Addressing Contamination Bias in Child Maltreatment Research: Innovative Methods for Enhancing the Accuracy of Causal Estimates

Contamination occurs when members of a control condition receive or are exposed to the treatment under scientific investigation. The presence of contamination violates assumptions within counterfactual models of causal inference and results in two systematic and sequential problems: (1) measurement error in the form of misclassification of units in a control condition, and (2) bias in statistical modeling that affects the direction, magnitude, and significance of causal estimates. Contamination has the potential to underestimate the true causal effect within an individual study while creating variation in causal estimates across studies based on different degrees of contamination present. Originally examined in experimental research, this chapter introduces the concept of contamination as applied to observational research and uses the substantive area of child maltreatment as an illustrative example. The paper also offers methodological solutions to improve the detection of contamination while describing statistical approaches that demonstrate the impact of contamination bias and estimate causal effects in observational research after it is controlled. The goal of this chapter is to orient child maltreatment scientists conducting observational research to the issue of contamination bias and current approaches for addressing it.