Developmental Foundations of a Pediatric Mental Health Risk Calculator for Young Children.

Abstract

Objective: To advance clinical utility of an emerging risk calculator for identifying when to worry and when to act when young children show signs of mental health concerns in pediatric care, we: (1) replicate an early childhood mental health risk algorithm (DECIDE); (2) determine preliminary predictive utility of additional child and parenting assets, advancing a strengths-based framework to reduce the likelihood of biased identification.

Methods: Data were from two independent studies: The national Future of Families and Child Wellbeing Study (FFCWS; N=2,763) and the regional Mental Health, Earlier Synthetic Cohort study (MHESC; N=323). Predictors were assessed in toddlerhood/early preschool age. Internalizing/externalizing problems were measured in older preschoolers, serving as outcomes. Epidemiologic risk prediction methods were applied to: (1) replicate the DECIDE risk algorithm domains comprised of demographics, child irritability, and adverse childhood experiences; and (2) examine the added predictive utility of child and parenting assets. Predictive utility was based on area under the curve (AUC) and/or the integrated discrimination improvement (IDI).

Results: The DECIDE algorithm was replicated in FFCWS and MHESC (AUC=.70 for both studies; IDI=.07 in FFCWS and.06 in MHESC). IDIs indicated predictive utility for child assets beyond the existing DECIDE algorithm in both studies, and for parenting assets in FFCWS.

Conclusions: Robust validation of predictive algorithms is critical for assessing generalizability. Reducing bias in early mental health risk algorithms via a strengths-based approach is key to equitable decision-making. This work lays the foundation for implementation of early mental health decision tools in routine care of young children.