Characterizing gastrointestinal disorder trajectories for autistic sub-groups: Machine learning prediction of risk profiles and response to treatment

PROJECT SUMMARY/ABSTRACT Gastrointestinal (GI) problems are one of the most common concerns reported by families of autistic children and youth and can have significant lifelong impacts on health, quality of life, and participation. Existing research, however, primarily relies on parent report and is cross-sectional, leaving a gap in our understanding

of GI trajectories, i.e., how the symptoms emerge, including clinical and behavioral indicators; and how they develop (and potentially change) over time. Further, existing research lacks standard approaches to measuring GI symptoms; and a dearth of research using large, diverse, real-world clinical datasets limits our

understanding of which sub-groups are at higher risk of which constellation of symptoms, and what factors predict response to standard of care treatments. The proposed multimethod research addresses these gaps by using qualitative, participatory, and machine learning approaches to build prediction models of risk of GI

symptom profiles and response to treatment among autistic sub-groups. Our study aims to: (1) Qualitatively describe autistic people’s GI experiences throughout the lifespan through analysis of narrative interviews with 25 autistic adults and 25 caregivers of autistic children/youth. (2) Quantitatively characterize GI symptom rates,

presentations, trajectories, and responses to treatment using electronic health records (EHRs) from Children’s Hospital Los Angeles (CHLA) (N=7,478 autistic children/youth ages 1 to 25) with both (a) structured data (e.g., diagnosis codes, prescriptions) and (b) unstructured data (i.e., keywords extracted from clinical notes via

natural language processing). (3) Build predictive models of risk of GI symptom profiles and response to treatment using both traditional and machine learning approaches with the Aim 2 dataset and a matched cohort of non-autistic children and youth (1:5). To ground our work in lived experience perspectives, and in

response to autism community advocacy for autistic representation in research, we will use a participatory research approach with a community advisory board made up of (a) autistic adults and (b) caregivers of autistic children and youth, who will contribute to data collection, analysis, interpretation, and dissemination.

The proposed study has the strong potential to contribute to a personalized medicine approach to GI disorders for autistic people, including targeted risk assessments across the lifespan, to improve effective, person- centered care.