Abstract
BACKGROUND: To advance early identification efforts, we must detect and characterize neurodevelopmental sequelae of risk among population-based samples early in development. However, variability across the typical-to-atypical continuum and heterogeneity within and across early emerging psychiatric/neurodevelopmental disorders represent fundamental challenges to overcome. Identifying multidimensionally determined profiles of risk, agnostic to DSM categories, via data-driven computational approaches represents an avenue to improve early identification of risk. METHODS: Factor mixture modeling (FMM) was used to identify subgroups and characterize phenotypic risk profiles, derived from multiple parent-report measures of typical and atypical behaviors common to autism spectrum disorder, in a community-based sample of 17- to 25-month-old toddlers (n = 1,570). To examine the utility of risk profile classification, a subsample of toddlers (n = 107) was assessed on a distal, independent outcome examining internalizing, externalizing, and dysregulation at approximately 30 months. RESULTS: FMM results identified five asymmetrically sized subgroups. The putative high- and moderate-risk groups comprised 6% of the sample. Follow-up analyses corroborated the utility of the risk profile classification; the high-, moderate-, and low-risk groups were differentially stratified (i.e., HR > moderate-risk > LR) on outcome measures and comparison of high- and low-risk groups revealed large effect sizes for internalizing (d = 0.83), externalizing (d = 1.39), and dysregulation (d = 1.19). CONCLUSIONS: This data-driven approach yielded five subgroups of toddlers, the utility of which was corroborated by later outcomes. Data-driven approaches, leveraging multiple developmentally appropriate dimensional RDoC constructs, hold promise for future efforts aimed toward early identification of at-risk-phenotypes for a variety of early emerging neurodevelopmental disorders.