Social barriers associated with pediatric no-shows, without improving prediction

Researchers evaluated whether non-medical drivers of health improved a transformer-based model for predicting missed pediatric appointments. The retrospective study, published in Pediatric Research, included 7,931 scheduled appointments at six UT Physicians general pediatrics clinics in Greater Houston during April–August 2024. Electronic health record data were linked to social-needs screening completed before appointments. Multivariable logistic regression assessed factors associated with no-shows.

Overall, 2,761 appointments (34.8%) were no-shows. Models with and without social-needs variables had identical areas under the receiver operating characteristic curve (AUROC), at 0.708; corresponding areas under the precision-recall curve (AUPRC) were 0.536 and 0.535. Adding demographic and insurance variables modestly improved performance, yielding an AUROC of 0.721 and AUPRC of 0.542 regardless of social-needs inclusion. Unconfirmed appointments, inactive patient portal status, transportation needs, and housing instability were associated with no-shows.

The findings identify potential targets for outreach and barrier-reduction strategies rather than demonstrating additional predictive value from social-needs data. The observational design cannot establish that addressing these factors reduces missed visits, and interventions were not tested. Inclusion was restricted to appointments with prior social-needs screening, which may limit generalizability to unscreened families and other settings.