Towards greater transparency in neurodevelopmental disorders research: use of a proposed workflow and propensity scores to facilitate selection of matched groups

BACKGROUND: Matching is one commonly utilized method in quasi-experimental designs involving individuals with neurodevelopmental disorders (NDD). This method ensures two or more groups (e.g., individuals with an NDD versus neurotypical individuals) are balanced on pre-existing covariates (e.g., IQ), enabling researchers to interpret performance on outcome measures as being attributed to group membership. While much attention has been paid to the statistical criteria of how to assess whether groups are well-matched, relatively little attention has been given to a crucial prior step: the selection of the individuals that are included in matched groups. The selection of individuals is often an undocumented process, which can invite unintentional, arbitrary, and biased decision-making. Limited documentation can result in findings that have limited reproducibility and replicability and thereby have poor potential for generalization to the broader population. Especially given the heterogeneity of individuals with NDDs, interpretation of research findings depends on minimizing bias at all stages of data collection and analysis. RESULTS: In the spirit of open science, this tutorial demonstrates how a workflow can be used to provide a transparent, reproducible, and replicable process to select individuals for matched groups. Our workflow includes the following key steps: Assess data, Select covariates, Conduct matching, and Diagnose matching. Our sample dataset is from children with autism spectrum disorder (ASD; n = 25) and typically developing children (n = 43) but can be adapted to comparisons of any two groups in quasi-experimental designs. We work through this method to conduct and document matching using propensity scores implemented with the R package MatchIt. Data and code are publicly available, and a template for this workflow is provided in the Additional file 1 as well as on a public repository. CONCLUSIONS: It is important to provide clear documentation regarding the selection process to establish matched groups. This documentation ensures better transparency in participant selection and data analysis in NDD research. We hope the adoption of such a workflow will ultimately advance our ability to replicate findings and help improve the lives of individuals with NDDs.