Comparison of network structures between autistic and non-autistic adults, and autism subgroups: A focus on demographic, psychological, and lifestyle factors

Abstract

Differences in (autism) characteristics are often reported between autistic and non-autistic adults but also between autistic adults. We aimed to determine whether mean differences correspond to differences in network structure of these characteristics in (1) autistic and non-autistic adults and (2) two previously identified autism subgroups. A total of 16 network variables related to demographic and psychological characteristics were included. First, Gaussian Graphical Models (GGMs) were used for network estimation in 261 autistic adults and 384 non-autistic comparisons aged 30–85 years. Second, we repeated this step within two previously identified autism subgroups ( N1 = 124, N2 = 130). Third, sex differences were explored in the networks of the autism subgroups. The networks of the autism and comparison groups differed on individual edges and visual inspection, although the Network Comparison Test (NCT) showed no overall differences. The networks of autism subgroups were similar based on visual inspection and statistical comparisons. Sex did not impact the subgroup networks differently. Networks were more similar than different, but observed edge differences could be informative for targeted support. Focusing on mean differences is not sufficient to determine which factors (and associations) are important for autistic people. Thus, network analysis provides a valuable tool beyond assessing mean differences for autistic adults. Lay Abstract There are large differences in the level of demographic, psychological, and lifestyle characteristics between autistic and non-autistic adults but also among autistic people. Our goal was to test whether these differences correspond to differences in underlying relationships between these characteristics—also referred to as network structure—to determine which characteristics (and relationships between them) are important. We tested differences in network structure in (1) autistic and non-autistic adults and (2) two previously identified subgroups of autistic adults. We showed that comparing networks of autistic and non-autistic adults provides subtle differences, whereas networks of the autism subgroups were similar. There were also no sex differences in the networks of the autism subgroups. Thus, the previously observed differences in the level of characteristics did not correspond to differences across subgroups in how these characteristics relate to one another (i.e. network structure). Consequently, a focus on differences in characteristics is not sufficient to determine which characteristics (and relationships between them) are of importance. Hence, network analysis provides a valuable tool beyond looking at (sub)group level differences. These results could provide hints for clinical practice, to eventually determine whether psychological distress, cognitive failures, and reduced quality of life in autistic adults can be addressed by tailored support. However, it is important that these results are first replicated before we move toward intervention or support.