Addressing contamination bias in non-randomized research on child maltreatment: A dual measurement strategy to enhance the estimation of causal effects

Abstract:

Contamination is a methodological phenomenon occurring in non-randomized research when units within a control condition receive or are exposed to the treatment under investigation. Conceptualized as measurement error, this presentation reviews the origins and prevalence of contamination in child maltreatment research as well as the impact resulting bias has on the significance and magnitude of causal effects informing public policy. A dual measurement strategy for detecting and controlling contamination will be presented and tested across two multi-wave independent cohorts, the National Survey of Child and Adolescent Well-being-II (N=5872) and the Longitudinal Studies of Child Abuse and Neglect (N=1354), using entropy-balanced propensity score matching and synthetic control modeling frameworks. Implications for the measurement of child maltreatment and the replicability of child maltreatment effects will be discussed.