Study 1
Chaffin, Hecht, and colleagues (2012a) conducted a 2 x 2 cluster experiment to examine the effectiveness of the SafeCare program on Child Protective Services (CPS) recidivism. This study took place in Oklahoma, including both rural and urban areas, between September 9, 2003, and October 1, 2006. Regions of the state (n = 6) were randomized to either implement SafeCare or continue with services as usual (SAU). SafeCare providers were trained to implement the program; SAU providers addressed similar goals and needs as SafeCare providers but were less structured and not driven by protocols. In addition, agencies within each region were randomized to receiving implementation support or coaching or to implement without coaching, thus forming a 2 x 2 design. Coaching for SafeCare providers focused on maintaining fidelity to the model via observation, scoring fidelity via standardized checklist, and providing feedback to providers. Coaching of SAU providers focused on the quality of services in general with little structure to the coaching sessions.
Participants were eligible if they were considered maltreating caregivers and referred by child welfare agencies. Typically, to be included in SafeCare, participants had to have 1) at least one preschool-aged child, and 2) an absence of an untreated substance abuse disorder. However, this study took place within a more inclusive services system, and families were considered eligible if they had at least one child younger than age 12 and without regard to substance use. The total study sample included 2,175 parents. Nearly all of the participants (91 percent) were women, 31 percent were married, and the average age was 29.4 years. A majority (67 percent) were white/non-Hispanic, 16 percent were American Indian, 9 percent were African American, and 5 percent were Hispanic. Of the total sample, 41 percent reported past sexual abuse, 40 percent reported past physical abuse, and 22 percent reported being removed from their own parents during their childhood. Most of the households (82 percent) fell below the federal poverty line.
In total, 53 percent of the parents were assigned to the SafeCare treatment group (n = 1,153), and 47 percent of the parents were assigned to the services-as-usual control group (n = 1,022). Twenty-eight percent of the parents in the total sample were assigned to the SafeCare uncoached condition, 25 percent were assigned to the SafeCare coached condition, 23 percent were assigned to the services-as-usual uncoached condition, and 24 percent were assigned to the services-as-usual coached condition. For the CrimeSolutions review of this program, the outcomes that were scored were those pertaining to the direct comparisons between all families who received SafeCare (both coached and uncoached) and all families who received services as usual (coached and uncoached). Due to significant differences in attributes between treatment and control groups that might affect the effectiveness of the intervention, the study authors controlled for the following variables: 1) a historical risk variable (derived from historical American Indian cases and 12 raw variables such as gender, education, and age), and 2) a county-level report for proneness. This allowed for a more direct comparison of recidivism outcomes.
Data for CPS recidivism were collected from past and follow-up CPS reports. The average follow-up time was 6 years, and a recidivism event was defined as any CPS report filed after study enrollment. Of the entire sample, 69 percent of participants had one or more recidivism events, 52 percent had two or more, and 37 percent had three or more. Two-level recurrent event survival models were used to model recidivism outcomes. The study authors did not conduct subgroup analyses.
Study 2
Chaffin, Bard, and colleagues (2012b) examined a subsample from the larger study (Study 1, Chaffin, Hecht, et al. 2012a, described above), but specifically looked at the effectiveness of the SafeCare program among American Indian parents. Of the original 2,175 parents involved in the larger sample of families, 354 reported American Indian ethnicity and were analyzed for this study. Of this sample, the mean age was 29, and 94 percent were female. The majority (68 percent) lived in small communities, 18 percent lived in urban communities, and 14 percent lived in rural communities. Nearly half (43 percent) self-reported being physically abused as a child, 41 percent self-reported being sexually abused as a child, and 23 percent reported being removed from their own parents at some point during their childhood. In their current households, 45 percent reported some history of domestic violence and had an average of three prior referrals to child welfare. Of those referrals, 93 percent involved child neglect, 19 percent involved physical abuse, and 5 percent involved sexual abuse.
The authors utilized the same treatment and control groups as Study 1. Of the subsample of American Indian parents, 60.5 percent of the parents were assigned to the treatment group (n = 214), and 39.5 percent were assigned to the control group (n = 140). In total, 18 percent were assigned to the services-as-usual/uncoached condition, 21.5 percent to the services-as-usual/coached condition, 30.5 percent to the SafeCare/uncoached condition, and 30 percent to the SafeCare/coached condition. For the CrimeSolutions review of this program, the outcomes that were scored were those pertaining to the direct comparisons between all families who received SafeCare (both coached and uncoached) and all families who received services as usual (coached and uncoached). Due to significant differences in attributes between treatment and control groups that might affect the effectiveness of the intervention, the study authors controlled for the following variables: 1) a historical risk variable (derived from historical American Indian cases and 12 raw variables such as gender, education, and age), and 2) a county-level report for proneness. This approach mirrored the methods to control for differences used in the full study and allowed for a more direct comparison of welfare recidivism outcomes (Chaffin, Hecht, et al. 2012a)
Data were collected in the participant households by an independent research assistant using Audio Computer Assisted Self-Interview (ACASI). Measures were collected from participants at baseline, around the end of services, and 6 months after service exit. The outcomes of interest were parents’ symptoms of depression and child abuse risk. Symptoms of depression were measured by the Beck Depression Inventory, a 21-item multiple choice questionnaire that is self-completed. Child abuse risk was measured by the Child Abuse Potential Inventory, a 160-item agree/disagree questionnaire that is self-reported. Piecewise latent growth curve models were used to analyze the data. The study authors did not conduct subgroup analyses.