BackgroundRisk assessment instruments used in child protection to guide intervention decisions should be equitable. To be considered equitable, two criteria are necessary: instruments should avoid discriminatory logic and display comparative validity for children of different races. Empirical evidence about the equity of risk assessment instruments is limited. Objective: To investigate the racial equity of a widely used risk assessment instrument, the Structured Decision Making Family Risk Assessment (FRA). Participants and setting: Administrative data were obtained for all children subject to investigation in Queensland, Australia in 2018, for whom an FRA was completed (n = 17,851). MethodsRelationships between FRA items, race, and recurrence (subsequent investigation within 12 months) were explored using crosstabulations and Cox Proportional Hazards. Accuracy of FRA recommendations and practitioner decisions were compared using sensitivity and specificity for Indigenous and non-Indigenous children. ResultsMost FRA items predicted race better than recurrence. Differences in recurrence by race were only partially explained by FRA items. The FRA produced high rates of false positives for Indigenous children (incorrectly classifying 48.9% of children who did not recur compared to 31.7% for non-Indigenous children) and high rates of false negatives for non-Indigenous children (incorrectly classifying 59% of children who did recur compared to 39% for Indigenous children). Practitioner discretion did not mitigate inequitable FRA recommendations. ConclusionsThe FRA failed both criteria of equity, discriminating indirectly by relying heavily on factors correlated with race, and demonstrating less accuracy for Indigenous children. High rates of false positives for Indigenous children exacerbate Indigenous overrepresentation in child protection.
Read full abstract