Printer Friendly

Reentry into out-of-home care: implications of child welfare workers' assessments of risk and safety.

This longitudinal analysis examined predictors of reentry to foster care among children and youths who entered foster care between 2001 and 2007. Three sources of administrative data (Chapin Hall Center for Children longitudinal files, National Child Abuse and Neglect Data System, and structured decision making) from one state were used to assess whether Child Protective Services workers' risk and safety assessment decisions are predictive of reentry into foster care. Cox regression modeling identified characteristics associated with reentry during this period. Findings suggest that current neglect assessment, problems with parenting skills, motivation to improve parenting, safety assessment decision, length of stay, substantiated allegations, and unsubstantiated allegations were associated with likelihood of reentry. The results suggest that child welfare systems could more effectively use assessment tools to match families with needed services.

KEY WORDS: child maltreatment; foster care; risk assessment; safety assessment; structured decision making


When social workers and other child welfare professionals reunify maltreated children and adolescents with their parents, it is with the expectation that known risks have been minimized and that homes are safe. Decisions to reunify families increasingly rely on child welfare workers' perceptions of both risk of maltreatment and household safety. However, there has been little examination of whether social workers' assessments are predictive of child welfare outcomes such as reentry into care.

Since passage of the Adoption and Safe Families Act (ASFA) in 1997 (H.R. 867, 1991), child welfare agencies have been required to seek more timely permanency for children through family preservation, time-limited reunification efforts, and, when reunification was not possible, through termination of parental fights and adoption promotion and support (Child Welfare League of America, 1997). While promoting family reunification through the provision of services, ASFA reduced the amount of time parents have to reunify with their children from 18 months to 12 months.

ASFA also required the development of performance indicators and outcome measures related to the achievement of timely permanency. These changes in policy and practice, and the outcome measures used to evaluate state performance, have led to increased efforts to quickly reunify families after a child has been removed from a home due to child maltreatment. There is general consensus that in some cases, these reunification efforts are unsuccessful and children reenter out-of-home care after being returned to live with their parents. In an effort to protect children from subsequent maltreatment and to standardize assessment practices, many states have implemented empirically based assessment tools for use in child welfare practice.


Decision Making in Child Welfare

Increasingly, child welfare and juvenile justice agencies are adopting assessment instruments as decision-making tools (DePanfilis & Scannapieco, 1994; Krysik & LeCroy, 2002; Shook & Sarri, 2007). In child welfare agencies, these tools generally focus on family-level risk and environmental factors, whereas in juvenile justice settings, they are used to assess individual behavior (Shook & Sarri, 2007). Literature has suggested that these tools can be used to measure or predict case outcomes (Lyle & Graham, 2000), but there have been few empirical studies using assessment instruments in this manner.

Rather, the literature related to these assessment tools has largely examined four domains: (1) definitions required for assessing risk and safety (DePanfilis & Scannapieco, 1994; English & Pecora, 1994); (2) implementation of specific decision-making systems (Doueck, English, DePanfilis, & Moote, 1993; K. Johnson, Wagner, Scharenbrock, & Healy, 2006); (3) challenges related to the use of existing tools and models, including reliability and validity of instruments (Baird & Wagner, 2000; Baird, Wagner, Healy, & Johnson, 1999; Baumann, Law, Sheets, Reid, & Graham, 2005, 2006; Camasso & Jagannathan, 2000; Dorsey, Mustillo, Farmer, & Elbogen, 2008; English & Graham, 2000; Gambrill & Shlonsky, 2000; W. Johnson, 2006a, 2006b; Knoke & Trocme, 2005); and (4) analyses of results using data collected from these instruments (English & Graham, 2000; English & Pecora, 1994; Lyle & Graham, 2000).

The state in which this analysis was conducted uses structured decision making (SDM) risk and safety assessment instruments as decision-making tools (Children's Research Center, 2008). Literature has identified a reduction in repeated incidents of maltreatment as one of the primary goals of SDM tools (Gambrill & Shlonsky, 2000; W. Johnson, 2006a) and suggested that the use of SDM risk assessment tools has been found to result in better rates of permanency for children reunified with their families (K. Johnson & Wagner, 2005).

Decision theory, based on the principle of breaking down complex decisions into smaller components, is generally considered the foundation for SDM and similar assessment systems (Children's Research Center, 2008; Schwalbe, 2004). Decision theory assumes that statistical models can be used to develop standards and that those standards can be the basis for consistent decision making (Lindley, 1985). In this context, SDM instruments provide child welfare workers with a standard that can be used for consistent decision making in child maltreatment assessments. Actuarial or empirically based instruments attempt to predict future maltreatment in a consistent, research-based manner, as compared with consensus-based instruments, which rely more on clinical judgment (Children's Research Center, 2008; D'Andrade, Austin, & Benton, 2008; Knoke & Trocme, 2005; Shlonsky & Friend, 2007).

SDM instruments include both empirically based and consensus-based measures. This analysis used data collected using two specific SDM instruments, the safety assessment tool and the risk assessment tool. The SDM safety assessment tool is a consensus-based instrument, used to evaluate current circumstances for children and families (Children's Research Center, 2008). The SDM risk assessment tool is actuarial, in the sense that it provides a score, and also empirically based, in that it was developed using statistical analyses of risk factors.

Safety Assessment Instruments. Safety assessments are designed to identify those children currently facing an imminent threat of harm (Children's Research Center, 2008). In this state, the safety assessment instrument is generally completed within 24 hours of the first contact with a family and provides a tool for workers to use in making decisions about immediate removal from the home. The SDM safety assessment tool includes a checklist designed to assist workers in assessing the current situation of children to ensure safety (Children's Research Center, 2008). In this state, child welfare workers can complete more than one safety assessment, but generally, one such instrument is completed for each child maltreatment assessment.

Risk Assessment Instruments. Once initial safety concerns have been addressed, generally at the conclusion of an assessment, child welfare agencies using SDM tools evaluate family risk. Risk is considered to be associated with the "long-term likelihood of child maltreatment" and subsequent placement in out-of-home care (Children's Research Center, 2008, p. 9). Therefore, these tools are logical indicators of potential for reentry to foster care.

Reentry into Out-of-Home Care

In 2005, over 500,000 children and youths were living in out-of-home care and of that population, the case plans for over half of these children and youths identified reunification with their biological family as the primary goal (U.S. Department of Health and Human Services, Administration for Children & Families, 2005). However, even when reunification does occur, children may subsequently return to care. In California, Courtney (1995) and Courtney, Piliavin, and Entner Wright (1997) found that, on average, about 19% of foster youths reentered out-of-home care within two to three years of being reunified with their parents. In similar studies, investigators found that around 14% of foster youths reentered foster care after being reunified with their biological parents (Shaw, 2006; Wells & Guo, 1999). Miller, Fisher, Fetrow, and Jordan (2004) conducted a small study and found that of the 16 family reunifications studied, five youths (31%) reentered. Festinger's (1994) study of reunified children in New York found that about 13% of children reentered care within one year.

Predictors of Reentry into Out-of-Home Care

Predictors of reentry to care vary and have included variables reflecting child characteristics, structural features of families, and case characteristics. However, few studies have examined how child welfare workers' perceptions of risk and safety may be associated with reentry into out-of-home care.

Child Characteristics. Some findings have suggested that race and ethnicity are strong predictors of reentry, with African American children and youths reentering at the highest rate (Courtney, 1995; Courtney et al., 1997; Jones, 1998; Shaw, 2006; Wells & Guo, 1999) or that being other than white is associated with reentry (Jones, 1998). Others have not found significant differences between African American children and white children (Terling, 1999). Poverty has also been associated with reentering out-of-home care, in that investigators found that children in Temporary Assistance for Needy Families- or Aid to Families with Dependent Children--eligible families returned to care at a much higher rate than children in noneligible families (Courtney, 1995; Courtney et al., 1997; Jones, 1998).

Research is unclear as to whether age is a predictor of reentry into foster care. Courtney (1995) originally found that children who were seven to 12 years old when they entered care were more likely to reenter care; however, these findings were not replicated in later research (Courtney et al., 1997). Research has suggested that infants have a high rate of reentry into out-of-home care postreunification (Courtney, 1995; Courtney et al., 1997; Shaw, 2006). Shaw noted that reentry rates begin to climb, after a consistent decline postinfancy, for youths ages 14 to 15 years.

Parental Characteristics. Courtney et al. (1997) suggested that parental substance abuse may influence both reunification and reentry to care, in part due to the impact of parental substance use on child health. In a qualitative analysis of cases ending in reentry, Terling (1999) found that the presence of untreated substance abuse issues, even in the secondary caregiver, were associated with reentry to care. Shaw (2006) found that referrals to drug or alcohol services increased the odds of reentry. In a study of infants, Frame, Berrick, and Brodowski (2000) found that maternal substance abuse was associated with reentry. Festinger (1994) identified low parenting skills and lack of social support as the strongest predictors of reentry to care.

Case Characteristics. Studies have found that several case characteristics predict reentry. Most consistently, research has suggested that length of stay (LOS) in out-of-home care affects reentry rates, with shorter LOS generally correlating positively with reentry (Courtney, 1995; Shaw, 2006; Wells & Guo, 1999).

In addition, some studies have found that children placed in care because of neglect were more likely to reenter care than were children placed as a result of abuse (Jones, 1998; Wells & Guo, 1999). Terling (1999) found that neglect cases were the most common type of cases reunified with biological parents and the most common type of maltreatment cases to reenter out-of-home care. Other studies have not found a correlation between maltreatment type and reentry (Miller et al., 2004).

Multiple reports to Child Protective Services (CPS) have been identified as a predictor of reentry. Using administrative data, Frame et al. (2000) researched reentry to out-of-home care for reunited infants and found that the number of CPS reports were higher among infants who reentered care, when compared with infants with successful reunification. Terling (1999) found similar results, with 67% of reentry cases having previous referrals, compared with 12% of non-reentry cases.

Finally, placement type at the time of reunification has also been identified as a predictor of reentry, such that children and youths who were placed in kinship care settings were less likely to reenter out-of-home care than youths in nonkin foster homes or group homes (Courtney, 1995; Courtney et al., 1997; Frame et al., 2000; Wells & Guo, 1999).

Risk and Safety Assessment Components. Despite ongoing discussion about the need to adequately assess risk to minimize reentry (Frame et al., 2000), little empirical research has examined risk and safety indicators as predictors of child welfare outcomes, including reentry outcomes. Festinger (1994) found that child welfare workers' identification of three specific risks (parenting skills, planned discharge, and caregiver attachment) predicted reentry to care. Terling (1999) included one six-point worker assessment of risk and found that workers' assessments were not predictive of reentry. Antle et al. (2007) used a consensus-based instrument to examine comorbidity of domestic violence and neglect and did find such a relationship.


This analysis examined the impact of child characteristics, risk assessment indicators, safety assessment decisions, and case characteristics on reentry into out-of-home care. First, on the basis of previous research, we analyzed whether child characteristics, including age and race or ethnicity affect reentry. Second, we assessed whether, as predicted in the risk assessment literature, SDM risk assessment measures predict a reentry to care after reunification. Third, to assess practical outcomes of SDM safety assessment measures, we explored whether child welfare workers' assessment of child safety are predictive of reentry into out-of-home care. Finally, to inform agency practice, we examined whether case characteristics, including primary placement setting, LOS in care, prior unsubstantiated allegations, and prior substantiated allegations predict reentry to care.


This analysis was conducted using three sources of administrative data from one state's child welfare agency. Child and placement characteristics were obtained from the state's longitudinal foster care placement file (Chapin Hall), which the Center for State Foster Care and Adoption Data--a partnership of the American Public Human Services Association and Chapin Hall Center for Children (2009) at the University of Chicago--created using state administrative child welfare data. Risk and safety assessment characteristics were exported from the state's SDM tools in the administrative data system. Total unsubstantiated and substantiated allegations of maltreatment were compiled from six years of National Child Abuse and Neglect Data System (NCANDS) files compiled from the state's NCANDS submissions.

Analytic File Creation

File Components. This analysis used a linked, longitudinal file developed using the three administrative data sources noted earlier. This analysis uses the Chapin Hall spell file, which includes child-specific and spell-related variables used to describe the experiences of children removed from their caretakers due to child abuse or neglect. A spell is defined as a length of time in out-of-home placement, with the spell start being the day the child is removed from her or his home and the spell end being the day the child is returned home, is placed with a permanent caregiver through adoption or guardianship, or exits the foster care system by reaching the age of majority. A child who reenters out-of-home placement after reunification or permanent placement with caregivers would have more than one spell.

The second data source was the NCANDS Detailed Case Data Component. NCANDS is a voluntary national data collection and analysis system that includes a compilation of child-level information for all child maltreatment investigations completed during the federal fiscal year. Key variables include child and perpetrator demographics, the number of unique allegations (both substantiated and unsubstantiated), types of alleged maltreatment, and the disposition of each investigation. The data are available at the county level. The third data source was SDM risk and safety assessments integrated into the state's administrative data management system. The child welfare agency in this state implemented SDM in December 2001.

File Development. The file used in these analyses was based on the Chapin Hall initial population of children and youths who were in their first spell in out-of-home care between the years 2001 and 2006. The children and youths included in the original Chapin Hall longitudinal file may have experienced more than one episode of out-of-home care during the study period. Therefore, the Chapin Hall file was modified so that it included only the first removal spell for unique children who first entered care between 2001 and 2006 (N = 2,507). Those cases were merged with a restructured NCANDS file reflecting the same six years of data, resulting in a file that included 1,942 cases in which a child's first spell in care corresponded with the NCANDS submission window. NCANDS data were missing for 565 cases or about 22% of the restructured Chapin Hall file. NCANDS data could be missing if, for example, a child's assessment was conducted prior to 2001 or a case was removed during the federal cleaning process for NCANDS files.

The resulting Chapin Hall and NCANDS file was merged with SDM data for the same group of children in the same window, resulting in 1,423 cases with data from all three sources. This sample of cases included some children with placement settings that were included in the Chapin Hall Center for Children data file but were not germane to this specific analysis. Specifically, cases were dropped if a child was primarily placed in a medical setting (14), some other setting (13), or an unknown setting (5). SDM data were missing for 519 cases, or about 27% of the sample, with over 95% of SDM data missing in 2001, the first year of SDM implementation. Once SDM had been officially implemented, the state's data entry system would not allow a worker to close a case without entering SDM information. However, given this data limitation, 16 cases with spell starts in 2001 were removed from the analytic file, leaving a sample of 1,375 children.

Given the tendency for some states to consider sibling groups in placement decisions (Herrick & Piccus, 2009) and studies suggesting that large numbers of children have a sibling in out-of-home care at the same time (Festinger, 1994; Shaw, 2006), likely sibling groups were identified using the NCANDS report identification number as a proxy for family units. There were unique NCANDS report identification numbers for 68% (n = 928) of the children, and the remaining 32% of children were identified as belonging to 183 sibling groups. To avoid violation of the assumption of independence in observations in multivariate analyses, one random child was selected from each of these groups. The sample used in multivariate analyses included these children randomly selected from a sibling group, combined with the children who did not have a sibling removed in the same spell. There were few differences (none statistically significant) in comparing the entire group of reunified children with the sample of children reflecting random siblings and others with no siblings removed.

This analysis reflects a subset of that sample, namely those children who were reunified with their original caretaker, generally a biological parent, after their first spell in care (N = 512). When comparing this analytic sample of 512 reunified children with the larger group of cases (N = 1,375), reunified children had statistically shorter stays in care and higher numbers of substantiated maltreatments but did not otherwise significantly differ from children who were not reunified in age, gender, race, placement type, or risk assessment safety decision characteristics.

Data Analysis

We used a longitudinal file reflecting one entry cohort to avoid possible bias inherent in cross-sectional analyses of foster care data (Wulczyn, Hislop, & Goerge, 2000). Cox proportional hazards modeling (Allison, 1995) was used to investigate models of risk of reentering foster care after being reunified with parents. Cox regression examines the likelihood of an event occurring at a specific point in time, given that an individual is at risk for that event and that the event time is unknown for some cases (Allison, 1995). Here, we examined the risk of foster care reentry, given that each reunified child is at risk for reentry and that the time until reentry was unknown for those children who had not reentered care at end of the Chapin Hall data collection period in 2007. This type of analysis is appropriate in that it also allows for analysis of increased or decreased risk of foster care reentry when some children or youths have reentered care and others have remained in their homes. One advantage of this analytic approach is that Cox modeling accounts for censored cases, or those children who have not yet experienced an event.


Before conducting analyses, we created covariate and outcome variables for Cox regression models from Chapin Hall, SDM, and NCANDS data files. Some variables were captured directly from the original coding, and others were recoded to reflect previous research or the state's child welfare practice.

Reentry. Reentry was noted in the Chapin Hall Center for Children file and is limited to the return of children to out-of-home care after reunification. Time to reentry was defined as the time from reunification to return to out-of-home care. For censored cases in which a reentry did not occur, time to censoring was calculated on the basis of the end of the 2007 Chapin Hall data collection window. This reflects reentry definitions in existing research (Courtney, 1995).

Child Characteristics. Developmental age categories were constructed using the child's age at the end of the initial placement spell, as identified in the Chapin Hall spell file. Child gender identified in the Chapin Hall file was used for this analysis. Child race/ethnicity was also reported in the Chapin Hall data file and was recoded into categories of nonwhite/Hispanic and white/non-Hispanic; more refined categorization of race/ethnicity was not possible because of the limited racial and ethnic variability in the sample (approximately 87% of the original sample was identified as of white/non-Hispanic ethnicity).

Case Characteristics. The Chapin Hall file provides various options for primary placement setting (greater than 50% of time in care), and this analysis trichotomized those settings into congregate care, foster family home, or kinship family home. LOS was defined as the duration of the specific placement spell, beginning with a removal from the home and ending with a discharge to reunification. A categorical measure of LOS (less than six months, six months to one year, and over one year) was used here to minimize bias from statistical outliers. NCANDS fries from 2001 to 2007 were used to calculate the total number of both substantiated and unsubstantiated child maltreatment allegations occurring prior to the initial removal spell. When possible, only allegations occurring within one year of the beginning of the placement spell were included in these indicators. However, it is possible that for some children removed earlier in the observation window (particularly in 2002), NCANDS data may have provided incomplete observations of prior allegations. Both substantiated and unsubstantiated allegation variables were recoded to reflect four categories: cases with no previous allegations, cases with one previous allegation, cases with two previous allegations, and cases with more than two allegations.

SDM Characteristics. Both safety and risk assessment data were pulled for the assessment closest to the removal date. Referrals after spells began were eliminated to use assessment data collected before the child's removal from the home.

The SDM safety assessment instrument asks child welfare workers to rate overall safety, related to the most vulnerable child or youth in the family. If workers select unsafe, the expectation is that a child is removed from the home. Conditionally safe assessments reflect cases in which some corrective intervention has taken place since the report was made. Safe assessments are those in which a worker does not believe that any child in the home is in "imminent danger of serious harm." This instrument is to be completed within 24 hours of the initial face-to-face meeting with the child. Although unsafe cases are expected to involve a removal, it is possible that children are removed when workers make other safety decisions. For instance, a worker may decide that a child is safe or conditionally safe because he or she was temporarily placed with a noncustodial parent or other family member. Subsequently, the child could be removed from that second household if it is determined to be unsafe. Any safety assessment of this second household would not be included in this state's child welfare data, as the administrative database includes only records of a case's first safety assessment tool, even if workers complete more than one as a result of changing safety decisions.

The SDM risk assessment instrument is completed within 60 days from the end of each assessment and asks child welfare workers to score the family's overall risk of maltreatment as low, moderate, high, or very high. The risk assessment tool includes a list of abuse items and a list of neglect items, which are summed to create a total risk score. This analysis examines eight of the abuse indicators and 11 of the neglect indicators from the instrument (see inclusive fist in Table 1). This instrument is completed at the end of the assessment period, which may be prior to removal. A structured risk reassessment is also completed as a component of SDM's Reunification Assessment, but this analysis focuses on the initial risk assessment results.


Description of Sample

This sample of children who reunified after their first spell in care was almost equally divided according to gender (52% female) and primarily white (81%). About one-tenth (13%) were under one year of age at the end of the spell; about 30% were between one and five years of age, about one-third (32%) were between five and 12 years of age, and 26% were over age 12. For all children included here, the median LOS in care was about 0.60 years (about seven months). Forty percent of cases had zero or one prior unsubstantiated allegation, and 73% had one or fewer substantiated allegations. Foster family homes were the primary placement setting for 63% of these children, with 13% in congregate care and 24% in kinship care. Twenty-two percent of the children and youths who were reunified with their parents in this window reentered out-of-home care (n = 115), a finding consistent with that found in research in other states (Terling, 1999).

SDM Indicators

During the most recent assessment that began prior to the child's removal from home, child welfare workers classified 42% of these families as unsafe, 33% conditionally safe, and 25% safe. Abuse risk assessment measures suggested that 34% of the assessments were for abuse, 49% had a prior abuse assessment, and 28% of the families had prior agency involvement. Eighteen percent of primary caregivers were thought to use excessive discipline, and 6% were considered overcontrolling. Over one-fourth (27%) of the children were identified as having special needs, and 11% had a delinquency history. Child welfare workers identified 39% of the primary caregivers as having an attitude consistent with the allegation.


The neglect risk assessment measures indicated that 72% of the assessments were for neglect and that 80% of the families had a prior neglect assessment. Over one-third (35%) of primary caregivers were under 30, 41% were identified as having a substance abuse problem, and 54% were considered to have a history of domestic violence. In more than half of the cases (53%), child welfare workers identified parenting skills as a major problem. Twenty-eight percent of caregivers were believed to be lacking in self-esteem and 23% were seen as apathetic or lacking hope. Over one-third (38%) were experiencing severe financial difficulty, and almost half (48%) were believed to have the motivation to improve parenting.

Predictors of Reentry to Out-of-Home Care

The results of a Cox proportional hazard model of reentry following reunification, with time to reentry as the dependent variable and including child characteristics, SDM measures, and case characteristics as covariates [[chi square](36, N = 398) = 60.34, p = .007] are presented in Table 2. This model was run using the analytic file reflecting one child per household (randomly selected siblings and all children not from a sibling group). This table shows hazard risk ratios and whether or not they vary significantly from the reference categories. A risk ratio of greater than one indicates a greater likelihood of reentry to care, whereas a ratio of less than one indicates a lower likelihood of reentry. A risk ratio of one suggests that there is no difference between the value and that of the comparison category.

Holding constant all of the other variables in the model, seven of the included independent variables were statistically significant. Current neglect assessment, problems with parenting skills, motivation to improve parenting, safety assessment decision, LOS, substantiated allegations, and unsubstantiated allegations were found to be statistically significant predictors of reentry. These findings were compared with findings of models using the initial analytic file (which included multiple siblings from 184 households). When sibling bias was not addressed, five of the variables found to be significant in the random sibling file remained significant. However, the analysis without the sibling adjustment identified two additional predictors as significant (prior abuse assessment and child race) and did not find neglect or parenting skills to be significant predictors of reentry. This comparison suggests that, as noted in the literature, patterns in sibling placements may introduce bias in unadjusted analyses (Guo & Wells, 2003).

Several risk assessment indicators were identified as significant predictors of reentry into out-of-home care. Children from families with current neglect assessments were 60% less likely to reenter than children from families with other assessment types. Children from families in which parents were identified as having major parenting skills were 40% less likely to reenter as compared with other children. Children from families with parents seen as motivated to improve parenting were 60% less likely to reenter care.

Child protection workers' assessment of safety was also identified as a statistically significant predictor of reentry, with children from households initially assessed as "conditionally safe" being 55% less likely to reenter care than those children from households assessed as unsafe. Children whose LOS in care was between six months and one year were 53% less likely to reenter than children in care for less than six months. Children with more than two unsubstantiated allegations were 276% more likely to reenter care than children with no unsubstantiated allegations. Any number of substantiated allegations was predictive of reentry. Children with one substantiated allegation were 85% more likely to reenter, those with two were about 150% more likely to reenter, and those with more than two were 310% more likely to reenter than children with no substantiated allegations. Other child, case, and SDM characteristics were not statistically significant predictors of reentry.

One minus survival curves of reentry and length of time between placement spells, with lines representing the proportion of cases that ended in reentry for each category of LOS in care during initial spell are illustrated in Figure 1. Among this sample of children and youths in out-of-home care in the 2002 through 2006 window, children with initial placement spells of more than one year reentered most slowly. At one year, about 17% of children in care for six months or less reentered care, compared with close to 12% of those with spells between six months and a year and about 7% of those with spells more than one year in length.


These results provide evidence that child welfare workers' assessments of risk and safety are predictive of reentry outcomes. Three of the individual risk assessment items and the overall safety decision were found to be predictive of reentry into care. In a general sense, this finding suggests that workers' assessments of risk and safety may have practical implications for reentry outcomes.

The finding that cases identified as neglect assessments were associated with less likelihood of reentry reflects previous results (Jones, 1998; Wells & Guo, 1999). As Connell et al. (2009) have noted, many neglect cases likely reflect poverty or other chronic elements that may require interventions with an emphasis on the entire family. Additional research is needed, given the wide range of conditions labeled as neglect, including lack of supervision, substance abuse, unsafe living conditions, and other factors. Future research could determine which specific components or dimensions of neglect are most predictive of reentry into care.

The finding that children from families in which parenting skills were a major problem were less likely to reenter out-of-home care raises interesting questions. Festinger (1994) found that low parenting skills predicted reentry, whereas the present results suggest that identification of problems with parenting skills is associated with a decreased likelihood of reentry. This may imply that when child welfare workers note challenges in parenting skills in their assessment of family risk, appropriate interventions are put in place and contribute to better outcomes for maltreated children. The data examined here do not include measures of services provided, so this possibility should be explored with intervention-level indicators.

The finding that children reunified with parents who were initially seen as motivated were less likely to reenter care supports ongoing child welfare efforts to maximize parental capacity and family strengths (National Association of Public Child Welfare Administrators, 2009). It is likely that parental motivation serves as a mediator that could explain some of the ways that parenting skills affect reentries. For instance, it is possible that when motivation to improve parenting is present, parenting skills training or similar interventions have a stronger impact on outcomes such as reentry into out-of-home care. An implication may be that child welfare agencies are effective at working with caregivers who are motivated to improve their parenting skills. However, only half of the parents in the present sample were considered to have such motivation, leaving practical questions about how to engage this other group of caregivers.

Safety assessment was predictive of reentry outcomes in this analysis, suggesting that children removed from homes that were initially assessed as conditionally safe are less likely to reenter than children removed from unsafe homes. Workers assess safety to make initial removal decisions, and although less than one-third of this sample of children were initially assessed as living in unsafe households, all of these children were removed from their homes within the study window. Child welfare workers' safety assessments could be reflective of not only immediate options to ensure children's safety, but also longer term safety within the home environment. Therefore, one implication for child welfare practice may be that safety should be continually reassessed as conditions change.

LOS in care, found to be a significant predictor of reentry here, is consistent with existing literature (Courtney, 1995; Shaw, 2006; Wells & Guo, 1999). Shorter LOS have consistently been associated with reentry to out-of-home care, but it is difficult to assess what this finding means in practical terms. As has been stated elsewhere, these results do not indicate that longer stays in care are desired (Shaw, 2006), and it is not clear that the amount of time children spend in care can be considered a measure of child well-being or of the effectiveness of the child welfare system (Courtney & Wong, 1996). These findings may imply that within this state, some short stays in care may not provide families with enough time to make lasting change. This may inform agency practice in cases involving substance abuse or other long-term challenges within families and support efforts to more quickly engage families in substance abuse or other treatment interventions.

The finding that the number of unsubstantiated and substantiated allegations were significant predictors of reentry reflects other research findings that new substantiations or investigations correspond with increase in risk level (Baird & Wagner, 2000). Families with multiple reports prior to a child's removal from home may present with unique challenges and could be already known to multiple human services agencies. In some instances, these prior reports may have been made by professionals from community agencies, suggesting that child welfare workers may have opportunities to collaborate with other systems, such as schools, mental health providers, and other supportive services in reunification efforts and follow-up services.


Although this study illustrates the research utility of risk and safety assessment tools, it is not without limitations. First, worker error in entering data or system error in creating indicators may have contributed to some error in these data files. This type of challenge is associated with most analyses of administrative data (English, Brandford, & Coghlan, 2000) but deserves note- here. Second, some cases were not included here, because of either missing records or challenges related to linking multiple data sources (Simpson, Imrey, Geling, & Butkus, 2000). Third, specific features of these data might have affected these analyses. However, in cases with potentially high consequences, these instruments may not directly affect workers' judgments (Baumann et al., 2005) or actions. Fourth, this analysis uses the SDM measures to assess whether the child experienced abuse or neglect but does not include more specific detail about maltreatment characteristics, which may conceal important associations among specific types of maltreatment (Knoke & Trocme, 2005). Fifth, these data reflect results from one state during one observation window. Using longitudinal rather than cross-sectional data does allow for a more comprehensive analysis of outcomes such as reentry to foster care, but a longer window may have provided additional detail regarding child welfare outcomes. Finally, the data used here do not include several important indicators, such as remaltreatment details (Connell et al., 2009) or provision of services.


This analysis provides some evidence that child welfare workers' assessments of risk and safety can identify families at risk of reentry into out-of-home care. SDM and similar assessment instruments can be valuable tools for decision making around removal and placement priorities. In addition, child welfare agencies should link these decisions with targeted interventions and services, specifically related to addressing identified risks in reunification planning and following reunification.

Future research should examine what occurs in high-risk families in the time period not only between removal and reunification but also from reunification to reentry. It may be that the services provided to these families are limited or do not adequately address underlying challenges. Child welfare agencies could consider whether best practice or evidence-based interventions effectively minimize specific risks or safety hazards for families with reunification as a goal.


Readers: Submit your reactions to and comments about an article published in Social Work Research or a contemporary issue on social work research. Send your letter (two double-spaced pages or fewer) as a Word document through the online portal at http:// (initial, one-time registration is required).

doi: 10.1039/swr/svs011


Adoption and Safe Families Act of 1997, H.R. 867, 105th Cong. (1991). Retrieved from fdsys/pkg/BILLS-105hr867enr/pdf/BILLS-105hr867 enr.pdf

Allison, P. D. (1995). Survival analysis using SAS: A practical guide. Cary, NC: SAS Institute, Inc.

Antle, B. F., Barbee, A. P., Sullivan, D., Yankeelov, P., Johnson, L., & Cunningham, M. R. (2007). The relationship between domestic violence and child neglect. Brief Treatment and Crisis Intervention, 7, 364-382.

Baird, C., & Wagner, D. (2000). The relative validity of actuarial and consensus-based risk assessment systems. Children and Youth Services Review, 22, 839-871. doi:10.1016/S0190-7409(00)00122-5

Baird, C., Wagner, D., Healy, T., & Johnson, K. (1999). Risk assessment in child protective services: Consensus and actuarial model reliability. Child Welfare, 78, 723-748.

Baumann, D.J., Law, R.J., Sheets, J., Reid, G., & Graham, J. C. (2005). Evaluating the effectiveness of actuarial risk assessment models. Children and Youth Services Review, 27, 465-490. doi:10.1016/j. childyouth.2004.09.004

Baumann, D.J., Law, R.J., Sheets, J., Reid, G., & Graham, J. C. (2006). Remarks concerning the importance of evaluating actuarial risk assessment models: A rejoinder to Will Johnson. Children and Youth Services Review, 28, 715-725. Retrieved from childyouth.2005.08.005

Camasso, M.J., & Jagannathan, R. (2000). Modeling the reliability and predictive validity of risk assessment in child protective services. Children and Youth Services Review, 22, 873-896. doi:10.1016/S0190-7409(00) 00121-3

Chapin Hall Center for Children. (2009). The Center for State Foster Care and Adoption data. Retrieved from

Child Welfare League of America. (1997). Summary of the Adoption and Safe Families Act of 1997. Retrieved from

Children's Research Center. (2008). The structured decision making model: An evidenced-based approach to human services. Retrieved from

Connell, C., Vanderploeg, J. J., Katz, K. H., Caron, C., Saunders, L., & Tebes, J. K. (2009). Maltreatment following reunification: Predictors of subsequent Child Protective Services contact after children return home. Child Abuse & Neglect, 33, 218-228. doi: 10.1016/j.chiabu.2008.07.005

Courtney, M. E. (1995). Reentry to foster care of children returned to their families. Social Service Review, 69, 226-241.

Courtney, M. E., Piliavin, I., & Entner Wright, B. R. (1997). Transitions from and returns to out-of-home care. Social Service Review, 71, 652-667.

Courtney, M. E., & Wong, Y.-L.I. (1996). Comparing the timing of exits from substitute care. Children and Youth Services Review, 18, 307-334. Retrieved from 0190-7409(96)00008-4

D'Andrade, A., Austin, M.J., & Benton, A. (2008). Risk and safety assessment in child welfare: Instrument comparisons. Journal of Evidence-Based Social Work, 5 (1-2), 31-56.

DePanfilis, D., & Scannapieco, M. (1994). Assessing the safety of children at risk of maltreatment: decisionmaking models. Child Welfare, 73, 229-241.

Dorsey, S., Mustillo, S. A., Farmer, E.M.Z., & Elbogen, E. (2008). Caseworker assessments of risk for recurrent maltreatment: Association with case-specific risk factors and re-reports. Child Abuse & Neglect, 32, 377-391. doi:10.1016/j.chiabu.2007.06.006

Doueck, H.J., English, D.J., DePanfilis, D., & Moote, G. T. (1993). Decision-making in child protective services: A comparison of selected risk assessment systems. Child Welfare, 72, 442-449.

English, D., & Graham, J. C. (2000). An examination of relationships between children's protective services social worker assessment of risk and independent LONGSCAN measures of risk constructs. Children and Youth Services Review, 22, 897-933. doi:10.1016/ S0190-7409(00)00120-1

English, D., & Pecora, P.J. (1994). Risk assessment as a practice method in child protective services. Child Welfare, 73, 452-466.

English, D.J., Brandford, C. C., & Coghlan, L. (2000). Data-based organization change: The use of administrative data to improve child welfare programs and policy. Child Welfare, 79, 499-516.

Festinger, T. (1994). Returning to care: Discharge and reentry to foster care. Washington, DC: Child Welfare League of America.

Frame, L., Berrick, J. D., & Brodowski, M. L. (2000). Understanding reentry to out-of-home care for reunified infants. Child Welfare, 79, 339-369.

Gambrill, E., & Shlonsky, A. (2000). Risk assessment in context. Children and Youth Services Review, 22, 813-837. doi:10.1016/S0190-7409(00)00123-7

Guo, S., & Wells, K. (2003). Research on timing of foster care outcomes: One methodological problem and approaches to its solution. Social Service Review, 77, 1-24. doi:10.1086/345702

Herrick, M. A., & Piccus, W. (2009). Sibling connections: The importance of nurturing sibling bonds in the foster care system. In D. N. Silverstein & S. L. Smith (Eds.), Siblings in adoption and foster care: Traumatic separations and honored connections (pp. 27-42). Westport, CT: Praeger.

Johnson, K., & Wagner, D. (2005). Evaluation of Michigan's foster care management system. Research on Social Work Practice, 15, 372-380. doi:10.1177/ 1049731505276312

Johnson, K., Wagner, D., Scharenbrock, C., & Healy, T. (2006). Minnesota Department of Human Services, Risk assessment validation: A prospective study. Madison, WI: Children's Research Center.

Johnson, W. (2006a). Post-battle skirmish in the risk assessment wars: Rebuttal to the response of Baumann and colleagues to criticism of their paper, "Evaluating the effectiveness of actuarial risk assessment models." Children and Youth Services Review, 28, 1124-1132. doi:10.1016/j.childyouth.2006. 04.002

Johnson, W. (2006b). The risk assessment wars: A commentary response to "Evaluating the effectiveness of actuarial risk assessment models" by Donald Baumann, J. Randolph Law, Janess Sheets, Grant Reid, & J. Christopher Graham, Children and Youth Services Review, 27, pp. 465-490. Children and Youth Services Review, 28, 704-714. doi:10.1016/j.childyouth.2005. 07.007

Jones, L. (1998). The social and family correlates of successful reunification of children in foster care. Children and Youth Services Review, 20, 305-323. doi:10.1016/ SO 190-7409 (98)00009-7

Knoke, D., & Trocme, N. (2005). Reviewing the evidence on assessing risk for child abuse and neglect. Brief Treatment and Crisis Intervention, 5, 310-327.

Krysik, J., & LeCroy, C. (2002). The empirical validation of an instrument to predict risk of recidivism among juvenile offenders. Research on Social Work Practice, 12, 71-81. doi:10.1177/104973150201200106

Lindley, D. V. (1985). Making decisions (2nd ed.). New York: John Wiley & Sons.

Lyle, C. G., & Graham, E. (2000). Looks can be deceiving: Using a risk assessment instrument to evaluation the outcomes of child protective services. Children and Youth Services Review, 22, 935-949. doi:10.1016/ S0190-7409(00)00119-5

Miller, K., Fisher, P., Fetrow, B., & Jordan, K. (2004). Trouble on the journey home: Reunification failures in foster care. Children and Youth Services Review, 28, 260-274. doi:10.1016/j.childyouth.2005.03.010

National Association of Public Child Welfare Administrators. (2009). A framework for safety in child welfare. Retrieved from FrameworkforSafety.pdf

Schwalbe, C. (2004). Re-visioning risk assessment for human service decision making. Children and Youth Services Review, 26, 561-576. doi:10.1016/j.childyouth.2004.02.011

Shaw, R. (2006). Reentry into the foster care system after reunification. Children and Youth Services Review, 28, 1375-1390. doi:10.1016/j.childyouth.2006.02.006

Shlonsky, A., & Friend, C. (2007). Double jeopardy: Risk assessment in the context of child maltreatment and domestic violence. Brief Treatment and Crisis Intervention, 7, 253-274.

Shook, J. J., & Sarri, R. C. (2007). Structured decision making in juvenile justice: Judges' and probation officers' perceptions and use. Children and Youth Services Review, 29, 1335-1351. doi:10.1016/j.childyouth. 2007.05.008

Simpson, D. G., Imrey, P. B., Geling, O., & Butkus, S. (2000). Statistical estimation of child abuse rates from administrative databases. Children and Youth Services Review, 22, 951-971. doi:10.1016/S0190-7409(00) 00118-3

Terling, T. (1999). The efficacy of family reunification practices: Reentry rates and correlates of reentry for abused and neglected children reunited with their families. Child Abuse & Neglect, 23, 1359-1370. doi: 10.1016/S0145-2134(99)00103-9

U.S. Department of Health and Human Services, Administration for Children & Families. (2005). The AFCARS report. Retrieved from http://www.acf.hhs. gov/programs/cb/stats_research/afcars/tar/report13.htm

Wells, K., & Guo, S. (1999). Reunification and reentry of foster children. Children and Youth Services Review, 21, 273-294. doi:10.1016/S0190-7409(99)00021-3

Wulczyn, F., Hislop, K. B., & Goerge, R. M. (2000). Foster care dynamics 1983-1998: A Report from the Multistate Foster Care Data Archive (pp. i-62). Chicago: Chapin Hall Center for Children.

Wulczyn (Chapin Hall Center for Children), Christian Connell (The Consultation Center, Yale University School of Medicine), Kathleen Lamphere, and Rebecca Austin.

Melissa Wells, PhD, LICSW, is associate professor, Department of Social Work, University of New Hampshire, 55 College Road, Durham, NH 03824; e-mail: melissa. Melissa Correia, MA, is systems improvement advisor, data advocacy, Casey Family Programs, Seattle. This research was previously presented in draft form at the Society for Social Work and Research Conference, January 2009, New Orleans, and the National Research Conference on Child and Family Programs and Policy, July 2009, Bridgewater, Massachusetts, and was supported by a Summer Faculty Fellowship Grant from the University of New Hampshire. The authors acknowledge the assistance of Fred

Original manuscript received March 10, 2009

Final revision received June 2, 2009

Accepted September 22, 2010

Advance Access Publication October 10, 2012
Table 1: Demographic and Case Characteristics for All
Reunified and Random Siblings and Only Children

Characteristic                            All Reunified      Random
                                          Children 100%     Sibling
                                            (N = 512)       (N = 398)
Child characteristics
  Female                                        52             50
  Male                                          48             49
Age at exit from initial spell
  Infancy (0-1)                                 13             11
  Early childhood (over 1-5)                    29             30
  Middle childhood (over 5-12)                  32             35
  Adolescence (over 12)                         26             24
  White/non-Hispanic                            81             82
  Nonwhite/Hispanic                             19             18
Abuse risk assessment characteristics
  Current assessment is for abuse
  No                                            66             68
  Yes                                           34             32
Any, prior abuse assessment
  No                                            51             51
  Yes                                           49             49
Prior agency involvement
  No                                            72             72
  Yes                                           28             28
Caregiver excessive discipline
  No                                            82             82
  Yes                                           18             18
Caregiver overcontrolling
  No                                            94             94
  Yes                                           6               6
Child has special needs
  No                                            73             72
  Yes                                           27             28
Child has delinquency history "
  No                                            89             91
  Yes                                           11              9
Caregiver attitude consistent with
  No                                            39             39
  Yes                                           61             61
Neglect risk assessment characteristics
  Assessment is for neglect
  No                                            28             26
  Yes                                           72             74
Any prior neglect assessment
  No                                            20             20
  Yes                                           80             80
Primary caregiver age (in years)
  30 or older                                   65             67
  Under 30                                      35             33
Any caregiver substance abuse
  No                                            59             57
  Yes                                           41             43
Any history of domestic violence
  No                                            46             45
  Yes                                           54             55
Parenting skills a major problem
  No                                            47             47
  Yes                                           53             53
Caregiver lacks self-esteem
  No                                            72             71
  Yes                                           28             29
Caregiver apathetic or lacking hope
  No                                            77             74
  Yes                                           23             26
Severe financial difficulty
  No                                            62             58
  Yes                                           38             42
Motivation to improve parenting
  No                                            52             52
  Yes                                           48             48
Safety assessment characteristics
  Safety assessment decision
  Unsafe                                        42             41
  Conditionally safe                            33             34
  Safe                                          25             25
Case characteristics
  Primary (over 50%) placement setting
  Congregate care                               13             12
  Foster family home                            63             64
  Foster home, kinship                          24             24
LOS (median years)                             0.61           2.33
  Less than six months                          45             45
  Six months to one year                        33             24
  Over one year                                 30             29
  Missing                                       <1             <1
Unsubstantiated allegations
  None                                          18             18
  One                                           22             23
  Two                                           22             22
  More than two                                 38             37
Total substantiated allegations
  None                                          29             30
  One                                           44             43
  Two                                           19             21
  More than two                                 8               6
Time to next spell (median years)              0.59           2.31

Note: LOS = length of stay.

Table 2: Cox Regression Models for Reentry to Foster Care (N = 398)

Characteristic                    B       SE      Risk        95% CI

Child characteristics
    Male (a)
    Female                      -0.17    0.24      0.84     0.53-1.34
  Age (in years) at exit
      from initial spell
    Infancy (0-1) (a)
    Early childhood (>1-5)      -0.33    0.37      0.72     0.35-1.48
    Middle childhood (>5-12)    -0.45    0.42      0.64     0.28-1.45
    Adolescence (over 12)       -0.58    0.48      0.56     0.22-1.42
    Nonwhite/Hispanic (a)
    White/non-Hispanic          -0.31    0.29      0.73     0.41-1.29
Risk assessment
  Current assessment for
    No (a)
    Yes                         -0.31    0.43      0.74     0.32-1.69
  Any prior abuse assessment
    No (a)
    Yes                         -0.41    0.28      0.67     0.39-1.14
  Prior agency involvement
    No (a)
    Yes                          0.19    0.27      1.20     0.70-2.06
  Caregiver excessive
    No (a)
    Yes                          0.23    0.33      1.26     0.65-2.43
  Caregiver overcontrolling
    No (a)
    Yes                         -0.87    0.59      0.42     0.13-1.34
  Child has special needs
    No (a)
    Yes                         -0.12    0.28      0.89     0.52-1.52
  Child has delinquency
    No (a)
    Yes                         -0.23    0.45      0.80     0.33-1.91
  Caregiver attitude
    consistent with
    No (a)
    Yes                         -0.04    0.34      0.96     0.50-1.85
  Current assessment for
    No (a)
    Yes                         -0.90    0.45     0.41 *     0.17-.99
  Any prior neglect
    No (a)
    Yes                          0.03    0.36      1.02     0.51-2.06
  Primary caregiver age
    (in years)
    30 or Older (a)
    Under 30                    -0.06    0.29      0.94     0.53-1.67
  Any caregiver substance
    No (a)
    Yes                         -0.21    0.25      0.81     0.49-1.33
  Any history of domestic
    No (a)
    Yes                          0.20    0.25      1.22     0.75-1.97
  Parenting skills a major
    No (a)
    Yes                         -0.58    0.28     0.58 *     0.32-.97
  Caregiver lacks self-esteem
    No (a)
    Yes                         -0.05    0.32      0.96     0.51-1.78
  Caregiver apathetic or
    lacking hope
    No (a)
    Yes                          0.25    0.33      1.29     0.68-2.43
  Severe financial
    No (a)
    Yes                         -0.11    0.27      0.89     0.53-1.50
  Motivation to improve
    No (a)
    Yes                         -0.91    0.32    0.40' *     0.22-.75
  Caregiver response
    No (a)
    Yes                         -0.08    0.36      0.92     0.46-1.86
Safety assessment
  Safety assessment decision
  Conditionally safe            -0.81    0.30    0.45 **     0.25-.80
  Safe                          -0.27    0.32      0.77     0.41-1.45
Case characteristics
  Primary (>50%) placement
    Congregate care (a)
    Foster family home          -0.29    0.41      0.75     0.34-1.68
  Foster home, kinship          -0.55    0.47      0.58     0.23-1.44
    Less than six months (a)
    Six months to one year      -0.75    0.33     0.47 *     0.25-.90
    More than one year            29     0.30      0.75     0.41-1.35
  Unsubstantiated allegations
    None (a)
    One                          0.49    0.45      1.63     0.68-3.91
    Two                          0.65    0.45      1.91      0.79-4.6
    More than two                1.33    0.41    3.76 ***   1.69-8.37
  Total substantiated
    None (a)
    One                          0.61    0.31     1.85 *    1.00-3.40
    Two                          0.90    0.36     2.47 *    1.21-5.02
    More than two                1.41    0.42    4.10 ***   1.77-9.53

Notes: CI = confidence interval; LOS = length of stay.

(a) Reference category for contrasts.

* p < 05. ** p < .01. *** p < .001.
COPYRIGHT 2012 Oxford University Press
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 2012 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Author:Wells, Melissa; Correia, Melissa
Publication:Social Work Research
Article Type:Report
Geographic Code:1USA
Date:Sep 1, 2012
Previous Article:Material assistance: who is helped by nonprofits?
Next Article:Constructing practitioner research.

Terms of use | Privacy policy | Copyright © 2021 Farlex, Inc. | Feedback | For webmasters |