Attachment of young foster children

Flemish foster care

In Flanders (Dutch speaking part of Belgium) as from 2009, family foster care is the first option when a child is in need for out-of-home care (Vanderfaeillie & Van Holen, 2010; Van den Bruel, 2012). Flanders consist out of five provinces, with each their own foster care agency. The number of foster care situations in Flanders is rising and in 2019, 7756 children were placed in Flemish foster care. Interestingly, both the proportion of placements in kinship care (both placements in broader family or social network) and the number of kinship foster families keeps increasing. In 2019, 72% of all placements were kinship families (Pleegzorg Vlaanderen, 2020).

Flemish foster children can be placed in short- or long-term placements. Short-term foster care has a maximum duration of one and a half years. During this period the possibility of reunification with the birth parents is examined. If reunification cannot be realized, another perspective for the child (i.e., long-term foster care or residential care) has to be developed. Although long-term foster care aims at creating permanency for the foster child, even for these placements reunification can be considered at any moment. Children are placed in Flemish family foster care on a voluntary (62.7%) or court-ordered basis (37.3%) (Pleegzorg Vlaanderen, 2018). Of the new started Flemish foster care placements in 2019, approximately 36% of the foster children were under six years old and 48% were under 8 years old (Pleegzorg Vlaanderen, 2020).

Attachment in foster care

The behavior children show towards an attachment figure when they are upset is an indication of the quality of the attachment relationship. In the regular population, 62% of the children are securely attached to their attachment figures, 15% avoidant, 9% ambivalent and 15% disorganized (Van Ijzendoorn et al., 1999). For secure attached children the caretaker is a safe base, who is sought for protection or comfort when they are upset. In a non-threatening situation, children who developed a secure attachment relationship find a balance between proximity seeking and exploring the environment (Ainsworth et al., 1978). Some children are insecurely attached. Children can for example display anxious-avoidant attachment behavior. In general, these children are more focused on exploration and seem to need little support or comfort form the attachment figure in an unknown or worrying situation (Ainsworth et al., 1978). Anxious-ambivalent attached children display more ambivalent behavior. They often struggle to explore and seek proximity to the attachment figure. They tend to stay upset after a stressful situation and have troubles regulating their emotions (Ainsworth et al., 1978). Disorganized attached children display incoherent attachment behaviors (Lieberman & Zeanah, 1995). They display a chaotic variation of the behaviors described above. They can for example seek proximity of their care-giver and then pull away strongly (Lieberman & Zeanah, 1995).

The history of a foster child is often characterized by traumatic experiences such as abuse or neglect (Oswald et al., 2010; De Baat et al., 2014). In a Flemish sample for example, besides separation from their primary caregivers, 85% of the foster children encountered at least one extra traumatic event (Vanderfaeillie et al., 2014). Children who had negative experiences with attachment figures can have more difficulties forming secure attachments afterwards (Ainsworth et al., 1978; Juffer, 2010).

Research on rates of secure and insecure attachment relationships of foster children is inconsistent, ranging from not significantly different than a normative group (Jacobsen et al., 2014; Oosterman & Schuengel, 2008; Ponciano, 2010), to foster children being more often insecurely attached (Gabler et al., 2014; Lamb et al., 1985; Bovenschen et al., 2016) or only showing more disorganized attachment behavior (Cole, 2006; Dozier et al., 2001). Two meta-analyses examined the security of attachment between foster children and their carers. They both found similar results. No differences in the rates of secure attachment between children and their foster carers and children living at home and their parents were found, but foster children were found to be more likely to have disorganizes attachment relationships than regular children living with their parents (Van den Dries et al., 2009; Vasileva & Petermann, 2018).

The quality of attachment between a child and attachment figures is an important factor in a child’s development. Children who developed a secure attachment relationship with at least one significant other tend to have better developmental outcomes (Van Ijzendoorn, 2008). A secure attachment relationship can even buffer other risk factors (Belsky & Fearon, 2002; Morisset et al., 1990). An insecure attachment relationship with attachment figures is associated with psychosocial problems (Bohlin et al., 2000; Groh et al., 2014; Pallini et al., 2014; Speltz et al., 1999; Verissimo et al., 2014), externalizing behavioral problems (Fearon et al., 2010), and internalizing behavioral problems (Groh et al., 2012). Especially an insecure disorganized attachment is considered a risk factor, as it is associated with difficulties in dealing with stress (Van Ijzendoorn et al., 1999), externalizing behavioral problems (Lyons-Ruth et al., 1993; Van Ijzendoorn et al., 1999), and psychopathology in later life (Van Ijzendoorn et al., 1999). Insecure attachment behavior is associated with a higher likelihood of foster placement breakdown (Stone & Stone, 1983; Strijker et al., 2008), which in turn has negative consequences for the foster child (Van den Bergh, 2010; Vanderfaeillie et al., 2012), the foster parents (Wilson et al., 2000) and foster care workers (Rostill-Brookes et al., 2011).

Associated factors with quality of attachment in foster care

In a recent review of West et al. (2020), the factors associated with the quality of attachment of foster children with their foster parents were listed. The characteristics were divided into four categories: factors related to the foster parents, the placement, the former home environment, and the foster child. Regarding foster parent factors, most studies found a positive association between quality of attachment and positive, sensitive parenting of the foster parents (Bovenschen et al., 2016; Gabler et al., 2014; Lang et al., 2016; Oosterman & Schuengel, 2008; Ponciano, 2010; Quiroga & Hamilton-Giachritsis, 2017). In addition, traumatic experiences during the childhood of foster parents as well as insecure attachment representations of foster parents lowered the likelihood of secure attachment between foster carers and foster children (Cole, 2005a; Dozier et al., 2001; Stovall-McClough & Dozier, 2004). The motivation to foster appeared to have an impact on the quality of attachment. Wanting to increase the family size or the wish to make a difference for the caregiver’s community were both associated with higher rates of secure attachment. Starting foster care out of religious reasons, to replace an older child who left home, and foster care with a view to adoption were all associated with lower rates of secure attachment (Cole, 2005a; Quiroga & Hamilton-Giachritsis, 2017). Cole (2005a) did not detect an association with social support. The association of the following foster parent related characteristics with quality of attachment between foster children and their foster carers remained uncertain: the number of foster children in the family (Lang et al., 2016; Ponciano, 2010; Quiroga & Hamilton-Giachritsis, 2017; Dozier et al., 2001), parenting stress of foster parents (Gabler et al., 2014; Cole, 2005a), age of foster parents (Gabler et al., 2014; Quiroga & Hamilton-Giachritsis, 2017) and the type of placement (kinship or no-kinship placement; Cole, 2006; Tarren-Sweeney, 2008).

Regarding the placement characteristics, the number of previous placements did not seem to affect the quality of attachment (Bernier et al., 2004; Dozier et al., 2001; Lang et al., 2016; Bovenschen et al., 2016). The association with duration of the current placement in foster care (Bovenschen et al., 2016; Gabler et al., 2014; Oosterman & Schuengel, 2008; Tarren-Sweeney, 2008; De Schipper et al., 2012), psychological problems of birth parents (Bovenschen et al., 2016; Lang et al., 2016) and the association with physical, sexual or emotional abuse prior to the placement (Gabler et al., 2014; Lang et al., 2016; Tarren-Sweeney, 2008) remained unclear.

Regarding child factors, foster children who have more negative affectivity and shyness were less likely to develop a secure attachment relationship (Bovenschen et al., 2016). The ethnicity of the foster children had no effect on attachment (Bernier et al., 2004) and the number of foster children’s medical problems had a negative effect on attachment (Cole, 2006). The influence of the following child factors was unclear: gender (Gabler et al., 2014; Lang et al., 2016; Oosterman & Schuengel, 2008; Bernier et al., 2004), contact with birth parents (Lang et al., 2016; Ponciano, 2010), age (Lang et al., 2016; Ponciano, 2010; De Schipper et al., 2012; Bernier et al., 2004) and age at the start of placement (Stovall-McClough & Dozier, 2004; Bovenschen et al., 2016; Gabler et al., 2014; Dozier et al., 2001; Tarren-Sweeney, 2008; De Schipper et al., 2012; Bernier et al., 2004).

Research question

Sufficient attention should be paid to the attachment relationships of foster children as the quality of attachment between foster children and significant others, for example foster carers, impacts significantly the overall development of a foster child (Van Ijzendoorn, 2008). Moreover, there is an association between placement breakdown and attachment behavior (Stone & Stone, 1983; Strijker et al., 2008). International research regarding associated factors with attachment quality is scarce and inconclusive. To our knowledge, there’s no Flemish research regarding associated factors with attachment quality of foster children.

The research question is: “What factors are associated with insecure attachment behavior between foster children and their foster mothers?” Drawing on results from the review of West et al. (2020), our hypothesis is that especially sensitive parenting of foster parents will be associated with the quality of attachment and the number of previous placements will not be associated with the quality of attachment. To get a better image of the quality of attachment behavior of young Flemish foster children perceived by their foster mothers, we also added an extra research question: “Have young Flemish foster children more often an insecure attachment relationship with their foster mother (as perceived by foster mothers) than children from the norm population?” Based on the two meta-analyses, our hypothesis is that there will be no difference in rates of insecure attachment relationships.

Procedure

The pretest data of an ongoing randomized controlled trial (RCT) investigating the effectiveness of Video-feedback Intervention to promote Positive Parenting and Sensitive Discipline in Foster Care (VIPP-FC) was used for this study. Recruitment of foster families was done by all Flemish foster agencies, which provided information on the study to the eligible families, provided the informed consents to foster mothers and returned the informed consents of the participating families. All families welcoming a new foster child in their family in the past year were contacted by the foster care agencies. Placements of children with a duration of more than 1 year can also be included in the RCT when the families or the foster care workers assesses the risk for attachment, parenting or behavior problems as high. For this study, only new placements were included. Together with the informed consent, foster care workers filled in a questionnaire regarding characteristics related to the foster child (age, gender), the foster care placement (kinship/non-kinship, age at placement), the case history (number of previous placements, reason for out-of-home placement), the parents (amount of contact with their child) and the foster parents (age, education).

When an informed consent was collected, the first author sent out the questionnaires by mail to the foster mothers. The foster mother participated together with one foster child. Children had to be between 18 months and six years old at time of inclusion and without a known cognitive or severe motor disability. When more than one foster child was eligible, the youngest child was included. Twins were excluded from the study. Because the foster mothers needed to fill in multiple questionnaires, they had to be able to read Dutch (criterium: ‘able to read and understand a Dutch paper’). All procedures were approved by the boards of the foster care agencies and by the ethical board of the faculty of Psychology and Educational sciences of the Vrije Universiteit Brussel (ECWH_146).

Questionnaires

Statistical analysis

First, descriptive analysis was done. Comparisons with norm groups were made using One-Sample t-tests. Next, bivariate correlations were performed to examine associations between independent variables and the AISI Total score. As there were no T-scores available for the dependent variable (AISI Total score) and one of the independent variables, we used the raw scores for the CBCL subscales and PSQ as well. All independent variables associated with the AISI Total score at a p ≤ .10 level were included in a multiple regression model (Bursac et al., 2008). Because pretest data was used, the amount of missing data was limited to some missing variables on the characteristics of the foster children and their foster families; 2 children whose history of abuse and/or neglect was unknown, 4 missing values regarding ‘number additional children in foster home’, ‘amount of contact with mother’ and ‘amount of contact with father’. We imputed the mean for the continuous variables and the median for data on interval level. In case of missing data for dichotomous data, listwise deletion was used per analysis resulting in numbers of cases differing per analysis. The largest Cook’s distance value was 0.24, suggesting that outliers were not problematic. Because the sample was rather small, a conservative approach was used and all observations with a Cook’s D of more than 3 times the mean were considered possible outliers (Boussiala, 2020). When the analysis were redone after five possible outliers were removed, results were similar, thus final analysis were conducted using the original dataset. There was evidence of multicollinearity between internalizing, externalizing and total problem behavior, with the highest VIF score being 36, 06. Only total problem behavior was included in the multiple linear regression to avoid multicollinearity. The remaining assumptions were met. All statistical tests were done two-tailed and an alpha level of p < .05 was used. All calculations were done using SPSS 27.

Description of the sample

The sample consisted of 68 foster children and their foster mothers, the descriptive statistics can be found in Table 1.

OPEN IN VIEWER

Table 1.

Sample characteristics.

Variable	n (%)/M (SD)	N
Age in months
(Range 18–74)	36.43 (14.67)	68
Age in months at start current placement
(Range 0–64)	21 (17.09)	68
Gender
Boys	34 (50%)	68
Girls	34 (50%)
Placement type
Kinship	26 (38.2%)	68
Non-kinship	42 (61.8%)
Number out-of-home placements
(Range 0–6)	1.03 (1.12)	68
Number additional children in foster home
(Range 0–4)	1.24 (1.24)	64
Abuse/neglect
Yes	31 (47%)	66
No	35 (53%)
Frequency of contact with mothers
Never	10 (15.6%)	64
Less than every 2 months	1 (1.6%)
Monthly	13 (20.3%)
Every other week	29 (45.3%)
Weekly	11 (17.2%)
Frequency of contact with fathers
Never	21 (33.3%)	63
Less than every 2 months	2 (3.2%)
Monthly	11 (17.5%)
Every other week	24 (38.1%)
Weekly	5 (7.9%)

In Table 2, the mean scores for the various questionnaires and (if available) corresponding T-scores are shown. The mean T-score for the total PSQ-score was a non-clinical T-score of 56 for the 3 years and younger group and a subclinical T-score of 60 for the 4 years and older group. All the mean T-scores of the CBCL scales and the AISI Total scale lie within the normal range.

OPEN IN VIEWER

Table 2.

Questionnaires’ discriptives.

Variable	Range	M (SD)	M t-score	N
Emotional availability–Self report	91–143	121.51 (12.769)	/	68
Parenting stress questionnaire	39–85	53.81 (10.122)	/	68
Parenting stress questionnaire ≤3 y	39–77	53.19 (9.399)	56	47
Parenting stress questionnaire ≥4 y	41–85	54.98 (12.376)	60	18
Child behavior checklist internalizing problems	0–35	10.18 (7.775)	55	68
Child behavior checklist externalizing problems	2–41	15.81 (10.126)	55	68
Child behavior checklist total problems	4–106	37.82 (24.790)	54	68
Total scale for attachment insecurity of the attachment insecurity screening inventory	21–76	46.38 (11.472)	/	68

In Table 3, results of the PSQ, CBCL and AISI were compared with norm scores that could be found in the manuals of the questionnaires (Polderman & Kellaert-Knoll, 2008; Wissink et al., 2016; Vermulst et al., 2012; Verhulst et al., 2000). For all questionnaires, the results were significantly higher than the norm scores. Twenty-six children (38%) had a clinical total insecure attachment score, which suggests an insecure attachment relationship between the foster child and the foster mother.

OPEN IN VIEWER

Table 3.

Comparisons to norm scores.

	M (SD)	df	t	p	Cohen’s D
Parenting stress questionnaire
Parenting stress questionnaire ≤3 y	53.81 (10.122)	46	3.342	.002	.487
Parenting stress questionnaire >4 y	54.98 (12.376)	17	2.327	.033	.549
CBCL internalizing problems	10.18 (7.775)	67	4.112	<.001	.499
CBCL externalizing problems	15.81 (10.126)	67	4.405	<.001	.534
CBCL total problems	37.82 (24.790)	67	4.365	<.001	.529
AISI total	46.38 (11.472)	67	3.538	.001	.429

Factors associated with attachment insecurity screening inventory total scores

As indicated in Table 4, the following factors were significantly associated with higher scores on the AISI Total scale: lower scores on the EA–SR and higher scores on the PSQ, CBCL internalizing problems scale, CBCL externalizing problem scale and CBCL total problems scale.

OPEN IN VIEWER

Table 4.

Associations of characteristics of foster children, foster parents and foster placements with insecurity of attachment.

Variable	Attachment Insecurity Screening Inventory Total scores n = 68
Age in months	−.054
Age in months at start current placement	−.015
Gender (0 = male, 1 = female)	−.168
Placement type (0 = kinship, 1 = non-kinship)	.032
Number out-of-home placements	.042
Number additional children in foster home	.018
Abuse/neglect (0 = no history, 1 = history of abuse and/or neglect)	.138
Frequency of contact with mothers	.136
Frequency of contact with fathers	.088
Emotional availability–Self report	−.273**
Parenting stress questionnaire	.573****

Subsequently, all significant factors (for the CBCL scales only the CBCL total problem score to avoid multicollinearity) were entered in one multiple linear regression model, resulting in a significant model (F (3, 63) = 19.461, p < .000), with an R² of .481. Higher scores on the PSQ and on the CBCL total problems scale were positively associated with AISI Total scores. Table 5

OPEN IN VIEWER

Table 5.

Multiple regression model with attachment insecurity screening inventory total score as dependent variable.

Effect	Estimate	SE	95% CI		p
			LL	UL
Intercept	17.94	14.03	−10.1	45.97	.21
Emotional availability–self report	.04	.09	−.14	.22	.67
Parenting stress questionnaire	.28	.14	.00	.551	.047
CBCL total problems	.24	.06	.13	.35	<.00
R2	.48
F	19.46

A post-hoc power analysis showed that statistical power of the Multiple Regression is adequate (0.99; G*Power 3.1.9.4) with a study sample of 68 foster families with α = 0.05.

Attachment of foster children

A total of 26 children (38%) scored clinical on the AISI, suggesting an insecure attachment relationship between the foster children and their foster mothers. This perentage is in line with the results of Ijzendoorn et al. (1999), who found that 38% of the children in the regular population have an insecure attachment relationship with an attachment figure. We can conclude that, following our hypothesis, the amount of insecure attachment relationships between foster children and their foster mothers (described by the foster mothers) is very similar to the regular population. This finding is supported by two meta-analysis, that found no differences in the rates of secure attachment between children and their foster carers and children living at home and their parents (Van den Dries et al., 2009; Vasileva & Petermann, 2018). According to these two meta-analyses, we would expect to find more disorganized attachment relationship in our foster care sample. However, the instruments used in this study do not allow to examine this question. Interesting is that the mean AISI Total score was non-clinical, but significantly higher than the mean of the norm group described in the manual (Polderman & Kellaert-Knoll, 2008), indicating that foster mother reported more insecure attachment behavior in the relationship with their foster children than parents of children in the regular population. This points to foster mothers experiencing more insecure attachment behavior in the attachment relationship with the foster child compared to mothers out of the regular population, but the rates of insecure attachment are according to foster mothers very similar to the regular population.

Associated characteristics

A number of child-related characteristics were not associated with attachment problems in this research. In this study, no association was found between the child’s age, both at the time of placement and the current age, and insecure attachment behavior. In the review on associated characteristics with security of attachment of West et al. (2020), the results regarding these included child variables (age at time of placement, current age, gender) were unclear.

Regarding placement and foster mothers’ characteristics, we expected no association with the number of previous placements. Moreover, there was no consensus regarding associations with the duration of the placement in foster care, abuse and/or neglect in the past, age of the foster parents, number of children in the home, and kinship or non-kinship care (West et al., 2020). In this study, placement type was not associated with quality of attachment, in line with Coles’ (2006) research, neither were number of additional children in the home (similar with previous research of Dozier et al. (2001) and Lang et al. (2016)). Being exposed to abuse and/or neglect was not associated with insecurity of attachment insecure attachment behavior in this study, as in the article of Lang et al. (2016) where no differences in attachment for children who were neglected of maltreatment prior to their placement were found. There was no association with the number of out-of-home-placements and being previously placed out-of-home, which is supported by a significant amount of research (Bernier et al., 2004; Bovenschen et al., 2016; Dozier et al., 2001; Lang et al., 2016). Age and education of foster mothers were also not associated with insecure attachment behavior. There was no association with insecure attachment behavior and frequency of contact with parents, in line with the results of Lang et al. (2016).

Other characteristics were indeed associated with insecure attachment behavior. Higher levels of parenting stress reported by foster parents were associated with higher scores of insecure attachment behavior. This is in line with research of Gabler et al. (2014), but in contrast to Cole (2005a) findings. Research in regular populations generally points to a positive association between parenting stress and insecure attachment relationships (e.g. Jarvis & Creasey, 1991). Jarvis and Creasey (1991) suggested that parenting stress might have a negative impact on creating a secure attachment relationship (through creating a consistent caregiving environment), but it is also possible that having a secure attachment relationship with their child leads towards lower stress levels.

Internalizing, externalizing and total problem behavior were associated with insecure attachment behavior on univariate level, but only total problem behavior was inputted in the regression model in order to avoid multicollinearity. More behavior problems were associated were more insecure attachment behavior. Gabler et al. (2014) and Hockaday (2018) found no association between security of attachment and behavior problems of the foster child, but research in the regular population points to insecure attachment relationships increasing the risk for behavioral problems (Fearon et al., 2010; Groh et al., 2012). We cannot rule out the possibility that it is more difficult to build a warm relationship with foster children with behavior problems, in that case behavior problems can also be a predictor of insecure attachment relationships.

It was hypothesized that sensitivity of foster carers was an important factor in the development of secure attachment (Gabler et al., 2014; Oosterman & Schuengel, 2008; Ponciano, 2010; Quiroga & Hamilton-Giachritsis, 2017). Self-reported sensitivity of foster carers, measured by the EA-SR, was indeed negatively associated with insecure attachment behavior, but only at a univariate level. When it was inputted in a multiple linear regression model, sensitivity was no longer significant. This result was surprising, but might be explained by the rather short average period of time the children were living with their foster families, as all children were included in the year following their placement. Sensitive parenting behavior of foster parents is expected to support foster children in adjusting their internal working model (Bakermans-Kranenburg et al., 2003), but this might take more time, especially in the challenging and stressful circumstances of a recent out-of-home placement. Apart from that, the EA-SR was developed for parents of infants, while the children included in our sample group were older. Finally, a self-reported questionnaire was used, requiring high levels of introspection and self-knowledge (Vliegen et al., 2009), which might make it difficult for foster parents to assess their own parental sensitivity.

Limitations and suggestions for further research

The studies included in the theoretical part about associated factors with quality of attachment between foster children and their carers, were mostly based on observational methods. Of the 11 studies included in the literature part, six used the Strange Situation Procedure (SSP), and five the Attachment Q-Set. Two studies (Bernier et al., 2004; Bovenschen et al., 2016) each used two methods (SSP and Attachment Diary; Attachment Q-set and Attachment Story Completion Task) to assess the quality of attachment of the foster children, but there was a lack of correlation between the two methods. This implies that caution is needed when comparing studies examining attachment.

The choice to use the AISI instead of the ‘gold standard’ Strange Situation Procure was based on the format of the RCT. The first reason was the age range of our sample, as the SSP was developed for children in the second year of their life. Second, the RCT has a repeated measures design. The second and third measurements only lie 2 months apart, so it would not be possible to register the SSP with such a short time gap as there is a risk the children may become sensitized to the procedure. Third, we were afraid coming to the laboratory setting would be too time-consuming and burdensome for the foster parents, and might prevent foster parents from participating to the study.

In future research, we would suggest to use multiple methods, each with their strengths and pitfalls. The Strange Situation Procedure provides more information on the types of attachment behavior, but limits the age range of the sample. The Attachment Q-Set can be used in a low-stress setting, but does provide less reliable information on the types of attachment behavior. The AISI adds information about the perception of the rater on the attachment relationship, but does not provide observational data. Using multiple methods would also enable to get more nuanced inside into the strengths and weaknesses and into the usability of the instruments.

In this sample 38.2% of the foster children were placed in kinship foster care, but in the total Flemish foster care sample, 72% of the foster families are kinship carers (Pleegzorg Vlaanderen, 2019). This raises concerns regarding the representativeness of the sample and the generalization of the results. It was very difficult to collect information on the families that refused to participate in the study. Since reliable data considering response-rate and non-response was not available, the representativeness of our sample could not be assessed. All foster families voluntarily participated in the study into intervention effects of VIPP-FC, so there could be a positive selection bias, especially as VIPP-FC involves filming the families which could be intimidating for families struggling with their foster children.

The sample size is small, so the generalizability of the results is limited. Additionally, as the study was not longitudinal, no causality of the associations can be determined. More longitudinal research is needed.

As mentioned before, the children in our sample were older than the intended age range of the EA-SR. Seven children were between 18 and 24 months old, so slightly under the age range for the AISI questionnaire.

Additionally, the use of self-reporting questionnaires limited the variety in answers and the amount of the obtained information. Some of the questionnaires require certain levels of introspection, potentially leading to different results than observations would have. Self-reporting questionnaires could also lead to response bias, when respondents answer more social desirable (Edwards, 1990). All data was collected by questionnaires that were all answered by one rater (foster mother), which might cause concern for common method variance (Podsakoff et al., 2003).

Further research using observations advisable. The findings regarding sensitivity were surprising, more longitudinal research regarding this matter is advisable, as research in the regular population clearly indicates the positive influence on the security of attachment between parents and their children.

Implications for practice

The introduction emphasizes the importance of attachment behavior of foster children. It is thus important to systematically monitor the quality of attachment between the foster carers and their foster children and provide additional support when needed.

Both parenting stress and behavior problems often lead to reduced well-being of the foster family and are known predictors of breakdown (Brown & Bednar, 2006; Farmer et al., 2004; Konijn et al., 2019; Leathers, 2006; Maaskant, 2016; Vanderfaeillie et al., 2018; Whenan et al., 2009). As higher levels of parenting stress were associated with more insecure attachment between foster parents and their foster children, it is important to pay sufficient attention to the parenting stress of foster parents. Behavior problem scores contributed most to the regression model and are known predictors of placement breakdown, so it is important to systematically monitor for behavior problems and provide additional support and interventions when these are indicated. Internalizing problems might be harder to detect, so monitoring them should be done carefully.