ABSTRACT
This study examines autism diagnosis prevalence within the New York City (NYC) Universal Pre-K for All (UPK) program expansion into racially, ethnically, and socioeconomically diverse NYC neighborhoods. Here, it is hypothesized that racial/ethnic differences in autism diagnoses may close as more children are referred for testing by UPK programs, which they have more thorough interactions with, instead of by public health clinics or other medical avenues. Using NYC Medicaid claim data from 2006 to 2016, descriptive analyses were conducted by estimating linear probability regression and generalized multiple logistic regression to examine whether the probabilities of being diagnosed with autism in comparison to two other disability types (as counterfactuals), learning disabilities (LD) and physical disabilities (PD), differ by race. Subsequently, a difference in difference (DID) strategy (with pre- and post-UPK expansion cohorts) was used to examine the effects of UPK on the probabilities of receiving disability diagnoses. Notably, Latinx and “Other” racially identified children have much higher odds than White children of being diagnosed with autism. By contrast, all non-White groups had much lower odds of being diagnosed with a LD. These findings offer important insight for future UPK and childhood program implementation.
Acknowledgments
The authors thank NYU Health Evaluation and Analytics Lab and the New York State Department of Health for making the Medicaid claims data available and gratefully acknowledge the funding for this research from the Robert Wood Johnson Foundation’s Policies for Action program.
Disclosure statement
No potential conflict of interest was reported by the author(s).
Disclaimer
The views and opinions expressed in this article are those of the authors and do not necessarily reflect the official policy or position of the New York State Department of Health. As per NYS Department of Health guidelines, the analyses reported in this article are only examples of analyses that could be run with these data. They should not be utilized in real-world analytic products.
Notes
1. In addition to the covariates listed above, the authors ran models with first three-digit zip codes instead of school districts. Further, the authors used Medicaid aid categories. None of the models changed the study results.
2. The authors also conducted a difference in regression discontinuity model. The authors used local linear regression with a triangular kernel with a bandwidth of 60 days to estimate the UPK effects separately for cohort youngest and cohort oldest (Model DRD a), then pooled the effects together for the overall UPK effect (Model DRD b). Model DRD a: yit = α + ρ1 Dit + ρ2(DOBi – ct) + ρ3(DOBi – ct) + 1{t = 2014}[β + θDit + ρ5(DOBi – ct) + ρ6 Dit(DOBi – Ct)] + ɛit