The United States child welfare system cares for nearly half a million foster children each year. The system is tasked with providing children with a safe and stable environment until a permanent arrangement is achieved. The stability of foster care arrangements has been a focal point of foster care practice and policy reforms given evidence that instability is associated with negative impacts on foster children's immediate and long term well-being (Newton, Litrownik, & Landsverk, 2000; Unrau, Seita, & Putney, 2008). Since the 1997 Adoption and Safe Families Act, the federal government, in their Child and Family Service Reviews, has assessed states on the placement stability of children in their foster care systems. In the most recently released findings, no states met the high performance target for stability across all reviewed cases.1 According to states, an insufficient number of foster care placements, lack of foster parent training, and limited resources to support foster parents present challenges to placement stability (U.S. Department of Health and Human Services, 2011). However, the review concluded that kinship care was a viable approach for improving stability outcomes. If true, this would appear to be a less expensive approach than allocating funds for improving recruitment, training, and support services.
Notably, stability and permanency are often considered together, but this study focuses only on stability, which refers to the movements of children during their time in foster care, irrespective of where they ultimately end up – returned home, adopted, long term foster care, or some other outcome. On the other hand, permanency is about whether children ultimately achieve a permanent living arrangement, and how long it has taken to achieve that outcome.
This study uses a statewide administrative database to explore differences in placement change experiences by placement type, focusing specifically on non-relative foster care (NRFC) and formal (court-ordered) kinship care (KC). Although there have been many studies on this topic, this study is able to address some limitations of prior work. Specifically, past research has been limited, to varying degrees, by non-representative samples, interval censored data, small lengths of observation, and an atheoretical approach to analysis. This study includes 8 years of data from an entire state, and includes exact dates of placement entries and exits rather than interval censored data. This permits for more generalizable and precise estimates. Secondly, the data used here explore additional questions to address not simply whether kinship care placements are more stable than non-relative foster care (as prior research broadly concludes), but also why that seems to be the case. That is, there are at least three reasons children in kinship care would have a lower risk of disruption than children in non-relative foster care. First, it may be an issue of selection bias. That is, the children who enter NRFC are less advantaged on a number of factors relative to children in KC—they are more likely to have a disability or health problem, and exhibit more behavior problems and cognitive deficits when entering care (Beeman et al, 2000; Font, 2014; Grogan-Kaylor, 2000). Thus, children who enter NRFC may be at higher risk of disruption, irrespective of the placement type. If true, then given similar children, NRFC and KC placements should be at similar risk of disruption. Second, it is often implicitly assumed that higher stability in kinship care is attributable to differences in foster parents. Specifically, it is believed that kinship foster parents may be more attached, or more committed to, a child in their care because of their shared lineage. If this explanation were true, then the gap in stability should be explained primarily by lower rates of foster parent-requested moves among children in KC. And third, it may be the case that children in NRFC move more often because of differences in how KC and NRFC placements are valued. That is, children may be intentionally moved from NRFC to KC because KC is a more preferred placement setting—to the contrary, there would be no reason to move children from KC to NRFC absent a deficit with the KC placement. If this explanation were true, then it should be observed that children in NRFC are more likely to experience moves to more-preferred placements, but equally likely to experience moves to less or equally preferred placements.
To examine these possible explanations, I examine 3 questions: (1) Among high-risk subgroups of children, are NRFC placements more likely to disrupt than KC placements?; (2) Is higher stability in KC driven by lower rates of foster-parent requested moves?; and (3) When children move from NRFC or KC, what are the respective probabilities of moving to more, equally, or less preferred placements? All of these questions have important implications for how to best use kinship care, and for efforts to expand KC.
Background
Most children do not experience frequent moves while in foster care, although the risk of multiple moves increases the longer a child remains in care (Wulczyn, et al 2003). James, Landsverk and Slymen (2004) suggest that foster children who spend at least 18 months in foster care experience an average of 4.4 placements, and whereas the majority of foster children are in a stable placement within 9 months, a third of children do not achieve long term stability.
Several studies have examined the association between placement type and placement stability; particularly examining whether children in formal kinship care (KC) have higher placement stability than children in non-relative foster care (NRFC). Studies differ in their generalizability, rigor in accounting for social selection, and length of observation. These differences in measurement and study methods lead to somewhat different conclusions, though nearly all studies suggest greater stability in kinship care, at least in the short term (Chamberlain et al., 2006; Koh & Testa, 2008; Koh, 2010; Koh, Rolock, Cross, & Eblen-Manning, 2014; Strijker, Knorth, & Knot-Dickscheit, 2008; Usher, Randolph, & Gogan, 1999; Webster, Barth, & Needell, 2000; Winokur, Holtan, & Valentine, 2009).
Other studies add nuance to these conclusions. The difference in the risk of placement disruption between kinship and non-relative foster care is generally highest for the first placement, and smaller for all subsequent placements (Koh & Testa, 2008; Koh, 2010), perhaps because most placement moves occur within 6 months of entering foster care (Wulczyn et al., 2003). However, children who remain in care longer are more likely to experience multiple placements, irrespective of the type of placement in which they were first placed (Usher et al, 1999). It is unclear whether children who stay in care longer are more likely to have multiple placements simply because they remain at risk of placement disruption for a longer period of time, or because the characteristics associated with long stays in foster care are also associated with placement instability. Nevertheless, studies that only consider the first placement change or follow children for a shorter period of time tend to find larger effects of kinship care, with some estimates suggesting that children's risk of any placement change or having a higher number of placement changes is exponentially higher in NRFC as compared with KC (Chamberlain et al, 2006; Koh et al, 2014; Webster et al, 2000).2
Prior work nearly always finds higher stability in kinship foster care as compared with non-relative foster care; yet, this body of work has been notably atheoretical. The broader literature on kinship care can provide insight, however. In the sections below, I consider the theoretical support for each of the three possible explanations for stability differences between KC and NRFC outlined above.
Child Selection Factors
Are differences in stability being incorrectly attributed to the type of placement rather than the type of child in the placement? Studies comparing stability outcomes of children in NRFC and KC require the general assumption that children in NRFC and KC are comparable, meaning that they are similar on all relevant characteristics, or that all relevant differences are known and accounted for in the model. However, assignment to a given type of placement is not random – placement decisions are made based on children's unique circumstances as well as policy priorities. As it pertains to children's circumstances, children going into kinship care tend to have more advantageous characteristics prior to entering out-of-home care (OHC), including fewer behavior problems and higher cognitive abilities (Font, 2014). Children entering kinship care are also less likely to have disabilities or health problems, and are more likely to have been removed for neglect (Beeman et al, 2000; Grogan-Kaylor et al, 2000). This suggests a process through which higher-functioning children enter kinship arrangements at a higher rate than non-relative placements. At the same time, these factors associated with selection into placements are often also associated with instability. Evidence suggests that age, history of sexual abuse, and behavior problems are the best predictors of placement instability (Barber, Delfabbro, & Cooper, 2001; Chamberlain et al., 2006; James, 2004; Oosterman, Schuengel, Wim Slot, Bullens, & Doreleijers, 2007). In addition, placement changes may exacerbate existing behavior problems (Newton et al, 2000), which is consistent with data suggesting that each placement change increases the risk of a subsequent placement change (Webster et al., 2000).
Moreover, setting aside known differences, it is likely to be the case that children entering kinship care differ on unobservable characteristics as well. Following a wave of policy changes in the last few decades, today's children generally only enter NRFC if a kinship placement is not available. Having a relative who is willing and able to provide care is not likely to be an isolated factor – the mere existence of a relative who is both able and willing to take a child into their home suggests that child may have stronger familial ties or a more involved extended family. Moreover, given evidence suggesting intergenerational patterns of maltreatment, substance abuse, and mental illness (Kendler, Davis, & Kessler, 1997; Kim, 2009; McCloskey & Bailey, 2000), all of which are common antecedents to out-of-home placement, having relatives that are able to be approved for placement by the relevant child welfare agency may be advantageous in itself, irrespective of whether the placement occurs. The percentage of foster children in kinship placements in 2012 was 28%, which, despite being an increase since bottoming out at 24% in 2002, is exactly the same proportion observed in 1998 (U.S. Department of Health and Human Services, 2006, 2013). Despite widespread efforts to increase kinship placements, there is a large swath of children for whom no relative placement was identified and approved by the CPS agency and family court. It is possible that, in practice, the legally-required efforts to identify an appropriate relative placement are not always made and a relative placement could have been identified through more intensive efforts. Nevertheless, children who do enter kinship care may constitute a unique subset of children entering the foster care system.
Foster Parent Factors
Theory offers some indication that kinship care should be more stable, at least to the extent that kinship foster parents may be less likely to request a child be moved from the home. Whereas little comparative work in this area has been theory-driven, a study of only kinship foster placements assesses a social exchange theory (Testa & Slack, 2002). Their findings suggest that several factors influence placement stability in KC, including the foster parent-child relationship, monetary compensation, sense of duty (obligation) and the extent to which the biological parents were making efforts toward reunification. Although there is variation in the quality of kinship foster parent-child relationships, kinship foster parents have some advantage over non-relative foster parents in bond formation. Even in the absence of a well-developed pre-existing relationship between the foster child and the kinship foster parent, shared family history, culture, and traditions may facilitate bond formation. Moreover, because most foster children are ultimately reunified, non-relative foster parents may be reluctant to build a strong bond with a child who is likely to leave soon, even if their reluctance increases the risk of loss.
Notably, however, theoretical expectations of higher placement stability in KC, to a large extent, assume that the kinship foster parent had a relationship with the foster child that preceded the placement, or minimally that the relative would feel some sense of obligation either to the child or the child's parents. However, the accuracy of such an assumption is not well-established empirically. Given explicit mandates to search for relatives when a relative is not immediately identified, caseworkers may be drawing from an increasingly distal pool of relatives. To the contrary, there may be a selection process through which kin with a pre-established bond are be more likely to select into fostering and kin lacking a bond with the child are more likely to decline to foster. This latter concept is, to some extent, supported by evidence maternal relatives (with whom families often have a higher degree of contact) are more likely to be kinship foster parents and provide higher placement stability than paternal relatives (Perry, Daly, & Macfarlan, 2014). Nevertheless, a pre-existing bond would be one possible mechanism through which KC could produce higher placement stability than NRFC, as explained in the ensuing paragraphs.
As behavior problems are suggested in prior research as an important antecedent to moves, how non-relative and kinship foster parents are likely to deal with children's behavior problems warrants consideration. Again, premised on a pre-existing bond or sense of familial obligation, kinship caregivers may be more willing to work with maladaptive behaviors. Although many studies have considered behavior problems to be an outcome affected by placement type, the data derived from such studies suggests something potentially more complex. That is, caregiver reports of behavior tend to suggest children have better behavior in kinship care (Holtan, Rønning, Handegård, & Sourander, 2005; Keller et al., 2001; Rubin et al., 2008), whereas teacher ratings of children's behavior on the same or similar measures tend to suggest no differences in the behavior of children in NRFC and KC (Hegar & Rosenthal, 2009; Shore, Sim, Le Prohn, & Keller, 2002). Although, children may simply behave differently at home versus at school, this may also suggest that there are no differences in actual behavior, but rather that kinship caregivers have a higher tolerance for behavior problems than non-relative foster parents. In support of this notion, in a study of only kinship foster placements, Testa and Slack (2002) did not find a significantly higher risk of disruption for children with a behavior problem; this finding conflicts with research on foster placements generally, which tend to indicate an association between behavior problems and placement moves. A higher tolerance for behavior problems among kinship foster parents may prevent, or at least delay, a disruption to the placement. In this case, it would be expected that the highest risk children would benefit most from kinship care.
Alternatively, kinship caregivers may be less likely to elicit problematic behaviors from children in the first place. That is, children behave differently in different settings, and they may be less inclined to act out when they are familiar with or bonded to their caregiver. Attachment theory is one lens through which the possible advantages of kinship care may be viewed. Although lineage is not requisite for such a bond (Dozier, Stoval, Albus, & Bates, 2001), the length of the caregiving relationship is positively associated with foster parent commitment (Bernard & Dozier, 2011). Thus, as kinship caregivers may have had a previous caretaking role in the child's life, there may already be, at the time of placement, a child-caregiver bond to ease adjustment into the new setting (Shlonsky & Berrick, 2001). An easier transition may prevent or delay disruption in the near term, whereas placements that endure past that transition period, thereby allowing for non-relative caregivers to forge a bond with the child, would be expected to have approximately equal risks of disruption. Again, bonds need not be biological to be influential, but there is some evidence to suggest that relatedness is a factor in how much families invest in children (Anderson, 2005). There may be biological reasons to believe that kin receive a higher level of investment than non-relative children (Lawler, 2008), and investment may include continuing to care for a child in the face of economic hardship, health or other ailments, or the child's maladaptive behaviors. However, it is important to understand that the majority of placement changes are intentional and not initiated by the caregiver or child (James, 2004; Wald, 2009), and there is limited understanding of why kinship care would have lower risk of other types of moves.
Conflicting Policy Priorities
Not all moves can be considered a negative outcome. First, whereas placement moves are generally framed in a negative way, moves may also occur to meet important policy objectives or case planning needs. For instance, a child may move in order to be placed in the same home as siblings, to be closer to one's school or birth family, or to be placed in an adoptive home. These moves are intentional, and not a result of any particular problem with the existing caregiver or placement, and thus should not be used to infer a deficit with the initial placement type. Rather, these types of moves may be necessary for long term stability needs, even if they create instability in the immediate term. For these sorts of moves not to bias a comparative estimate, it must be the case that they are randomly distributed across placement type. However, this is not likely to be the case, as relatives are more likely to live in the same neighborhood from which the child originated (i.e., closer to family of origin and school) (Testa, 1997).
Potentially more problematic for the interpretation of stability differences, however, are intentional moves that may occur as a result of policy preferences for kinship care. Consider a case wherein a child enters OHC and is placed in non-relative foster care because a kinship caregiver had not been identified or needed to be screened further. When a kinship caregiver is identified and approved, that child will be moved into the kinship home. This type of move is not the result per se of any deficit in the NRFC placement, but rather, the deficit is that the placement is non-relative. These types of moves should not be used to indicate a lower level of stability in NRFC, not only because they are intentional, but also because there is no similar situation which occurs in KC. If a child is in a kinship home, the only reason a child is moved to NRFC is because the KC placement was not tenable for some reason. Thus, these cross-type moves must be considered differently.
Method
Data
This study uses Wisconsin administrative data for years 2005 through 2012.3 There were 52,752 foster care episodes (43,184 children) in which the first placement was on or after January 1, 2005. Episodes beginning within six months of the end of observation (December 31, 2012) were excluded, unless the placement ended within six months (3,000 placements excluded). A small number of cases (<1% of observations) were also excluded due to imprecision in the matching algorithm used to identify children in placements and match them with CPS records and demographic records4, along with exclusions of duplicate or erroneous placements. This left nearly 50,000 episodes (over 41,000 children and 106,763 placements) remaining. No other exclusions were made for some of the descriptive analyses. However, some descriptive statistics and the regression models focus on placements in NRFC (35,954 placements) or KC (20,712 placements), and sometimes only on placement episodes that begin in NRFC or KC. The relevant sample sizes and inclusion criteria for all models are reported in the applicable table.
Measures
Placement type
The data include 7 basic types of placements (notably, some cannot be categorized): non-relative foster care (NRFC), formal kinship care (KC), shelter placement, residential or group home, detention facility, TANF-funded informal (voluntary) kinship care, or a pre-adoptive home. Analyses are primarily concerned with NRFC and KC placements, each of which is indicated by a dichotomous indicator of current placement.
Reasons for moving
Children may move from one foster care placement to another as a result of any number of factors. In this study, I separate foster parent-requested moves from all other move reasons. Other reasons include moves occurring due to request of the agency, request of parent or other pre-placement caregiver, (rarely) request of child, or foster parent quality issues, as well as moves for which the reason is not clearly documented. Adoption, reunification, and guardianship are not considered moves, and rather are censored as being an exit from foster care.
Types of moves
I considered three types of placement moves, categorized based on where the child was moved to, relative to the placement from which they came: (1) moves to more-preferred placements, (2) moves to equally-preferred placements, and (3) moves to less-preferred placements. These categorizations reflect how a child is moving within a hierarchy of placements, with the hierarchy reflecting the perceived desirability of each type of placement, as explicitly expressed in law or clearly implied by policy goals. Notably, these policy preferences may or may not reflect the best interests of any given child; however, policy priorities are ostensibly made with regard to perceived child interests.
Policy states that children should be in the most family-like, least restrictive setting possible. Moreover, policy goals suggest that placements should be made with an expectation that they will create stability and permanency for a child. Lastly, kinship placements are given explicit preference in several federal laws, most recently in the Fostering Connections to Success and Increasing Adoptions Act of 2008. Accordingly, at the top of this hierarchy of placements would be pre-adoptive homes and informal kinship placements, because they are less restrictive, more family-like, and/or more permanent (and, in the case of informal kinship placements, do not require the ongoing oversight of the child welfare system). Thus, moves from NRFC or KC to informal kinship care or a pre-adoptive home are considered moves to more-preferred placements. Given explicit policy preferences for kinship care, a move from NRFC to KC is categorized as a move to a more-preferred placement. Likewise then, a move from KC to NRFC is coded as a move to a less-preferred placement.
At the bottom of the hierarchy of placements would be residential or group homes and detention centers (because they are more restrictive and less family-like) as well as shelter placements (because they are short-term/non-permanent). Accordingly, all moves from NRFC or KC to those placement types are considered moves to less-preferred placements. This hierarchy is generally consistent with acknowledged policy priorities for placement of children. Lastly, within-type moves (e.g., NRFC to NRFC) are considered neutral, or moves to equally-preferred placements.
Covariates
Available child demographic data include age, sex and race. Age is measured in developmental segments of 0 to 2, 3 to 5, 6 to 10, and 11 to 14, and 15 years or older. Sex is measured as is typical (male=1, female =0). For race, mutually exclusive categories are constructed as follows: white only, black only, Hispanic – any race, American Indian only, and other race/multiracial. I also include an indicator of whether the placement is intended to be long term. This is a dichotomous item entered by the caseworker, with 1 indicating the placement is intended to be long term, and 0 otherwise. In addition, to account for the extent and type of maltreatment experienced prior to entering an OHC, official records of maltreatment investigations pertaining to these children that occurred in the six months prior to their initial entry into OHC were retrieved5. From these records, a series of non-exclusive dummy variables were created to indicate whether children, in the six months prior to entering OHC, were alleged to have experienced neglect (supervisory or physical), physical abuse, or sexual abuse.6 County level demographics include county population, child poverty rate, and the percent of the county that is nonwhite.7 These measures are derived from 2010 Census data for Wisconsin. County population is split into three categories: rural (< 20,000) midsize (reference group; 20,000 to 100,000), and large/urban (>100,000).
Analytic Approach
For all analyses, survival models are used to estimate the effect of placement type (NRFC or KC) on the hazard of experiencing a placement change. Survival analysis is useful in this context because it takes into account that children are observed for different lengths of time, and they both enter and exit out-of-home care at different times. The flexible piecewise exponential model is used, as it does not assume that the hazard changes at a constant rate. Rather, it allows the baseline hazard to vary at predetermined time points. I use a single time cut-point of 2 months, based on descriptive data (not shown) that suggested a large reduction in the probability of a placement change after 2 months in the placement. Additionally, time-varying factors can be incorporated easily into this framework. These models use the OHC episode as the unit of observation, which is the same as the child if they only entered OHC one time. If the child experienced an exit and reentry into foster care, the second and subsequent OHC episodes are considered new observations.
As children may experience multiple placement changes, survival models can be modeled as single or multiple failure models. In the former, only the hazard of a first placement change is considered. This limits the sample more specifically to OHC episodes that began in NRFC or KC. However, there are several problems with this approach. The first is that episodes may begin in one placement type and move to another type. Limiting the analysis to the first placement in an episode excludes many NRFC and KC placements from consideration. Moreover, given that many OHC episodes involve 2 or more placement changes, it cannot be assumed that a disparity in the hazard rate for a first placement is equal across all subsequent placement changes. Indeed, prior research suggests that the disparity between NRFC and KC in the time to a first placement change is much larger than in subsequent placement changes (Koh, 2010). To account for the occurrence of multiple failures (multiple placement changes per episode), and the likelihood that the baseline hazard is affected by past failures, the estimates are stratified by risk set (placement number). That is, all episodes are initially observed in the first stratum, but once a placement change occurs, the second placement in an episode is estimated within the second stratum, and so forth. Given the skewed right tail of number of placement changes, all placement changes that occur subsequent to the sixth placement change are estimated within stratum 6.
The analyses intend to test the three primary explanations for higher stability in KC: child selection, foster parent commitment, and policy preferences and priorities. These explanations, the hypotheses they imply, and my approaches to testing are detailed in Table 2.
Table 2.
Possible explanations for less instability in kinship foster care, implied hypotheses, and approach to testing
| Possible explanation | Implied Hypothesis | Test |
|---|---|---|
| Child selection factors: Children who enter KC have fewer behavioral and cognitive deficits. Consequently, children in KC may be easier to foster, and thus would be less likely to have a placement change irrespective of where they were placed. | Given similar children, children in NRFC and KC would have similar risks of a placement move. That is, the stability gap would be smaller for the higher risk children than for all children. | Compare the difference between NRFC and KC in risk of a placement move for all children vs. higher-risk subgroups of children |
| Foster parent commitment: Kinship foster parents may be more committed to child, feel a greater sense of obligation to maintain placement irrespective of any personal or economic hardships. In addition, if they had a pre-existing bond with the child, kinship foster parents may be less likely to evoke, or more likely to tolerate, problematic behaviors or temperaments. | Children in KC would be less likely to be moved at the request of a foster parent than children in NRFC. The stability gap would be predominantly driven by fewer foster parent-requested moves in KC. | Compare the difference between NRFC and KC in risk of a placement move for foster-parent requested moves vs. moves that occur for any other reason. |
| Policy preferences and priorities: Placement of removed children with kin is an explicit priority for the child welfare system. Consequently, children in NRFC may be intentionally moved to facilitate a KC placement, even if there were no inherent flaws of the NRFC placement. Notably, there is no comparable scenario in KC. | The stability gap is largely driven by the movement of children in NRFC into placements given more preference in policy, including KC. | Compare NRFC to KC on the risks of a move to (1) a more-preferred placement, (2) an equally-preferred placement, and (3) a less-preferred placement. |
First, to test whether child selection factors are driving estimated differences in stability, I estimate the hazard of a placement change for the full sample of children in NRFC and KC in the first 2 months and after 2 months. I then compare these estimates to several subgroups of children in NRFC and KC that, based on past research, should be at higher risk of experiencing a move. The purpose of these comparisons is to test two countervailing hypotheses. First, it may be the case that children with a readily available relative placement are advantaged overall compared with NRFC children, and that, given similarly disadvantaged children, NRFC and KC placements would be equivalent. (Alternatively, if KC caregivers were more invested than NRFC caregivers in a child, children at greater risk of a placement disruption may benefit more from KC.) To do this, I identify subgroups of children who would theoretically be at higher or lower risk for a placement disruption and may have been more difficult to place initially. The higher risk subgroups include older children, children with a prior placement in a restrictive placement (group, residential or detention center), and children who were initially placed in a shelter (emergency) home. Older age and a history of residential care are both identified as risk factors for placement disruption (Oosterman, et al., 2007). Lastly, I argue that children in NRFC or KC who began in shelter placements are likely to be more similar on unobserved factors to one another than children who began in NRFC or KC, for two reasons. First, it means that a relative was not immediately identified or approved for placement, which undermines the possibility of an innate advantage. Second, these are children for whom a family-based placement was more difficult to find – meaning they are more likely to be children that are older, and may be more likely to have behavioral or emotional disturbances.
Second, I assess the foster parent commitment explanation – that is, whether stability differences are attributable to lower rates of foster parent-requested moves in KC. I estimate survival models of the same variety as above, but the dependent variables are the identified reason for the move (either foster parent-requested or other reason).8
Third, I assess whether stability differences are explained by policy-based preferences for placements. The initial step in this analysis is to assess the types of moves children experience. I first describe, by initial placement type, the types of placements children experience thereafter. Then I categorize all placement changes as moves to more preferred, equally preferred or less preferred placements, based on established policy and practice preferences for various placement arrangements. I then estimate the hazard, again using survival analysis, of experiencing each type of move.
Results
Sample Description
A description of the NRFC and KC sample is found in Table 1. NRFC and KC placements are statistically significantly different on nearly all comparisons, but given the large sample size, significance is found for some differences that are too small to be meaningful. Differences of note include the following: NRFC placements are more likely to involve children ages 11 to 18 and children with histories of alleged physical or sexual abuse, and less likely to be intended as long term placements, to involve children with histories of alleged neglect, or children under 5. Lastly, many children who ultimately spend time in family-based care (either NRFC or KC) do not begin there. In this sample, 25% of NRFC placements had prior placements in other types of care (most commonly a shelter placement) and 33% of KC placements (most commonly NRFC).
Table 1.
Descriptive Statistics
| NRFC | KC | ||
|---|---|---|---|
| N Placements | 35,954 | 20,712 | |
| M or % | M or % | Sig. | |
| White | 43.95 | 47.46 | *** |
| Black | 30.78 | 28.74 | *** |
| American Indian | 4.88 | 5.83 | *** |
| Hispanic | 9.60 | 9.52 | |
| Other/multiracial | 10.81 | 8.45 | *** |
| Male | 50.44 | 49.72 | |
| Long term arrangement expected | 68.80 | 76.11 | *** |
| Age 0 to 2 | 28.03 | 30.95 | *** |
| Age 3 to 5 | 17.12 | 21.12 | *** |
| Age 6 to 10 | 20.55 | 22.47 | *** |
| Age 11 to 14 | 16.48 | 13.91 | *** |
| Age 15+ | 17.83 | 11.55 | *** |
| History of neglect | 52.86 | 59.40 | *** |
| History of sexual abuse | 7.26 | 5.83 | *** |
| History of physical abuse | 19.58 | 19.29 | |
| Rural county | 5.55 | 4.94 | ** |
| Large/urban county | 68.68 | 68.80 | |
| Percent of county non-white | 17.09 | 18.20 | *** |
| County child poverty rate | 20.77 | 21.14 | *** |
p<.01
p<.001
Explanation 1: Child Selection Factors
Table 3 presents multiple failure models for the full sample as well as for select high-risk subgroups. For the oldest children (ages 15 to 18), the NRFC-KC gap is 3.4 percentage points (PP) in the first 2 months, which is equivalent to a 23% lower risk in KC. This is about half the size of the gap for the full sample. This pattern persists after 2 months, where the gap for the full sample is smaller but still notable, and the gap for the oldest children is essentially zero. I find no statistically significant difference in the hazard of placement change in KC as compared with NRFC for children with a prior restrictive placement. Among children whose OHC episode began in a shelter placement, subsequent KC placements have an 18% higher risk of placement change in the first 2 months than subsequent NRFC placements. This pattern does not hold after 2 months, however. What is noteworthy here is that NRFC placements, but not KC placements, tend to have a lower risk of disruption for children in the high risk subgroups than in the overall sample.
Table 3.
Estimated Hazards for Full Sample and Select Subgroups
| All | Ages 15 to 18 | Prior restrictive placement | Began in Shelter Placement | |||||
|---|---|---|---|---|---|---|---|---|
| N moves | 30,892 | 4,308 | 2,605 | 3067 | ||||
| h(t) | SE | h(t) | SE | h(t) | SE | h(t) | SE | |
| Months 0-2 | ||||||||
| NRFC | .177 | .002 | .151 | .004 | .146 | .006 | .065 | .002 |
| FKC | .103 | .002 | .117 | .006 | .143 | .011 | .077 | .005 |
| Absolute difference | −074*** | −.034*** | −.003 | .012* | ||||
| Relative difference | −42% | −23% | −2% | 18% | ||||
| After 2 months | ||||||||
| NRFC | .081 | .001 | .072 | .002 | .104 | .003 | .082 | .002 |
| FKC | .060 | .001 | .070 | .003 | .094 | .006 | .073 | .004 |
| Absolute difference | −.021*** | −.002 | −.010 | −.009* | ||||
| Relative difference | −26% | −3% | −10% | −11% | ||||
Notes: Multiple failure models. Estimates are the predicted hazard rates based on piecewise exponential survival models, controlling for child demographics, maltreatment history, and county characteristics.
p<.05
p<.001
Explanation 2: Foster parent commitment
Table 4 shows the results of multiple failure models predicting the hazard of a foster parent-requested move, as well as a move for any other reason. Move reasons were not further categorized due to some ambiguities with how data were entered. KC-placed children have a 1.2 PP (33%) lower probability of a foster parent-requested move in the first 2 months than NRFC placed children. However, after 2 months, the difference in hazards (though still statistically significant) falls to 0.3 PP (20% lower risk in KC). For moves that were not foster parent-requested, we see a larger disparity between NRFC and KC. KC-placed children have a 6 PP (43%) lower probability of moving for any reason other than foster parent request in the first 2 months, and a 1.8 PP (28%) lower probability thereafter.
Table 4.
Predicted Hazards by Move Reason
| Foster parent request | Other reason | |||
|---|---|---|---|---|
| N moves | 6,041 | 24,290 | ||
| h(t) | SE | h(t) | SE | |
| Months 0-2 | ||||
| NRFC | .036 | .001 | .138 | .002 |
| FKC | .024 | .001 | .078 | .002 |
| Absolute difference | −.012*** | −.060*** | ||
| Relative difference | −33% | −43% | ||
| After 2 months | ||||
| NRFC | .015 | .000 | .065 | .001 |
| FKC | .012 | .000 | .047 | .001 |
| Difference | −.003*** | −.018*** | ||
| Relative difference | −20% | −28% | ||
Notes: Multiple failure models. Estimates are the predicted hazard rates based on piecewise exponential survival models, controlling for child demographics, maltreatment history, and county characteristics.
p<.001
Explanation 3: Policy-based placement preferences
Figure 1 depicts the later placement experiences of foster care episodes that began in NRFC or KC. Overall, children, if they move at all, are most likely to experience placements of the same type in which their foster care episode began. Half of children whose foster care episode began in NRFC had a subsequent NRFC placement, whereas just over 40% of episodes beginning in KC had subsequent KC placements. Nearly 1 in 5 episodes that began in NRFC later moved to KC, and the opposite is true as well. Episodes beginning in NRFC are slightly more likely to have subsequent placements in residential/group homes, detention facilities, shelters, or pre-adoptive homes, and less likely to have subsequent informal kinship care placements.
Figure 1.
Subsequent placement experiences of children initially placed in non-relative and kinship care
The estimated hazard rates for moves to more, equally and less preferred placements are shown in Table 5. Results suggest that KC-placed children have a lower hazard of moving to a more preferred placement and an equally preferred placement. The difference in moves to equally preferred or more preferred placements diminishes in magnitude after the first 2 months, but remains statistically significant. Compared with KC-placed children, NRFC-placed children have a 2.8 PP (40%) lower probability experiencing a move to a less preferred placement in the first 2 months, and a 1.5 PP (35%) lower probability thereafter.
Table 5.
Estimated Hazards of Moves to Less, Equally, and More Preferred Placements
| To less-preferred placement | To equally-preferred placement | To more-preferred placement | To less or equally preferred placement | |||||
|---|---|---|---|---|---|---|---|---|
| N moves | 5,825 | 16,227 | 4602 | 22,053 | ||||
| h(t) | SE | h(t) | SE | h(t) | SE | h(t) | SE | |
| Months 0-2 | ||||||||
| NRFC | .019 | .001 | .100 | .001 | .048 | .001 | .118 | .002 |
| FKC | .047 | .001 | .052 | .001 | .002 | .000 | .100 | .002 |
| Absolute difference | +.028*** | −.048*** | −.046*** | −.018*** | ||||
| Relative difference | 147% | −48% | −96% | −15% | ||||
| After 2 months | ||||||||
| NRFC | .008 | .000 | .042 | .001 | .015 | .000 | .051 | .001 |
| FKC | .023 | .001 | .030 | .001 | .002 | .000 | .050 | .001 |
| Absolute difference | +.015*** | −.012*** | −.013*** | −.001 | ||||
| Relative difference | 188% | −29% | −87% | −2% | ||||
Notes: Multiple failure models. Estimates are the predicted hazard rates based on piecewise exponential survival models, controlling for child demographics, maltreatment history, and county characteristics.
p<.001
Given that moves to more preferred placements are (ostensibly) not a cause for concern, I also estimate the risk of a move to less or equally preferred placement. Results suggest that, in the first 2 months, KC placements have a 1.8 PP (15%) lower probability of a move to an equally or less preferred placement compared with NRFC, but the difference thereafter is nonsignificant.
Discussion
The federal government suggests that kinship care placements are a promising strategy for decreasing placement disruption rates, based on prior research identifying lower rates of placement change among children in kinship care. However, evidence is scant on why, and under what circumstances, this is true. This study explores several reasons why kinship care might result in higher placement stability and assesses the extent to which higher stability, as defined by placement moves, is a useful metric for evaluating foster care outcomes. Overall, results suggest that differences in stability may be largely driven by factors that cannot be attributed to the placement itself, including child characteristics and underlying policy issues.
However, this study has a few caveats that warrant consideration. This study is limited to a single state, and thus results may not be generalizable to other states. For example, the Adoption and Foster Care Analysis and Reporting System (AFCARS) data report that approximately 28% of formal foster care placements are in kinship care, whereas 21% of formal placements in WI are identified as kinship placements. Yet, in the National Survey of Child and Adolescent Well-Being II, the vast majority of formal kinship care placements are estimated to be unlicensed, but the majority of kinship care placements in Wisconsin are identified as licensed.
Second, this study made every effort to maintain the full sample size, but some exclusions had to be made due to inconsistencies, errors, or missing data. Children excluded for these reasons may differ from the overall sample. However, these exclusions account for a very small portion of observations. In addition, this study was not able to control for some important factors –including children's mental and physical health and aspects of children's pre-placement environments (e.g., childhood poverty and family structure). I suggest, however, that omitted variable bias stemming from characteristics of children and their pre-placement environments would likely bias estimates in favor of kinship care (i.e., suggest higher stability in kinship care). That is, prior research has demonstrated that children who enter kinship care have better cognitive scores, fewer behavior problems, lower rates of disability, and fewer biological family risk factors than children entering non-relative foster care (e.g., Font, 2014). Thus, to the extent that those factors contribute to children's probability of success in a given placement, models that fail to account for them would attribute some amount of higher stability to kinship care that is actually attributable to pre-existing differences in children and their circumstances. Relatedly, the type of kinship caregiver is not known in these data, and the benefits of kinship care may differ by degree of relatedness. However, Testa and Slack (2002) do not find that degree of relatedness is associated with the risk of a placement change among kinship placements.
Lastly, aside from placement moves documented as having been requested by the foster parent, the reasons underlying placement moves are fairly ambiguous. James (2004) suggests that 70% of placement changes are due to system or policy-related factors, with 15% of those moves being intentionally done to place a child in a kinship placement and 24% to place a child in another long term care arrangement. Although I cannot precisely assess this in my data, it is consistent with my findings that foster parent-requested moves are not particularly common, and moves from non-relative foster care are often moving a child to a more preferred type of placement.
Nevertheless, this study has several findings that warrant further consideration. First, the purported benefits of kinship care are premised on the attributes of the caregiver (e.g., their sense of familial obligation, cultural similarity to child, etc.) and the child-foster parent relationship. Thus, we might expect the highest risk children to benefit most from kinship care, in that those are the placements in which the foster parents’ patience may be most tried, and their commitment to the child may be most relevant. Results from this study, however, suggest that among high risk subgroups such as teenagers, children with prior restrictive placements, and children initially placed in emergency care, the risk of a placement change is nearly equivalent between NRFC and KC. This is largely driven by better performance among NRFC placements with high risk children. This may be indicative of more training and experience on the part of non-relative foster parents in parenting at-risk youth.
Moreover, if it is true that lower risk of placement change in KC is explained by higher commitment or attachment to the child among kinship foster parents, for example, then a differences in the risk of moving should be primarily observed in foster parent-requested moves. However, only about 20% of moves are documented as occurring due to foster parent request. This is consistent with a finding from James (2004) that the majority of placement changes are attributable to system or policy mandates (e.g., moving children to be placed with siblings). Though, it is certainly possible (arguably probable) that some moves, although not initiated by the foster parent, occurred due to conflict between the child and foster parent. Nevertheless, the results of this study suggest that foster parent-requested moves are indeed lower in KC than in NRFC—but this does not account for a substantial portion of the overall gap. Whereas KC-placed children have a 33% lower risk of a foster parent-requested move in the first 2 months, and a 20% lower risk thereafter, this is a smaller relative difference than was observed in overall moves. Moreover, these moves account for a small portion of all moves. The absolute difference in the probability of a foster parent-requested move is about 1.2 PP in the first 2 months, and 1/3 PP thereafter.
Observed differences in stability are also very time limited. All differences, even in the full sample, decrease substantially or disappear entirely after the first 2 months. This finding is consistent with the idea that non-relative foster parents and foster children need time to forge a bond, and over time a foster parent's commitment to a child increases. In addition, because the risk of a placement move is highest during those first 2 months, this suggests that all placements and especially NRFC placements, may benefit from more intensive support from caseworkers early on. Alternatively, it may be the case that placements that disrupt, given that the majority disrupt quickly, were not a good match to start. This is bolstered by the observation that, whereas children originally placed in shelter placements may be more difficult to place, those who subsequently enter NRFC or KC have a lower risk of disruption than those who began in family-based care. That is, the shelter placement may have bought caseworkers some time to identify the most suitable longer term arrangement, with foster parents who have the necessary skills and training to meet the needs of the child. Caseworkers and agencies have a limited amount of time to identify a suitable placement once children are removed from their homes, often only a few hours. Thus it is not particularly surprising that some mismatched placements will occur; however, additional research may be able to identify risk factors for placement disruption based on combined child and caregiver characteristics and use such an algorithm to identify placement matches. States are implementing a variety of strategies to improve placement stability, including increased support services and efforts to match placements better, but the effectiveness of these approaches lack rigorous evaluation (Blakey et al., 2012).
Are Placement Moves the Right Measure?
Placement stability is, and will continue to be, an important aspect of foster care systems’ performance evaluation. The number of placement moves a child experiences, or the risk of experiencing a move, in foster care is an easy approach to measuring stability. However, the results of this study call to question the extent to which the measure should be used to make inferences about the quality of a placement or to compare the quality of placement types. In part, this study sought to identify the extent to which various moves should be considered a negative outcome. When considering moves in terms of where children go next (moves to more, equally, or less preferred placements) the stability gap looks much different than it does when simply comparing moves generally. In the traditional model, it was estimated that children in KC had a 7.5 PP lower probability of failure in the first 2 months. However, if considering only moves to less preferred placements, KC placements are at higher risk, and when looking at moves to equally preferred or less preferred placements, children in NRFC have a 1.8 PP higher risk in the first 2 months and equivalent risk thereafter, far less concerning than the generic estimates.
Furthermore, there is a significantly higher risk of entering a restrictive placement or detention center following NRFC as compared with KC. This suggests that at least some children who initially entered NRFC were inappropriate for family foster care in the first place, a claim that has been noted elsewhere in this field of study (Barber et al, 2001). Similarly, qualitative data suggest that, in the event of caregiver-requested moves, many placements may have been maintained if the child welfare system had provided appropriate services, but the majority were considered unsalvageable due to safety concerns, verbal abuse, or criminal activity (Gilbertson & Barber, 2003). When safety concerns are present, it is not clear that either NRFC or KC is appropriate or sustainable, particularly when there are other children present in the home. That is, in these instances, it is difficult to imagine how a kinship caregiver would have been in a position to prevent such disruptions.
Implications for Expanding Kinship Care
Related to the use of stability to make inferences about placement quality is the extent to which stability differences logically suggest an expansion of kinship care. This study finds a reasonably high risk of moving from KC to NRFC, which highlights an important problem. Of all children whose OHC episode began in KC, nearly 1 in 5 will end up in NRFC (more than 1/3 of all KC movers). This suggests that a portion of kin placements may not have been appropriate in the first place.
These findings in some ways highlight a conundrum for child welfare policy and practice. Specifically, there are a set of priorities for placements – safety, stability, well-being, placement with relatives, placement with siblings, keeping children in the same school district, and so forth. These priorities may find themselves in conflict, and thus all priorities cannot be considered of equal importance. However, it is unclear where kinship care falls in the hierarchy of placement priorities. For instance, when a relative is identified after a child has been in care, and prioritizing kinship care requires moving a child from an existing placement, is the disruption created justified by benefits of kinship care? This study cannot answer this question, but must pose it as a critical area of future research.
Conclusion
This study has several implications for kinship care research, policy and practice. First, it suggests that research has yet to explain why kinship care is more stable. If kinship care is to be preferred to non-relative foster care, in part based on assumptions about higher rates of stability, it is important that research understand the mechanisms through which kinship care produces higher rates of stability. Second, findings suggest that at least some portion of the stability gap that is being considered a benefit of kinship care, and used to promote kinship preferences, may be better explained by the policy preferences themselves. That is, children in non-relative foster care are more likely to experience positive placement changes (i.e., moves to a more desirable placement, as defined through policy priorities), whereas placement changes in kinship care tend to be to less desirable placements. Future research must be explicit about the extent to which placement changes are uniformly identified as negative outcomes. Lastly, that the highest risk of placement change is in the first 2 months suggests a need to improve child-placement matching to prevent placements that are likely to disrupt from occurring in the first place, and to provide increased support to foster parents and children early in placement.
Acknowledgements
The author thanks the Robert Wood Johnson Foundation for providing a dissertation grant for this work. In addition, this work was supported by the Training Program in Population Studies, grant No.5 T32 HD007081 and the Population Research Center, grant No. 5 R24 HD042849, both awarded to the Population Research Center at the University of Texas at Austin by the Eunice Kennedy Shriver National Institute of Child Health and Human Development.
Footnotes
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Despite states failing to meet the high federal standards for stability, it remains the case that the majority of foster children do not experience frequent moves (Wulczyn, Kogan, & Harden, 2003). The majority of children exit care before two years (U.S. Department of Health and Human Services, 2013), and frequent moves only become the typical foster care experience for children who remain in care past that time. Forty percent of children in care for 12 to 24 months, and 66% of those in care for more than 24 months, experienced more than 2 placements (U.S. Department of Health and Human Services, 2011).
This finding did not hold for treatment foster care (Fisher et al, 2011).
About 450 placements note an end date of January 2013 and thus are observed slightly past the 2012 year.
Observations are matched using a combination of available identifiers, including names, dates of birth and social security numbers where available. However, sometimes this results in mismatches due to data entry errors within systems (where names may be misspelled or dates of birth may be a digit off, for example), or in cases where people have similar names (i.e., siblings or children named after their parents).or identical dates of birth (i.e., twins).
Only the prior six months of maltreatment history are considered due to incomplete records of maltreatment prior to mid-2004.
Maltreatment history is believed to affect children's socio-emotional and behavioral well-being, and in turn, behavioral problems are associated with a higher risk of being maltreated (Font & Berger, 2014). The effect of maltreatment on these outcomes is believed to vary both by type and chronicity.
Prior studies (e.g., Grogan Kaylor 2000; Font, 2014) have suggested geographic factors are associated with the probability of entering kinship care vs. non-relative foster care. It is less clear that these variables would be associated with stability. However, it is possible that geographic factors would impact the types of placement to which a child would move (conditional on moving), given that the availability of different types of placements is likely to vary geographically. Ideally, geographic characteristics would be controlled on a smaller scale (i.e., zip code), but this is not possible in this study given data constraints.
In these analyses, the counterfactual includes other types of moves, in addition to not moving and censoring (i.e., exiting foster care entirely). Although these risks are mutually exclusive in that only one reason can be given for the end of the placement, there is some question as to whether they are independent. That is, an agency may request a move based on contact with a child or caregiver who expresses concern about the placement, and thus the agency request may have pre-empted an impending foster parent request. Typical survival models assume independence of these risks, however, thus primary estimates may be biased.
References
- Anderson KG. Relatedness and investment in children in South Africa. Human Nature. 2005;16(1):1–31. doi: 10.1007/s12110-005-1005-4. [DOI] [PubMed] [Google Scholar]
- Barber JG, Delfabbro PH, Cooper L. The predictors of unsuccessful transition to foster care. Journal of Child Psychology and Psychiatry. 2001;42(6):785–790. [PubMed] [Google Scholar]
- Beeman SK, Kim H, Bullerdick SK. Factors affecting placement of children in kinship and nonkinship foster care. Children and Youth Services Review. 2000;22(1):37–54. [Google Scholar]
- Bernard K, Dozier M. This is my baby: Foster parents' feelings of commitment and displays of delight. Infant Mental Health Journal. 2011;32(2):251–262. doi: 10.1002/imhj.20293. doi:10.1002/imhj.20293. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Blakey JM, Leathers SJ, Lawler M, Washington T, Natschke C, Strand T, Walton Q. A review of how states are addressing placement stability. Children and Youth Services Review. 2012;34(2):369–378. doi:10.1016/j.childyouth.2011.11.007. [Google Scholar]
- Chamberlain P, Price JM, Reid JB, Landsverk J, Fisher PA, Stoolmiller M. Who disrupts from placement in foster and kinship care? Child Abuse & Neglect. 2006;30(4):409–424. doi: 10.1016/j.chiabu.2005.11.004. doi:10.1016/j.chiabu.2005.11.004. [DOI] [PubMed] [Google Scholar]
- Dozier M, Stoval KC, Albus KE, Bates B. Attachment for infants in foster care: The role of caregiver state of mind. Child Development. 2001;72(5):1467–1477. doi: 10.1111/1467-8624.00360. doi:10.1111/1467-8624.00360. [DOI] [PubMed] [Google Scholar]
- Flynn R. Kinship foster care. Child & Family Social Work. 2002;7(4):311–321. doi:10.1046/j.1365-2206.2002.00253.x. [Google Scholar]
- Font SA. Kinship and nonrelative foster care: The effect of placement type on child well-being. Child Development. 2014 doi: 10.1111/cdev.12241. (in press). doi:10.1111/cdev.12241. [DOI] [PubMed] [Google Scholar]
- Font SA, Berger LM. Child maltreatment and children's developmental trajectories in early to middle childhood. Child Development. 2014 doi: 10.1111/cdev.12322. in press. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Gilbertson R, Barber JG. Breakdown of foster care placement: Carer perspectives and system factors. Australian Social Work. 2003;56(4):329–340. [Google Scholar]
- Grogan-Kaylor A. Who goes into kinship care? The relationship of child and family characteristics to placement into kinship foster care. Social Work Research. 2000;24(3):132–141. doi:10.1093/swr/24.3.132. [Google Scholar]
- Hegar RL, Rosenthal JA. Kinship care and sibling placement: Child behavior, family relationships, and school outcomes. Children & Youth Services Review. 2009;31:670–679. doi:10.1016/j.childyouth.2009.01.002. [Google Scholar]
- Holtan A, Rønning JA, Handegård BH, Sourander A. A comparison of mental health problems in kinship and nonkinship foster care. European Child & Adolescent Psychiatry. 2005;14(4):200–207. doi: 10.1007/s00787-005-0445-z. doi:10.1007/s00787-005-0445-z. [DOI] [PubMed] [Google Scholar]
- James S. Why do foster care placements disrupt? An investigation of reasons for placement change in foster care. Social Service Review. 2004;78(4):601–627. [Google Scholar]
- James S, Landsverk J, Slymen DJ. Placement movement in out-of-home care: Patterns and predictors. Children and Youth Services Review. 2004;26(2):185–206. [Google Scholar]
- Keller TE, Wetherbee K, Le Prohn NS, Payne V, Sim KE, Lamont ER. Competencies and behavior problems of children in family foster care: Variations by kinship placement status and race. Children & Youth Services Review. 2001;23(12):915–940. doi:10.1016/S0190-7409(01)00175-X. [Google Scholar]
- Kendler KS, Davis CG, Kessler RC. The familial aggregation of common psychiatric and substance use disorders in the National Comorbidity Survey: A family history study. British Journal of Psychiatry. 1997;170:541–548. doi: 10.1192/bjp.170.6.541. doi:10.1192/bjp.170.6.541. [DOI] [PubMed] [Google Scholar]
- Kim J. Type-specific intergenerational transmission of neglectful and physically abusive parenting behaviors among young parents. Children and Youth Services Review. 2009;31(7):761–767. doi:10.1016/j.childyouth.2009.02.002. [Google Scholar]
- Koh E, Testa MF. Propensity score matching of children in kinship and nonkinship foster care: Do permanency outcomes still differ? Social Work Research. 2008;32(2):105–116. doi:10.1093/swr/32.2.105. [Google Scholar]
- Koh E. Permanency outcomes of children in kinship and non-kinship foster care: Testing the external validity of kinship effects. Children and Youth Services Review. 2010;32(3):389–398. [Google Scholar]
- Koh E, Rolock N, Cross TP, Eblen-Manning J. What explains instability in foster care? Comparison of a matched sample of children with stable and unstable placements. Children & Youth Services Review. 2014;37:36–45. [Google Scholar]
- Lawler MJ. Maltreated children's emotional availability with kin and nonkin foster mothers: A sociobiological perspective. Children & Youth Services Review. 2008;30:1131–1143. doi:10.1016/j.childyouth.2008.02.007. [Google Scholar]
- McCloskey LA, Bailey JA. The intergenerational transmission of risk for child sexual abuse. Journal of Interpersonal Violence. 2000;15(10):1019–1035. doi:10.1177/088626000015010001. [Google Scholar]
- Newton RR, Litrownik AJ, Landsverk JA. Children and youth in foster care: Disentangling the relationship between problem behaviors and number of placements. Child Abuse & Neglect. 2000;24(10):1363–1374. doi: 10.1016/s0145-2134(00)00189-7. [DOI] [PubMed] [Google Scholar]
- Oosterman M, Schuengel C, Wim Slot N, Bullens RA, Doreleijers TA. Disruptions in foster care: A review and meta-analysis. Children and Youth Services Review. 2007;29(1):53–76. [Google Scholar]
- Perry G, Daly M, Macfarlan S. Maternal foster families provide more stable placements than paternal families. Children & Youth Services Review. 2014;46:155–159. [Google Scholar]
- Rubin DM, Downes KJ, O'Reilly ALR, Mekonnen R, Luan X, Localio R. Impact of kinship care on behavioral well-being for children in out-of-home care. Archives of Pediatrics and Adolescent Medicine. 2008;162(6):550–556. doi: 10.1001/archpedi.162.6.550. doi:10.1001/archpedi.162.6.586. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Shlonsky AR, Berrick JD. Assessing and promoting quality in kin and nonkin foster care. Social Service Review. 2001;75(1):60–83. doi:10.1086/591882. [Google Scholar]
- Shore N, Sim KE, Le Prohn NS, Keller TE. Foster parent and teacher assessments of youth in kinship and non-kinship foster care placements: Are behaviors perceived differently across settings? Children and Youth Services Review. 2002;24(1–2):109–134. doi:10.1016/s0190-7409(01)00170-0. [Google Scholar]
- Strijker J, Knorth EJ, Knot-Dickscheit J. Placement history of foster children: A study of placement history and outcomes in long-term family foster care. Child Welfare. 2008;87(5) [PubMed] [Google Scholar]
- Testa M. Kinship foster care in Illinois. In: Barth RP, Berrick JD, Gilbert N, editors. Child Welfare Research Review. Vol. 2. Columbia University Press; NY: 1997. pp. 101–129. [Google Scholar]
- U.S. Department of Health and Human Services . The AFCARS report: Final estimates for FY 1998 through FY 2002. Author, Administration for Children and Families, Administration on Children, Youth and Families, Children's Bureau; Washington, DC: 2006. [Google Scholar]
- U.S. Department of Health and Human Services . Federal Child and Family Services Reviews Aggregate Report: Round 2, Fiscal Years 2007-2011. U.S. Department of Health and Human Services, Administration for Children and Families, Administration on Children, Youth, and Families, Children's Bureau; Washington, DC: 2011. [Google Scholar]
- U.S. Department of Health and Human Services . The AFCARS report: Preliminary FY 2012 estimates as of November 2013. Author, Administration for Children and Families, Administration on Children, Youth and Families, Children's Bureau; Washington, DC: 2013. [Google Scholar]
- Unrau YA, Seita JR, Putney KS. Former foster youth remember multiple placement moves: A journey of loss and hope. Children and Youth Services Review. 2008;30(11):1256–1266. [Google Scholar]
- Usher CL, Randolph KA, Gogan HC. Placement patterns in foster care. Social Service Review. 1999;73(1):22–29. [Google Scholar]
- Ward H. Patterns of instability: Moves within the care system, their reasons, contexts and consequences. Children and Youth Services Review. 2009;31(10):1113–1118. [Google Scholar]
- Webster D, Barth RP, Needell B. Placement stability for children in out-of-home care: A longitudinal analysis. Child Welfare. 2000;79(5) [PubMed] [Google Scholar]
- Winokur M, Crawford CA, Longobardi RC, Valentine DP. Matched comparison of children in kinship care and foster care on child welfare outcomes. Families in Society. 2008;89(3):338–347. doi:10.1606/1044-3894.3759. [Google Scholar]
- Winokur M, Holtan A, Valentine D. Cochrane Collaboration, editor. Kinship care for the safety, permanency and well-being of children removed from their home for maltreatment (a review) 2009;1 doi: 10.1002/14651858.CD006546.pub2. [DOI] [PubMed] [Google Scholar]
- Wulczyn F, Kogan J, Harden BJ. Placement stability and movement trajectories. Social Service Review. 2003;77(2):212–236. [Google Scholar]