Placement of children in out-of-home care in Québec, Canada: When and for whom initial out-of-home placement is most likely to occur

Children and Youth Services Review 35 (2013) 2031–2039

Contents lists available at ScienceDirect

Children and Youth Services Review journal homepage: www.elsevier.com/locate/childyouth

Placement of children in out-of-home care in Québec, Canada: When and for whom initial out-of-home placement is most likely to occur Tonino Esposito a,⁎, Nico Trocmé b, Martin Chabot c, Aron Shlonsky d, Delphine Collin-Vézina e, Vandna Sinha e a

McGill University, Centre for Research on Children and Families, 3506, University, suite 106, Montreal, Quebec, H3A 2A7, Canada McGill University, Faculty of Social Work, Centre for Research on Children and Families, Canada McGill University, Centre for Research on Children and Families, Canada d University of Toronto, Faculty of Social Work, Canada e McGill University, Faculty of Social Work, Canada b c

a r t i c l e

i n f o

Article history: Received 22 May 2013 Received in revised form 10 October 2013 Accepted 10 October 2013 Available online 22 October 2013 Keywords: Out-of-home placement Child maltreatment Neighborhood effects Clinical-administrative data Census data Longitudinal analysis

a b s t r a c t This study contributes to the growing child protection placement literature by providing the first Canadian provincial longitudinal study examining when and for whom initial out-of-home placement is most likely to occur. Anonymized clinical-administrative child protection data were merged with the 2006 Canadian Census data for the province of Québec, and the final dataset included 127,181 children investigated for maltreatment for the first time between April 1, 2002 and March 31, 2010. Cox proportional hazard results indicate that the vast majority of investigated children do not experience a placement, but for the others, placement tends to occur immediately following the maltreatment investigation with only a slight increase in risk over time. The increased risk of placement for younger children aged 0 to 9 years was statistically explained by a combination of male gender, behavioral problems, parents' high risk lifestyles, hospital referral, the number of investigations and neighborhood area socioeconomic disadvantages. The increased risk of placement for older children aged 10 to 17 years was statistically explained by a combination of behavioral problems, police reporting, the number of investigations and neighborhood area socioeconomic disadvantages. Neighborhood area socioeconomic disadvantages significantly contributed to the increased risk of out-of-home placement for all children, but this factor is most influential when it comes to younger children. © 2013 Elsevier Ltd. All rights reserved.

1. Introduction The placement of children in out-of-home care occurs in principle when children cannot be safely supported in their own homes. This notion is entrenched in many national and international child protection statutes, such as Section 4 of the recent amendments to the Québec Youth Protection Act (2007) which states that, when children are in need of child protection services, every effort must be made to keep them with their families. At the international level, Sections 2 and 3 of the United Nations (2009) for the alternative care of children also state that support efforts should primarily be directed to keep children in the care of their families. These legal tenets reinforce the notion that the best environment for children is ideally with their natural families. In spite of these efforts, out-of-home care rates have been increasing in Canada. The Canadian Incidence Study of Reported Child Abuse and Neglect (CIS) shows a 21.7%1 relative increase in the placement rate, from 2.67 per 1000 in 1998 to 3.25 per 1000 in 2008 (Trocmé et al., 2010). ⁎ Corresponding author. Tel.: +1 514 691 6517. E-mail address: [email protected] (T. Esposito). 1 Percentage represents the relative change between 1998 CIS and 2008 CIS but it does not provide statistical evidence that the increase is clinically significant. 0190-7409/$ – see front matter © 2013 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.childyouth.2013.10.010

While at the individual level, the removal of a child from an unsafe family environment is not a negative outcome – indeed for some children it may be the only feasible option available – at a program level, every effort should be made to develop and provide family support services needed to improve family circumstances while keeping children living at home. Yet, it is difficult to know who would benefit without first knowing who is most at risk, and when placement is most likely to occur. This knowledge is meant to assist child protection authorities in developing and implementing support services needed to improve family circumstances while keeping children living at home. No Canadian provincial longitudinal studies have attempted, up to now, to examine when and for whom placement is most likely to occur. As a result, Canadian child protection authorities must often make difficult program decisions while relying on research evidence from jurisdictions where child protection policies and the structure of child protection services are notably different. Quebec's child protection statutes, for example, include situations where a child's behavior might put the child's safety or well-being at risk as a ground for intervention, a category that is not included in most other jurisdictions in North America. There also is significant variation regarding the extent to which child protection services target younger children versus adolescents: in the United States and United Kingdom younger children represent the majority of those admitted to out-of-home care, whereas

2032

T. Esposito et al. / Children and Youth Services Review 35 (2013) 2031–2039

in Canada older children represent the majority of first admissions (Thoburn, 2010; Wulczyn, Barth, Yuan, Harden, & Landsverk, 2005). Additionally, studies often do not differentiate sufficiently between the placement trajectories of the youngest compared to older children, thereby masking age-specific differences that may explain why placement changes occur. The current study addresses this knowledge gap by providing the first Canadian provincial longitudinal analysis of when and for whom initial out-of-home placement is most likely to occur. The study uses age-specific models to compare the youngest children to early school age children, and adolescents to middle school age children, and separates behavioral problems from other reasons for requiring child protection placement services. 2. Background studies Studies examining why children are admitted to out-of-home care vary in objective and methodology. While these studies are useful in informing child protection authorities about the specific needs of the unique population examined, they do not as a whole lead to conclusions regarding other factors most likely to influence placement. For example, certain studies report that infants are more likely to be placed, while others report that risk of placement increases with age. Wulczyn, Hislop, and Harden (2002) and Thoburn (2010) found that the youngest cohorts of children were more likely to enter out-of-home care; in contrast, James et al. (2006) found that as children aged, each year increased their odds of placement by 32%. James et al. also reported that the primary reason children were removed from their caregivers was supervisory neglect (46%). However, Farmer, Mustillo, Burns, and Holden (2008) did not find that any of the family-level factors – including supervisory neglect – were significantly related to out-of-home placement. Like James et al. however, they reported that older children, primarily males, were more likely to be placed in out-of-home care. Other studies have also reported that males and older children, specifically those manifesting behavioral problems, have a higher likelihood of placement (Berger, Bruch, James, & Rubin, 2009; Brook & McDonald, 2009; Walrath & Liao, 2005). Institutions with which children are in contact during certain periods of their lives can influence the risk of placement (Wulczyn, Barth, Yuan, Harden, & Landsverk, 2005). These institutions may include the hospital at birth, the school for middle school age children and the police for older youth. These institutions may be indicative of family and child functioning. For example, infants referred by front-line health services may come from families with co-occurring problems (i.e. substance abuse, material and physical neglect etc.). Runyan, Gould, Trost, and Loda (1982) report that referrals from hospital physicians and law enforcement agencies significantly contributed to the increased likelihood of placement. However, Lindsey (1991) examined factors affecting out-of-home care placement using discriminant analysis of various age groups, and reports that referrals from law enforcement agencies, as well as other legal referrals, were predictive of placement for children aged 13 to 15 years and no other age group. Evidence generally supports the notion that maltreatment and youth criminal justice services are associated. A study by Yampolskaya, Armstrong, and McNeish (2011) examined risk factors for children in placement aged 7 to 17 years using Florida administrative data and reported that the chronicity, but not the severity of maltreatment, increases the risk that children will become involved with the youth criminal justice system. Regarding the association between placement and youth criminal justice services, early studies such as Runyan and Gould (1985) – which used a historical matched cohort design to compare rates of subsequent youth criminal justice service requests between maltreated children in long-term out-of-home care and children who remained in their homes – found that maltreated children in out-ofhome care were more likely to have committed a crime. The study also reported a positive correlation between the number of placements and youth criminal justice convictions. Jonson-Reid and Barth (2000a,

2000b) examined children in California using administrative data, and reported higher youth criminal justice involvement for children in outof-home care compared to those living at home as had Runyan & Gould. Doyle (2007) also examined youth criminal justice data linked to child protection data in Illinois, and reported that children placed in out-ofhome care are two to three times more likely to enter the youth criminal justice system. In contrast, in a retrospective study of foster children, Widom (1991) reported that placement did not increase the risk of youth criminal behavior. Most research in the area of youth criminal justice, such as the studies mentioned above, attempts to demonstrate a one-way association between the placement and youth criminal behavior. While these studies contribute to our understanding of the placement– youth criminal justice link, no studies to date considered the influence of youth criminal justice on the decision to place children. Many studies in the United States report a disproportionate representation of African American and Native American children in out-ofhome care (Hill, 2007; Texas Health Human Services Commission, 2006; Wulczyn & Lery, 2007). The overrepresentation of First Nations children placed in out of home care has been an alarming problem in Canada, where, in some jurisdictions, over two-thirds of children in care are First Nations. In a re-analysis of the 2008 Canadian Incidence Study of Reported Child Abuse and Neglect (CIS), Sinha et al. (2011) found that, for every 1000 First Nations children served by the sampled agencies, there were 10.3 investigations leading to informal kinship care placement and another 12.6 investigations leading to formal child welfare placements. For non-Aboriginal children, however, there were 0.9 investigations leading to informal kinship care, and 1.1 investigations leading to formal placements per 1000 non-Aboriginal children (Sinha et al., 2011). In a compilation of child welfare research published by Lindsey and Shlonsky (2008), they report that a disproportionate number of children come to the attention of child protection services as a result of socioeconomic disadvantages alone. Lindsey (1991) found that the income level of the parents was the best predictor of whether children were placed in out-of-home care. Similarly, Berger and Waldfogel (2004) examined the influence of family structure and economic disadvantages on the likelihood of placement for children using the National Longitudinal Survey of Youth, and concluded that those from low-income families are more likely to be placed in out-of-home care. Lery's (2009) study examining the role of neighborhood structure and placement using three different spatial scales – (1) census tract, (2) block groups and (3) zip codes – found that the different spatial scales produced similar results in that placement was significantly higher in disadvantaged neighborhoods. Lery (2009) concluded, “no matter how neighborhoods are delineated, areas with high levels of poverty tend to border other high poverty areas, and areas with high risk for out-of-home care entry tend to be located near other areas with high entry rates” (p. 335). There has been an extensive effort to document the risk factors associated with out-of-home placement. Yet, applying current research evidence within a Canadian (or specifically a Québec) child protection context is difficult as a result of variations in the structure of child protection services across jurisdictions and samples used across studies, which often results in a variation in the factors found to be most likely to influence placement. Building on the existing literature, this study will: (1) use age-specific models to compare the youngest children to early school age children and adolescents to middle school age children; (2) separately analyze behavioral problems and other reasons for child protection services; (3) examine whether a request for youth criminal justice services influences the risk of placement; and (4) examine if neighborhood area socioeconomic disadvantages further contribute to the unique age-specific risk of placement. 3. Method For this study, two different data sources were merged together to create a provincial dataset. The first data source consists of anonymized

T. Esposito et al. / Children and Youth Services Review 35 (2013) 2031–2039

longitudinal clinical–administrative child protection data from all sixteen mandated child protection jurisdictions across the province of Québec. A detailed accounting of children's maltreatment history, along with all covariates used in this study – except for neighborhood area socioeconomic disadvantages – was constructed using these clinical– administrative data. Family composition, employment and income are not yet systematically collected across child protection jurisdictions and are therefore not available in the clinical–administrative child protection data. The second data source, therefore, is provincial data extracted from the 2006 Canadian Census public archive at McGill University, used to create a neighborhood area socioeconomic disadvantage composite index. The cohort used for this study consists of 127,181 children investigated for maltreatment for the first time within a child protection jurisdiction between April 1, 2002 and March 31, 2010. Initial out-ofhome care is defined as any placement lasting longer than 72 h following initial investigation. Placements are only considered if they last longer than 72 h in order to control for respite placements and emergency placements, which are not part of a child's long term plan (for details see Esposito, Trocmé, Chabot, Duret, & Gaumond, in press). Out-of-home placement includes the following placement classifications: (1) kinship foster care — a formal subsidized placement with an extended family member; (2) family foster care — a formal subsidized placement in family-based care; (3) group home placement — a formal subsidized placement in a structured group living setting; and (4) residential treatment — a formal subsidized placement in a therapeutic residential treatment facility. The follow-up period starts from the date of initial investigation within a child protection jurisdiction to the date of initial placement or end of follow-up period – September 31, 2011 – or the child's 18th birthday, whichever comes first. 3.1. Covariates The covariates examined in this study include age at initial investigation, ethno-racial background2 of the child, gender, reason for investigation, number of investigations, source of the referral, request for youth criminal justice services and neighborhood area socioeconomic disadvantage index. These covariates were used in block multivariate Cox proportional hazard regression analyses to obtain the independent effect of each covariate on the risk of experiencing an initial out-of-home placement. Age at entry is a nominal variable with children aged 2 to 5 acting as the reference group for comparisons with children aged 0 to 1 and children aged 6 to 9. For models with older children, youth aged 10 to 13 acted as a reference group for 14 to 17 year olds. Gender is a nominal variable with female acting as the reference group for male. Reason for investigation includes the following dichotomous constructs: (1) psychological & emotional abuse, which includes rejection, denigration, exposure to intimate partner violence and exploitation; (2) physical, material & health neglect, which includes physical neglect, medical neglect and material deprivation; (3) parent high risk lifestyle, which represents parents' lifestyle resulting in a failure to supervise or protect the child, including abandonment due to parental absence, substance abuse, refusal to assure child care and risk of neglect; (4) school truancy & school neglect, which includes failure to attend school or failure to ensure that the child attends school; (5) physical abuse; (6) sexual abuse; (7) behavioral problems such as harming behavior, violence towards self and others,

2 Including children's ethno-racial background in the final age-specific Cox proportional hazard regression models poses a particular methodological challenge given that for 40.3% of investigated children the ethno-racial background was not identified. Additional analysis on the missing information also showed that missing ethno-racial identity information was not random and that these children were much less likely to receive ongoing child protection services.

2033

child substance abuse, school behavioral problems, runaway behavior and destruction of property; (8) risk of sexual abuse; and (9) risk of physical abuse. Youth criminal justice service request is a nominal variable measuring whether children received a request for services under the Québec Youth Criminal Justice Act (LSJPA—Loi sur la justice pénale des adolescents) prior to placement. Number of investigations is a continuous variable calculated by examining the number of times children are investigated for maltreatment prior to placement or end of follow-up period. Source of referral includes the following nominal values: (1) community health and social services clinics (CLSC); (2) child protection agency; (3) extended family and neighbors; (4) school staff; (5) police; (6) hospital staff; (7) other professional institutions; and (8) unknown. Neighborhood area socioeconomic status includes six socioeconomic indicators for each Québec specific census dissemination area: (1) total population 15 years and over who are unemployed or not in the labor force; (2) median income in 2005 for population 15 years and over; (3) total persons in a private household living alone; (4) total population 15 years and over who were separated, divorced or widowed; (5) family median income in 2005; and (6) median household income in 2005. The three income indicators were transformed by subtracting the individual score by its maximum value so that each unit increase represents an increase in socioeconomic disadvantages. They were then normalized using Log10. A principal component analysis with varimax rotation was performed on the transformed and normalized census-based indicators in order to construct single index of socioeconomic neighborhood disadvantages for each dissemination area. Principal component analysis reduces the six indicators mentioned above into a single socioeconomic disadvantage construct. Construct scores were calculated for each dissemination area (10,907 dissemination areas in Québec) with the lowest score representing low-risk socioeconomic disadvantage and high score representing high-risk socioeconomic disadvantage. This composite index was then merged with the child protection clinical-administrative data matched by children's postal codes at initial investigation, representing 42,989 unique geographic areas with 10,778 unique socioeconomic index estimates. The index has a minimum score of −3.37 representing the lowest socioeconomic risk and a maximum score of 3.51 representing the highest socioeconomic risk. The index has a mean score of 0.2898 (Std. 0.92037) and median of 0.2931.

3.2. Analytic model Cox proportional hazard regression analysis was used to examine the risk of initial placement from the moment when children's initial maltreatment investigation begins. It identifies the probability of placement at time t given that children were investigated for maltreatment and at risk of placement until time t. One of the reasons for using Cox proportional hazard regression modeling is the ability to make use of censored observations (Walters, 1999), that is, children who were not placed within the follow-up period. It is unknown when or if these children experienced an initial out-of-home placement after the follow-up period expires, and leaving out these observations would represent an important loss of information. In a Cox proportional hazard regression model, subjects are considered at risk until they either experience the placement, are censored, that is, the follow-up time – September 31, 2011 – expires, or turn 18 years old and are no longer at risk of placement. The block function was used to test the possible ecological interactions between age, individual factors and neighborhood area socioeconomic disadvantages on the risk of initial placement in out-of-home care. The Cox proportional hazard regression model is specified as:

H ðt Þ ¼ H0 ðt ÞX exp b1 X 1þ b2 X 2þ b3 X 3þ …… ::þ bk X k




T. Esposito et al. / Children and Youth Services Review 35 (2013) 2031–2039

where, X1……Xk represent the individual covariates, and H0(t) is the baseline hazard of placement at time t. By taking the logarithms after dividing both sides of the equations above by H0(t), we obtain: ln½H ðt Þ=H 0 ðt Þ ¼ b1 X 1þ b2 X 2þ b3 X 3þ …… ::þ bk X k : H(t)/H0(t) represents the risk of experiencing a placement. The coefficients b1…..bk are estimated by Cox proportional hazard regression function and the expb1 represents the risk (expressed as a hazard ratio) of placement for the independent variable X1, at any time, holding all other covariates constant. It provides an estimate by which the risk of placement will increase (greater than 1) or decrease (less than 1) when the condition of the independent variable is 1 — nominal or when the independent variable increases by 1-unit — continuous. Statistical tests were conducted at 95% level of confidence. The dataset has been built and transformed using SPSS version 19 and analyzed using both SPSS 19 and STATA 11.

Days to placement from initial investigation Fig. 2. Hazard rate of placement from the point of initial maltreatment investigation.

age at initial investigation and the time that elapsed since the initial investigation in order to produce a baseline −2 Log likelihood for use in comparing the goodness-of-fit between blocks. The −2 Log likelihood ratios produce a p-value for model significance and are often negative, with values closer to zero indicating a better fitting model (Frees, 2004). For each of the age-specific models predicting out-of-home placement, three blocks of covariates were added in a sequential and cumulative fashion starting with (1) age; (2) gender, reason for the investigation, source of referral, request for youth criminal justice services (only for 10 to 17 year olds) and number of investigations; and (3) the addition of neighborhood area socioeconomic disadvantages. The final block 3 models include independent variables that were not significantly associated to placement in out-of-home care in block 1 or 2. This enables us to assess whether neighborhood area socioeconomic disadvantages would change the significance of the estimate. Multivariate Cox proportional hazard regression analysis was then used to examine the risk of placement at any given time for each independent variable while holding the value of all the other variables in the model constant. Tables 2 & 3 report estimates of the Cox proportional hazard regression models for both age-specific groups — children aged 0 to 9 years (see Table 2, N = 72,071) and children aged 10 to 17 years (see Table 3, N = 55,110).

4. Results The population of children studied includes 127,181 children investigated for maltreatment for the first time between April 1, 2002 and March 31, 2010. Of the 127,181 children investigated by the Québec child protection system in the last 9 years, 22.8% (N = 29,040, see Table 1) were admitted to out-of-home care at some point during the follow-up period.

Rate of out-of-home placement

3.2.1. Analytic process Fig. 1 illustrates a significant curvilinear effect (X2 = 7498.265, df = 17, p = b 0.001) regarding children's age at initial investigation and rates of out-of-home placement in Québec. Of all the children investigated for the first time in the last 9 years, children younger than 1 year and older than 12 years represent the highest proportion of those placed in out-of-home care. Similar trends are noted in the Canadian Incidence Study of Reported Child Abuse and Neglect, with 27% of children aged 0 to 1 and 42% of children 14 and older entering out-of-home care during investigation in 2003 (Trocmé et al., 2005). Given that placement rates appear to be non-linear and distinct, a decision was made to analyze older and younger children separately. The analysis was composed of several steps. First, a hazard function was used to examine the risk of placement over time by age at the moment of initial investigation (see Fig. 2). Second, descriptive analyses were performed between all independent covariates and placement (see Table 1). In order to ensure that there is no linearity among covariates, an ordinary least squares linear regression was conducted with the same covariates used in the final hazard models in order to determine the variance inflation factor estimates (VIF). This allows us to examine the amount of variance of the corresponding covariate estimate that is increased due to multicollinearity compared to what it would be if there were no multicollinearity. There is no formal cut-off value for determining multicollinearity using VIF, however, if the values of VIF exceed 10, they are regarded as indicating multicollinearity (Kutner, Nachtsheim, & Neter, 2004). For children aged 0 to 9 years, the VIF estimates ranged from a low of 1.020 to a high of 1.881, and, for children aged 10 to 17 years, the VIF estimates ranged from a low of 1.010 to a high of 2.157, indicating no issues of “linearity” between covariates in either of the age-specific models. Next, an age-only model was estimated for the purpose of comparing each age-specific block model. The age-only model includes the

Observed Hazard Rate

2034

Children’s age at initial investigation Fig. 1. Placement rate by age at initial investigation (N = 127, 181).

T. Esposito et al. / Children and Youth Services Review 35 (2013) 2031–2039

2035

Table 1 Descriptive factors. Individual factors

Children investigated 0–17 (N = 127,181)

Children placed 0–17 (N = 29,040)

Children placed 0–9 (N = 12,349)

Child age at investigation 0–1 2–5 6–9 10–13 14–17

15.1% 20.0% 21.6% 21.6% 21.8%

18.5% 11.1% 12.9% 22.5% 35.0%

43.5% 26.1% 30.4%

Gender Male Female

50.6% 49.4%

52.6% 47.4%

55.1% 44.9%

50.8% 49.2%

Reason for investigation Psychological & emotional abuse Physical, material & health neglect Parent high risk lifestyle School truancy & neglect Risk of sexual abuse Sexual abuse Behavioral problems Risk of physical abuse Physical abuse

7.1% 5.8% 38.0% 4.6% 1.6% 8.3% 14.8% 2.1% 17.6%

5.2% 4.8% 37.8% 3.4% 0.4% 3.8% 31.6% 0.8% 12.2%

5.9% 8.9% 63.3% 3.5% 0.6% 3.0% 1.9% 1.4% 11.5%

4.6% 1.8% 19.0% 3.4% 0.3% 4.3% 53.5% 0.4% 12.8%

Source of referral at investigation CLSC Youth protection agency Police Other professional institutions School Hospital staff Unidentified Family

10.2% 10.4% 16.5% 8.3% 20.4% 6.8% 3.8% 23.4%

11.3% 10.0% 17.4% 6.1% 16.7% 9.5% 3.4% 25.6%

12.7% 13.7% 12.8% 7.8% 12.1% 16.4% 4.9% 19.7%

10.4% 7.2% 20.8% 4.9% 20.1% 4.3% 2.3% 29.9%

Request for youth criminal justice services

–

–

–

17.4%

Number of investigations Neighborhood socioeconomic disadvantage

Children placed 10–17 (N = 16,691)

39.1% 60.9%

Mean (S.D.)

Mean (S.D.)

Mean (S.D.)

Mean (S.D.)

1.51 (0.96) .29 (.92)

1.63 (1.07) .36 (.92)

1.90 (1.31) .55 (.86)

1.44 (0.78) .22 (.94)

Percentages are column percentages for each category and may not add to 100% because of rounding.

The hazard curve in Fig. 2 illustrates the risk of placement over time from the point of initial maltreatment investigation by age at investigation. As shown in Fig. 2, the hazard function portrays an early peak for all age groups indicating that placement, irrespective of age at initial investigation, tends to occur during or shortly after the maltreatment investigation with gradual increases over time — half of all placed children, irrespective of age, were placed within the first 100 days of initial investigation. Half of all placed children aged 14 to 17 years were placed within 24 days of initial investigation, followed by days for 0 to 1 year olds and 207 days for 10 to 13 year olds. Children aged 2 to 9 years at initial investigation have the lowest hazard rate of placement. 4.1. Descriptive factors As reported in Table 1, there is an equal proportion of male (50.6%) and female (49.4%) children investigated for maltreatment. The average number of investigations for all children was 1.51 (Std. 0.96) investigations per child (see Table 1). The average number of investigations is higher for placed children aged 0 to 9 years (1.90 Std. 1.31) compared to placed children aged 10 to 17 years (1.44 Std. 0.78). Although 0 to 1 year olds represent the smallest proportion (15.1%) of all investigated children, they make up the third largest proportion of all children placed (18.5%) as well as the largest proportion of children placed aged 0 to 9 years (43.5%). Children aged 14 to 17 years represent the highest proportion of all children investigated (21.8%) and placed in out-of-home care (35.1%) as well as the highest proportion of children placed (60.9%) between 10 and 17 years.

Thirty-eight percent of children were investigated because of the parents' high risk lifestyle followed by physical abuse (17.6%) and behavioral problems (14.8%). Although 14.8% of children were investigated for behavioral problems, they represent 31.6% of all children placed in out-of-home care. When we divide placed children into two age-specific groups, the need for age-specific models becomes evident: 63.3% of the placed 0 to 9 year olds were investigated because of their parents' high risk lifestyle as a primary concern, while 19.0% of the placed 10 to 17 year olds were investigated for the same reason. On the other hand, the highest proportion of the placed 10 to 17 years old were investigated for behavioral problems (53.5%) as a main concern compared to 1.9% of the placed 0 to 9 year olds. A difference in distribution of placed children between age-specific groups is also noted for all other reasons for investigation, except for school truancy and school neglect. A request for youth criminal justice services was made prior to placement for close to 1 out of 5 children aged 10 to 17 years and placed in out-of-home care. This allows for modeling the probability of placement as a function of whether children aged 10 to 17 years had a request for youth criminal justice services. Lastly, the composite estimate of the neighborhood area socioeconomic disadvantages for investigated children was .29 (Std. 0.92), and was higher [.36 (Std. 0.92)] for placed children. We also find that, of all children placed in out-of-home care, neighborhood area socioeconomic disadvantage composite estimates were much higher for children aged 0 to 9 years compared to the ones aged 10 to 17 years, going from.55 (Std. 0.86) for 0 to 9 year olds .22 (Std. 0.94) for 10 to 17 year olds. Therefore, children's chances of removal,

2036

T. Esposito et al. / Children and Youth Services Review 35 (2013) 2031–2039

Table 2 Cox proportional hazard models of initial out-of-home placement for children aged 0 to 9 years. Number of events and censored values Total

Event

Censored

% Censored

72,071

12,349

59,722

82.8% Block 2

Block 1

SE

Wald

Adj. HR (95% CI)

Adj. HR (95% CI)

Adj. HR (95% CI)

.654

0.23

775.3

1.924⁎⁎⁎(1.837, 2.014)

1.966⁎⁎⁎(1.878, 2.059)

2.549⁎⁎⁎(2.440, 2.663)

.097

0.25

14.7

1.100⁎⁎⁎(1.048, 1.157)

1.085⁎⁎⁎(1.033, 1.140)

1.079⁎⁎⁎(1.030, 1.131)

Block3 (final model) Beta Age at initial investigation 0–1 2–5(ref) 6–9 Child sex Male (female ref)

.097

.018

28.0

1.101⁎⁎⁎(1.063, 1.141)

1.099⁎⁎⁎(1.060, 0.139)

Reason for investigation Psychological & emotional abuse Physical, material & health neglect Parent high risk life School truancy & neglect Risk of sexual abuse Sexual abuse Behavioral problems Risk of/or physical ab.(ref)

.220 .374 .532 .333 −.540 −.460 .738

.045 .040 .028 .055 .122 .058 0.71

23.9 87.7 352.1 37.1 19.36 62.3 107.8

1.246⁎⁎⁎(1.141, 1.361) 1.454⁎⁎⁎(1.344, 1.572) 1.703⁎⁎⁎(1.611, 1.800) 1.395⁎⁎(1.253, 1.553) .583⁎⁎⁎(.459, .740) .631⁎⁎⁎(.563, .707) 2.092⁎⁎⁎(1.870, 2.405)

1.274⁎⁎⁎(1.167, 1.391) 1.513⁎⁎⁎(1.399, 1.636) 1.744⁎⁎⁎(1.650, 1.844) 1.436⁎⁎⁎(1.290,1.599) .584⁎⁎⁎(.450, .741) .627⁎⁎⁎(.559, .702) 2.125⁎⁎⁎(1.849, 2.443)

Source of referral CL5C Youth protection Police Other prof. institutions School Hospital staff Unidentified family (ref) Number of investigations Socioeconomic dis.

.272 .305 −.056 −.066 .047 .586 .099 .081 .210

.033 .032 .032 .038 0.35 0.31 .046 .007 .010

69.3 91.5 3.01 2.97 1.86 347.6 4.72 150.0 408.3

1.312⁎⁎⁎(1.231, 1.399) 1.356⁎⁎⁎(1.274, 1.443)

1.311⁎⁎⁎(1.230, 1.398) 1.359⁎⁎⁎(1.277, 1.477)

.945(.887, 1.007) .936(.868, 1.009) 1.049(.980, 1.122) 1.796⁎⁎⁎(1.689, 1.910) 1.104⁎(1.010, 1.207) 1.084⁎⁎⁎(1.070, 1.098) 1.233⁎⁎⁎(1.209, 1.259)

.946(.887, 1.008) .924⁎(.857, .996) 1.046(.977, 1.120) 1.788⁎⁎⁎(1.681, 1.901) 1.097⁎(1.003, 1.199) 1.089⁎⁎⁎(1.075, 1.103)

⁎ P b 0.05. ⁎⁎ P b 0.01. ⁎⁎⁎ P b 0.001.

particularly for the children aged 0 to 9 years, appear to be related to neighborhood area socioeconomic disadvantages. 4.2. Hazard analysis for children 0 to 9 years old Table 2 presents estimates of the effect of case level factors and neighborhood area socioeconomic disadvantages on the risk of experiencing an initial out-of-home placement for children aged 0 to 9 years at investigation. For children aged 0 to 9 years, the block 1 model produced a −2 Log probability statistic of 268,164 (df = 2), the block 2 model produced a Log statistic of 266,391 (df = 18) and the final model produced a Log statistic of 265,979 (df=19). The decreasing Log probability estimates suggest that the final model for children aged 0 to 9 years is a better model fit. The chi-square (χ2) changes between blocks suggest that individual factors and neighborhood socioeconomic disadvantages contribute to the overall risk of placement. The neighborhood area socioeconomic disadvantage variable was added to the model in the final block and is much higher for children 0 to 9 years old (χ2 = 411.9, df = 1, P b .001) compared to children 10 to 17 years old (χ2 = 47.6, df = 1, P b .001—as shown in Table 3). This indicates that neighborhood area socioeconomic disadvantages, although associated with placement for all children, are most influential for younger children aged 0 to 9 years. Factors that increase the risk of placement include: being aged 0 to 1 or 6 to 9 years at the time of the initial investigation; being male; being investigated because of psychological or emotional abuse, physical, material or health related neglect, school truancy or school neglect, behavioral problems, and having parents' with high risk lifestyles; being referred by a community health and social services clinic and child protection agency, a hospital staff member or an unspecified source; being referred for multiple investigations; and the child's

neighborhood's level of socioeconomic disadvantages. Although children aged 6 to 9 years who are investigated for behavioral problems are at an increased risk of experiencing a placement, they represent less than 2% of placed children aged 0 to 9 years at initial investigation (see Table 1). Those investigated for sexual abuse (risk of or confirmed) are less at risk of experiencing a placement compared to children investigated for physical abuse (risk of or confirmed). A referral for investigation by the police, school or other professional institution did not significantly contribute to the risk of placement. While children aged 0 to 1 and 6 to 9 years are significantly more at risk of experiencing a placement than children aged 2 to 5, those aged 0 to 1 year at initial investigation are at an increased risk of experiencing an out-of-home placement. For these same children, however, the risk of placement significantly decreases between blocks, going from a hazard ratio of 2.549 to 1.924 in the final model. In contrast, the risk of placement increases for children aged 6 to 9 years, going from a hazard ratio of 1.079 to 1.100 in the final model. It seems that the risk of placement for 0 to 1 year olds is reduced in part as a function of the person who is referring the child for investigation, child's gender, the reason for the investigation, the number of investigations, and the neighborhood area socioeconomic disadvantages. For children aged 6 to 9 years, however, there is a compounding effect as these same factors contribute to an increased risk of placement. Neighborhood area socioeconomic disadvantage was considered to be a statistically significant predictor of placement and seems to increase the individual risk of placement for all factors except for the reasons children were investigated. For every one-unit increase over the average neighborhood socioeconomic disadvantage estimate, children were 1.233 times more likely to experience a placement. Within each category, the most influential factors predicting an increased risk of placement were: children aged 0 to 1 year (Wald = 775.3), males (Wald = 28.0), children investigated

T. Esposito et al. / Children and Youth Services Review 35 (2013) 2031–2039

2037

Table 3 Cox proportional hazard models of initial out-of-home placement for children aged 10 to 17 years. Number of events and censored values Total

Events

Censored

% Censored

55,110

16,691

38,419

69.7% Block 2

Block 1

Wald

Adj. HR (95% CI)

Adj. HR (95% CI)

Adj. HR (95% CI)

1.574⁎⁎⁎(1.519, 1.631)

1.569⁎⁎⁎(1.515, 1.626)

2.178⁎⁎⁎(2.109, 2.249)

Blocks (final model) Beta Age at initial investigation 10–13 (ref) 14–17 Child sex Male (female ref) Reason for investigation Psychological & emotional abuse Physical material & health neglect Parent high risk life School truancy & neglect Risk of sexual abuse Sexual abuse Behavioral problems Risk of/or physical ab.(ref) Source of referral CLSC Youth protection Police Other prof, institutions School Hospital staff Unidentified family (ref) Request for youth criminal justice services Number of investigations Socioeconomic dis.

.453

SE

.018

634.0

−.008

.016

.268

.992(.961, 1.074)

.991(.960, 1.023)

−.003 −.246 −.080 .022 −1.261 −.499 .999

.042 .062 .028 .048 .146 .043 .025

.005 15.3 8.00 .214 74.4 131.5 1569

.997(.918, 1.083) .782⁎⁎⁎(.692, .884) .924⁎⁎(874, 976)

1.002(.922, 1.088) .794⁎⁎⁎(.702, .897) .931⁎(881, .983)

1.022 (.933, 1.122) .283⁎⁎⁎(.213, .377) .607⁎⁎⁎(.558, .661) 2.716⁎⁎⁎(2.585, 2.853)

1.035(.943, 1.136) .284⁎⁎⁎(.213, .378) .611⁎⁎⁎(.561, .665) 2.723⁎⁎⁎(2.592, 2.861)

.001 −.238 .120 −.244 −.359 −.048 −.262 −.354 .197 .058

.028 .032 .022 .038 .023 .040 .053 .022 .010 .008

.002 53.9 28.9 41.8 246.9 1.44 24.4 268.3 414.7 47.6

1.001(.948, 1.058) .788⁎⁎⁎(.727, .843) 1127⁎⁎⁎(1.079, 1.178) .783⁎⁎⁎(.727, .843) −698⁎⁎⁎(.668, .730)

1.003(.949, 1.059) .794⁎⁎⁎(.743, .844) 1130⁎⁎⁎(1.082, 1.180) .782⁎⁎⁎(.726, .842) .700⁎⁎⁎(.669, .732)

.953(.882, 1.031) .770⁎⁎⁎(.694, .854) .7 02⁎⁎⁎(.673, .732) 1217⁎⁎⁎(1.195, .1.241) 1.059⁎⁎⁎(1.042, 1.077)

.950(.878, 1.027) .773⁎⁎⁎(.697, .857) .704⁎⁎⁎(.675, .735) 1.221⁎⁎⁎(1.198, 1.244)

⁎ P b 0.05. ⁎⁎ P b 0.01. ⁎⁎⁎ P b 0.001.

because of their parents' high risk lifestyle (Wald = 352.1), children reported by a hospital staff member (Wald = 347.6), children who experience a higher number of investigations (Wald = 150.0) and who come from neighborhoods with more socioeconomic disadvantages (Wald = 408.3). 4.3. Hazard analysis for children 10 to 17 years old Table 3 presents estimates of the effect of case level factors and neighborhood area socioeconomic disadvantages on the risk of experiencing an initial placement for children aged 10 to 17 years at investigation. The block 1 model produced a −2 Log probability statistic of 353,401 (df = 1), the block 2 model produced a Log statistic of 347,896 (df = 18) and the final model produced a Log statistic of 347,848 (df = 19). The decreasing Log probability estimates suggest that the final model for children aged 10 to 17 years is a better model fit. Factors that increase the risk of placement include: being aged 14 to 17 years at the time of the initial investigation; being investigated because of behavioral problems; being referred for investigation by the police; being referred for multiple investigations; and the child's neighborhood level socioeconomic disadvantages. There is no significant gender effect on the risk of placement and those investigated for physical and health neglect, their parents' high risk lifestyle, and sexual abuse (risk of or confirmed) are significantly less at risk of experiencing a placement than those investigated for physical abuse (risk of or confirmed). Children referred for investigation by a child protection agency, school, and other professional institution or by an unidentified entity are less at risk of experiencing a placement than those referred by a family member. Children aged 10 to 17 years who received a request for youth criminal justice services are significantly less at risk of placement compared to those with no request for youth

criminal justice services. Neighborhood area socioeconomic disadvantages were considered to be a statistically significant predictor of placement. For every one-unit increase over the average neighborhood area socioeconomic disadvantages, older children were 1.059 times more likely to experience a placement. Within each category, the most influential factors enabling us to predict an increased risk of initial placement were the following: children aged 14 to 17 years (Wald = 634.0), children investigated because of behavioral problems (Wald = 1569), children referred for investigation by the police (Wald = 28.9), children who experience a higher number of investigations (Wald = 414.7) and who come from neighborhoods with more socioeconomic disadvantages (Wald = 47.6). 5. Discussion The results of this study are rather encouraging and in-line with child protection objectives of family preservation.3 The majority of children served by the Québec child protection system in the last decade are not placed in out-of-home care. Of those who are, the oldest followed by the youngest are admitted to out-of-home care the most often and the most rapidly. This study has identified who these children are. In line with Wulczyn et al. (2002), Runyan et al. (1982) and James et al. (2006), the current study finds that the high risk lifestyle of parents and hospital referrals contribute significantly to the risk of 3 “Family preservation” does not exclusively refer to the short-term child protection services often referred to as family preservation services but rather to family-centered practice that promotes overall family functioning by accessing a plethora of child protection and community resources to support and preserve families in order to prevent the removal of children from the home and encourage family reunification for placed children.

2038

T. Esposito et al. / Children and Youth Services Review 35 (2013) 2031–2039

placement for the very young. Additionally, it echoes the earlier studies of Berger et al. (2009), Walrath and Liao (2005), James et al. (2006), Brook and McDonald (2009), and Lindsey (1991) in that behavioral problems and police interventions seem to be the driver of out-ofhome placement for older children. It may be that hospital staff and police referrals are an indication of the severity of the situation, or that these organizations unduly influence caseworkers' decisions to remove older children from the home. Although source of referral plays an important role in predicting the risk of initial placement, future studies should examine such possible explanations in detail. This study also reports that older children with a request for youth criminal justice services were not at increased risk of placement. It is possible, then, that many children served by both the child protection and youth justice system are engaged with multiple services (i.e. extra judiciary services), which may reduce the risk of placement. Although this was a first attempt to examine the relationship between youth criminal justice services and placement, future research can benefit from examining whether there is a multi-directional relationship between youth criminal justice services and out-of-home placement. Consistent with other studies, as the number of times a child is investigated for maltreatment increases, so does the probability of placement — a result that may be attributed to the lack of early detection and evaluation of the child's risk of compromised security and development. Special efforts should be made to intervene early and support families from the point of initial contact with the child protection system. Neighborhood area socioeconomic disadvantages significantly increase the risk of out-of-home placement and emerge as more influential for the removal of children aged 0 to 9 years compared to older children. These families struggle financially, are often overwhelmingly poor, with limited resources and supports (Baumann et al., 2009; Coulton, Crampton, Irwin, Spilsbury, & Korbin, 2007; Drake & Pandey, 1996; Freisthler, Merritt, & LaScala, 2006; Lindsey, 1991; Pelton, 1989). Addressing the socioeconomic challenges is often seen as outside the purview of what child protection services can or should manage. While the risk of placement increases as a function of neighborhood area socioeconomic disadvantages, it is not acceptable to remove children from difficult family situations that result from inadequate socioeconomic resources. As suggested by scholars such as Coulton et al. (2007) and Lindsey and Shlonsky (2008), among others, maltreatment incidences and ultimately placement can best be reduced if socioeconomic disadvantages are mitigated. 6. Practice implications Québec's child protection statute stipulates that parents and children are entitled to receive health and social support services that are deemed both helpful for child development and for improving parent's capacity to assume the care of their children (Youth Protection Act, 2007, art. 8). While scholars such as Swift and Callahan (2006) refer to Québec as a Canadian example when it comes to efforts to provide a range of community health and social services, the extent to which these services are available to children and youth involved with child protection services is questionable. The suggestion here is not necessarily to increase the availability of community support services; rather it is to ensure priority access for these children and youth. Specifically, mental health services for adolescents with behavioral problems, and substance abuse counseling and therapeutic mental health services for high-risk parents of younger children. Additionally, income and housing supports must be a priority given that socioeconomic disadvantage increases the risk of placement, especially for younger children. 7. Limitations While the methodology of this study is unique in allowing for a broad provincial epidemiological analysis of factors that increase a

child's risk of experiencing an initial placement, it is not without limitations. One such limitation is that the neighborhood area socioeconomic disadvantage index was created as a composite of several socioeconomic indicators. These indicators are based on the neighborhoods in which children live, which may not directly correspond to the socioeconomic disadvantages of the families involved in this study as a result of aggregation biases. We attempted to address this neighborhood bias by using the smallest geographic unit in the census, with a population ranging from 400 to 700 persons (Statistics Canada, 2011). Geronimus and Bound (1998), Geronimus, Bound, and Neidert (1996) and Soobader, LeClere, Hadden, and Maury (2001) all suggest that, compared to studies using neighborhood socioeconomic status aggregated at the broad neighborhood level, using zip code and census tract data, or even smaller geographic units as measures of neighborhood socioeconomic status, reduces that bias. There are a number of other limitations to this study. First, each child is assigned an identifier within each of the sixteen child protection jurisdictions in Québec. This identifier, however, is unique within a child protection jurisdiction, but not across child protection jurisdictions. It is quite possible that families have moved between child protections jurisdictions, yet we are not able to identify the families that have received prior services from other jurisdictions. Second, this study used clinical–administrative data that may underestimate the actual prevalence of placement, since the children's informal unsubsidized placements with kith or kin are not captured here. Finally, this study did not adjust for the correlation that may arise because of siblings. The clinical–administrative data does not allow us to identify siblings at this point. This study, however, is the first to use longitudinal provincebased data in Canada allowing for an ecological examination of the factors influencing when out-of-home placement is most likely to occur and for whom. It is also unique in that it uses age-specific multivariate Cox proportional hazard regression models allowing us to explore age-specific factors associated with placements. Lastly, this study is unique as far as exploring the influences of youth criminal justice services and neighborhood area socioeconomic disadvantages on the risk of out-of-home placement for the province of Québec. Together, the results of this study can lead to policies and interventions aimed at minimizing, when possible, the reliance on out-of-home placement while ensuring the security and optimal development of children. References Baumann, D. J., Fluke, J., Graham, J. C., Wittenstrom, K., Hedderson, J., Riveau, S., et al. (2009). Disproportionality in child protective services: The preliminary results of statewide reform efforts. : Texas Department of Family and Protective Services. Berger, L. M., Bruch, E. I. J., James, S., & Rubin, D. (2009). Estimating the “impact” of out-of-home placement on child well-being: Approaching the problem of selection bias. Child Development, 80(6), 1856–1876. Berger, L. M., & Waldfogel, J. (2004). Out-of-home placement of children and economic factors: An empirical analysis. Review of Economics of the Household, 2, 387–411. Brook, J., & McDonald, T. (2009). The impact of parental substance abuse on stability of family reunifications from foster care. Children and Youth Services Review, 31(2), 193–198. Coulton, C., Crampton, D., Irwin, M., Spilsbury, J., & Korbin, J. (2007). How neighborhoods influence child maltreatment: A review of the literature and alternative pathways. Child abuse and Neglect, 31, 117–1142. Doyle, J. (2007). Child protection and adult crime: Using investigator assignment to estimate causal effects of foster care. Journal of Political Economy, 116(4), 746–772. Drake, B., & Pandey, S. (1996). Understanding the relationship between neighborhood poverty and specific types of child maltreatment. Child abuse and neglect, 20, 1003–1018. Esposito, T., Trocmé, N., Chabot, M., Duret, A., & Gaumond, C. (in press). Gestion fondée sur les indicateurs de suivi clinique en protection jeunesse. Jeunesse en tête: au-delà des risques, les besoins de développement des enfants. : Presses du l'Université du Québec (in press). Farmer, E. M. Z., Mustillo, S. M., Burns, B. J., & Holden, E. W. (2008). Use and predictors of out-of-home placements within systems of care. Journal of emotional and behavioral disorders, 16(5), 5–13. Frees, E. W. (2004). Longitudinal and panel data: analysis and applications in the social sciences. Cambridge University Press.

T. Esposito et al. / Children and Youth Services Review 35 (2013) 2031–2039 Freisthler, Merritt, & LaScala (2006). Understanding the ecology of child maltreatment: A review of the literature and directions for future research. Child maltreatment, 11(3), 263–280. Geronimus, A. T., & Bound, J. (1998). Use of census-based aggregate variables to proxy for socioeconomic group: Evidence from national samples. American Journal of Epidemiology, 148(5), 475–486. Geronimus, A. T., Bound, J., & Neidert, L. J. (1996). On the validity of using census geocode factors to proxy individual socioeconomic factors. Journal of the American Statistical Association, 91, 529–537. Hill, R. B. (2007). An analysis of racial/ethnic disproportionality and disparity at the national, state, and county levels. North Carolina: Casey-CSSP alliance for racial equity in child welfare. James, S., Leslie, L. K., Hurlburt, M. S., Slymen, D. J., Landsverk, J., & Davis, I. (2006). Children in out-of-home care: Entry into intensive or restrictive mental health and residential care placements. Journal of Emotional and Behavioural Disorders, 14, 196–208. Jonson-Reid, M., & Barth, R. P. (2000a). From maltreatment report to juvenile incarceration: The role of child welfare services. Child Abuse & Neglect, 24(4), 505–520. Jonson-Reid, M., & Barth, R. P. (2000b). From placement to prison: The path to adolescent incarceration from child welfare supervised foster or group care. Children and Youth Services Review, 22(7), 493–516. Kutner, M. H., Nachtsheim, C. J., & Neter, J. (2004). Applied linear regression models (4th ed.): McGraw-Hill Irwin. Lery, B. (2009). Neighborhood structure and foster care entry risk: The role of spatial scale in defining neighborhoods. Children and Youth Services Review, 31, 331–337. Lindsey, D. (1991). Factors affecting the foster care placement decision: An analysis of national survey data. American Journal of Orthopsychiatry, 61(2), 272–281. Lindsey, D., & Shlonsky, A. (2008). Child welfare research. Advances for practice and policy. London, UK: Oxford University Press. Pelton, L. H. (1989). For reasons of poverty: A critical analysis of the public child welfare system in the United States. New York: Praeger. Runyan, D. K., & Gould, C. (1985). Foster care for child maltreatment: Impact on delinquent behavior. Pediatrics, 75, 562–568. Runyan, D. K., Gould, L. C., Trost, C. D., & Loda, A. F. (1982). Determinants of foster care placement for the maltreated child. Child Abuse and Neglect, 6, 343–350. Sinha, V., Trocmé, N., Fallon, B., MacLaurin, B., Fast, E., Thomas Prokop, S., et al. (2011). Kiskisk Awasisak: Remember the children. Understanding the overrepresentation of first nations children in the child welfare system. Ontario: Assembly of First Nations.

2039

Soobader, M. J., LeClere, F. B., Hadden, W., & Maury, B. (2001). Using aggregate geographic data to proxy individual socioeconomic status: Does size matter? American Journal of Public Health., 91(4), 632–636. Statistics Canada (2011). Dissemination area. Retrieved from. http://www12.statcan.ca/ census-recensement/2006/ref/dict/geo021-eng.cfm Swift, K., & Callahan, M. (2006). Problems and potential of Canadian child welfare. In N. Freymond, & G. Cameron (Eds.), Towards positive systems of child and family welfare. Toronto: University of Toronto Press. Texas Health Human Services Commission (2006). Disproportionality in child protective services: Statewide reform effort begins with examination of the problem. Department of Family and Protective Services. Thoburn, J. (2010). International perspectives on foster care. In E. Fernandez, & R. P. Barth (Eds.), How does foster care work. London, UK: Jessica Kingsley Publishers. Trocmé, N., Fallon, B., MacLaurin, J., Daciuk, C., Felstiner, T., Black, L., et al. (2005). Canadian incidence study of reported child abuse and neglect. : Public Health Agency of Canada. Trocmé, N., Fallon, B., MacLaurin, B., Sinha, V., Black, T., Fast, E., et al. (2010). Canadian incidence study of reported child abuse and neglect. Public Health Agency of Canada. United Nations (2009). Guidelines for the alternative care of children: A United Nations framework. Retrieved from: http://www.iss-ssi.org/2009/index.php?id=25 Walrath, C., & Liao, Q. (2005). The clinical and psychosocial factors of children with serious emotional disturbance entering system-of-care services. In M. Epstein, K. Kutash, & A. Duchnowski (Eds.), Outcomes for children and youth with behavioral and emotional disorders and their families: Programs and evaluation best practices (pp. 46–68) (2nd ed.). Austin, TX: PRO-ED. Walters, S. J. (1999). What is a Cox model? Hayward Medical Communications. Widom, C. S. (1991). The role of placement experiences in mediating the criminal consequences of early childhood victimisation. American Journal of Orthopsychiatry, 6, 195–209. Wulczyn, F., Barth, R., Yuan, Y., Harden, J., & Landsverk, J. (2005). Beyond common sense: Child welfare, child well-being, and the evidence for policy reform. New Brunswick, NJ: Transaction. Wulczyn, F., Hislop, K. B., & Harden, B. J. (2002). The placement of infants in foster care. Infant Mental Health Journal, 23(5), 454–475. Wulczyn, F., & Lery, B. (2007). Racial disparity in foster care admissions. Chicago: Chaplin Hall Center for Children at the University of Chicago. Yampolskaya, S., Armstrong, M. I., & McNeish, R. (2011). Children placed in out-of-home care: Risk factors for involvement with the juvenile justice system. Violence and Victims, 26(2), 231–245. Youth Protection Act (2007). Law as amended by Bill 125. R.S.Q. (chapter P-34.1).