The Child Opportunity Index and Pediatric Hospitalizations: Are ZIP Codes Good Enough?

Abstract

To quantify the misclassification of Child Opportunity Index (COI) quintiles when using ZIP codes instead of census tracts and to compare the strength of associations with ambulatory care sensitive condition (ACSC) hospitalization rates when using COI linked to these 2 geographies. This retrospective, cross-sectional study analyzed pediatric ACSC hospitalizations from 2 Midwest children's hospitals between 2013 and 2018. Patient home addresses were geocoded and linked to the COI 3.0 at ZIP code and census tract levels. COI scores were assessed as nationally-normed quintiles, percentiles, and z-scores. Misclassification was defined as ZIP code COI differing by ≥ 2 quintile levels from census tract quintile assignment. Geospatial and regression analyses assessed the impact of misclassification on associations with ACSC hospitalization rates. There were 26 512 ACSC hospitalizations for youth from 2 metropolitan areas comprised of 604 943 children. Misclassification occurred in 8.1% of ACSC hospitalizations. Regression analyses indicated lower hospitalization rates for areas with higher opportunity. Between the census tract and ZIP code geographies, there was no significant difference in the percent change of ACSC hospitalization rates for every COI quantile increase when using COI quintiles (-21.7% vs -22.0%; P = .51) and percentiles (-1.2% vs -1.2%; P = .6); however, there was a significant difference between the geographies when using COI z-scores (-35.1% vs -28.5%; P < .001). Using ZIP codes to assign COI scores with the most common approaches - quintiles and percentiles - is likely acceptable for correlational analyses in large datasets. However, ZIP code COI scores may underestimate true associations between neighborhood opportunity and health outcomes when using z-scores.