Intensive family preservation services: Demonstrating placement prevention using event history analysis

This study re-examined the ability of intensive family preservation services (IFPS) to prevent out-of-home placements of children in abusive or neglectful families. A retrospective, population-based design was used. The sample comprised a statewide, six-year, archival population of high-risk child protective services children. The study ensured a high degree of treatment fidelity among service providers, controlled risk factors that may have adversely affected findings in earlier studies, and used event history analysis to examine treatment effects. IFPS significantly reduced placement rates or delayed placements of children compared with children of the same risk level but who received traditional child welfare services. Treatment effects increased as risk increased. In contrast to previous research, IFPS is shown to be effective in reducing out-of-home placements when model fidelity is high and the service is appropriately targeted. Key words: event history analysis; family preservation; outcomes; placement; placement prevention; risk factors ********** Intensive family preservation services (IFPS) are time-limited (usually four to six weeks), intensive, in-home services designed to prevent the removal of children from home as a result of abuse or neglect (Kinney, Haapala, & Booth, 1991). This study responds to studies that challenged the effectiveness of IFPS in preventing out-of-home placements. Many of those studies used experimental designs and attempted to achieve a high degree of scientific rigor (Feldman, 1991; Shuerman, Rzepnicki, Littell, & Chak, 1993; U.S. Department of Health and Human Services [DHHS], 2001; Yuan, McDonald, Wheeler, Struckman-Johnson, & Rivest, 1990). Studies using experimental designs have produced equivocal findings, begging a review of the designs and methods used. Evidence exists that the research may have failed to detect treatment effects rather than demonstrate a lack of treatment effects (Fraser, Nelson, & Rivard, 1997). Design, model fidelity, and implementation issues may have compromised findings in these studies (Fraser, Nelson, & Rivard, 1997; Heneghan, Horwitz, & Leventhal, 1996; Pecora, Fraser, Nelson, McCroskey, & Meezan, 1995; Rossi, 1992). Rossi (1991, 1992) suggested that equivocal findings in early evaluations might have been due to differences in experimental and control groups with respect to true risk of placement before receiving IFPS. Inadequate attempts by workers to judge risk or refer only high-risk cases resulted in lower-risk cases being served. The most recent federal study (DHHS, 2001) attempted to resolve this issue using a specially designed risk-referral instrument, but the authors reported that the new tool did not identify high-risk families for random assignment (DHHS). Theoretically, IFPS is intended for the highest risk families. However, low placement rates for both experimental and control groups in most earlier studies suggest that lower-risk families were actually receiving IFPS. Under these circumstances, treatment effects would be mitigated when placement prevention is used as the dependent variable. A number of reviewers (Berry, 1992; Fraser, Walton, Lewis, Pecora, & Walton, 1996; Meezan & McCroskey, 1996; Rossi, 1991; Wells & Whittington, 1993) called for analysis of questions relating to family issues, family functioning, and multiple family outcomes to clarify the basis for placement prevention rather than relying solely on the placement prevention statistic. However, addressing criticism in the literature concerning the effect of family preservation services requires that placement prevention be included as an outcome. Problems have also been noted with respect to statistical and analytic approaches used in past research. Fraser and colleagues (1997) conducted a meta-analysis of treatment effects in the treatment literature relating to mental health, juvenile services, and child welfare (including IFPS) and concluded that the studies might have failed to detect treatment effects rather than determine that treatment effects did not occur. …