The impact of contamination bias when replicating maltreatment effects on child behavior problems

Contamination, when members of a comparison condition receive or are exposed to a treatment under investigation, can bias the direction, statistical significance, and magnitude of treatment effects.

This pre-registered study examined whether contamination biased estimates when replicating child maltreatment effects on two established outcomes, externalizing and internalizing behavior problems.

A multi-wave, U.S. national probability sample of children and families involved with child protective services (N = 4000; Age = 0-17.5; Sex = 48% Female; Race = 60% White; Ethnicity = 30% Hispanic). Child maltreatment and comparison conditions were established using official case records and contamination was identified as caregiver-reported maltreatment for those children assigned to the comparison condition. Outcomes were 6-month frequencies of externalizing and internalizing behaviors, represented in T-scores, three years post-maltreatment exposure.

The prevalence of contamination in this study was 96%. Propensity score analyses that ignored contamination revealed that maltreatment exposure had no impact on externalizing behaviors (b = 0.28, p = .746) or internalizing behaviors (b = 0.27, p = .744), failing to replicate prior research. However, propensity score analyses that corrected contamination replicated prior research by generating estimates that were in the expected directions, statistically significant, and moderate to large in effect magnitude for externalizing (b = 8.72, p = .002, SMD = 0.74) and internalizing behavior problems (b = 9.96, p < .001, SMD = 0.90).

Correcting contamination has the potential to improve the replication of child maltreatment effects, strengthening the generalizability of results.