How well can we measure chronic pain impact in existing longitudinal cohort studies? Lessons learned

Multiple large longitudinal cohorts provide opportunities to address questions about predictors of pain and pain trajectories, even when not anticipated in design of the historical databases. This focus article uses two empirical examples to illustrate the processes of assessing the measurement properties of data from large cohort studies to answer questions about pain. In both examples, data were screened to select candidate variables that captured the impact of chronic pain on self-care activities, productivity and social activities. We describe a series of steps to select candidate items and evaluate their psychometric characteristics in relation to the measurement of pain impact proposed. In UK Biobank, a general lack of internal consistency of variables selected prevented the identification of a satisfactory measurement model, with lessons for the measurement of chronic pain impact. In the English Longitudinal Study of Ageing, a measurement model for chronic pain impact was identified, albeit limited to capturing the impact of pain on self-care and productivity but lacking coverage related to social participation. In conjunction with its supplementary material, this focus article aims to encourage exploration of these valuable prospectively collected data; to support researchers to make explicit the relationships between items in the databases and constructs of interest in pain research; and to use empirical methods to estimate the possible biases in these variables.

Perspective: This focus article outlines a theory-driven approach for fitting new measurement models to data from large cohort studies, and evaluating their psychometric properties. This aims to help researchers develop an empirical understanding of the gains and limitations connected with the process of re-purposing the data stored in these datasets.