PROJECT SUMMARY Late talkers are at risk for language impairment and have high rates of co-occurring mental health impairments, estimated to range from 30-70%. Our group has shown that as early as 12 months, children’s dimensional irritability is related to their vocabulary size, and late talkers have significantly more frequent and severe temper tantrums than typical talkers. Despite the frequency of late talking, language impairment and internalizing and externalizing (INT/EXT) mental health impairments, few studies have examined the multifaceted trajectories that lead to them. Further, few studies examine multi-faceted predictors of outcome or promotive factors which may help increase reliability of outcome prediction for children from under-resourced and historically marginalized environments, who experience inequities in early identification. Via our group’s expertise and currently funded supplement parent study, the Mental Health, Earlier Synthetic Cohort (MHESC, NIMH R01MH121877, PIs Wakschlag, Luby, Rogers) whose feeder studies are Northwestern’s irritability- and late talker-enriched W2W samples, and WashU’s adversity-enriched eLABE sample), we are poised to deepen and accelerate the aims of the MHESC, via supplementary study of the precursors and consequences of late talking. To do so, we enrich consideration of language development and ecological environment within the MHESC risk models designed to predict preschool psychopathology, using computational approaches, and employing a health equity lens. We pursue the following supplementary aims, within scope of MHESC grant’s Aim 2a: Supp Aim 1a. Apply the MHE-RiskCalc algorithm with deeper phenotyping measures of child language at 24 months (e.g., language sample analysis) and joint consideration of language and irritability measures to improve precision of prediction (beyond the base model) for three outcomes at age 4+ years: a) INT/EXT, b) language, and c) co-occurring language and INT/EXT impairments. 1b. Test whether a RiskCalc based on 12-month measures can predict 24-month late talking status. 1c. Test whether a RiskCalc based on 12-month measures can improve prediction of (a) INT/EXT (b) language & (c) INT/EXT+language impairments at age 4+. Supp Aim 2. Test whether adding promotive factors, including observed measures of dyadic language environment (caregiver responsiveness to child communication, language input quality) and child social-emotional competence improves RiskCalc precision and reduces bias of language and INT/EXT prediction, especially increasing accuracy for children from historically marginalized/under-resourced communities. Supp Aim 3. Use computational machine learning methods to determine data-driven indicators of risk of language and mental health impairments in the MHESC cohort, with emphasis on bias reduction and equity for diverse demographic sub-groups. In sum, we are poised to advance the objectives of the MHESC to accelerate pace of translation and equita...