Duke AI Health Director of Data Science Benjamin Goldstein, PhD, and Interim Director for Duke Center for Autism and Brain Development  Lauren Franz, MBChB, have recently been awarded a grant from the National Institutes of Health to develop approaches for using existing sources of health data to better understand late language emergence in children, or “late talking.” Late talking often co-occurs with autism and is frequently “undercoded” in medical records, which can affect whether children and families receive referrals for early intervention. The R21 grant, which is part of the NIH’s cross-institute Tackling Acquisition of Language in Kids (TALK) Initiative, will support efforts at Duke to apply machine learning to compare coding data in electronic health records (EHRs) with free-text information in EHR notes to determine whether indications of late talking recorded there are matched by appropriate coding.

The work will also leverage longitudinal health data and NC-statewide sources of claims data to shed additional light on developmental pathways, including methods for analyzing co-occurring factors that may predispose children to late talking, as well as downstream effects of delayed language acquisition later in life. Another focus for the project will be the health equity dimensions of late talking, including differences across gender, language, and race and ethnicity.

The 2-year grant adds to a body of work underway through the Duke Center for Autism and Brain Development to develop and apply machine learning methodologies to existing sources of health data to better understand developmental issues in children on the autism spectrum. In addition to co-principal investigators Goldstein and Franz, North Carolina Central University speech-language pathology specialist Danai Fannin, PhD, and Duke AI Health Data Science Fellowship Director Matthew Engelhard, PhD, will also serve as co-investigators on the project, which will run through August of 2026.