was increased when headache intensity was introduced as a factor. We speculated that analyses based on headache frequency alone as the independent variable, eliminating both the subjectivity of intensity estimates and the uncertainties of duration, might show stronger associations. Methods: Focusing on migraine, we used individual participant data from 16 countries surveyed either in population-based studies or in the Eurolight project. These data included frequency (headache days/month), usual attack duration (hours), usual headache intensity (“not bad”, “quite bad”, “very bad”) and lost productivity from paid and household work according to enquiries using the Headache-Attributed Lost Time (HALT) questionnaire. We used multiple linear regressions, calculating regression equations along with unstandardized and standardized regression coefficients. We made line and bar charts to visualize relationships. Results: Both frequency and intensity were significant predictors of lost productivity in all multiple linear regressions, but duration was a non-significant predictor in several of the regressions. Predicted productivity in paid work decreased among males by 0.75–0.85 days/3 months for each increase of 1 headache day/month, and among females by 0.34–0.53 days/3 months. In household chores, decreases in productivity for each added day/month of headache were more similar (0.67–0.87 days/3 months among males, 0.83–0.89 days/3 months among females). Visualizations showed that the impact of duration varied little across the range of 2–24 h. The standardized regression coefficients demonstrated that frequency w