Chaining transitions in short longitudinal datasets to estimate the lifetime health benefits of childhood interventions

Presenter: Alex J. Turner, University of Manchester

Author: Alex J. Turner

Co-author(s): Eleonora Fichera and Matt Sutton

The benefits of childhood health interventions should be assessed in terms of lifetime health consequences, but intervention studies often consider only short-term changes. Previous studies have attempted to match trial data to birth cohort datasets in order to proxy outcome trajectories across the life course. However, cohort datasets are inevitably dated and, even with the longest follow-up, measure incomplete lifetimes. This study aims to describe and demonstrate a method for estimating the full lifetime health returns to childhood interventions via matching of data from intervention studies to short longitudinal datasets collecting data on all ages. Using coarsened exact matching, child-level data on outcomes and other characteristics from an intervention study are matched to longitudinal data on children of the same age. Lifetime trajectories are generated by chaining one-year transitions from consecutive starting ages, using only the two most recent waves of longitudinal data. The trajectories for each individual ends when death occurs in a transition between waves. Confidence intervals are generated using bootstrapping.

Data for 953 children from a randomised controlled trial of a school-based social and emotional wellbeing intervention are linked to 790 children from a large longitudinal dataset, Understanding Society. The mean 0.852-unit [CI:-1.46,-0.24] improvement in the Strength and Difficulties Questionnaire score generated by the intervention is associated with a mean reduction of 0.018 [CI:-0.294, 0.257] discounted lifetime Quality-Adjusted-Life-Years. This study illustrates that matching between intervention studies and consecutive one-year transitions in longitudinal datasets offers a feasible method for estimating lifetime outcomes using the most up-to-date information on changes over the life course.