Policy and programmatic initiatives of the last decade has focused on increasing the coordination of services and the expansion of prompt substance abuse treatment for mothers involved with child welfare services. Yet, little evidence has been amassed about the implications of this approach on future recurrence of child welfare services. Data are from a large national probability sample of children who are the alleged victims of child abuse and neglect and their caregivers—the National Survey of Child Abuse and Neglect (NSCAW)--and include the recurrence of maltreatment for this group of children who remain at home. Preliminary analyses of NSCAW show that children whose parents receive substance abuse treatment are more likely than those who do not to have a recurrence of child maltreatment within 18-months. Yet, this analysis shows that there is much selection into treatment, as the caregivers who receive treatment are different on many variables. To mitigate the effects of selection, a propensity score matching (PSM) approach was used. Among 1101 caregivers with an indicated substance abuse problem a group of 224 clients who did not receive services were selected for comparison to 224 treated clients, using PSM. Prior to matching, all cases without one of six indicators of need for substance abuse treatment (the indicators came from child welfare worker and client self-report) were excluded. The logistic regression used to generate the PSM was included measures of variables previously identified by substance abuse treatment (SAT) researchers as related to treatment use and that distinguished between the SAT and non-SAT groups. The propensity scores for receiving SAT had a mean of .36 for the treated cases and .38 for the untreated cases, with nearly identical ranges from .33 to .86 and .89, respectively.

Mitigation of selection bias did not change the finding that receipt of SAT is associated with new referrals for child maltreatment. Event history analyses showed that clients who received substance abuse treatment were about twice as likely (18.7% vs. 10%) to have another child abuse report in the next 18-months, in contrast to untreated cases. A confirmatory logistic regression on the matched group shows that receipt of SAT decreased the odds of a child abuse report by 2.6 times. Child and caregiver age, prior child welfare services, an open child welfare case, urbanicity, and family cumulative risk were not associated with the odds or re-reports. Reasons why participation in substance abuse treatment may result in greater involvement with child welfare services are posited and framed within a discussion of the possibility that the PSM-based approach is contributing to the result.