Revisiting Conventional Wisdom: What Do We Know from 30 Years of Research on Sibling Placement in Foster Care?

Abstract

Background:

Efforts to place sibling groups together in foster care have long been considered best practice and is required under federal law. Practice and policy guidance are based in part on the belief that sibling placement is in the best interests of children. In this article, we first review literature reviews on this topic to assess the extent to which prior efforts to characterize this body of research are thorough, objective, and based on research specific to the foster care population. We then assess the quality and volume of empirical evidence on the effects of sibling placement for the stability, permanency, and wellbeing of children in foster care to ascertain whether existing reviews accurately reflect the empirical evidence or extend beyond it without adequate empirical support.

Methods:

We conducted a scoping search of reviews of published research from the year 1990 to 2019 using Google Scholar, PsycInfo, EBSCO, and PubMed (Medline). From the search results, we extracted all review articles, quantitative studies, and qualitative studies on sibling placement in foster care. We identified 16 literature reviews on sibling placement in foster care. In addition, we identified 27 longitudinal quantitative studies measuring the association between sibling placement and child placement stability, permanency or wellbeing in foster care, and nine qualitative studies on the role of sibling placement or relationships for child wellbeing in foster care.

Results:

Many of the literature reviews relied on research evidence from non-foster care samples and studies with weak methodologies. Moreover, although the research evidence is inconclusive – with studies reporting positive, negative, and null effects –most literature reviews concluded that the practice of sibling placement was supported by research evidence.

Conclusion:

Although there are moral reasons to support sibling placement, the research evidence does not consistently support the practice of placing siblings together. Further research is needed to identify when sibling placement poses a risk to children and when sibling placement is likely to facilitate positive outcomes.

Introduction

The U.S. foster care system provides temporary care for over 400,000 children at any given time (Children’s Bureau, 2018). Estimates suggest that 60–70% of children in foster care have at least one sibling in care with them (Child Welfare Information Gateway, 2013). While in foster care, states are required to make efforts to place siblings in the same home (Fostering Connections to Success & Increasing Adoptions Act, 2008). This requirement is, in part, a response to research on the importance of sibling relationships for child wellbeing. Dozens of individual research studies and literature reviews have been published on the topic of sibling placement for children in foster care, and a vast majority of research reviews assert that keeping siblings together is beneficial for children. Yet, a critical and extensive assessment of the quality of evidence, and the inferences that can be drawn from it, remains lacking. In this article, we first conducted a scoping search of the published literature reviews on siblings in foster care between published between 1/1/1990 and 12/31/2019. We addressed two questions about the existing body of literature reviews: (1) To what extent do conclusions about the effects of sibling placement in foster care rely on data or theory developed from general (non-foster care) samples?; (2) On the whole, how well does the volume of literature reviews on sibling placement represent the quality of research, with regard to generalizability, measurement, and methodology? We then evaluated the evidence from individual quantitative and qualitative research studies that addressed the role of siblings in the stability, permanency, or wellbeing of children in foster care. Focusing on the most rigorous and directly applicable studies, we addressed two additional questions: (1) Is there a general consensus across studies that sibling placement in foster care is positively associated with stability, permanency, or wellbeing?; (2) Is the volume and quality of evidence sufficient to draw clear conclusions about the effects of sibling placement? Lastly, we discussed current policy and practice pertaining to sibling placement and provide implications for future research.

Methods

Search Procedures

On March 27, 2020, searches were conducted in Google Scholar, PsycInfo, EBSCO, and PubMed (Medline) for the phrase “foster care” and any of the following phrases: “sibling placement” or “sibling placements” or “placement with siblings” or “sibling separation” or “separation from siblings”. We restricted the search to English-language works with dates between 1/1/1990 and 12/31/2019. Google Scholar returned 1,310 results; all other databases returned fewer than 100 results. The search results were reviewed separately by each author to identify all articles that met our inclusion criteria for (1) reviews of the literature; (2) longitudinal quantitative studies; (3) qualitative studies. We screened each search result for inclusion in any of the three groups (review, quantitative-longitudinal, qualitative) simultaneously. Common reasons that a search result was excluded from all three groups were: (1) it was not published in a book or journal (e.g., thesis, working paper, dissertation, conference presentation/abstract); (2) it was not a research study or research review (e.g., theory paper, law review, case description); and (3) it focused on a population other than children in foster care (e.g., studies of adopted children, retrospective studies of adults previously in foster care). The specific inclusion criteria for each type of study is described in later sections.

The first author reviewed all search results first and compiled articles that met one of the three included study types. The second author then reviewed all search results. Discrepancies (where a study was identified for inclusion by only one of the authors) were few in number but generally due to lack of clarity in studies’ description of methods or different conceptualizations of what types of outcome measures could be categorized as child wellbeing. In those cases, studies were included if at least one author identified it for inclusion. We also reviewed the reference list of each review article and included in our analysis any relevant articles that were referenced in the reviews, even if they were published prior to 1990. Included articles are detailed in Table 1. A majority of articles in each category were identified in Google Scholar, with few additional studies identified through other databases or the article reference lists.¹

Table 1.

Articles Included in Systematic Review

Authors	Article Type	Outcomes/Focus			Located in
		Well-being	Permanency	Stability	Google scholar	PsycInfo	Pubmed/Medline	EBSCO	Other
Affronti et al, 2015	Qual	X							X
Ainsworth & Mallucio, 2002	Review	X			X
Akin 2011	Quant		X		X	X		X
Albert & King, 2008	Quant		X		X
Angel, 2014	Qual	X			X
Barth et al, 2007	Quant			X	X
Berridge & Cleaver 1987	Quant			X					X
Child Welfare Information Gateway, 2019	Review	X	X	X	X
Fernandez & Lee 2013	Quant	X	X		X
Folman, 1998	Qual	X			X				X
Font et al, 2018	Quant			X	X	X	X	X
George, 1970	Quant			X					X
Groza et al, 2003	Review	X		X	X	X
Gustavsson & MacEachron, 2010	Review	X	X	X	X	X		X
Harrison, 1999	Qual	X							X
Hegar & Rosenthal, 2009	Quant	X			X			X
Hegar & Rosenthal, 2011	Quant	X			X	X		X
Hegar, 2005	Review	X		X	X	X		X
Holtan et al, 2013	Quant			X	X
James et. al, 2008	Qual	X			X
Jones, 2016	Review	X	X	X	X
Koh et al, 2014	Quant			X	X
Konijn et al, 2019	Review			X	X
Kothari et al, 2018	Quant	X			X
Leathers, 2005	Quant		X		X	X		X
Leathers, 2006	Quant			X	X
Leathers et al, 2019	Quant			X	X
Leichtentrit, 2013	Qual	X			X
Linares et al, 2007	Quant	X			X		X
McCormick, 2010	Review	X	X	X	X
McDowall, 2015	Quant			X	X
Meakings et al, 2017	Review	X	X	X	X
O’Neil, 2002	Review	X		X	X
Oosterman et al, 2007	Review			X	X
Parker, 1966	Quant			X					X
Rock, 2015	Review			X	X	X
Sattler et al, 2018	Quant			X	X		X
Seale & Damiani-Taraba, 2017	Review	X	X	X	X
Shlonsky et al, 2005	Review	X	X	X	X	X		X
Staff & Fein, 1992	Quant			X	X	X		X
Sting, 2013	Qual				X
Thorpe & Swart, 1992	Quant	X				X	X
Trasler, 1960	Quant			X					X
van Rooij et al, 2015	Quant			X	X
Vinnerljung et al, 2017	Quant			X
Waid, 2014	Review	X	X	X	X
Waid et al, 2016	Quant			X	X	X	X	X
Waid et al, 2017	Quant	X							X
Washington, 2007	Review	X	X		X	X		X
Webster et al, 2005	Quant		X		X
Wojciak et. al, 2018	Qual	X			X
Wojciak, 2017	Qual	X			X

Note: Qual = Qualitative; Quant = Quantitative

The data extraction process and elements are described separately for each type of article (review, quantitative-longitudinal, qualitative). The study protocol was not pre-registered. We relied solely on the information provided in the published works and did not obtain any unofficial/external information directly from study authors. To ascertain potential explanations for differences in study findings or conclusions, we also identify and discuss differences in methodologies, samples, and measures as applicable. To identify potential weaknesses or limitations, we drew on the work of prior review articles (e.g., Shlonsky et al., 2005), though we discuss additional limitations or concerns as identified from our review.

Research Review

We identified 16 review articles that were focused on or included a section on placement of siblings in foster care; of these, 14 were peer-reviewed by an academic journal (for exception, see: Children’s Bureau, 2019; Meakings et al., 2017). Reviews focused on sibling relationships rather than sibling placement were excluded. Fifteen of the sixteen reviews were found in Google Scholar. The review by Child Welfare Information Gateway (2019) was not identified in any of the search databases; however, we opted to include it because it was issued through a prominent government agency (the U.S. Children’s Bureau) and is a widely used resource by researchers, policymakers, administrators, and practitioners. (Note: Child Welfare Information Gateway issued a similar review in 2013; we included only the 2019 version). Three of the 16 reviews focused on foster care placement instability for children in the general foster care population, but each included a section on sibling placement (Konijn et al., 2019; Oosterman, Schuengela, Slot, Bullens, & Doreleijers, 2007; Rock, Michelson, Thomson, & Day, 2015). We only examined the sections on sibling placement for these reviews.

An initial set of information was extracted from the reviews pertaining to the use of theory and evidence and excerpts of the conclusions drawn by the authors. After reviewing the initial set of information extracted, we refined the data elements to better characterize the type of evidence cited by the reviews. Each review article was reviewed multiple times by the first author to extract key information about the use of evidence.

For each review, we first evaluated whether and how theories or concepts tested in the general population were applied to the foster care population. We were particularly interested in understanding whether the authors discussed the unique contexts of foster care in applying general theory or research evidence. We also sought to understand whether issues that may lead to sibling separation or result in negative outcomes from sibling placement, such as intra-sibling abuse, parentification, or negative competition, were considered.

Second, for each review, we identified all works cited to determine which citations were specific to understanding the effects of co-placement or separation of siblings in foster care on child stability, permanency, or wellbeing. We sought to assess the extent to which review articles’ conclusions were based on research on sibling or familial relationships in non-foster care populations versus empirical evidence specific to the foster care context. From each review, we extracted the references that were cited as evidence on the effects of sibling placement or separation. We then categorized those references based on whether the cited work (1) was an empirical research study; (2) was based on a foster care sample; (3) used sibling placement as an independent/predictor variable; and (4) was quantitative and longitudinal.

Notably, due to limited empirical evidence discussed, two reviews were excluded from our citation extracting process (F. Ainsworth & Maluccio, 2002; Shlonsky, Bellamy, Elkins, & Ashare, 2005). Ainsworth and Maluccio (2002) only cited descriptive statistics of sibling placement at first entry; their discussion of the effects of sibling placement did not reference empirical evidence. Additionally, the review by Shlonsky and colleagues (2005) cited the empirical literature largely to highlight the limitations and inconsistencies of prior work, rather than to assert the effects of sibling placement. Thus, this aspect of our article includes only 14 reviews.

Quantitative Research Studies

We used four inclusion criteria for quantitative research studies. First, included studies drew on a sample/population of children in foster care (not children in the general population or post-adoption). Second, one of the study outcomes needed to be placement stability, permanency, or child wellbeing. We defined child wellbeing broadly, to include mental and behavioral health indicators, school performance, self-reported wellbeing, and adjustment / satisfaction with foster home. Third, we only included studies where placement with siblings was used as one of the predictor (independent) variables. Lastly, studies were included if they collected data longitudinally – cross-sectional studies were not included. Of course, studies on stability and permanency are usually longitudinal as the outcomes constitute events (e.g., placement change, exit from care) which must be observed over time. However, this criterion was important for the review of wellbeing outcomes, as many studies measured sibling placement and child wellbeing concurrently.

After processing the search results for Google Scholar, PsycInfo, Pubmed (Medline), and EBSCO, as well as the reference lists of all review articles, we identified 27 quantitative articles that met our inclusion criteria. As shown in Table 1, most quantitative articles that met inclusion criteria were found in Google Scholar, though many were found in multiple databases. Articles published prior to 1990 were included only if cited in a prior review. There was also one article (Waid et al., 2017) that was not identified by our search terms nor cited in any of the review articles.

Information was extracted from each study by the first author and then reviewed by the second author for accuracy and completeness. Key information extracted from all studies was: years and location for data collection (also noting where multiple studies used the same raw data); sample size and age composition, length of study follow-up, primary method of analysis, statistical controls used, and results. In our tables, we report the results of the study’s ‘final model’ (with all included statistical controls) where available. In some studies, sibling placement was non-significant in initial stages of the analysis and not reported on in the final models; in those instances, we report only data (method, sample, controls) relevant to the analysis that included sibling placement as a predictor. Also, some studies reported only the results of models with interaction terms if the main effects of sibling placement were non-significant. In those instances, we extracted the results of the interaction models. The initial extraction was completed by the first author, and then replicated by the second author to check for errors. Then, the information extracted was edited for consistency of terminology and level of specificity/detail across studies.

Qualitative Research Studies

We included qualitative studies that met the following criteria: (1) sample was comprised of children in foster care or foster families; and (2) goal of the research was (in full or in part) to assess the risk or benefits of sibling placement or inquired about the experiences of children with siblings in foster care. Studies were excluded if they focused solely on post-adoption outcomes of former foster children (Berge et al., 2006; Cossar & Neil, 2013; Saunders & Selwyn, 2012; Selwyn, 2019), parent-child relationships, sibling contact, sibling policy, or sibling groups not in foster care (Church & Moe, 2015; Cooper, 2013; Gardner, 1996, 2004; Grigsby, 1994; Meyers, 2014; Mosek, 2013; P. Parker & McLaven, 2018). A total of nine qualitative studies met these inclusion criteria.

The synthesis of qualitative research is a difficult undertaking due to the wide variety of samples and interview questions. Similar to the quantitative study process, the first author extracted data (where reported) on the site and time frame of data collection, sample size and role (foster parent, child), and reported the main findings/common themes as characterized by the study authors. All extracted information was entered into the table template and reviewed by the second author.

Results

Review of Reviews

Purpose of Reviews

The review articles included in this study were aimed toward a variety of audiences: some were research-focused, whereas others had a practical or clinical emphasis. Although all 16 reviews purported to provide an evidentiary review, nine reviews also asserted a clinical or practical purpose. Although reviews targeted toward a non-academic audience would not be expected to include the same degree of methodological detail, it is critical that readers understand how articles were selected for inclusion in the review and the quality of evidence provided by the reviewed literature. Yet, only eight reviews explained their methods for identifying and including empirical studies. Therefore, we were unable to assess the potential for bias or missing information in the authors’ search criteria. In addition, half of the reviews (n=8) did not clearly address methodological and sampling limitations of extant research on siblings in foster care, which may leave the impression the evidence is conclusive and without significant caveats.

Use of Theory to Inform Research on Siblings in Foster Care

Although a theory section is not pro forma in literature reviews, two reviews focused on attachment theory (McCormick, 2010; Washington, 2007) and a third (Waid, 2014) emphasized social learning and bioecological theory.

Attachment theory, which is commonly cited to support sibling placement, supports only a claim that separating a child from a secure attachment figure can cause distress and psychological harm and that having a secure attachment is important for healthy development. (M. Ainsworth, 1989; Bowlby, 1973) Yet, attachment theory does not prove, or even assert, that all sibling relationships are secure attachments that are beneficial to child development. Moreover, attachment theory does not suggest that separation of any connection regardless of quality is harmful. Attachment theory heavily emphasizes the safety and security that consistent nurturing relationships provide children, which may lead to two conclusions. First, children may experience harm if safe, secure, and mutually beneficial sibling relationships are disturbed. Second, when sibling relationships are not safe, secure, or mutually beneficial, separation or intervention to improve the relationship may be warranted. Although supporting and sustaining positive or neutral sibling relationships through co-placement may benefit children’s development, clinicians have cautioned that abusive, competitive, or parentified sibling relationships may prevent a child from forming secure and healthy attachments with caregivers or other children/peers in a new environment (Whelan, 2003).

McCormick (2010) discussed the role of attachment deficiencies in children’s socioemotional development, citing work by prominent attachment theorists Ainsworth (1989), Bowlby (1973), and Howe and colleagues (2000). Yet, readers lacking a strong background in psychology may be unaware that these studies largely focus on parent-child relationships and do not consider the special context of foster care. McCormick (2010) further asserted “In fact, depending on the conditions, some sibling separation can be more traumatic and difficult than separation from parents” citing a single agency report (Connor, 2005). McCormick then made the overarching conclusion, without citation, that, “Separating siblings who have been removed from their parents only seems to intensify the pain, grief, and trauma that they have already experienced when they were initially removed from their parents.” A distinction needs to be made between the evidence suggesting separation of siblings intensifies trauma versus evidence suggesting positive sibling relationships mediate and decrease already-experienced trauma. Whereas some research suggests that positive sibling relationships may alleviate the effects of trauma (Linares, Li, Shrout, Brody, & Pettit, 2007), there is inconsistent evidence as to whether siblings placed together exhibited fewer internalizing behavior problems (Fernandez & Lee, 2013; Hegar & Rosenthal, 2009, 2011; Linares et al., 2007; Thorpe & Swart, 1992).

Waid (2014) presents two alternative theories—bioecological and social learning theory, but failed to explain how the theories apply to sibling placement; however, other researchers have used both theories to address sibling placement, specifically James and colleagues (2008). James and colleagues (2008, p. 112) explained how the social learning perspective may suggest that sibling relationship quality, especially between an older and younger sibling, may be vital when making placement decisions. If siblings have a negative relationship or if the older sibling is a poor role model, then placement of siblings together may encourage negative behavior from both siblings. If siblings have a positive relationship and the older sibling is a good role model, then placement together may encourage better behavior from both siblings. Second, James and colleagues (2008, p. 12) discussed ecological systems theory (similar to bioecological theory) relevance to sibling placement. Ecological systems theory suggests examining all of a child’s relationships when considering placement decisions, rather than solely the sibling relationship.

Use of Research Evidence in Review Articles

Table 2 summarizes the evidence presented in the review articles. Columns C1-C8 delineate the types of articles cited in each review. C1 shows the total number of references in each review and C2 reports the number of references that each review cited as evidence on the effects of sibling placement. C3 shows the number of studies cited as evidence (from C2) that were empirical research studies (qualitative or quantitative). Reviews included between 21 and 121 citations, less than half of which used these as evidence on the effects of sibling placement (range 7 to 23 citations). The majority of studies used as evidence were empirical research studies (range 7 to 21 citations).

Table 2.

Review Articles Summary

Author, Year	(C1). Total References	(C2). Of (C1), # Used as Evidence	(C3). Of (C2), # of Research Studies1	(C4). Of (C3), # met QL. Criteria2	(C5). Proportion QL Studies (C4)/(C2)	(C6). # QL Studies Available to Cite	(C7). Proportion of Available QL Studies Cited (C4/C6)	(C8). Review article citation count	Conclusion Statements
Ainsworth and Maluccio, 2002	36	03	0	0	0	6	0	12	“A paramount goal of child care and protection authorities should be to preserve family and sibling ties as appropriate. Ideally, this should be through joint placement of siblings.” Pg. n/a
O’Neil, 2002	103	10	9	2	0.20	6	0.33	3	“What is clear is that sibling relationships are very likely to be of lifelong importance…separation should only be a consideration of last resort.” Pg. 14
Groza et. al., 2003	35	16	14	2	0.13	6	0.33	49	“Operating from the basic philosophy that siblings should be placed together unless there is a compelling reason to separate them is consistent with best practice.” Pg. 489
Hegar, 2005	102	15	15	2	0.13	6	0.33	132	“More important, findings of the studies support the tentative conclusion that joint sibling placements are as stable as or more stable than placements of single children or separated siblings, and several studies suggest that children do as well or better when placed with their brothers and sisters.” Pg. 731
Shlonsky et. al., 2005	83	n/a4	n/a	n/a	n/a	6	n/a	86	“While we advocate for a constitutionally protected right of association among siblings based on our read of the scarce extant literature, case law, and on moral principle, it is unknown whether such an approach will result in optimal outcomes for children. We simply need more information…” Pg. 713
Oosterman et. al., 2007	47	75	7	5	0.72	5	1	478	“In sum, although most of the findings concerning siblings in care were in the same direction, that is placement with siblings was associated with less placement breakdown, there was considerable diversity in types of placements…which hampers comparisons across the studies.” Pg. 69
Washington, 2007	27	11	11	2	0.18	8	0.25	45	“If at all possible, children in foster care should be placed with their siblings, unless compelling reasons exist for their separation. Researchers have utilized a wide range of techniques and approaches and arrived at similar conclusions, offering strong support for this finding.” Pg. 431
McCormick, 2010	52	14	12	2	0.17	12	0.17	33	“The evidence presented in this review suggests that in most cases children benefit significantly from having access to their siblings.” Pg. 213
Gustavsson and MacEachron, 2010	41	15	10	3	0.30	12	0.25	26	“Reasonable efforts to enable siblings to maintain their ties regardless of agency actions or placement plans is the cornerstone of best practice.” Pg. 44
Waid, 2014	53	13	12	8	0.67	15	0.53	25	“Much has been learned about the relationship between the co-placement of siblings and placement stability and permanency outcomes. Despite this, the relationship between sibling co-placement stability, and child and family well-being remains unclear.” Pg. 294
Rock et. al., 2015	79	85	8	6	0.75	8	0.75	100	“Separation of siblings was associated with instability in a majority of studies.” Pg. 194
Jones, 2016	40	22	21	5	0.23	19	0.26	20	“Given that research on siblings in adoption and fostering since 2004 broadly confirms Hegar’s conclusion that co-placement of siblings appears to be protective in terms of placement stability, achieving permanence and child well-being ...” Pg. 332
Seale and Damiani-Taraba, 2017	62	23	18	7	0.30	21	0.33	1	“Results of the review partially reflect the results of past research, which stated that siblings benefit from being placed with their siblings and thus co-placement should be an overriding goal when making placement decisions.” Pg. 19–20
Meakings, Sebba, and Luke, 2017	51	18	18	8	0.44	21	0.38	13	“Despite the acknowledge policy and practice imperative to place siblings together in foster care, evidence on the outcomes for sibling placements remains relatively sparse. Taken together, the evidence…suggests that the outcomes for children placed with siblings in foster care are mostly better than for those placed apart from siblings.” Pg. 13
Konijn et. al., 2019	121	95	9	7	0.78	13	0.54	11	“Finally, whenever possible, children should preferably be placed with kin and together with their siblings.” Pg. 495
Children’s Bureau, 2019	21	10	9	3	0.30	23	0.13	n/a6	“Maintaining and strengthening sibling bonds is a key component to child well-being and permanency outcomes. It is also central to meeting the requirements of the Fostering Connections Act.” Pg.8
Total Unique Studies	--	98	83	19	--	--	--	--	--

¹Quantitative and qualitative research studies are included at this point.

²QL = Quantitative and Longitudinal. The criteria for inclusion consists of studies that had sibling placement as the independent variable, were quantitative and longitudinal, and included only children in foster care. All qualitative studies are dropped at this point; however, relevant qualitative studies are examined separately in Table 5. Note that some studies may not distinguish between siblings placed alone versus children placed alone who do not have any siblings. Studies that include children without siblings are specified as footnotes in Tables 2–4.

³Ainsworth and Maluccio (2002) literature review consists of descriptive statistics on initial sibling placement and size of sibling groups. An outcome section beyond these descriptive statistics is not included.

⁴Shlonsky et. al. (2005) was largely a methodological review. Although it included a short outcome section; however, noted that “Much of this literature, however, is conflicting and based on practice experience rather than empirical studies, and this is particularly true in the area of assessment of sibling relationships.” (pg700). Being that Shlonsky et. al. did not specify which studies are based on practice versus empirical research, no further attempt to count empirical studies is included.

⁵This review is on foster care stability with a section on sibling placement included. We examined only the section on sibling placement.

⁶Citation count data not available.

C4 counts the number of studies from C3 that met the following criteria (referred to as “QL criteria” in Table 2): (1) were quantitative and longitudinal (not cross-sectional); (2) sampled from the foster care population; (3) sibling placement was an independent variable (predictor) for at least one of three possible outcomes: stability, permanency, or child wellbeing (broadly defined). These are characterized as studies providing the highest quality of evidence available. C5 shows the proportion of all studies cited as evidence that were QL studies (C4 divided by C2), which ranged from .13 to .78 across reviews. In the median review, about 30% of articles cited as evidence met QL criteria.

C6 reports the approximate number of QL studies that the review could have included, out of the 27 QL studies that we identified. This provides a rough assessment of the thoroughness of prior reviews; however, we caution that none of these reviews claimed to be systematic and thus it is not, per se, incumbent on the review authors to cite all of these studies. Nonetheless, the thoroughness of prior reviews is relevant to whether the conclusions drawn are likely to represent the available evidence. Given delays between when a review was completed and when it was published, we focused on studies that were published two or more years prior to the publication of the review. That is, if the review came out in 2017, we calculated how many of the quantitative longitudinal studies published in 2015 or earlier were cited in the review. For the three reviews focused on foster care placement stability only (Konijn et al., 2019; Oosterman et al., 2007; Rock et al., 2015), quantitative studies not on stability were excluded from their count in C6. Again, we excluded Ainsworth and Maluccio (2002) and Shlonsky and colleagues (2005) from the citation range count because the descriptive statistics provided by Ainsworth and Maluccio (2002) were not part of the outcomes we considered, and Shlonsky and colleagues (2005) cited literature to assert the lack of evidence rather than to provide an outcome. We found that studies could have cited anywhere between 5 and 23 QL studies. C7 reports the proportion of QL studies that could have been cited that were cited in each review; this value ranges from 0.13 to 1, with the reviews focused on placement stability citing the greatest proportions of available research. The median review cited one-third of available QL studies.

Lastly, C8 reports how often each review was cited (if available) from Google Scholar. This gives some indication of how the reviews conclusions have been disseminated in the field (range = 1 to 478).

Review Article Conclusion Statements

Review articles have the difficult job of summarizing findings across multiple studies. Researchers and policymakers who are not familiar with academic literature may rely on conclusion paragraphs to understand the consensus and key takeaways of research findings. Thirteen reviews reached a general conclusion in support of sibling placement, regardless of evidence or lack thereof from empirical studies. Three were more circumspect about the ability to draw conclusions from available evidence (Oosterman et al., 2007; Shlonsky et al., 2005; Waid, 2014). In some cases, reviews dismissed null or inconclusive findings, based on the premise that sibling placement is morally or ethically right in the absence of strong evidence that it is harmful. Reviews may acknowledge that some sibling relationships are harmful or that separation is sometimes necessary as a last resort but provide little clarification on how such cases may be identified. Research evidence on when sibling placement is harmful is critical, given that even in cases of sibling abuse, the sibling relationship may be prioritized over child safety. For example, James and colleagues (2008) wrote about a case where the caregiver considered separating a violent older sibling from the younger but ultimately decided that the trauma of losing a sibling was worse than the risk the older sibling may pose to the younger child. Yet, the amount of trauma experienced from sibling separation has not been rigorously studied, and there is some evidence that maintaining a negative sibling relationship is associated with worse outcomes (Linares et al., 2007).

Three reviews emphasized the lack of evidence supporting sibling placement and the difficulty of drawing conclusions based on methodological limitations (Oosterman et al., 2007; Shlonsky et al., 2005; Waid, 2014). Whereas Waid (2014) simply stated the evidence is unclear, Shlonsky and colleagues (2005) stated more information is needed before any conclusion can be made, and Oosterman and colleagues (2007) noted the difficulties in comparing results across studies due to research design differences.

Qualitative Studies of Siblings in Foster Care

Table 3 describes the nine qualitative studies addressing protective or beneficial factors—or lack thereof—of sibling relationships in foster care. Notably, sample characteristics varied widely across studies. Many individual studies sampled solely or predominantly sibling groups that were placed together (Harrison, 1999; Leichtentritt, 2013; Sting, 2013) or apart (Wojciak, 2017). Two studies only interviewed caseworkers or foster parents (James et al., 2008; Wojciak et al., 2018) and the remaining seven studies only interviewed the children. Lastly, four studies took place in countries outside of the United States: Norway (Angel, 2014), Israel (Leichtentritt, 2013), Austria (Sting, 2013) and Britain (Harrison, 1999). Although aspects of sibling separation and sibling relationships may be generalizable across countries, the significant differences in social context and foster care systems warrants caution in applying the findings of non-U.S. studies to the U.S. foster care system.

Table 3:

Qualitative Studies on Siblings in Foster Care

Author, Year	Sibling Separation	Sample/Procedures1	Summary
Folman, 1998	Unclear2	N=903, ages 8–14; interview with children	Sibling relationships perceived as positive. Some children were not told they would be separated from their siblings until they arrived at the foster homes.
Harrison, 1999 (in Mullender44	One pair interviewed together; remainder had siblings living elsewhere	N=9, ages 11 to 20; interview with children	Siblings perceived as protective, easing transition into care, and providing a sense of identity. Feelings of loss and grief were associated with separating siblings. No negative experiences of sibling placement/relationships discussed.
James et. al., 2008	Mixed	N=14, ages 6 to 14; interviews with caregivers5	Caregivers mostly viewed sibling relationships positively. In cases of sibling conflict/abuse, caregivers kept siblings together due to fear of additional trauma from loss and placement instability.
Leichtentrit, 2013	All lived with at least one sibling	N=12, ages 7 to 14; interviews with children	Sibling relationships were viewed as both positive and negative; often, children had contradictory views of their siblings.
Sting, 2013	Lived with at least one sibling	N=varied6; ages 12 to 14; interviews with children	The sibling relationship was described as very important and held a central role in a child’s social network when in foster care.
Angel, 2014	Mixed	N=10, ages 4 to 12 at entry; interviews with children	Children define children as their siblings or not based on who they live with and how they perceive their social interactions. Biological siblings can be the most meaningful relationship, but not always.
Affronti, Rittner, and Jones, 2015	Unclear	N=18, ages 18 to 25 interviews with foster children alumni	Sibling separation perceived as positive and negative. Some said placement with siblings helped them to adjust and feel connected to birth families. One respondent felt separation from sibling was why they did not do well in foster care. Some stated separation was positive because they received extra attention from their foster families.
Wojciak, 2017	Goal of the camp was to reunite separated sibling groups, but some groups may have been placed together	N=178, ages 6 to 19; children answered open-ended survey questions	Siblings were characterized as people that children can rely on, are connected to, and fulfills them. No negative experiences of sibling placement/relationships discussed.
Wojciak et. al., 2018	Did not specify	N=15, foster parents, ages n/a; interviews with foster parents	Foster parents reported that sibling relationships should be maintained and if split, contact should be priority. Reports of negative consequences when split: acting out, shutting down, anxiety, worrying/wondering about sibling. Authors emphasized in discussion to educate parents that even with negative relationships, sibling relationships could still be beneficial/lifelong.

Nevertheless, qualitative studies investigated and discussed both benefits and harms of sibling placement. Common benefits discussed for maintaining sibling placement were avoiding additional loss and grief, providing a lifelong bond and connection, and creating a sense of fulfilment (Affronti et al., 2015; Folman, 1998; Harrison, 1999; James et al., 2008; Sting, 2013; Wojciak, 2017; Wojciak et al., 2018). When negative outcomes were mentioned—such as in cases of sibling abuse, parentification of the older sibling, and the older sibling as a poor role model –the benefits of maintaining the sibling relationship were often described as outweighing any risks. However, not all qualitative studies strongly supported sibling relationships. Leichtentritt (2013), Angel (2014), and Affronti and colleagues (2015) reported cases where children were conflicted, showed no desire for connection with their siblings, or enjoyed separate placements due to receiving increased attention from caregivers. Although biological siblings were commonly cited as important, Angel (2014) directly noted that sometimes children create closer bonds with their foster siblings.

We also note that one of the most commonly cited articles in support of sibling placement is not research, but rather a semi-autobiographical account from adults formerly in the foster care system (Herrick & Piccus, 2005). Their article has been cited 174 times as of March 2020 and was cited as evidence in two of the review articles. We do not wish to discredit the voices of children separated from their siblings in foster care; rather, we caution that their experiences may not be generalizable. Herrick and Piccus described feelings of loss and intense desire for their old life with their siblings. They mentioned gratitude at being placed with at least some of their siblings. They also suggested changing policies for a consistent understanding across states on the importance of maintaining sibling ties. Yet, two of the nine qualitative studies’ samples included cases in which there was indifference or no attachment between siblings (Angel, 2014; Leichtentritt, 2013). Similarly, foster parents and caseworkers also reported situations in which sibling placement or maintaining sibling relationships could be counterproductive (Angel, 2014; Boer & Spiering, 1991; Smith, 1996). These qualitative and descriptive studies recommended monitoring of sibling relationships and that individualized data be collected to inform the sibling placement decision as well as other forms of sibling contact.

Although the purpose of this paper is not to discuss factors influencing sibling placement and we did not conduct a thorough review of the literature on this topic, the included articles described a variety of factors outside the control of individual social workers or agencies may cause sibling separation. Some of the common factors contributing to the decision of sibling placement included constraints on the availability of foster homes willing and able to provide for sibling groups, with common barriers including insufficient space in the house, rules regarding bedroom sharing for children of different age/gender combinations, difficulty providing care and supervision for large sibling groups, ability to attend to special needs cases; behavioral problems or conflict between siblings; and biological families’ requests to keep siblings together (Boer & Spiering, 1991; James et al., 2008; Kosonen, 1996; McDowall, 2015; Smith, 1996). In addition, based on questionnaires filled out by social workers on behalf of 297 children, Kosonen (1996) found the top two reasons for separation had to do with systematic barriers where 52% of cases were due to siblings not in care at the same time and 18% of cases were due to siblings entering care through separate admissions. Notably, Kosonen (1996) study took place in Scotland and findings may not generalize to the U.S. context.

Longitudinal Studies of Siblings in Foster Care

Sibling Placement and Child Well-Being

Table 4 describes nine studies on sibling placement and child wellbeing. Notably, measurements of child wellbeing differ across the nine studies and are therefore not directly comparable. The identified studies on child wellbeing focused on one or more of the following domains: behavior, internalizing behavior/mental health, academics, and relationships. Seven of the nine studies examined internalizing and externalizing behavior problems using the CBCL (Child Behavioral Checklist), the ECBI (Eyberg Child Behavior Inventory), the North Carolina Family Assessment Scale-Reunification (NCFAS-R), self-created checklists, or administrative data (Fernandez & Lee, 2013; Hegar & Rosenthal, 2009, 2011; Kothari et al., 2018a; Leathers, 2005; Linares et al., 2007; Thorpe & Swart, 1992).

Table 4.

Longitudinal Studies of Sibling Placement and Child Wellbeing

Author, Year	Data type/Years	Sample1	Analytical Approach	Controls Used	Results
Thorpe and Swart, 1992	Child’s Aid Society (CAS); administrative records: location/year not specified	N=115, ages 0 to 15 at entry	T-tests	None	If siblings placed together, greater number of negative symptoms while in foster care and at discharge including diagnoses of mental disorder, behavioral issues, or criminal behavior.*2
Leathers, 2005	Cook County, Illinois; survey & administrative records: 1997/1998–2002	N=196, ages 12/13	Hierarchical linear regression	Demographics, placement history, behavior problems at interview, foster home integration, no. of siblings, placement type, frequency of maternal visits	If siblings placed consistently together, child more likely to experience foster home integration than if alone in all placements (ns), alone at time of interview with history of placement together*, or together at time of interview with history of separation* (based on foster parent and caseworker answers)
Leathers, 2006	Cook County, Illinois; interviews and agency records: 1997/98–2003	N=179, ages 12 to 13	Logistic regression; tracked for 5 years	Demographics, behavioral issues, placement history	No differences by sibling placement in probability of negative placement outcome once leaving foster care (placement in residential treatment centers, juvenile detention centers, prison, or runaway)
Linares et. al., 2007	New York City survey (NYU Longitudinal Study): 2002/2005–2007	N=156, ages 3–14	Linear mixed effects regression; average 14.6 month follow-up3	Baseline value of dependent variable, sibling relationship quality, child age	No significant differences by sibling placement in depressive symptoms/loneliness (child reports)
					For children with few baseline behavior problems: sibling separation (after initially together) associated with worse Time 2 birth parent-reported behavior problems, compared with either consistently together or consistently separated*. Associations between sibling placement and wellbeing did not differ by sibling relationship quality (though positive relationship quality was associated with better outcomes overall).
Hegar & Rosenthal, 2009	National survey (NSCAW I): 1999/2000–2003	N=1,415, ages 4–14	Linear, binary logistic, and ordinal logistic regression; 48–96 month follow-up4	Demographics, kinship care, sample type5, caregiver demographics, county poverty, wave 1 outcomes, CPS investigation	If any sibling in same home as child (versus not, including children with no siblings in foster care): Fewer child-reported internalizing behavior problems*
					Fewer teacher-reported externalizing behavior problems for Hispanic* and Black youth+.
					No statistically significant differences in: child-reported, externalizing behavior, teacher-reported internalizing behavior, foster parent-reported internalizing or externalizing behavior, feeling part of family, or parent involvement Child more likely to report feelings of emotional support from caregiver*
					Among those who recently entered care: child more likely to report feeling close to caregiver, liking those with whom they live* Among white children in kinship care: lower school performance
Hegar & Rosenthal, 2011	National survey (NSCAW I): 1999/2000–2004/2008	N=1,113, ages 6–14	Linear and ordinal logistic regression; 48–96 month follow-up	Demographics, kinship care, sample type5, caregiver demographics, county poverty, wave 1 outcomes, CPS investigation	If siblings placed fully intact (versus fully separated): No statistically significant differences in internalizing or externalizing behavior problems as reported by foster parents, youth, or teachers
					Higher teacher-reported school performance* No difference in child’s feelings of closeness or likes living with foster parents If placed partially together (versus fully separated): Children more likely to report feelings of closeness and enjoy living with foster parents * More teacher-reported externalizing behavior problems* (specifically for non-kinship placements and recent foster care entries) No differences in behavior or school performance
Fernandez and Lee, 2013	Temporary Family Care (TFC) programs in Barnardos, Australia: not specified	N=145, ages 0 to 12 at entry	OLS regression; measured at entry and at exit; tracked for 18 months3	Demographics, biological mother age/educ, reason in care, NCFAS-R scores at placement entry/exit6	If placed with at least one sibling: No statistically significant differences in (reunification family) environment, parental capabilities, family interactions, family safety, and child/caregiver ambivalence children had better wellbeing at intake*, but no statistically significant difference for child wellbeing at exit
Waid et. al., 2017	Unnamed U.S. North Pacific Region; survey and administrative data: 2010–2015	N=185, ages 7–15	Latent growth curve modeling, tracked for 18 months4	Demographics, age order, treatment group, child behavior, placement history, kinship care	No statistically significant difference by sibling placement in feeling positively towards foster caregivers or feeling integrated into foster household
Kothari et. al., 2018	3 Oregon Metropolitan regions; survey and administrative education data: 2009–2015	N=315, age 7–15	Hierarchical linear or logistic regression; tracked for 18 to 24 months4	Demographics, learning disability, school mobility, kinship care, treatment (larger study was RCT)	If siblings placed together (versus apart), children had fewer discipline events at school, including in-school suspension, truancy/attendance violation, out-of-school suspension, and expulsion*

ns=not statistically significant;

⁺statistically significant at p<.01

^*statistically significant at p<.05

^1.Ages based on Time 1 of study observation period unless otherwise specified

^2.Information on academic functioning in this study not included due to conflict between tables and text of article. Study reports using logistic regression to predict sibling separation, but appears only bivariate comparisons made to identify differences in outcomes by sibling placement

^3.Study sample included multiple siblings but did not distinguish between fully separated and partially intact sibling groups.

^4.Study sample included multiple siblings but did not distinguish between fully intact and partially intact sibling groups.

^5.NSCAW includes a sample of children who recently entered care (CPS) and children who had been in care for 1 year at first observation (LTFC). The two samples had different lengths of follow up.

^6.NCFAS-R: North Carolina Family Assessment Scale-Reunification assessed by practitioners includes 7 domains: 1) environment—housing/financial/food stability, community safety, habitability of housing, personal hygiene, transportation, learning environment 2) parental capabilities—child supervision, disciplinary practice, parent/caregiver mental/physical health/drug or alcohol use, development/enrichment opportunities, 3) family interactions—bonding with child, expectation of child, family support/parent relationship with each other, 4)family safety—physical, emotional, sexual abuse, neglect, intimate partner violence, 5)child well-being—mental health, child behavior/school performance relationship with caregiver/siblings/peers, motivation/cooperation to maintain family, 6)care giver/child ambivalence, and 7)readiness for reunification—resolution of CPS risk factors, completion of case service plans, resolution of legal issues, parent/child understanding child’s treatment needs, established back-up support/service plan

Overall, studies produced inconsistent conclusions. Thorpe and Swart (1992) used administrative records from a children’s aid society and found that children placed together had more negative behavioral and mental health symptoms by the time they exited foster care than children placed together (p<.05), but the analysis included no statistical controls. When children, foster parents, and teachers reported on internalizing and externalizing behavior, two studies (both using the U.S. National Survey of Child and Adolescent Wellbeing I) found no statistically significant associations between sibling placement and externalizing behavior when reported by youth, teachers, or foster parents in the full sample (Hegar & Rosenthal, 2009, 2011) though there appeared to be some benefits of sibling placement within select subgroups for teacher-reported externalizing behavior problems (Hegar & Rosenthal, 2009).

Linares and colleagues (2007) found that sibling separation after initial joint placement may result in greater behavior problems for children who had few behavior problems previously, whereas separation may reduce behavior problems for children with high levels of initial behavior problems. Yet, among children placed continuously together or continuously apart, Linares and colleagues (2007) found no differences in behavior problems. In addition, a positive sibling relationship was associated with better child wellbeing, regardless of whether siblings were placed together. Notably, behavior for this study was reported by the children’s biological parent, who had approximately three visits per month with their children; thus, their reports may not reflect children’s usual behavior.

Outcomes of mental wellbeing and internalizing behavior problems were also inconsistent across studies. Linares and colleagues (2007) found no differences by sibling placement in child-reported depressive symptoms or loneliness and Hegar and Rosenthal (2011) found no differences in child-, foster parent-, or teacher-reported internalizing behavior problems. Yet, Thorpe & Swart (1992) reported that siblings placed together exhibited more mental health symptoms than siblings placed apart, whereas Hegar and Rosenthal (2009) found fewer child reported internalizing behavior problems for children placed with at least one sibling, but no differences in foster parent or teacher-reported internalizing behavior problems. Inconsistencies may reflect differences in benefits or harms associated with partially-intact sibling groups relative to being placed alone (Hegar & Rosenthal, 2011), such that studies with binary measures of sibling placement draw different conclusions than studies that distinguish among fully intact/together, partially together, and fully separated sibling groups, as well as children without siblings in foster care.

Other studies used alternative measures to indicate child behavior or mental functioning. Fernandez and Lee (2013) used an overall measure of child wellbeing and found that, although separated siblings had lower well-being scores at intake, there were no differences by sibling placement at exit. This suggests that aspects of child well-being may have affected the likelihood of sibling separation, but separation did not produce persistent differences in well-being. Leathers (2006), indirectly studied child’s behavior by examining whether the child exited into a residential treatment center, juvenile detention center, prison, or ran away. Leathers found no association between sibling placement and negative placement outcomes upon exiting foster care. However, Kothari and colleagues (2018) found that placing siblings together was associated with fewer school discipline events.

Two studies examined academic outcomes (Hegar & Rosenthal, 2009, 2011). Hegar and Rosenthal (2009) found that for some subgroups of children, specifically white children in kinship care, sibling placement was associated with lower school performance. Yet, in a follow-up study using the same dataset, Hegar and Rosenthal (2011) found that placement together versus fully separate improved school performance, but that children placed with one but not all siblings had equivalent school performance as children placed alone. Although both studies used the same dataset, Hegar and Rosenthal (2009) included children without siblings in the placed alone group and combined partially and entirely intact sibling groups, whereas Hegar and Rosenthal (2011) excluded children without siblings and distinguished between fully intact and partially intact sibling groups. Both studies also examined different interaction terms in their models. These differences in study design likely explain differences in findings.

Lastly, four studies examined associations between sibling placement and measures of adjustment and child-foster parent relationship. Hegar and Rosenthal (2009) examined several measures of adjustment: liking their current caregiver, caregiver involvement, feeling like part of the family, feeling close to their caregiver and perceived emotional support from caregiver. In most models, no statistically significant associations were found. However, they did find children reported feeling more supported by the foster parent when placed together than apart, and among children who had been in foster care for a shorter amount of time at the beginning of the study, sibling placement was associated with greater likelihood of feeling close to their caregivers and liking living with their foster parents as compared with sibling separation. In contrast, among children in foster care for a longer amount of time at the beginning of the study, those placed with siblings were equally or less likely to report feeling close to or liking living with foster parents. Notably, however, Hegar and Rosenthal (2011) found siblings partially placed together were more likely to report feelings of closeness and enjoying living with foster parents than children placed completely alone. However, there was no difference in the same outcomes for children placed completely together versus completely alone. In addition, Leathers (2005) found children placed continuously with siblings were more integrated into their foster family than children placed apart some or all of the time. In contrast, Waid and colleagues (2017) found no differences by sibling placement in children’s positive feelings toward the foster family or feelings of being integrated into their foster home.

Notably, the time frame for the data used in the studies ranged from the 1990s to 2015. To our knowledge, two studies (Kothari et al., 2018; Waid et al., 2017) have assessed associations between sibling placement and child wellbeing using U.S. data from after the enactment of the federal policy requiring reasonable efforts to keep siblings together (Fostering Connections to Success and Increasing Adoptions Act, 2008). Because that policy may have increased rates of sibling placement for children who would previously been separated – specifically for difficult-to-place sibling groups or siblings who have a negative relationship– it is possible that the average effects of sibling placement might have changed consequent to this policy. Of the two studies, one identified protective effects of sibling placement (Kothari et al., 2018) and one found no differences (Waid et al, 2017). Given the lack of studies using recent data – and that the two studies identified were small samples (185 and 315 children, respectively)—we cannot ascertain whether current policy has improved, had no effect, or worsened the wellbeing of siblings in foster care.

Sibling Placement and Permanency

Table 5 summarizes five studies on the association between sibling placement and permanency. Three forms of permanency are examined: reunification, adoption, and guardianship. Under federal law, agencies are typically required to pursue reunification before allowing other forms of permanency, and reunification is the most common form of permanency (Children’s Bureau, 2018). All five studies examined reunification as one of their outcomes. Notably, because reunification often reflects parents’ compliance with the case plan, the outcome of reunification would typically not vary among siblings removed in the same time period. Three of five studies found that siblings placed together in foster care were more likely to reunify and reunified more quickly than when placed alone (Akin, 2011; Albert & King, 2008; Webster et al., 2005); however, Albert and King (2008) did not report statistical significance. In the remaining two studies, neither Leathers’ (2005) nor Fernandez and Lee (2013) found stastically significant associations between sibling placemen and reunification. However, Fernandez and Lee found parents were more likely to be ready for reunification (completion of case plan, legal issues, understanding of child’s treatment needs, etc.) when siblings were in the same placement. Notably, when comparing siblings placed completely together versus partially together (Albert & King, 2008) and siblings placed partially together versus completely alone (Akin, 2011), no differences were found in exits to reunification.

Table 5.

Sibling Placement and Permanency

Author, Year	Location/Years	Sample1	Analytical Approach2	Outcome	Controls Used	Results
Leathers, 2005	Cook County, Illinois; survey & administrative records: 1997/1998–2002	N=167 (adoption/guardianship), N=195 (reunification), ages 12/13 at time 1 of study	Logistic regression, 5 year follow-up3	Reunification, adoption or guardianship	Demographics, placement history, behavior problems at interview, foster home integration, no. of siblings, frequency of maternal visits	If siblings consistently placed together: higher odds of adoption/guardianship than if placed alone at interview* or placed alone in all placements* lower odds of reunification than if placed alone at interview, or placed alone in all placements (ns). No difference in adoption/guardianship or reunification compared to children with current sibling placement but prior separations
Webster et. al., 2005	California; administrative records: 2000–2001	N=15,517, ages 0 to 17	GEE and logistic regression; up to 12 months follow-up	Reunification	Demographics, removal reason, placement type, no. of moves, no. of siblings, initial sibling placement, region	If siblings placed completely or partially together (versus alone), higher odds of reunification*
Albert and King, 2008	Nevada; administrative records: not specified	N=602, ages 0 to at least 16	Survival analysis; up to 19 months follow-up	Remaining in care vs reunification	None	If siblings placed together (versus alone), less likely to still be in foster care at end of observation period. Similar likelihood of remaining in care if placed fully together versus partially together. (Did not specify statistical significance)
Akin, 2011	Unspecified Midwestern state; administrative records: 2006~2010	N=3,351, ages 0 to 17	Survival analysis; 30–42 month follow-up	Reunification, adoption, guardianship	Demographics, disability, mental health, removal reason, removal history, initial placement type, early placement stability, runaway	If placed with all siblings (versus alone), higher odds of reunification* and adoption+; no difference in guardianship
						No difference in exits to reunification, guardianship or adoption if partially together versus completely separated
Fernandez and Lee, 2013	Temporary Family Care (TFC) programs in Barnardos, Australia: not specified	N=145, ages 0 to 12	Logistic regression; measured at entry and at exit; tracked for 18 months4	Reunification	Demographics, biological mother age/educ, reason in care, NCFAS-R scores at placement entry/exit5	If placed together (versus apart): No difference in reunification rate Higher scores on readiness for reunification at time of exit*

^1.Age is age at entry unless otherwise specified

^2.Sibling placement included sibling groups placed completely together, partially together, or completely split unless otherwise specified. All children had at least one sibling.

³Sibling placement categorized as alone in all placements, currently alone with history of being together, currently together with history of being alone, always together. All children had at least one sibling.

^4.Sibling placement categorized as sibling in same placement (yes or no). Study sample included multiple siblings but did not distinguish between fully separated and partially intact sibling groups.

^5.NCFAS-R: North Carolina Family Assessment Scale-Reunification assessed by practitioners includes 6 domains: 1) environment—housing/financial/food stability, community safety, habitability of housing, personal hygiene, transportation, learning environment 2) parental capabilities—child supervision, disciplinary practice, parent/caregiver mental/physical health/drug or alcohol use, development/enrichment opportunities, 3) family interactions—bonding with child, expectation of child, family support/parent relationship with each other, 4)family safety—child maltreatment, intimate partner violence, 5) child well-being—mental health, behavior/school performance, relationship with caregiver/siblings/peers, motivation/cooperation to maintain family, 6) readiness for reunification.

Ns= not statistically significant;

^*p<.05;

⁺p<.1

Although efforts to keep sibling groups together for adoption and guardianship are typically required, sibling groups that are not reunified may exit to adoption or guardianship together or separately, and they may exit foster care at different times. However, the two studies that included adoption or guardianship as a permanency outcome did not address whether children achieved permanency together or separately. Akin (2011) found no statistically significant association between siblings placed partially together versus completely alone for either adoption or guardianship, but children had (marginally significantly) higher odds of adoption if placed completely together versus alone. Leathers (2005) combined adoption and guardianship into a single category and found higher odds of adoption/guardianship among intact (versus separated) sibling groups. However, Leathers’ (2005) model demonstrated poor fit and was based on a smaller and narrower sample than Akin (2011).

Sibling Placement and Stability

Table 6 includes 18 studies on sibling placement and placement stability in foster care. Of these, 11 reported only bivariate analyses (no statistical controls). Studies used a variety of measures of stability, including number of placement changes and risk of placement disruption, with some studies differentiating between planned/intentional moves to more preferable settings versus unplanned moves or moves to less preferable (e.g., more restrictive) settings. Across studies, the results were mixed and not all studies specified statistical significance.

Table 6:

Sibling Placement and Stability

Author, Year	Locations/Years	Sample1	Analytical Approach2	Controls Used	Results
Trasler, 1960 3	Unspecified London location; agency records and interviews: not specified	N=138, ages 0 to 11+ at entry	Chi-Square test of significance; analysis based no. of placements4	None	If placed with at least one sibling (versus alone), more likely to have satisfactory placement.*5
Parker, 1966 3	England; agency records: 1952/53–1957/58	N≈1446, ages 0 to 11+ at entry	Bivariate analysis; tracked for 5 years; analysis based case records	None	Children who were always separated from siblings (versus consistently together or inconsistently together) were more likely to have a successful placement (ns)7
George, 1970 3	Two unspecified England towns and a county; questionnaires and agency records: 1961/63–1968	N=86, ages <13 at entry	Bivariate analysis; tracked for 5 years; based on no. of placements	None	If placed together: less likely have successful placement than if placed partially or completely separate. *,7 If placed together: more likely to have failed placement than if placed partially or completely separate*
Berridge and Cleaver, 1987 3	3 unspecified England agencies; survey: 1983	N=145 long-term care; N=112 short-term care, ages 0 to 18	Bivariate analysis; retrospective for 5 years	None	Child separated from all siblings were more likely to experience breakdown than children placed partially together or completely together (significance not specified).
Staff and Fein, 1992	Four New England States; staff interviews and agency records: 1976–1990	N=134, 0 to 13+ at entry	Bivariate analysis8	None	If placed initially together (versus initially separate): more likely that neither sibling changes placements* less likely one sibling changes placements and the other does not* more likely that both siblings change placements.*
Thorpe and Swart, 1992	Child’s Aid Society (CAS); administrative records: location/year not specified	N=115, ages 0 to 15 at entry	T-tests3	None	Fewer total placements if siblings placed together.*
Leathers, 2005	Cook County, Illinois; survey & administrative records: 1997/1998–2002	N=196, ages 12/13 at time 1 of study	Logistic regression	Demographics, placement history, behavior problems at interview, foster home integration9, no. of siblings, frequency of maternal visits	In final model (all controls), lowest odds of disruption if consistently placed with all siblings (versus alone, or inconsistently placed with siblings; ns).
Leathers, 2006	Cook County, Illinois; interviews and agency records: 1997/98–2003	N=179, ages 12/13	Logistic regression; tracked for 5 years4	Demographics, behavioral issues, placement history, foster home integration9	In final model (with controls), higher odds of disruption post interview if placed with at least one sibling (versus alone)+
Barth et. al., 2007	National survey (NASCAW: CPS sample): 1999/2000–2004	N=362 with Emotional/Behavioral Disorder; N=363 without, ages 7 to 14	Poisson Regression; followed up at 48 months4	Demographics, initial placement, criminal court appearance, special needs, mental health, trauma, expects reunification, no. prior moves	Among those with behavioral issues, if placed together (versus alone): less likely to experience multiple moves.* No difference in number of moves experienced when children have no behavioral issues (ns).
Holtan et. al., 2013	Norway; questionnaire: 2000–2008	N=136, ages 4 to 13 at entry	Generalized linear mixed model; followed up 8 years later4	None	If placed together (versus alone): lower odds of disruption experienced between time 1 to time 2 (ns).
Koh et. al., 2014	Illinois; agency records: 2006–2007	N=121, ages 0 to 12+ at entry	Hierarchical linear logistic regression; propensity score matching; tracked for 18 months4	Time spent in kinship care, caregiver willingness to commit to permanence, DSM diagnosis	No statistically significant difference in number of moves based on proportion of time spent in placement with at least one sibling (ns).
Vinnerljung, Sallnas, & Berlin, 2014	12 municipalities in southern Sweden; agency records: 1980/1992 birth cohort	N=136, age 12+	Only bivariate analysis presented; tracked from age 12 to 184	None	Placement with a sibling increased risk of disruption*
McDowall, 2015 3	Australia, CREATE’s Report Card, survey: 2013	N=1,160, ages 8 to 17	Univariate One-Way ANOVA; length of study not reported	None	Siblings placed with all siblings or with at least one sibling (versus alone) experienced fewer mean number of placements.*
van Rooij et. al., 2015	2 regional Dutch Foster Care Organizations agency records: not specified	N=164, ages 0 to 18 at entry	Bivariate; length varied4	None	No bivariate differences between planned and unplanned termination of placement; sibling placement not included in multivariate logistic model
Waid et. al., 2016	Northwestern State, administrative data and SIBS-FC study: not specified	N=328, age 7–15	Logistic regression; tracked for 18 months10	Sibling relationship quality, home integration, behavioral issues, no. of placements prior to study, treatment conditions, age	If siblings placed together (versus apart), placement is less likely to disrupt* Placed together in kinship care least likely to disrupt.*
Font, Sattler, & Gershoff, 2018	Texas: administrative data: 2008/09–2016	N=23,760, ages 0 to 18 at entry	Multilevel logistic regression; tracked until exit or mid-2016 (whichever later)	Demographics, removal reason, CPS history, initial placement, no. prior moves	If siblings placed together (versus separated from some or all siblings): less likely to end in non-progress11 move if kinship placement* for other placement settings, associations between sibling placement and non-progress move depended on outcome measure
Sattler, Font &, Gershoff, 2018	Texas; administrative data: 2008/09–2016	N=23,765, ages 0 to 13+ at entry	Multilevel survival analysis; tracked until exit or mid-2016 (whichever later)	Demographics, mental health, behavior issues, disability, CPS history, removal reason, time in care, placement type	If siblings placed together (versus separated from some or all siblings): less likely to experience child-initiated disruption* or mismatch disruption* within 2 years13 more likely to experience substandard care disruption within 2 years*13
Leathers et. al., 2019	Large Midwestern US state: interviews occurred in 2014–2015	N=139, ages 8 to 14	Not included in final models, initial model type not specified; follow-up 2 years post interview4	None	No significant differences by sibling placement in initial regression; sibling placement not included in final models

ns=not statistically significant at p<.1

^*p<.05

^1.Age at time of study unless otherwise specified

^2.Analysis based on number of children (at lowest level if multilevel model) unless otherwise specified

^3.This study may not be peer reviewed

^4.Sibling placed alone may include children without siblings

^5.All children in the sample had experienced at least one removal. Satisfactory placements were judged by caseworkers based on four criteria given by the researcher. A placement was satisfactory if caseworkers rated it as “excellent” on a 5 point scale and unlikely to experience a disruption. Failed placements consisted of children in long-term care who were removed from the foster home at least once during a three year period.

^6.209 children were in the total sample. 69% had at least one sibling in care with them. The exact count of the number of children with siblings was not given; based on provided information, we estimate ~144 children had siblings in care. Only children with siblings were included in results

^7.A placement was successful if the child did not experience a removal during the study period. A placement failed if the child was removed from the foster home during the study period.

^8.Sibling pairs (dyads) examined exclusively

^9.Foster home integration was a measure to test foster family-child attachment and asked by caseworkers and foster parents.

^10.Sample focused on sibling pairs (dyads).

^11.Non-progress moves are when a child moves to a less-preferred setting (e.g., from kinship care to non-relative care; from non-relative care to a restrictive setting) or equally-preferred setting (i.e., to a new placement of the same setting type).

^12.Restrictive settings included examples such as group homes or residential facilities

^13.Child-initiated disruption = child refused to stay/ran-away from placement; mismatch-disruption = children needed more specialized care or request move due to behavior; substandard care disruption = foster caregivers did not follow state standards for care

Four of the 18 studies found that children placed with siblings were significantly less likely to experience placement disruptions, or experienced fewer total placements, than siblings who were placed apart (Berridge & Cleaver, 1987; McDowall, 2015; Thorpe & Swart, 1992; Trasler, 1960). Six additional studies (Holtan et al., 2013; Koh et al., 2014; Leathers, 2005; Leathers et al., 2019; R. Parker, 1966; van Rooij et al., 2015) found no significant decrease in disruption risk. Where findings were not statistically significant, the direction of coefficients was sometimes consistent with benefits of sibling placement (Holtan, 2013; Leathers, 2005) and other times consistent with benefits of separation (Parker, 1966; Koh et al., 2014). (Not all studies reported coefficients where differences were not statistically significant).

Three of the 18 studies found that sibling separation was associated with a lower risk of disruption (George, 1970; Leathers, 2006; Vinnerljung et al., 2017). However, two of the three studies (George, 1970; Vinnerljung et al., 2017) included no controls, and differences reported in Leathers (2006) were marginally statistically significant (p<.10).

Five studies considered whether associations between sibling placement and stability varied by child or placement characteristics (Barth et al., 2007; Font et al., 2018; Waid et al., 2016) or how placement moves were measured (Staff & Fein, 1992; Font, Sattler, & Gershoff, 2018; Sattler, Font, & Gershoff, 2018). Waid and colleagues (2016) found that siblings placed apart in non-kinship foster care had a higher risk of placement disruption than children co-placed in kinship care, children co-placed in non-kinship care, and children placed apart in kinship care. That is, the benefits of sibling placement appeared to be stronger in kinship care than non-kinship care. Similarly, Font and colleagues (2018) found that placement with all siblings reduced the risk of a non-progress move (move to a less-preferred setting or due to problems with the existing placement) for children placed with kin, but associations for children in other types of placement settings were sensitive to measurement and categorization of placement moves. Barth and colleagues (2007) found that for children with known behavioral or emotional issues, children placed with siblings were less likely to experience multiple moves, but there was no difference among children with no known behavioral or emotional issues

In a bivariate analysis, Staff and Fein (1992) examined placement disruptions in which one sibling moved versus all siblings moving; siblings placed together had fewer total and individual moves but more joint moves. Lastly, Sattler and colleagues (2018) found that, if placed with all known siblings, children were less likely to refuse to stay in placement or run away and less likely to move due to their own behavioral issues or a mismatch between children’s needs and caregivers’ abilities. However, children placed with all siblings were more likely to move placements due to risk of abuse.

Barriers to Understanding the Effects of Sibling Placement

In our review of the quantitative literature, we observed that many of the existing research studies on sibling placement were limited in their generalizability and internal validity. There are a variety of barriers to identifying the effects of sibling placement. Large survey datasets used to study sibling placement (such as the National Surveys of Child and Adolescent Wellbeing) only include a single ‘focal child’ rather than observing a complete sibling group. Thus, most studies cannot consider whether benefits or harms to the focal child are similarly observed in their siblings. It is certainly plausible that joint placement could benefit one sibling while harming another, but the existing body of research cannot assess this issue. Moreover, by focusing on a focal child, studies are not observing the characteristics of siblings that may affect the outcomes of sibling placement, such as whether their sibling has a disability or serious behavioral problems.

In addition, there are serious concerns about selection bias that go unaddressed. A substantial number of studies reported only bivariate analyses. Even with multivariate models, it is difficult to understand the true effects of non-random events, such as sibling placement – there are a variety of factors, often unobserved or unrecorded, that may lead to sibling separation and also affect the stability, permanency, and wellbeing of siblings. These factors may include inter-sibling abuse, negative sibling relationships, or different levels of need within the sibling group (i.e., one sibling requires a more restrictive placement). No studies we located accounted for sibling abuse and only three studies included measures of sibling relationship quality (Fernandez & Lee, 2013; Linares et al., 2007; Waid et al., 2016).

Lastly, it is possible that publication bias affects the quality and quantity of studies available for review. Given that authors are substantially less likely to submit null results for publication (Franco et al., 2014), better-controlled studies, which are less likely to produce statistically-significant findings, may be underrepresented in the extant literature.

Discussion

In this article, we reviewed published literature reviews, qualitative research studies, and longitudinal quantitative research studies on the effects of sibling placement in foster care published since 1990 to 2019. We sought to understand the quality and quantity of evidence that has been used to inform policy and practice on sibling placement in foster care. We found that literature reviews on siblings in foster care relied on research evidence to draw conclusions, but that research was frequently methodologically weak or based on children in the general population. Our extensive search of the literature identified 27 quantitative longitudinal studies of sibling placement and placement stability, child wellbeing, or permanency. However, the review articles typically cited only a small proportion of these studies. On average, review articles cited less than half of the quantitative longitudinal studies published on their topic two or more years prior to the publication of the review. It is also possible that review articles used more narrow selection criteria than used in this review. However, it is also possible that authors of review articles emphasize findings that support sibling placement, due to humanitarian, theoretical, political, and best practice reasons, despite limited and often contradictory results from empirical studies. Support for sibling placement is explicit in the writing of many authors on this topic. For example, as McCormick (2010, p. 213) wrote, “Few would argue against the fact that maintaining and supporting sibling relationships is important from both a humanitarian and philosophical perspective”. We also note, as others have, that children themselves often place substantial value on sibling connections and their perspective should not be discounted.

Nevertheless, it is critical to have rigorous, unbiased analysis of the effects of sibling placement, even if the political and societal support for keeping siblings together remains unchanged. From the 27 longitudinal quantitative research articles we reviewed for this study, it was difficult to draw meaningful conclusions. Many studies presented bivariate analyses only (i.e., included no statistical controls to address confounding variables) and none of the studies considered the interdependency of outcomes among siblings (i.e., that a child’s outcomes are likely to be correlated with their siblings’ outcomes) and besides from total number of siblings and sibling relationship quality, none controlled for sibling group effects (i.e., how the characteristics of a child’s siblings or the sibling group affect the child’s outcomes). Lastly, in some cases, the researchers did not consider whether policies around sibling placement may themselves lead to differences in outcomes. For example, children separated from their siblings may be moved (experience a placement change) once a provider who is able to accommodate all siblings is identified (Font et al., 2018) so higher rates of instability among separated siblings could in part reflect efforts to achieve joint placement.

Not only are there serious limitations to the samples, methods, and measures of the existing research, the quantity of research is also lacking, especially on permanency and child wellbeing. Only two of the five studies on permanency consider permanency outcomes other than reunification. In addition, we identified only nine longitudinal studies addressing sibling placement and child wellbeing. Policy may continue to favor maintaining sibling placement on the basis of moral principle; however, we cannot conclude from nine observational studies (four of which used the same two datasets) whether placing children together promotes child wellbeing. Moreover, findings varied across the nine studies and across models within each study; thus, a generalized conclusion about the effects of sibling placement on wellbeing remains elusive.

Yet, issues of research generalizability, measurement, and omitted variable bias are not well-represented in the literature reviews. The majority of reviews focused on measurement issues such as how to define a sibling or sibling placement (F. Ainsworth & Maluccio, 2002; Children’s Bureau, 2013; Groza et al., 2003; Hegar, 2005; Jones, 2016; Meakings et al., 2017; Shlonsky et al., 2005; Washington, 2007), and fewer discussed issues of generalizability and omitted variable bias (Hegar, 2005; McCormick, 2010; Shlonsky et al., 2005; Waid, 2014). Review articles typically did not discuss the limitations of state or county level administrative data, or the extent to which studies from outside the United States could be generalized to children in the U.S. foster care system, despite the fact that 14 of the 27 longitudinal quantitative studies that we identified used state or county data and ten used data from non-U.S. samples. Due to significant differences across countries, states, and counties with regards to systems, policies, and foster care population composition, generalizability cannot be assumed. In addition, many studies relied on small or highly selective samples, which can produce biased estimates and underpowered analyses. Although there is no standard definition of a “small sample”, foster children have a variety of characteristics, experiences, and outcomes. Where there is substantial population heterogeneity, larger sample sizes are needed to produce precise estimates of association. Several studies included in this review reported imprecise estimates (large confidence intervals) or were unable to include a robust set of covariates due to sample size. In addition, all but eleven studies restricted their samples based on child age (see Tables 3–6). Overall, studies using multivariate analysis did not adequately consider sibling group effects. Sixteen studies had data on a focal child with no or limited information about the siblings or used sibling dyads (Barth et al., 2007; Fernandez & Lee, 2013; Hegar & Rosenthal, 2009, 2011; Holtan et al., 2013; Koh et al., 2014; Kothari et al., 2018b; Leathers, 2005, 2006; Leathers et al., 2019; Linares et al., 2015, 2015; Thorpe & Swart, 1992; van Rooij et al., 2015; Vinnerljung et al., 2017; Waid et al., 2016, 2017). Four studies observed all of the siblings in care but did not adjust their estimates for interdependency or account for sibling characteristics – that is, control variables were at the individual level only (Akin, 2011; Font et al., 2018; Sattler et al., 2018; Webster et al., 2005).

In addition, review articles paid scant attention to measurement bias: specifically, the source of information on outcome measures. For example, the findings of Hegar and Rosenthal (2009, 2011) suggest that the associations of sibling placement with behavioral outcomes may depend on who is reporting the behavior. Although caregiver reports of children’s behavior are widely used in research, the validity of these measures for children in foster care is less established, especially for short-term placements (Tarren-Sweeney et al., 2004). Nevertheless, differences in child behavior were typically statistically non-significant when behavior was reported by foster parents, youth, or teachers.

Many studies, due to small sample size or reliance on administrative data, were not able to adjust for a wide range of factors that may distort estimates of sibling placement effects. First, as shown in our review of qualitative studies, siblings are separated in care for a variety of reasons—foster parents and caseworks may make a conscientious decision to separate siblings due to toxic sibling relationships or considering special needs of individual siblings. Yet, no quantitative study accounted for why siblings were separated as part of a multivariate regression (likely because such information is unavailable or poorly documented), so it is possible that negative associations between sibling separation and child outcomes are driven by the experiences or events that led to sibling separation.

An additional concern is the heterogeneity of effects. That studies identify positive, negative, and null effects of sibling placement suggest that some children benefit from being placed with their siblings, whereas others may be harmed. Sibling relationship may be among the most important moderators. Linares and colleagues (2007) found that when the sibling relationship was strong and positive, sibling placement may promote child wellbeing, but where the relationship was negative, joint placement was associated with more negative outcomes. Although such findings may seem obvious or expected, it is not clear that the quality of the sibling relationship (aside from any evidence of inter-sibling abuse) is consistently evaluated as part of the placement decision. Other potential moderators warrant attention from researchers, such as children’s level of need (e.g., disability, degree of emotional or behavioral problems), the number of siblings in the group, and the skills and capacities of the foster caregivers (e.g., the ability of caregivers to manage and reduce negative/harmful sibling interactions).

Lastly, we caution that the effects of sibling separation may depend on how children exit foster care. For children who are reunified or exit to permanency together, temporary separation may have temporary effects. However, for children who age out of care or are adopted by different families, the sibling relationship may be permanently weakened or severed. Additional research that accounts for the duration of separation and the degree of contact during separation is needed. Moreover, sibling relationships may be especially important for children who age out of foster care, as such relationships may compensate for inconsistent or impermanent caregivers.

Limitations

This study did not include unpublished work because it is sometimes preliminary or incomplete, and not subject to peer review. Due to publication bias, our exclusion of unpublished work may underrepresent null research findings relative to positive or negative research findings.

Second, although we emphasize key limitations of studies’ samples, measurements, and methods we did not use a standardized tool (e.g., Robin-I; Sterne et al., 2016) to evaluate risk of bias. Third, although we limited our time frame to 1990 onward for our search criteria, foster care policy and practice have changed significantly since 1990 and thus findings of older studies may not apply to today’s youth. Fourth, we limit our review of quantitative research to longitudinal studies. Many of the longitudinal studies still had significant limitations, with many lacking any baseline measures that would adjust for differential selection into sibling placement. In such cases, such studies may not be more rigorous than some of the cross-sectional studies that were excluded.

Lastly, we do not use meta-analytic methods to compute average effects. Although meta-analyses are powerful tools for aggregating and averaging study findings, they may not produce reliable results when based on heterogenous and small studies or include very few studies (Walker, Hernandez, & Kattan, 2008). In some outcome areas, such as permanency, there were simply too few studies to draw meaningful conclusions. In other outcome areas, such as wellbeing, the domains measured varied widely, with few studies in any one domain. In the area of placement instability, there were multiple studies using similar outcome measures, meta-analyses have already been conducted (e.g., Konjin et al., 2019). Yet, we questioned whether there is adequate data to draw conclusions even in more widely studied areas, like placement instability. Of the studies reviewed in this article, a sizable proportion reported only bivariate comparisons or results from interaction models, making effect sizes incomparable. Similarly, studies measured sibling placement in a range of ways, resulting in different categories that cannot be readily compared.

Implications

Based on our review, we suggest three considerations for future research and policy. First, we suggest making it easier to track siblings using administrative data such as the Adoption and Foster Care Analysis and Reporting System (AFCARS). This will provide researchers the opportunity to better understand the dynamics of sibling groups and how those may moderate associations between sibling placements and child outcomes. Identifying the sources of heterogeneity in sibling placement effects will provide agencies guidance for how to assess the risks and benefits of sibling placement, and to provide support and monitoring to sibling placements where children are at risk for adverse outcomes. In addition, larger numbers of observations available for research is critical to improving the state of knowledge. Many of the studies reviewed in this article reported coefficients that were modest or large in magnitude but not statistically significant. The imprecision in studies’ estimates due to sample size or heterogeneity may result in false negatives, where no clear conclusion can be drawn. Consequently, our review can largely conclude only that there is no consistent evidence of an effect of sibling placement. We cannot conclude, however, that there is consistent evidence of no effect of sibling placement. Improvements to AFCARS to facilitate the study of sibling groups would contribute substantially to resolving this question.

Second, pursuant to the Fostering Connections to Success and Increasing Adoptions Act of 2008, agencies must now document reasons for sibling separation. A requirement to incorporate this documentation as a categorical data field in state and federal administrative databases would allow for estimates of how often sibling separation is necessary for safety or wellbeing reasons, and how often separation occurs due to difficulty identifying placements that will accept sibling groups. Moreover, such data collection would allow for better estimates of the effects of unnecessary sibling separation. Similarly, agencies are allowed to make exceptions to standard licensure rules about the amount of space the home must have or the number of children that can be placed in a single foster home can care, in order to accommodate joint placement of sibling groups. Yet, because crowded housing can induce stress and negatively affect psychological wellbeing for children and caregivers (Evans et al., 2003; Leventhal & Newman, 2010), household size or space-related exemptions could lower the quality of care environment that children receive. Most concerning, children with relatively high levels of emotional or behavioral needs may receive inadequate individualized attention in an environment where caregivers’ capacities are stretched thin. Future research must consider the potential for unintended consequences and whether any trade-offs needed to facilitate a sibling placement still produces a net benefit for the children.

Third, current policy requires reasonable efforts to be made to maintain sibling relationships unless it is not in the best interest of a child. Yet, there is no uniformly accepted definition of best interests (Administration for Children and Families, 2010) nor is it clear whether best interests are evaluated before or while siblings are placed together. A standard definition of best interests, and a verifiable process for evaluating best interests, would introduce consistency and accountability into the decision to place siblings together or apart. At the very least, creating valid and reliable measures of child well-being would help to inform best interest practices (Font & Gershoff, 2020). As with all decision-making for children in foster care, substantial consideration should be given to children’s own preferences and opinions.