From real-world data to causal claims: Target trial emulation, sensitivity and secondary analyses of ADHD interventions’ effects on academic achievement

The present thesis is part of the project led by Guido Biele, 'Effects of pharmacological treatment & special education on school performance of children with ADHD', and thus aims to answer two main questions: (i) What is the long-term effect of pharmacological treatment on academic outcomes in children diagnosed with ADHD? and (ii) What is the long-term effect of special education provision on academic outcomes in children diagnosed with ADHD? The problem is thus one of causal inference based on observational data sourced from Norwegian national registries and the Norwegian Mother, Father and Child Cohort Study (MoBa). The use of 'real-world data' is particularly challenging as it can provide relevant evidence only when coupled with strong assumptions. Thus, the present work focuses on the variety of analyses and approaches that can strengthen our inferences by weakening or complementing our reliance on these assumptions.
The thesis consists of an introduction and three articles. The introduction examines key epistemological and substantive issues relevant to this work, particularly the field of causal inference and the nature of ADHD. Paper I answers the first question via a target trial emulation of a pretest-posttest control group design, leveraging negative control analysis, proximal causal inference, a flexible machine learning semi-parametric model and secondary analyses on different subsets of the population. Paper II provides an extension of an existing sensitivity analysis for unobserved confounding to allow the use of bias prior distributions in a Bayesian model, applying this method to answer both questions while weakening the no unobserved confounding assumption. Paper III introduces an approach that leverages multiversal and meta-analytic tools to reconcile conflicting results between studies with different designs, as well as replicating a previous study from the literature in the Norwegian population. The thesis concludes with a review of treatment effects including an extended background on the main hypotheses about the effect of ADHD interventions, an extended discussion of these articles and future research directions.