Examining predictors of re-reports and recurrence of child maltreatment using two national data sources

This study examined predictors of child maltreatment re-reports and recurrence (substantiated re-reports) using two nationally representative data sets: the National Survey of Child and Adolescent Well-Being (NSCAW II) and the National Child Abuse and Neglect Data System (NCANDS). Maltreatment data from the survey and administrative data sources were merged for children who participated in NSCAW II and had an NCANDS record. Re-report and recurrence data were examined among the sample of 4715 with an NCANDS record and a subsample of 1873 who received a caseworker interview at the NSCAW II 18-months' follow-up because the child was in out-of-home care and/or the family had been involved with child welfare services since the baseline interview. Many of the regression and hazard estimates of the predictors of re-report and recurrence differed across the two samples due to endogenous selection bias that arises from conditioning on the “collider” variable of caseworker interviews which is itself partially an outcome of re-report and recurrence. Similar estimates, on the other hand, were obtained: Children with prior involvement with child protective services were more likely to have re-reports and recurrence than children without prior involvement. Children that the child welfare agency reported as receiving services at the time of investigation were also less likely to have re-reports and recurrence when compared with children not receiving services. The threats to the validity of inferences that are drawn from statistical models and samples that condition on collider variables are reviewed, and the use of administrative data to supplement survey information that is unavailable from caseworkers are discussed.