Contamination, when members of a comparison or control condition are exposed to the event or intervention under scientific investigation, is a methodological phenomenon that downwardly biases the magnitude of effect size estimates. This study tested a novel approach for controlling contamination in observational child maltreatment research. Data from The Longitudinal Studies of Child Abuse and Neglect (LONGSCAN; N = 1354) were obtained to estimate the risk of confirmed child maltreatment on trajectories of internalizing and externalizing behaviors before and after controlling contamination. Baseline models, where contamination was uncontrolled, demonstrated a risk for greater internalizing (b = .29, p < .001, d = .40) and externalizing (b = .14, p = .040, d = .19) behavior trajectories. Final models, where contamination was controlled by separating the comparison condition into subgroups that did or did not self-report maltreatment, also demonstrated risks for greater internalizing (b = .37, p < .001, d = .51) and externalizing (b = .22, p = .028, d = .29) behavior trajectories. However, effect size estimates in final models were 27.5%–52.6% larger compared to baseline models. Controlling contamination in child maltreatment research can strengthen effect size estimates for child behavior problems, aiding future child maltreatment research design and analysis.