social work

Children and Youth Services Review 58 (2015) 227–235

Contents lists available at ScienceDirect

Children and Youth Services Review

journal homepage: www.elsevier.com/locate/childyouth

Functioning patterns among older adolescents in foster care: Results from a cluster analysis

Svetlana Shpiegel a,⁎, Kerrie Ocasio b a Robert D. McCormick Center for Child Advocacy and Policy, Montclair State University, 1 Normal Ave., Dickson Hall, Room 370, Montclair, NJ 07043, USA b Institute for Families, School of Social Work, Rutgers University, 55 Commercial Ave., New Brunswick, NJ 08901, USA

⁎ Corresponding author. E-mail addresses: shpiegels@mail.montclair.edu (S. Sh

kocasio@ssw.rutgers.edu (K. Ocasio).

http://dx.doi.org/10.1016/j.childyouth.2015.09.024 0190-7409/© 2015 Elsevier Ltd. All rights reserved.

a b s t r a c t

a r t i c l e i n f o

Article history: Received 9 June 2015 Received in revised form 28 September 2015 Accepted 28 September 2015

Keywords: Foster care Adolescents Cluster analysis National Youth in Transition Database Aging-out

Older adolescents in foster care represent a heterogeneous population, though such heterogeneity is often underemphasized in research and practice. This study employed a cluster analysis to identify subpopulations in a large, national sample of 17-year-old youth based on the following indicators: educational attainment, con- nection to a supportive adult, adolescent parenthood, homelessness, substance abuse referral and incarceration. Data from the National Youth in Transition Database (NYTD) and Adoption and Foster Care Analysis and Reporting System (AFCARS) were used in the analysis. Results revealed five subpopulations defined by specific strengths, vulnerabilities and child welfare experiences. The largest group identified (39%) functioned successful- ly in all domains, whereas an additional group (15%) exhibited consistent maladaptation. The remaining groups evidenced variable adaptation patterns, with strengths in some domains and challenges in others. Entry to foster care for reasons other than child's problem behaviors, and placement in stable, family-based settings were asso- ciated with belonging to the most adaptive group. Findings emphasize heterogeneity among older adolescents in foster care, and call for better design and targeting of child welfare services and programs as appropriate to the needs of specific subgroups.

© 2015 Elsevier Ltd. All rights reserved.

1. Introduction

Every year, about 25,000 youth emancipate from foster care in the United States after reaching the age of maturity (U.S. Department of Health and Human Services, 2014). This period comes with the expecta- tion that the youth are able to negotiate adult responsibilities and be- come self-sufficient (Keller, Cusick, & Courtney, 2007). Mastering such tasks, however, may be challenging for adolescents who abruptly tran- sition out of foster care and into young adulthood (Lemon, Hines, & Merdinger, 2005). Unlike counterparts in the general population, foster youth must negotiate this transition suddenly and with limited or no support from family members (Collins, Spencer & Ward, 2010; Keller et al., 2007; Stott, 2013). In addition, many are underprepared for as- suming adult roles in terms of educational completion, job readiness and basic skills needed for independent living (Courtney, 2009; Keller et al., 2007; Stott, 2013).

Given such disadvantage, it is not surprising that foster youth tend to struggle as they transition to independence (Courtney, 2009; Stott, 2013). Nearly 50% fail to obtain a high school diploma

piegel),

by the age of 18, only 30% enroll in higher education institutions and less than 10% complete a four-year degree (Brandford & English, 2004; Stott & Gustavsson, 2010; Yates & Grey, 2012). Many experience unemployment, underemployment and homelessness, and receive need-based government assistance (Courtney, 2009; Dworsky & Courtney, 2009; Hughes et al., 2008; Naccarato, Brophy, & Courtney, 2010; Stott & Gustavsson, 2010). In addition, foster youth exhibit higher rates of mental illness, substance abuse, teen pregnancy and criminal justice involvement compared to peers in the general population (e.g. Hughes et al., 2008; McMillen et al., 2005; Narendorf & McMillen, 2010; Svoboda, Shaw, Barth, & Bright, 2012).

Nevertheless, not all youth exhibit dysfunctional outcomes during this vulnerable time period. Some demonstrate relatively uncompromised, or “resilient”, functioning as they leave the child welfare system and begin to live on their own (e.g. Daining & DePanfilis, 2007; Hass & Graydon, 2009; Hines, Merdinger, & Wyatt, 2005; Jones, 2012; Samuels & Pryce, 2008). Others function successfully in domains such as education and employment, but struggle with mental health difficulties, low self-esteem and com- promised peer relationships (e.g. Keller et al., 2007; Yates & Grey, 2012). Overall, available evidence suggests that different subgroups may exist within this population, calling for a “nuanced” approach to research, policy and practice (Courtney, Hook, & Lee, 2012).

228 S. Shpiegel, K. Ocasio / Children and Youth Services Review 58 (2015) 227–235

1.1. A person-oriented approach to understanding the functioning of foster youth

Existing studies on the functioning of foster youth typically used a variable-oriented approach to analysis, which examines how certain predictor variables relate to outcomes in specific domains (Keller et al., 2007; Yates & Grey, 2012). While this approach is useful for iden- tifying the correlates of individual outcomes, it fails to capture the mul- tidimensional nature of youths' adaptation (Keller et al., 2007). As noted by Courtney et al. (2012), “an alternative to the variable-oriented ap- proach…is the person-oriented approach, which assumes that develop- ment cannot be understood by examining single factors in isolation from their relationships with other interacting factors” (p. 410). Person-oriented methods aim to detect meaningful subgroups in a given population who share similar characteristics and experiences in multiple domains. Identifying subgroups of foster youth characterized by specific strengths and vulnerabilities offers important implications for practice, including better design and targeting of child welfare services and programs (Courtney et al., 2012).

Several existing studies employed person-oriented methods to ex- amine the functioning of older youth in foster care. In a study by Keller et al. (2007), data from 17 and 18-year-olds residing in three Midwestern states were used to identify four subpopulations. The larg- est group identified, “distressed and disconnected”, represented about 43% of the sample and included youth with high rates of behavioral problems and non-optimal employment and education outcomes. The second largest group, “competent and connected”, represented about 38% of the sample and included youth with positive education and em- ployment experiences and no significant problem behaviors. The last two groups, “struggling but staying” and “hindered and homebound”, presented variable adaptation patterns, with strengths in some domains and challenges in others. The authors concluded that identifying sub- groups who share similar characteristics holds promise for improving service delivery to this population.

In another study, Yates and Grey (2012) identified four profiles of functioning among emancipated foster youth in California between the ages of 17 and 21. The outcome domains included educational and vocational competence, civic engagement, inter- personal relationships, self-esteem and mental health. The largest group identified (47%) presented a “resilient” profile, fairing rea- sonably well in all domains. Two other groups exhibited “discor- dant” patterns of adjustment, where some youth demonstrated psychological health despite functional difficulties (“internally resilient”, 30%), while others presented emotional problems despite apparent functional competence (“externally resilient”, 6.5%). An additional group exhibited a “maladaptive” profile (16.5%) characterized by problematic functioning in all domains.

Several other studies also employed person-oriented methods to ex- plore the outcomes of current and former foster youth (e.g. Courtney et al., 2012; Yampolskya, Sharrock, Armstrong, Strozier, & Swanke, 2014). Most identified unique subpopulations requiring different levels of support and supervision on the part of child welfare officials. Taken together, existing evidence suggests that foster youth are a heteroge- neous population, pointing to the need for developing intervention strategies tailored specifically to different subgroups. Utilizing a “nuanced” approach to service delivery may enhance youths' motiva- tion and engagement in services and facilitate long-term competence (Courtney et al., 2012).

1.2. Factors influencing group membership

Although person-oriented studies consistently identified mean- ingful subgroups among older adolescents in foster care, they failed to identify a reliable set of factors differentiating between well- functioning youth and their more challenged peers. For instance, Keller et al. (2007) noted that members of the least adaptive group

in their sample reported increased child maltreatment, residence in non-family care arrangements and considerable placement instabil- ity. In contrast, Yates and Grey (2012) failed to detect similar differ- ences, noting that their subgroups were comparable with respect to child welfare experiences. The contribution of demographic factors was similarly inconsistent, with Keller et al. (2007) pointing to fe- male gender and African–American race as related to more adaptive profiles, but Yates and Grey (2012) finding no significant differences. Research is needed to understand how various subgroups differ from one another so that appropriate services can be provided to mitigate risks and facilitate competent functioning. Demographic factors and reasons for out-of-home placement need further investigation, as these represent pre-existing risks which may relate to variations in youths' functioning (e.g. Lee, Courtney, & Tajima, 2014; Stott, 2013). The contribution of post-removal factors – especially place- ment type and stability – should also be explored, as these are more amenable to intervention and can increase or mitigate existing risks (Newton, Litrownik, & Landsverk, 2000; Ryan & Testa, 2005; Rubin, O'Reilly, Luan, & Localio, 2007; Stott, 2013). Prior research re- vealed that residence in stable, family-based settings is associated with better functioning, regardless of youths' pre-exiting conditions (e.g. Barber & Delfabbro, 2003; Newton et al., 2000; Rubin et al., 2007).

1.3. The present study

The present study employed cluster analysis as a person- oriented method to identify distinctive profiles of functioning in a large, national sample of 17-year-old foster youth. Prior studies were based on relatively small samples confined to one or few states; therefore, utilizing a national sample is an important next step for the field. By limiting our study to 17-year-olds, we aimed to assess youths' functioning while they still had several years be- fore formal emancipation. As most states currently allow youth to remain in foster care until the age of 21, examining their function- ing at 17 provides a period of time for intervention to remediate risk and facilitate competent functioning.

This study included six outcome indicators relevant to youths' prospects for a successful transition to adulthood: educational at- tainment, connection with a supportive adult, teen parenthood, and a history of homelessness, substance abuse referral and incar- ceration. The domains selected considered the developmental tasks most relevant to this age group. Indicators related to employ- ment, independent living and economic self-sufficiency were ex- cluded, as they did not apply to the majority of 17-year-olds still under the care and supervision of child welfare agencies. In con- trast, educational attainment, connections to adults and avoidance of risky behaviors were deemed developmentally appropriate as markers of successful adaptation at this age. Following the identifi- cation of the clusters, we compared the obtained subgroups on var- ious child welfare factors which may relate to variations in youths' functioning.

The specific goals of the present study were to:

(1) Identify unique profiles of functioning in a large, national sample of 17-year-olds based on the outcome domains described above (i.e. education, connection to adult, childbirth, homelessness, substance abuse referral and incarceration).

(2) Examine whether the obtained clusters relate in meaningful ways to youths' pre-removal factors, including gender, race/eth- nicity, number of removal episodes, age at the most recent removal and reasons for out-of-home placement.

(3) Examine whether the obtained clusters relate in meaningful ways to youths' post-removal factors, including length of the current foster care episode and placement type and stability.

229S. Shpiegel, K. Ocasio / Children and Youth Services Review 58 (2015) 227–235

2. Methods

2.1. Dataset and procedure

This research is based on a secondary analysis of data from the National Youth in Transition Database (NYTD). Created by the John H. Chafee Foster Care Independence Program (CFCIP), NYTD is designed to (1) track various services provided through CFCIP; and (2) collect certain outcome measures to assess the effective- ness of the program. All 50 states, the District of Columbia and Puerto Rico are required to submit information to NYTD during the designated reporting periods (NDACAN, 2014).

The present investigation focused solely on the outcome component of NYTD, which included information on all youth who were in foster care at age 17, examining their educational, vocational, and general well-being indicators during the period of transition to adulthood. States are required to collect three phases of outcome data for every co- hort of youth – a baseline survey during the year in which they turn 17, and two follow-up surveys when they turn 19 and 21.

The present study analyzed baseline data from the first cohort of youth established in federal fiscal year (FY) 2011 (N = 15,601). All youth who reached their 17th birthday in FY2011, and were in foster care within a 45-day period beginning on their birthday, were eligible to complete the outcome survey. States could choose to administer the survey in person, via the Internet or over the phone, provided that it was administered to the youth directly. Youth participation was vol- untary, with freedom to refuse without adverse consequences, or to de- cline to answer specific survey questions. Those youth who at least partially completed the survey during the designated 45-day window were included in FY2011 cohort. The national response rate for the sur- vey was 53%, ranging from 12% in Arizona to 100% in Rhode Island and Vermont. A weighting procedure was implemented to correct potential non-response bias. For detailed information about the NYTD weighting procedures, see NDACAN (2014).

To obtain information about youths' demographics and child welfare histories, NYTD data were combined with data from the Adoption and Foster Care Analysis and Reporting System (AFCARS) for FY2011. AFCARS is a federally mandated data collection system that provides case level information on all children for whom the state child welfare agencies have responsibility for placement and supervision, as well as on children who are adopted under the auspices of the state's public child welfare agency. Data includes demographic information on chil- dren and caregivers, and episode-level information, such as removal reasons, placement types, and number of previous placements. All states are required to submit information to AFCARS on a semi-annual basis (NDACAN, 2013a).

2.2. Sample

The final sample for the present study consisted of all youth in the NYTD FY2011 cohort for whom valid information on child welfare variables was available in AFCARS. Youth from all states were represent- ed in the final sample, with the exception of Connecticut, which was excluded due to incompatibilities in the format of the child's unique identifier, preventing the merge of NYTD and AFCARS datasets. In addi- tion, youth with missing information on any of the six indicators used to form the clusters were excluded, as the clustering method employed did not permit missing data. The final sample consisted of 14,402 youth (92% of the NYTD FY2011 cohort) – 6732 males and 6600 females. The majority of the youth were White (N = 7421), followed by African Americans (N = 4141), American–Indians or Alaska Natives (N = 260), Asian (N = 118), Native Hawaiian or Other Pacific Islander (N = 28) and multiracial (N = 1700). In addition, 2230 youth, irrespective of race, identified as Hispanic or Latino.1

1 Unweighted count is presented here.

2.3. Measures

Three sets of variables were included in the analysis: (1) outcome in- dicators used to form clusters of functioning; (2) pre-removal factors; and (3) post-removal factors. Information about outcome indicators has been obtained from the NYTD dataset; information about pre- removal and post-removal factors has been obtained from AFACRS. As with other large administrative datasets, missing data were present for several variables, resulting in a modest decrease in sample size for some analyses.

2.3.1. Outcome indicators Six outcome indicators were used to identify patterns of functioning

among the participating youth. Each indicator was coded as (0) absent; or (1) present.

2.3.1.1. Current school enrollment. Current school enrollment was de- fined as attending high school, GED classes, post-secondary vocational training, or college at the time of the interview.

2.3.1.2. Connection to adult. Participants were asked if they knew at least one adult who they can go to for advice or guidance when there is a decision to make or a problem to solve, or for companionship when celebrating personal achievements. This could include, but was not lim- ited to, adult relatives, parents and foster parents; however, it excluded spouses, partners, boyfriends or girlfriends and current caseworkers.

2.3.1.3. Childbirth. To determine childbirth status, participants were asked if they had ever given birth or fathered a child that was born.

2.3.1.4. Homelessness. Homelessness was assessed by asking if the youth ever had no regular or adequate place to live, such as living in a car, on the street, or staying in a homeless or other temporary shelter.

2.3.1.5. Substance abuse referral. Youth were asked if they had ever been referred for an alcohol or drug abuse assessment or counseling, includ- ing either a self-referral or a referral by a social worker, school staff, phy- sician, mental health worker, foster parent or another adult.

2.3.1.6. Incarceration. Participants were asked if they had ever been confined in a jail, prison, a correctional facility, or juvenile or community detention facility in connection with allegedly committing a crime (a felony or a misdemeanor).

2.3.2. Pre-removal factors These factors included youths' demographics and reasons for out-

of-home placement. The demographic variables were gender, race and ethnicity, as well as number of removal episodes and youths' age at most recent removal. Youth reported their gender as either male or female. Ethnic identity was defined as Hispanic or non- Hispanic, and youth reported their race as White, Black/African American, American–Indian/Alaska Native, Asian, and Native Hawai- ian/Other Pacific Islander. An additional category (“multiracial”) was created to represent youth reporting two or more racial categories. Furthermore, all minority race categories (including “multiracial”) were combined into one category labeled “non-White” for use in some analyses. Youths' age at most recent removal was measured in years, and total number of removal episodes was measured continuously.

The AFCARS dataset includes fifteen reasons for out-of-home place- ment: physical abuse, sexual abuse, neglect, parental alcohol abuse, parental drug abuse, child's alcohol abuse, child's drug abuse, child's disability, child's behavioral problems, parental death, parental incar- ceration, inability to cope, abandonment, relinquishment and inade- quate housing. Each reason was coded as (0) absent, or (1) present,

230 S. Shpiegel, K. Ocasio / Children and Youth Services Review 58 (2015) 227–235

and more than one could be recorded for each youth. For a detailed description of each removal category, see NDACAN (2013b).

2.3.3. Post-removal factors Post-removal factors included duration of the current foster care ep-

isode, number of placement settings during this episode, current place- ment type and the duration of the current placement. Duration of the current foster care episode, as well as current placement, were mea- sured in days. Placement type was coded as: (1) relative foster home, (2) non-relative foster home, (3) group home or institution, and (4) other setting (i.e. pre-adoptive home, supervised independent living, trial home visit or runaway). Number of placement settings during the current episode was measured continuously.

2.4. Analytic strategy

Data analysis was conducted in several steps. First, a two-step cluster analysis was performed to organize youths' outcomes (i.e. education, connection to adult, childbirth, homelessness, substance abuse referral and incarceration) into mutually exclusive groups. The two-step cluster method is preferred for very large datasets, and is appropriate for cate- gorical variables (Fava et al., 2012; Tsai, Edens, & Rosenheck, 2011). The number of clusters was established by the two-step algorithm, though we imposed a maximum of seven clusters to optimize results' interpret- ability and utility in subsequent analyses. The log likelihood distance measure was used to determine cluster membership; Bayesian informa- tion criterion (BIC) as well as clusters' interpretability considerations were used to judge the adequacy of the final solution.

Following the identification of the clusters, descriptive bivariate analyses (i.e. one way ANOVA tests and chi-square tests) were used to examine the associations between pre-removal and post-removal fac- tors and youths' cluster membership. Weights were incorporated in all analyses to produce national estimates for the full NYTD baseline popu- lation (i.e. all 17-year-olds in foster care). To reduce the likelihood of committing a Type 1 error due to multiple comparisons, the Bonferroni correction was applied and p value of less than .01 was used. Analyses were conducted with SPSS 21 Complex Samples software.

3. Results

3.1. Cluster analysis

The two-step cluster analysis produced five clusters of functioning. The frequencies of the domains that formed the clusters are presented in Table 1. The largest of the obtained clusters (39%) was labeled resil- ience, and characterized by positive functioning in all domains. Youths in this cluster were enrolled in school, had a supportive adult, and did not have any of the risk behaviors studied (i.e. early parenthood, home- lessness, substance abuse referral, or incarceration). The second cluster (19%), labeled substance abuse, was characterized by substance abuse referrals for all members, as well as incarceration histories for some. The third cluster (15%), labeled multiple problems, was characterized

Table 1 Functioning indicators by cluster (N = 14,402).

Variable Resilient (N = 5778^)

Substance abuse (N = 2604^)

In school 100% 100% Supportive adult 100% 100% Has children 0% 0% Ever homeless 0% 0% Substance abuse referral 0% 100% Incarceration 0% 58.1%

Note: The number in the table represents the percentage of cases with the targeted outcome o ^Unweighted count. Homeless. = homelessness.

by compromised functioning in all domains. Nearly half of its members had children, one-third did not have a supportive adult, over 30% were not enrolled in school, 20% had been homeless, 45% had been incarcer- ated and about 30% reported a substance abuse referral. The fourth clus- ter (14%), labeled incarceration only, included youth who had been incarcerated; however, all were enrolled in school, had a supportive adult and did not have any other risks. Lastly, the smallest cluster (13%), labeled homelessness, was characterized by homelessness histo- ries for all members, as well as substance abuse and incarceration histo- ries for some.

3.2. Pre-removal factors

3.2.1. Demographics In the overall sample, 51% of youths were male, 53% were White and

20% were Hispanic. Differences by cluster are summarized in Table 2. Significant gender differences emerged between the groups (χ2 = 485.51, p b .001), such that multiple problems and resilience clusters were characterized by increased number of females, whereas incarcera- tion only and substance abuse clusters were characterized by increased number of males. Significant differences by race (χ2 = 128.34, p b .001) and ethnicity (χ2 = 32.13, p b .01) were also noted, such that substance abuse and homelessness clusters were characterized by in- creased number of Whites, whereas multiple problems cluster was char- acterized by increased number of non-Whites and Hispanics.

On average, participants were about 13.5 years-old during the latest removal episode, though differences by cluster were noted (F = 109.29, p b .001). Youths in resilience cluster were somewhat younger during the latest removal (M = 12.73), while those in substance abuse cluster were older (M = 14.40). The average number of removal episodes for the overall sample was 1.49. Cluster differences were significant (F = 5.26, p b .01), however, they were relatively small.

3.2.2. Reasons for removal Cluster differences in removal reasons are summarized in Table 3.

The most commonly reported reason for removal in the overall sample was neglect (41%), followed by a child's behavioral problem (38%), and caretaker's inability to cope (23%). Cluster differences were significant for the majority of reasons, with the exception of parental alcohol abuse, imprisonment, relinquishment and death. Youths in resilience cluster presented higher rates of physical abuse, sexual abuse and ne- glect, but lower rates of most child-related factors, with the exception of disability. Youths in the substance abuse cluster presented an opposite pattern, with decreased rates of child abuse and neglect but increased rates of child-related factors. Youths in the homelessness cluster were noted for relatively high rates of inadequate housing, parental substance abuse and parental inability to cope, while those in the incarceration only cluster were noted for high rates of child disabilities and problem behaviors. Interestingly, youths in the multiple problems cluster present- ed average rates of both child-related and parent-related reasons for out-of-home placement.

Multiple problems (N = 2178^)

Incarceration only (N = 1957^)

Homeless. (N = 1885^)

65.9% 100% 100% 68.4% 100% 100% 44.6% 0% 0% 21.7% 0% 100% 31.4% 0% 38.8% 45.0% 100% 43.7%

ut of all cases in that cluster.

Table 2 Demographic characteristics by cluster^.

Demographic Resilient Substance abuse Multiple problems Incarceration only Homeless. X2 or F

Gender 485.5** Male 44.4% 60.2% 41.7% 69.0% 50.6% Female 55.6% 39.8% 58.3% 31.0% 49.4%

N^^=13,332 Race 128.3**

White 52.0% 57.7% 45.4% 49.5% 61.0% Non-white 48.0% 42.3% 54.6% 50.5% 39.0%

N^^=13,668 Ethnicity 32.1*

Non-Hispanic 81.2% 80.8% 76.4% 82.1% 77.6% Hispanic 18.8% 19.2% 23.6% 17.9% 22.4%

N^^=12,730 Age at last removal 109.29**

M 12.7 14.4 14.1 13.8 13.9 (SE) (.05) (.06) (.07) (.07) (.07)

N^^=13,331 Number of removals 5.26*

M 1.45 1.51 1.53 1.53 1.52 (SE) (.01) (.02) (.02) (.02) (.02)

N^^=13,325

Notes: ^Weighted analysis. ^^ Unweighted count. Homeless. = homelessness. *p b .01; **p b .001.

231S. Shpiegel, K. Ocasio / Children and Youth Services Review 58 (2015) 227–235

When examining all reasons for removal, more than two-thirds (73%) of youths in the resilience cluster were removed exclusively due to parent-related difficulties (i.e. physical abuse, sexual abuse, neglect, parental alcohol or drug use, inability to cope, inadequate housing, imprisonment, abandonment, relinquishment and death), while just 11% were removed solely due to child-related factors (i.e. child's alcohol or drug use, disability and behavioral problem). In contrast, in the substance abuse and incarceration only clusters, 42% and 35% respectively were removed exclusively due to child- related factors, while only about 40% were removed solely due to parent-related difficulties. In the multiple problems and homelessness clusters, the majority of youths (59% and 63% respectively) were re- moved exclusively due to parent-related difficulties, though a sizable portion (about 18%) was removed due to a combination of parent- related and child-related factors.

Table 3 Removal reasons by cluster^ (N = 13,127).

Removal reason~ Resilient Substance abuse Multip

Physical abuse 17.5% 8.2% 11.8% Sexual abuse 10.9% 5.2% 7.2% Neglect 48.1% 29.6% 40.8% Parent alcohol abuse 6.1% 4.8% 4.6% Parent drug abuse 13.7% 10.2% 12.9% Child alcohol abuse .8% 3.9% 2.1% Child drug abuse 1.1% 9.6% 5.1% Child disability 4.9% 3.2% 2.9% Child behavior 23.7% 57.5% 38.1% Parent death 1.5% 1.1% 1.1% Parent incarceration 4.4% 2.9% 4.2% Caregiver coping 24.0% 17.9% 20.8% Abandonment 8.8% 6.6% 10.7% Relinquishment 2.2% 1.7% 1.8% Inadequate housing 7.9% 3.4% 7.5% Primary reason(s)

Parent only 73.2% 38.7% 58.9% Child only 11.7% 42.0% 23.2%

Parent and child 15.1% 19.3% 17.9%

Notes: ^Weighted analysis. ~More than one removal reason can be listed for each child. Homeless. = homelessness. **p b .001.

3.3. Post-removal factors

At the next step, youths' post-removal factors were examined, in- cluding placement type, time spent in current placement, number of placements during the current spell in foster care, and duration of the current foster care episode. In the overall sample, 38% of youths resided in non-relative foster homes, 36% resided in group homes or institu- tions, 10% lived with relatives and 16% were placed in other settings. The average time spent in current placement was 367 days, whereas av- erage duration of the current foster care episode was 1214 days. Place- ment instability was high for all participants, with an average of 5.42 placements since the latest removal.

Cluster differences in placement characteristics are summarized in Table 4. Significant differences emerged for current placement type (χ2 = 1027.26, p b .001), such that members of the resilience cluster

le problems Incarceration only Homeless X2

10.1% 11.9% 156.6** 7.3% 7.0% 87.5**

32.3% 44.4% 304.5** 4.5% 6.4% 15.1

10.1% 14.0% 32.4⁎⁎

.8% 3.4% 119.4** 2.3% 5.4% 332.1** 5.8% 3.6% 34.6**

57.1% 34.3% 1117.5** 1.0% 1.3% 4.7 4.0% 4.7% 11.8

22.2% 27.7% 67.6** 8.9% 10.8% 33.0** 1.3% 1.6% 7.7 5.1% 9.7% 86.3**

1304.2** 40.5% 69.9% 35.4% 18.7% 24.1% 18.4%

Table 4 Placement characteristics by cluster ^.

Variable Resilient Substance abuse Multiple problems Incarceration only Homeless. X2 or F

Current placement 1027.2** Relative home 15.0% 6.1% 8.5% 5.9% 10.1% Non-relative home 48.3% 26.0% 33.4% 26.8% 40.0% Residential 24.2% 50.0% 39.0% 51.9% 33.3% Other 12.6% 17.8% 19.2% 15.4% 16.6%

N^^=13,257 Number of placements 16.5**

M 5.0 5.0 5.9 6.1 5.7 (SE) (.07) (.12) (.16) (.17) (.14)

N^^=13,321 Days in placement 109.4**

M 529.3 238.1 259.9 258.2 299.8 (SE) (12.1) (8.7) (10.4) (9.9) (10.1)

N^^=13,031 Days in care 109.7**

M 1525.7 902.4 996.4 1113.1 1092.5 (SE) (21.7) (24.2) (27.9) (29.7) (26.8)

N^^=13,331

Notes: ^Weighted analysis. ^^ Unweighted count. Homeless. = homelessness. Residential = group homes and/or institutions. **p b .001.

232 S. Shpiegel, K. Ocasio / Children and Youth Services Review 58 (2015) 227–235

were likely to reside in relative or non-relative foster homes, while members of the substance abuse and incarceration only clusters were likely to reside in group homes and institutions. Youths in the multiple problems and homelessness clusters resided mostly in non-relative foster homes, as well as group homes and institutions.

Significant differences also emerged in the time spent in current placement, as well as the number of placements since the latest removal (F = 109.47, p b 001 and F = 16.52, p b 001). Youths in resilience cluster spent more time in current placement (M = 529.39 days) and had less placement settings (M = 5.03) compared to members of the other groups. In contrast, youths in the substance abuse cluster spent less time in current placement (M = 238.16 days), and those in the incarcer- ation only cluster had more placement settings (M = 6.12) compared to others. Duration of the current foster care episode also differed signifi- cantly (F = 109.75, p b .001), with members of the resilience cluster spending more time in care (M = 1525.79 days), and members of the substance abuse cluster spending less time (M = 902.49 days). The re- maining clusters differed little in the duration and stability of place- ments, as well as duration of the current foster care episode.

4. Discussion

The goal of the present study was to identify patterns of function- ing among older adolescents in foster care, and examine the factors associated with such patterns. Findings revealed five subpopulations characterized by specific strengths, vulnerabilities and child welfare experiences. Each subpopulation may have different prospects for a successful transition to adulthood, requiring different levels of sup- port and supervision on the part of child welfare staff.

4.1. The “resilience” cluster

The largest cluster in the present sample (39%) was characterized by competent, or “resilient”, functioning in all domains. Adolescents in this cluster were enrolled in school, had supportive adults, and avoided problematic outcomes, such as homelessness, teen parenthood and incarceration. This pattern is consistent with several prior studies indi- cating that foster youth may function successfully as they transition to independence (e.g. Daining & DePanfilis, 2007; Jones, 2012; Keller et al., 2007; Yates & Grey, 2012). Person-oriented research consistently identified well-functioning subpopulations among these youth (e.g. Keller et al., 2007; Yates & Grey, 2012), and variable-oriented studies

also corroborated such findings (e.g. Daining & DePanfilis, 2007; Jones, 2012). Overall, existing evidence suggests that resilient functioning may be quite common among young people leaving foster care, supporting the view of resilience as developing “via the operation of normal developmental processes…rather than from exceptional indi- vidual capacities” (Yates & Grey, 2012, p. 476).

Members of the resilience cluster were more likely to be female, though differences were relatively small. The observed pattern is con- sistent with prior studies reporting better outcomes for female foster youth (e.g. Daining & DePanfilis, 2007; Keller et al., 2007). Nevertheless, some of the indicators used in this analysis are inherently less prevalent among females (e.g. incarceration), which may bias the study findings. In fact, in studies that include internalizing, as well as externalizing, markers of successful adaptation, gender differences tend to be less pro- nounced (e.g. Yates & Grey, 2012).

Youths in the resilience cluster were younger during the latest out- of-home placement, and spent substantially more time in foster care compared to other youth. Their reasons for removal were also some- what different, with higher rates of caregiver-related factors (e.g. child maltreatment) and lower rates of child-related difficulties (e.g. behav- ioral problems). Many were placed in relative or non-relative foster homes and spent nearly twice as much time in current placement com- pared to members of the other groups.

Overall, findings suggest that young people who are removed in ear- lier stages of adolescence, for reasons other than severe externalizing problems, and who are placed in relatively stable, family-based settings, exhibit better functioning at age 17. Although these patterns are strictly correlational, prior research shows that children and adolescents placed in family-based settings fare better than those placed in residential care (e.g. Cusick, Courtney, Havlicek, & Hess, 2011). The stability of place- ments also promotes competent functioning (e.g. Stott, 2012), and sta- ble placements are easier to find for children without severe emotional and behavioral problems (e.g. Ryan & Testa, 2005; Stott, 2012).

4.2. The “substance abuse” cluster

The second largest cluster in the present sample (19%) was charac- terized by substance abuse referrals for all members, as well as incarcer- ation histories for some. Nevertheless, all youth were enrolled in school, had a supportive adult, and did not have histories of homelessness or adolescent parenthood. The emergence of this cluster is supported by prior research, reporting elevated rates of substance abuse among

233S. Shpiegel, K. Ocasio / Children and Youth Services Review 58 (2015) 227–235

older adolescents in foster care (e.g. McDonald, Mariscal, Yan, & Brook, 2014; Narendorf & McMillen, 2010). In addition, two person-oriented studies indicated that subgroups characterized by substance use also tended to include delinquent behaviors (Keller et al., 2007; Yates & Grey, 2012).

The substance abuse cluster was characterized by increased number of male and White youth, consistent with prior research reporting sim- ilar trends (e.g. Aarons et al., 2008; Vaughn, Ollie, McMillen, Scott, & Munson, 2007). Youths in this cluster were also older during the latest out-of-home placement and spent less time in foster care compared to members of the other groups. Many youth were placed in group homes or institutions, possibly due to difficulties related to substance abuse behaviors. Furthermore, given that over 40% were placed in out- of-home care exclusively due to child-related difficulties, many may have entered foster care with unmet behavioral and/or emotional needs, making it harder to find stable, family-based placements (Stott, 2012).

Noteworthy, all youth in this cluster were enrolled in school and had a supportive adult – these are important strengths which can be built upon when designing intervention strategies. Timely permanency planning that incorporates supportive adults, as well as school-based programming to increase independent living preparedness, can be par- ticularly beneficent.

4.3. The “multiple problems” cluster

The multiple problems cluster was characterized by a constellation of difficulties considered “typical” for emancipating foster youth, though it represented merely 15% of the sample. About one-third of these youth were not enrolled in school, a similar proportion had no supportive adults, and many reported homelessness, substance abuse referrals and incarceration. Furthermore, nearly 50% had at least one child. These findings are consistent with both variable-oriented and person- oriented studies, reporting an array of difficulties presented by some foster youth (e.g. Courtney, 2009; Keller et al., 2007; Yates & Grey, 2012). In the study by Keller et al. (2007), such youth accounted for over 40% of the sample, though in Yates and Grey's study (2012), they represented merely 17%. These differences may at least partially relate to variations in methodology, such as sample size, recruitment strate- gies, and the selection of specific outcomes included in the analysis.

Members of the current cluster were somewhat more likely to be fe- male, non-White and Hispanic. Given their high rates of childbirth and the established link between minority status and teen parenthood (Leathers & Testa, 2006), such patterns are aligned with existing litera- ture. African–American and Hispanic foster youth, in particular, were found to have high rates of pregnancy and childbirth (e.g. King, Putnam-Hornstein, Cederbaum, & Needell, 2014), possibly driving the increased representation of minorities in this cluster. Furthermore, mi- nority youths may have higher likelihood of placement in residential settings, which can impede engagement in age-appropriate tasks (e.g. attending school, developing supportive relationships), and increase the likelihood of problematic behaviors (e.g. criminal involvement) (Courtney, 2009; Cusick et al., 2011; Stott, 2012).

Youths were placed in out-of-home care for a variety of reasons, including both parent-related and child-related difficulties. At the time of data collection, nearly 40% were placed in group homes or in- stitutions, one-third resided in non-relative foster homes and less than 10% lived with relatives. The difficulties presented by these youth are noteworthy, especially as many of them may parent minor children. It is unknown to what extent having children con- tributed to challenges presented by these youth, though prior re- search identified teen parenthood as a risk factor for behavioral/ emotional difficulties, educational underachievement and poverty (Barnet, Liu, & DeVoe, 2008; Boden, Fergusson, & Horwood, 2008). However, other research suggested that parenthood may be a posi- tive factor in the lives of foster youth, increasing their motivation

for education and employment, and decreasing engagement in vari- ous problematic behaviors (e.g. Chase, Maxwell, Knight, & Aggleton, 2006). Overall, some youths in the multiple problems cluster may re- quire extensive supports as they transition to independent adult- hood, while others may need more targeted services designed to support adequate parenting, as well as economic independence.

4.4. The “incarceration only” cluster

The fourth cluster, incarceration only, represented 14% of the sample. Youths in this cluster were enrolled in school, had a supportive adult and did not report childbirth, homelessness or substance abuse, howev- er, all had been incarcerated at some point in their lives. Consistent with existing research (e.g. Cusick, Havlicek, & Courtney, 2012), nearly 70% were male, with race and ethnicity resembling the overall sample. Age of removal also matched the overall sample; however, many youth were removed exclusively due to child-related difficulties. Furthermore, over 50% were placed in group homes or institutions and placement in- stability was relatively high. This may be the result of movement in and out of correctional facilities, or due to problem behaviors which made it difficult to find stable arrangements (Stott, 2012). During the period of transition to adulthood, this group is likely to require interventions spe- cifically targeting emotional and/or behavioral difficulties. The absence of other risk factors in this group (e.g. substance abuse, childbirth), and the presence of strengths, such as school enrollment and supportive relationships, indicate that the period between ages 17 and 21 may present a unique opportunity to address existing challenges and facili- tate competent functioning.

4.5. The “homelessness” cluster

The homelessness cluster made up the smallest portion of the sample, at 13%. All members of this cluster had been homeless at some point in their lives, and some also had substance abuse and incarceration histo- ries. Nevertheless, all were enrolled in school, had a supportive adult and did not have children. These youths' gender resembled the overall sample, but they were likely to be White. In addition, they tended to be slightly older during the latest removal, with reasons for removal predominantly parent-related. Many were placed in non-relative foster homes and group homes/institutions, with average rates of placement disruption. Prior research identified homelessness as a significant prob- lem among emancipating foster youth (Dworsky & Courtney, 2009), and demonstrated that it often coincided with other risks (e.g. Courtney et al., 2012). While it is unknown if homelessness preceded or followed out-of-home placement in the current sample, some mem- bers of this cluster may require intensive supports as they transition to adulthood due to other difficulties (i.e. substance abuse and criminal in- volvement). As with the substance abuse group, services should build upon existing strengths, such as school enrollment and relationships with supportive adults.

5. Recommendations for policy and practice

The heterogeneity of foster youth in this sample and in similar stud- ies (e.g. Courtney et al., 2012; Keller et al., 2007; Yampolskya et al., 2014; Yates & Grey, 2012), suggests significantly different service needs. Youths in the resilience cluster seem to require less tangible supports, though they may still exhibit internalizing problems not accounted for in the present investigation. This is especially likely given that child maltreatment rates were higher in this cluster com- pared to any other group. Child welfare officials should continue to monitor these youths' condition, including periodic, age appropriate assessment of needs, informed by their abuse and neglect histories. During this time, caseworkers should reduce their reliance on foster parent reports of problems, instead seeking out opportunities to interact directly with the youth. Evaluating and addressing difficulties such as

234 S. Shpiegel, K. Ocasio / Children and Youth Services Review 58 (2015) 227–235

mental health problems, disturbed peer-relationships and low self- esteem is particularly important, as these could be easily overlooked due to apparent competence exhibited by these youth (Yates & Grey, 2012).

In contrast, youths in the substance use, incarceration only and home- lessness clusters may require more intensive supports, with an emphasis on remediation of problem behaviors, such as substance abuse and de- linquency. Criminal involvement should receive special attention, as it is represented in all three clusters and its impact on subsequent function- ing may be particularly detrimental. Some studies suggest that place- ment in residential settings is a predictor of future delinquency (e.g. Cusick et al., 2011), therefore, youth should be placed in family-based care to the extent possible. For youth requiring residential care, inter- ventions to minimize externalizing problems may be beneficent.

Youths in the multiple problems cluster may have high risk of prob- lematic outcomes, though members are likely to vary widely in the ex- tent of difficulties presented. In general, these youth can benefit from educational supports, housing assistance and counseling services to ad- dress problem behaviors. Connecting youths with positive adult figures willing to serve as mentors is also important, as about one-third report- ed no existing connections. A potential avenue for increasing connect- edness may be assisting youth to reestablish relationships with extended family members (e.g. grandparents, aunts and uncles, sib- lings), providing that they can serve as positive role models. Meaningful relationships can also be formed through participation in educational and vocational settings, as well as in extracurricular activities.

Although this study examined several child welfare factors possibly associated with the emergence of specific clusters, their predictive value is difficult to assess due to the cross-sectional nature of the analysis. Nevertheless, emerging trends suggest that youth who are removed at earlier age, for reasons other than severe behavioral problems, and who are placed in relatively stable, family-based settings fare better as they approach the age of emancipation. In contrast, those who are removed later, placed in residential settings and experience placement disruption may evidence more problematic functioning, such as school drop-out and early childbirth. This interpretation is consistent with existing literature indicating that stable, family-based settings are opti- mal for facilitating positive outcomes, while also noting that such place- ments may be difficult to achieve for older, behaviorally disturbed youth (Newton et al., 2000; Rubin et al., 2007; Ryan & Testa, 2005; Stott, 2012).

There is an urgent need to address the availability of family-based placements for youth with emotional and behavioral difficulties. In some cases, residential treatment settings may be appropriate for these youth; however, there is also emerging research indicating that family-based placements for troubled children can be promoted through specialized training – such as Multidimensional Treatment Fos- ter Care (Fisher, Burraston, & Pears, 2005) and Keeping Foster and Kin- ship Parents Trained and Supported (Price et al., 2008). Further research is also needed to determine why relative placements are underutilized among these youth. Kinship placements have been noted to be as much as 70% less likely to disrupt than non-kin placements (Webster, Barth, & Needell, 1999). Although relatives asked to provide care may have greater reluctance to take in a youth they know to have emotional and/or behavioral challenges, agencies might also be less likely to place these youth with relatives due to safety and well-being concerns. Fur- ther research is needed to understand these dynamics, and examine how placement decisions are made for youth with special needs.

Finally, the emergence of the substance abuse and incarceration only clusters may reflect an effort by the child protection system to provide services to youth for reasons other than protection from their parents. Both groups were more likely to be male, and be placed in residential settings. The substance abuse group spent less time in foster care and was older at the latest removal, while the incarcerated only group exhib- ited increased behavioral disturbance and disabilities. Most important, both groups had relatively high rates of child-related reasons for

removal and out-of-home placement. In combination with the profile characteristics described above, this may suggest that the child protec- tion system was providing substance abuse and mental health treat- ment to youth who might have been able to receive those services another way. It is also possible that in some cases, parents were per- ceived to be ineffectual or uncommitted in addressing these youths` challenges, and the state agency intervened on their behalf. States should consider developing children's behavioral health systems, such as those in place in New Jersey, where treatment services for children are available to the general public through a single access point (NJ Department of Children and Families, n.d.).

6. Limitations and direction for future research

The results of the present study should be interpreted in light of its limitations. First, the response rate to the NYTD survey was slightly over 50%, and while weighing procedures were implemented to in- crease generalizability, biases may still occur due to the specific proce- dures used. Second, both NYTD and AFCARS variables are limited in the amount of detail they provide. For instance, the circumstances under which youths were referred for a substance abuse evaluation, as well as reasons for incarceration and homelessness are not assessed. Furthermore, the wording of the NYTD variables may have included a wide range of behaviors, especially for substance abuse and incarcera- tion measures. Some youth referred for a substance abuse evaluation may have actually abused substances, while others could have been re- ferred erroneously. Similarly, some youth reporting incarceration may have spent a single night in jail, while others could have been incarcer- ated for long periods of time in connection with more serious crimes. The timing and frequency of these behaviors are also not evaluated, as only lifetime incidence is documented at age 17. In addition, there is limited detail about youths` removal circumstances, as well as reasons for the disruption of placements. The accuracy of large administrative datasets such as NYTD and AFCARS is also difficult to assess, as states may report partial or conflicting information in some cases. Most impor- tant, the findings of this study are limited by the cross-sectional nature of the analysis. Specifically, several indicators used to form the clusters (i.e. substance abuse referral, homelessness, incarceration) may also represent removal reasons in certain cases, as the timing of their occur- rence is unknown.

Future research should employ longitudinal designs to examine the link between child welfare factors and the functioning patterns exhibit- ed by the youth. Subsequent waves of NYTD data will allow conducting such investigations with a large, national sample. Future efforts should also follow youth until the age of 21 and, if possible, longer, to assess the stability and predictive utility of the obtained patterns, and evaluate the role of service provision in this regard. Such investigations should include additional indicators of successful adaptation as applicable to the age range studied. For instance, assessing employment status, inde- pendent living, earnings and dependence on public assistance may be appropriate for older youth who already left the child welfare system. Including indicators related to internalizing problems (e.g. mental health concerns, self-esteem) can be beneficent at any age, as some youth may function successfully in domains such as education and em- ployment, while still struggling with the aftermath of child maltreat- ment and years in out-of-home placement.

Acknowledgments

The data used in this publication were made available by the Nation- al Data Archive on Child Abuse and Neglect. Data from the National Youth in Transition Database were originally collected by the states and provided to the Children's Bureau. Funding was provided by the Children's Bureau, U.S. Department of Health and Human Services. The collector of the original data, the funder, the Archive, Cornell University

235S. Shpiegel, K. Ocasio / Children and Youth Services Review 58 (2015) 227–235

and their agents or employees bear no responsibility for the analyses or interpretations presented here.

We are particularly grateful to Michael Dineen, Holly Larabee and Elliott Smith from the National Data Archive on Child Abuse and Neglect at Cornell University for their technical assistance and support.

References

Aarons, G. A., Hazen, A. L., Leslie, L. K., Hough, R. L., Monn, A. R., Connelly, C. D., et al. (2008). Substance involvement among youths in child welfare: The role of common and unique risk factors. American Journal of Orthopsychiatry, 78(3), 340–349.

Barber, J. G., & Delfabbro, P. H. (2003). Placement stability and the psychosocial well- being of children in foster care. Research on Social Work Practice, 13(4), 415–431.

Barnet, B., Liu, J., & DeVoe, M. (2008). Double jeopardy: Depressive symptoms and rapid subsequent pregnancy in adolescent mothers. Archives of Pediatrics & Adolescent Medicine, 162(3), 246–252.

Boden, J. M., Fergusson, D. M., & Horwood, L. J. (2008). Early motherhood and subsequent life outcomes. Journal of Child Psychology and Psychiatry, 49(2), 151–160.

Brandford, C., & English, D. (2004). Foster youth transition to independence study. Seattle: WA: Office of Children's Administration Research.

Chase, E., Maxwell, C., Knight, A., & Aggleton, P. (2006). Pregnancy and parenthood among young people in and leaving care: What are the influencing factors, and what makes a difference in providing support? Journal of Adolescence, 29, 437–451.

Collins, M. E., Spencer, R., & Ward, R. (2010). Supporting youth in the transition from foster care: Formal and informal connections. Child Welfare, 89(1), 125–143.

Courtney, M. E. (2009). Describing the problem: Outcomes for older youth exiting the foster care system in the U.S. In B. Kerman, A. B. Maluccio, & M. M. Freundlich (Eds.), Achieving permanence for older children and youth in foster care. New York: Columbia University Press.

Courtney, M. E., Hook, J. L., & Lee, J. S. (2012). Distinct subgroups of former foster youth during young adulthood: Implications for policy and practice. Child Care in Practice, 18(4), 409–418.

Cusick, G. R., Courtney, M. E., Havlicek, J., & Hess, N. (2011). Crime during the transition to adulthood: How youth fare as they leave out-of-home care. Chicago: Chapin Hall at the University of Chicago.

Cusick, G. R., Havlicek, J. R., & Courtney, M. E. (2012). Risk for arrest: The role of social bonds in protecting foster youth making the transition to adulthood. American Journal of Orthopsychiatry, 81(1), 19–31.

Daining, C., & DePanfilis, D. (2007). Resilience of youth in transition from out-of home care to adulthood. Children and Youth Services Review, 29, 1158–1178.

Dworsky, A., & Courtney, M. (2009). Homelessness and the transition from foster care to adulthood. Child Welfare, 88(4), 23–56.

Fava, G. A., Guidi, J., Porcelli, P., Rafanelli, C., Bellomo, A., Grandi, S., et al. (2012). A cluster analysis-derived classification of psychological distress and illness behavior in the medically ill. Psychological Medicine, 42, 401–407.

Fisher, P. A., Burraston, B., & Pears, K. (2005). The early intervention foster care program: Permanent placement outcomes from a randomized trial. Child Maltreatment, 10, 61–71.

Hass, M., & Graydon, K. (2009). Sources of resiliency among successful foster youth. Children and Youth Services Review, 31(4), 457–463.

Hines, A. M., Merdinger, J., & Wyatt, P. (2005). Former foster youth attending college: Resilience and the transition to young adulthood. American Journal of Orthopsychiatry, 75(3), 381–394.

Hughes, D. M., Condron, B., Jackson, N., Pitchal, E., Garton, N., & Elliott, S. P. (2008). Prepar- ing our kids for education, work and life: A report of the task force on youth aging out of DSS care. The Boston Foundation.

Jones, L. (2012). Measuring resiliency and its predictors in recently discharged foster youth. Child and Adolescent Social Work Journal, 29, 515–533.

Keller, T. E., Cusick, G. R., & Courtney, M. E. (2007). Approaching the transition to adult- hood: Distinctive profiles of adolescents aging out of the child welfare system. Social Service Review, 81, 453–484.

King, B., Putnam-Hornstein, E., Cederbaum, J. A., & Needell, B. (2014). A cross-sectional examination of birth rates among adolescent girls in foster care. Children and Youth Services Review, 36, 179–186.

Leathers, S., & Testa, M. (2006). Foster youth emancipating from care: Caseworkers' reports on needs and services. Child Welfare, 85, 463–493.

Lee, J. S., Courtney, M. E., & Tajima, E. (2014). Extended foster care support during the transition to adulthood: Effect on the risk of arrest. Children and Youth Services Review, 42, 34–42.

Lemon, K., Hines, A. M., & Merdinger, J. (2005). From foster care to young adulthood: The role of independent living programs in supporting successful transitions. Children and Youth Services Review, 27(3), 251–270.

McDonald, T. P., Mariscal, S. E., Yan, Y., & Brook, J. (2014). Substance use and abuse for youths in foster care: Results from the communities that care normative database. Journal of Child & Adolescent Substance Abuse, 23(4), 262–268.

McMillen, J. C., Zima, B., Auslander, W., Scott, L., Munson, M. R., Ollie, M., & Spitznagel, E. (2005). Prevalence of psychiatric disorders among older youths in the foster care sys- tem. Journal of the American Academy of Child and Adolescent Psychiatry, 44(1), 88–95.

Naccarato, T., Brophy, M., & Courtney, M. E. (2010). Employment outcomes of foster youth: The results from the Midwest evaluation of the adult functioning of foster youth. Children and Youth Services Review, 32(4), 551–559.

Narendorf, S. C., & McMillen, J. C. (2010). Substance use and substance use disorders as foster youth transition to adulthood. Children and Youth Services Review, 32, 113–119.

National Data Archive on Child Abuse and Neglect (2013a). Adoption and foster care anal- ysis and reporting system. User's guide FY 2011. Ithaca, NY: Bronfenbrenner Center for Translational Research, Cornell University.

National Data Archive on Child Abuse and Neglect (2013b). Adoption and foster care analysis and reporting system. Codebook FY 2011. Ithaca, NY: Bronfenbrenner Center for Translational Research, Cornell University.

National Data Archive on Child Abuse and Neglect (2014). National youth in transition database: Outcome file. User's guide FY 2011. Ithaca, NY: Bronfenbrenner Center for Translational Research, Cornell University.

Newton, R. R., Litrownik, A. J., & Landsverk, J. A. (2000). Children and youth in foster care: Disentangling the relationship between problem behaviors and number of place- ments. Child Abuse & Neglect, 24(10), 1363–1374.

Price, J. M., Chamberlain, P., Landsverk, J., Reid, J. B., Leve, L. D., & Laurent, H. (2008). Effects of a foster parent training intervention on placement changes of children in foster care. Child Maltreatment, 13(1), 64–75.

Rubin, D. M., O'Reilly, L. A., Luan, X., & Localio, A. R. (2007). The impact of placement sta- bility on behavioral wellbeing for children in foster care. Pediatrics, 119(2), 336–344.

Ryan, J. P., & Testa, M. F. (2005). Child maltreatment and juvenile delinquency: Investigat- ing the role of placement and placement instability. Children and Youth Services Review, 27(3), 227–249.

Samuels, G. M., & Pryce, J. M. (2008). ‘What doesn't kill you makes you stronger’: Surviv- alist self-reliance as resilience and risk among young adults aging out of foster care. Children and Youth Services Review, 30(10), 1198–1210.

Stott, T. (2012). Placement instability and risky behaviors of youth aging-out of foster care. Child and Adolescent Social Work Journal, 29, 61–83.

Stott, T. (2013). Transitioning youth: Policies and outcomes. Children and Youth Services Review, 35, 118–227.

Stott, T., & Gustavsson, N. (2010). Balancing permanency and stability for youth in foster care. Children and Youth Services Review, 32, 619–625.

Svoboda, D. V., Shaw, T. V., Barth, R. P., & Bright, C. L. (2012). Pregnancy and parenting among youth in foster care: A review. Children and Youth Services Review, 34, 867–875.

Tsai, J., Edens, E. L., & Rosenheck, R. A. (2011). A typology of childhood problems among chronically homeless adults and its association with housing and clinical outcomes. Journal of Health Care for the Poor and Underserved, 22(3), 853–870.

U.S. Department of Health and Human Services (2014). The AFCARS report: Preliminary estimate for FY 2013. (Washington: D.C. Retrieved May 1, 2015 from https://www. acf.hhs.gov/sites/default/files/cb/afcarsreport21.pdf).

Vaughn, M. G., Ollie, M. T., McMillen, C., Scott, L., & Munson, M. (2007). Substance use and abuse among older youth in foster care. Addictive Behaviors, 32, 1929–1935.

Webster, D., Barth, R. P., & Needell, B. (1999). Placement stability for children in out-of- home care: A longitudinal analysis. Child Welfare, 79(5), 614–632.

Yampolskya, S., Sharrock, P., Armstrong, M. I., Strozier, A., & Swanke, J. (2014). Profiles of children placed in out-of-home care: Association with permanency outcomes. Children and Youth Services Review, 36, 195–200.

Yates, T. M., & Grey, I. K. (2012). Adapting to aging-out: Profiles of risk and resilience among emancipated foster youth. Development and Psychopathology, 24, 475–492.