Development of atypical parental behavior during an inpatient family preservation intervention program.

Since failed reunification is a detrimental outcome for children, particularly infants and toddlers, the aim of this study was to gain insight into support to families in multiple-problem situations to help them achieve sustainable good-enough parenting. Therefore, we examined outcomes of an assessment-based inpatient family preservation program. We prepared a thorough target-population description (n = 70) using file analysis. Next, we examined atypical parental behavior during the intervention using the Atypical Maternal Behavior Instrument for Assessment and Classification with a repeated measures design (n = 30). The family files revealed a great number of issues at the family, parent, and child levels, such as practical matters, problems in parent functioning and between parents, and difficulties in the broader environment. We found a significant decline in three dimensions of atypical parental behavior over time. This program has great potential in supporting vulnerable families in their pursuit of family preservation.

INTRODUCTION

Placing a child in out‐of‐home care is one of the most extreme child protection measures available to ensure a child's safety. In accordance with an international movement underscoring the family preservation ideal in the child protection field (Lindsey, 1994), this measure is intended to be temporary. Reunification of child and birth family is considered the most favorable outcome in the context of permanency planning (McCroskey, 2001; Tilbury & Osmond, 2006). Hence, the purpose of an out‐of‐home placement is to support families in accomplishing necessary changes in the family situation to enable the safe return of the child. This type of support is typically provided by child welfare services and often referred to as family preservation (FP) or reunification services. Prevention of out‐of‐home placement is the initial goal of FP services (Tully, 2008). When reunification is the outcome of permanency planning, ideally children return to a stable and safe home environment, resulting in permanency with their birth family (Kimberlin, Anthony, & Austin, 2009). Unfortunately, this is not always realistic. This is problematic, as failed reunification is detrimental to children, particularly infants and toddlers, since the disruption impacts their development of attachment security (Ainsworth, Blehar, Waters, & Wall, 1978; Bowlby, 1979; Mikulincer, Shaver, & Pereg, 2003; Sroufe, 1988).

Failed reunification

Rates of reentry into care after reunification vary considerably (Festinger, 1996; Lee, Jonson‐Reid, & Drake, 2012; Mc Grath‐Lone, Dearden, Harron, Nasim, & Gilbert, 2017; Taussig, Clyman, & Landsverk, 2001). Research has revealed high percentages of failed reunifications, indicating undesirable outcomes of care provided by FP‐related services. Failed reunification might be associated with poor practice, such as lack of sufficient assessment and service provision during and after the reunification process (Wilkins & Farmer, 2015). Therefore, it seems imperative to develop a thorough understanding of effective strategies for supporting families to achieve sustainable good‐enough parenting. To this end, we evaluated a unique Dutch FP intervention program (described later). In this study, we described good‐enough parenting as: “The parenting situation is considered ‘good enough’ when consensus is reached between the team of the Expertise Center, the case manager,1 and the parents that the quality of parenting (as operationalized by the Expertise Center) has been improved during the intervention program in such a way that the risk for adverse development of the child, which has led to the (planned) out‐of‐home placement, is eliminated” (Vischer, 2013, p. 7).

The severity of issues related to failed reunification can be explained from an attachment theoretical perspective. The experience of multiple placements, resulting in changing caregivers and re‐abuse after reunification (Lutman & Farmer, 2013) may be especially harmful to infants and toddlers, given their rapid and critical physical, affective, and cognitive development (Chinitz, Guzman, Amstutz, Kohchi, & Alkon, 2017; Frame, Berrick, & Brodowski, 2000; Harden, Buhler, & Parra, 2016). An extensive body of research built on attachment theory (Bowlby, 1982) has confirmed the association between the quality of attachment of children with their primary caregivers and developmental outcomes (e.g., Bernier, Beauchamp, Carlson, & Lalonde, 2015; Cyr, Euser, Bakermans‐Kranenburg, & van IJzendoorn, 2010; Thompson, 1999). Attachment theory indicates that the first 5 years of life is a key period in which young children form secure attachments, underpinning the importance of continuous adequate care without disruption in their attachment figures. For this reason, it is vital to provide children the opportunity to form a secure attachment early in life.

Accordingly, within a limited time frame after an out‐of‐home placement or in cases where a child is being cared for by birth parents experiencing severe parenting problems, an informed decision needs to be made on where the young child should stay on a permanent basis (Vischer, Grietens, Knorth, & Mulder, 2017). In the context of permanency planning, this process is referred to as the assessment of parenting. An important element in the assessment process is the capacity to change; that is, the ability of parents to make significant behavioral changes, in some cases following an out‐of‐home placement of their child. To demonstrate this capacity to change, parents need to be provided effective support aimed at improving the quality of their parenting to promote secure attachment (Harnett, 2007).

Disorganized attachment

Related to attachment theory, strategies aiming to preserve families have been developed specifically targeting the improvement of parental sensitivity behavior, particularly parents’ ability to accurately perceive their child's signals and respond in a prompt and adequate manner to fulfill the child's needs (Tully, 2008). Indeed, care disturbed in the first years of life due to insensitive parental behavior has been identified as a risk factor for the development of disorganized attachment strategies (Cyr et al., 2010; Lyons‐Ruth, Bronfman, & Parsons, 1999; Madigan et al., 2006). When a caregiver fails to serve as a source of protection, the infant does not develop a consistent strategy to cope with stress (Lyons‐Ruth et al., 1999; Main & Solomon, 1986). Disorganized attachment is more closely associated with psychopathological outcomes later in life than are other types of attachment that encompass a strategy (i.e., secure, avoidant, or resistant attachment) (e.g., Carlson, 1998; Fearon, Bakermans‐Kranenburg, van IJzendoorn, Lapsely, & Roisman, 2010; Lyons‐Ruth & Jacobvitz, 1999; van IJzendoorn, Schuengel, & Bakermans‐Kranenburg, 1999).

In the United States, researchers have identified atypical parental behaviors that seem to be displayed significantly more often by mothers of children who lacked an organized attachment strategy. Related to this, they developed the Atypical Maternal Behavior Instrument for Assessment and Classification (AMBIANCE) to assess the quality of caregiver behavior (Bronfman, Parsons, & Lyons‐Ruth, 1992; Lyons‐Ruth et al., 1999). Benoit, Madigan, Lecce, Shea, and Goldberg (2001) demonstrated this instrument's applicability in clinical research on intervention effects; it proved highly sensitive to differences in caregiving behaviors between two groups of parents who were experiencing problems in feeding their infants. Based on their results, these authors cautiously concluded that the instrument could make a positive contribution to treatments aimed at improving parental sensitivity by reducing atypical parental behaviors related to disorganized attachment.

Another study using the AMBIANCE examined the effect of a home‐visitation‐based, mixed‐intervention model including enhancement of sensitive parenting for parents from pregnancy through the child's second year of life (Tereno et al., 2017). Findings have indicated a significant reduction in infant disorganization and disrupted maternal communication, as compared to a control group. Furthermore, the reductions in infant disorganization were attributed in part to declines in disrupted maternal communication.

The results of a study by Forbes, Evans, Moran, and Pederson (2007) on changes in atypical maternal behavior and attachment disorganization in children from 12 to 24 months underscore the conclusions of the aforementioned studies in which such changes were identified. They reasoned that unlike patterns of adequate parent–child interactions, which appear to be stable and thus natural, self‐sustaining systems, “… a substantially atypical, disrupted interaction within a disorganized relationship may be more susceptible to change and, thus, intervention aimed at improving the relationship” (p. 966). In other words, atypical parental behaviors are not trait‐like features, which is a promising conclusion for interventions aiming to promote adequate parental behavior.

In conclusion, failed reunification is highly undesirable. Since FP services play a key role in supporting parents toward adequate parenting and reunification success, we need to know “what works” to prevent failed reunification among children and families. This topic has been studied broadly, for instance, by examining factors associated with failed reunification (for an overview see Shaw, 2006) and with effect studies on interventions targeting the quality of parenting (Landers et al., 2018; Tully, 2008). However, our understanding remains insufficient to prevent many children from further harm due to dysfunctional parenting and multiple placements into and out of care. Furthermore, Landers et al. (2018) argued that we also need to understand what is at work for specific populations. Moreover, we need to look beyond prevention of out‐of‐home placement as the sole indicator of FP program success (Cash & Berry, 2003).

Aim and research questions

To gain insight into support for parents of young children toward FP, we evaluated an FP intervention program run by the Expertise Center for Treatment and Assessment of Parenting and Psychiatry2 (henceforth, EC) in the Netherlands. The central aim of our study was to increase knowledge regarding the characteristics of the outlined target population, including (changes in) the ability to parent 3; this with the intention to (a) contribute to improvement of the FP program by further tailoring it to the needs of the families referred and (b) evaluate the impact of the intervention program on the ability to parent. We operationalized ability to parent by examining atypical parental behavior using the AMBIANCE.

Our first research question (RQ1) was: “What are the characteristics of the families referred to the EC upon intake?” (Time 0). With this question, we sought to assess whether the EC reached its intended target population and to identify the treatment emphasis. Considering the target population of the EC, we expected to find evidence that the referred families had experienced multiple and complex problems at different levels, including mental health issues, implying the presence of multiple risk factors for impaired parenting. Our second research question (RQ2) was: “What and to what extent are atypical parental behaviors displayed during the clinical phase of the program, and do these behaviors diminish during the intervention, indicating the ability to change?”

We expected to find an overall improvement in the ability to parent over time for two reasons: (a) The intervention aimed to improve sensitive parental behavior which, if effective, would result in a decline of atypical behavior; and (b) the intervention included three evaluation points (at Week 4, Week 10, and Week 14) when the trajectories of parents who did not seem to make sufficient progress in the program were terminated with a negative recommendation on FP.

METHOD

Intervention

Within the EC, parenting assessments are conducted to underpin placement decisions. The EC aims to be a “last resort” intervention for families seeking either to be reunited with their young child (0–2 years) or to avoid an out‐of‐home placement following confirmed or suspected child maltreatment (GGZ Drenthe, n.d.). EC intervention is grounded in attachment theory and attachment‐related principles, family‐system therapy, and trauma‐recovery therapy (GGZ Drenthe, n.d.). Mentalization‐based treatment (Bateman & Fonagy, 2011) is one of the methods utilized to promote a secure attachment between child and parents through improvement of parents’ ability to accurately “read” their child's signals and respond appropriately, thus improving sensitive parenting. Parents are approached as the experts of their children, and guided by family coaches toward a higher awareness and understanding about the needs of their children, through a questioning strategy. A range of activities and group sessions is provided, in which the families are supported toward improvement of their parenting skills. Parents are invited to contribute subjects related to their interests and needs. In addition, individual treatment is provided by using video feedback (Fukkink, 2007) (for a more extensive description of the intervention, see Vischer et al., 2017). Research has verified the link between such fundamental parental behavior, termed sensitive caregiving, and attachment security (De Wolff & van IJzendoorn, 1997; Fuertes, Santos, Beeghly, & Tronick, 2006; Moss et al., 2011).

The intervention includes a residential phase lasting up to 16 weeks, during which parents and children live in a clinic from Sunday afternoon through Friday afternoon. During this inpatient part of the program, the functioning of the family is evaluated at three set points, as noted earlier. The trajectory may be ended if an evaluation returns a negative recommendation on FP. This may occur, for example, if the EC team considers the capacity to change toward good‐enough parenting insufficient to safeguard child safety or if the change process is too slow, considering the limited time frame in which a decision must be made.

The international literature uses various terms to characterize the EC's target population, such as multiproblem families (Ghesquière, 1993), multistressed families (Sousa & Eusébo, 2007), and vulnerable families with complex and enduring needs (Morris, 2013). Examples of the issues that these families may experience are substance abuse, domestic violence, and problems with housing, authorities, and mental health, while having few resources (Marsh, Ryan, Choi, & Testa, 2006). The problems that these families experience are multiple, serious, and complex (Ghesquière, 1993), interwoven (Bodden & Deković, 2010), and chronic, and these families seem to lack an ability to solve the issues that they face (Bakker, Bakker, Van Dijke, & Terpstra, 1998). Persistence of problems may also be attributed to lack of effective and appropriate service delivery, perhaps caused in part by fear and mistrust of professionals by families due to bad prior experiences in the coercive context of child protection (Schout, Meijer, & De Jong, 2017; Waterhouse & McGhee, 2009). Therefore, one of the keystones of EC intervention is to establish a trustful relationship between parents and professionals, done using techniques from, among others, De Shazer's (1985) solution‐focused brief therapy. To refer to the EC target population, we prefer the term families in multiproblem situations (Tausendfreund, Knot‐Dickscheit, Schulze, Knorth, & Grietens, 2016), as most of the families’ problems were related to their environment and living situation.

Design

The first part of this study (addressing RQ1) is descriptive, reflecting administrative data available from the EC. To answer RQ2, we used a one‐group repeated measures design. This part of the study can be considered exploratory because to our knowledge, no other evaluation study of an intervention program similar to that of the EC, in terms of its combination of both inpatient treatment and decision‐making, has been conducted.

Figure 1 presents an overview of the study. The large arrow represents the intervention and its primary aim: improvement of the ability to parent to achieve good‐enough parenting in the long‐term; in other words, demonstration of the capacity to parent.

Figure 1

Overview of the project, with the main concepts in boxes

Participants

RQ1

Inclusion criteria regarding the first part of the study were (a) being referred to the EC from March 2013 through October 2014 (since we considered a referral time frame of 1½ years sufficient in terms of a representative sample) and (b) having subsequently had at least one intake interview at the EC. Seventy families complied with these criteria. Since RQ1 referred to the characteristics of the target population, the results section presents basic background data on the participants.

RQ2

The second part of the study, on the quality of parenting, sought to include all families admitted to the EC clinic for the inpatient part of the intervention from March 2014 onward. Inclusion then continued until 30 families had volunteered to participate in the study, a number which was reached in February 2016. During the inclusion period, 33 families were admitted to the clinic, translating into a participation rate of 91%. The three nonparticipating families all said that they did not feel comfortable being filmed. Figure 2 shows the flow of participants through this part of the study.

Figure 2

Families’ flow through stages of the parenting study, including the outcome of the decision on family preservation

For each participating family, data on parent–child dyads were analyzed.4 The index parent–child dyad was selected using two criteria: (a) the parent being a primary caregiver and (b) the child being under 3 years of age. If a participating family had two children in this age group, the oldest child was selected. At Time 1 (T1), we obtained data on 30 parent–child dyads, at Time 2 (T2), we obtained data on 21 dyads, and at Time 3 (T3), we obtained data on 19 dyads. Missing data at T2 were due to termination of a family's participation in the trajectory after the first evaluation with negative advice on FP (n = 7; “short trajectory” group) or a family's decision to drop out (n = 2). Missing data at T3 were the result of termination of two trajectories just before the final part of the residential phase. Table 1 presents background data on the participants.

Table 1

Parenting study participant characteristics upon referral

	M	SD	Range
Parent age at Time 1a	25.9	5.5	18–44
Child age (months) at Time 1b	15.6	10.5	1–32

n = 30.

n = 30.

Procedure

The first author compiled family files from EC's digital administrative system. These consisted of reports provided by the case manager of the family in the context of the referral procedure, an application form, and a report of the intake conversation. Twenty files were coded by two coders to calculate Cohen's κ. The remaining files were coded by the same coders and one extra coder who had been trained by the other two coders.

Data for the parenting study were collected during the residential phase of the intervention: in Week 2 (T1), in Week 6 or 7 (T2), and just before the final evaluation in Week 13 or 14 (T3). Parents were asked to notify the main researcher (who worked in an office in the clinic on data‐collection days) when they were ready to carry out one of the core parenting tasks that were part of the data‐collection protocol: feeding, caring, and putting to bed. The parents were requested to “act usual” and pretend that the camera was not there. The interaction was filmed for a minimum of 10 min. Some families were also observed by a family coach5 during filming when this was indicated in the family treatment plan. As part of the protocol, the coach was not to interfere unless the safety of the child was at risk. There were no such occurrences during data collection. After filming, the parents received a voucher for a local supermarket and a digital copy of the videos. In addition, parents could request that the researcher provide the videos to the family coaches for use in video‐feedback sessions (a method regularly used within the EC). Almost all participating families consented to using the videos in this way.

The procedure complied with the ethical guidelines of the University of Groningen, Department of Pedagogical and Educational Sciences. The Medical Ethical Board of the University Medical Center Groningen concluded that no further assessment of the ethical protocol was needed.

Instruments

The characteristics of 70 families were investigated through an extensive file analysis using a coding system based on two Dutch categorizing instruments related to treatment goals and problem types among children and adolescents in youth care (Konijn et al., 2009; Reith, Hofman, Stams, & Van Yperen, 2008). We revised the coding system several times, repeatedly analyzing a set of files until all relevant variables were covered. In addition, we developed a coding protocol to ensure similar and systematic coding across coders. Interrater reliability was checked by asking two coders to code 20 files and calculating Cohen's κ. The outcome for most variables was “good” (0.61–0.80) or “very good” (0.81–1), following the classification of Altman (1991). Three variables were coded with less agreement (κ = <0.61) and therefore were adjusted or removed from the coding system. The codes covered characteristics at the family, parent, and child levels.

The ability to parent was studied by naturalistic observation of three core parenting situations: feeding the child, physically caring for the child (e.g., bathing, dressing), and putting the child to bed. It was assumed that by filming parents while they were being assessed by the EC (in most cases, involuntarily), their best efforts and ability to parent would be observable and used as an indicator in our study. As parenting is relational, the ability to parent is not merely a characteristic of an individual parent but rather a characteristic of the relationship between a specific parent and child (Crittenden, 2005). We used the AMBIANCE to measure the ability to parent. This instrument is based on the construct that parental behavior toward a child is a major determinant of multiple child outcomes. As noted earlier, the AMBIANCE is a coding system for assessing parental behaviors associated with disorganized attachment. It includes items from “Frightening, Frightened, Dissociated or Disorganized Behavior on the Part of the Parent: A Coding System for Parent‐Infant Interactions” (Main & Hesse, 1992), and has been further developed into a revised version including rating scales (Bronfman, Madigan, & Lyons‐Ruth, 2014).

The AMBIANCE system consists of the following dimensions: (a) affective communication errors (e.g., contradictory signaling to infant or failure to appropriately respond to infant cues), (b) role/boundary confusion (e.g., treating the infant as a spousal partner or role reversal), (c) fearful/disoriented behavior (e.g., appearing frightened in interaction with the infant or generally disoriented), (d) intrusiveness/negativity (e.g., verbally or physically intrusive behavior, inappropriate attribution of negative feelings to the child), and (e) withdrawal (avoidance, maintaining distance from the child). Each dimension is operationalized by two to four subcategories, all of which have been given codes (for an illustration of behaviors coded with the AMBIANCE pre‐ and postpsychotherapy, see Baradon & Bronfman, 2010).

A transcript of a 5‐min, video‐recorded, parent–child interaction was used to assign the codes to the parental behaviors. Based on the number and severity of the coded behaviors, each dimension was rated on a scale from 1 (no evidence of the concerning behavior) to 7 (persistent evidence of the concerning behavior). Finally, a score was assigned from the Parental Level of the Disrupted Communication Scale based on the ratings of the five dimensions. Rating scores 1 and 2 were considered optimal, 3 and 4 were considered nonoptimal but not disrupted, and scores of 5 and higher were considered disrupted. In case of a disrupted score, two subtypes could be assigned: “intrusive/self‐referential” and “helpless/fearful.” Some parents exhibited features of both subtypes.

A certified AMBIANCE coder coded the video‐recording after completion of training provided by one of the developers and the AMBIANCE reliability test. As the AMBIANCE requires a 5‐min fragment of parent–child interaction, a selection procedure was used to select 6 min of video material (For each family, a minimum of 30 min of filmed interaction was available.) For the “feeding” and the “caregiving” situations, the last 2½ min were coded; the first minute of the “putting to bed” situation was coded, starting exactly when the parent put the child in bed. The coder strictly followed the AMBIANCE coding protocol. The trainer was regularly consulted, especially concerning (a) fearful/disoriented behaviors (as the reliability test returned a low intraclass correlation coefficient for the ratings on this dimension); and (b) application of the coding system during the core parenting situations, which differed from the training and reliability test, as these were based on the Strange Situation Procedure (Ainsworth et al., 1978). The coder was blind to time (T1, T2, or T3) of the measurement.

Data analysis

Data extracted from the family files were analyzed using descriptive statistics generated with SPSS Version 24 (IBM, New York, NY).

Descriptive statistics

First, we calculated means, SDs, and minimum and maximum scores for both the frequency and rating scores of each AMBIANCE scale, including the overall level of disruption at T1, T2, and T3. Further, we calculated the percentages of families with a rating score in the disrupted range (>4).

Testing differences

Using the Friedman test, we compared the rank means for each dimension over time across the group of parents with three measurements (n = 19) to identify relevant scales for further analysis. With reference to the small sample and the explorative character of the study, a statistical significance level of P ≤ .10 was employed. Significant differences were then tested pairwise using the Wilcoxon signed rank test.

Type of change

We examined whether parents with at least two measurements available (n = 21) changed between the nondisrupted range and the disrupted range on the dimension rating scores and the rating score of the overall level of disruption between T1 and the last measurement (TL) before the end of their trajectory. Cases were assigned to one of four categories representing the following types of change: (a) “no change, nondisrupted” (scores in the nondisrupted range at T1 and in the nondisrupted range at TL), (b) “negative change” (scores in the nondisrupted range at T1 and in the disrupted range at TL), (c) “positive change” (scores in the disrupted range at T1 and in the nondisrupted range at TL), and (d) “no change, disrupted” (scores in the disrupted range at T1 and in the disrupted range at TL).

The proportion of families in the last two categories can be regarded as an indicator of the potential for change within the sample toward a good‐enough level of the measured parenting aspects, as these families scored in the disrupted range at T1. Finally, we identified which families fell into the positive change category for each dimension to determine if positive change in one dimension was related to positive change in other dimensions.

RESULTS

Target population description

Family level

Most of the 70 families referred to the EC were two‐parent households (n = 50; 71.4%). In three fourths of these families (75.7%), there was one child under the age of 3 years, 16 families had two children in this age group (22.9%), and one family had three children under age 3. Eight of the referred families (11.4%) also had children older than age 3. Table 2 presents the most often reported problem areas at the family level.

Table 2

Reported problem areas at the family level

N = 70	N	%	N = 70	n
Parenting abilities	70	100.0
Parent–child interaction	29	41.4
Partner relation	54	77.1
Housing (current & past)	63	90.0
Financial	56	80.0
Related to pregnancy	47	67.1
Social network	67	95.7
Informal	62	88.6
Professional	51	72.9
Long history of service usea	66	94.3
Type of service use
Parenting	70	100.0
Mental health	53	75.7
Addiction service	14	20.0
Housing	43	61.4
Financial	51	72.9
Crisis help	33	47.1
Probation	18	25.7
Other	59	84.3
No. of types of service use	M	SD	Min	Max
	4.8	1.55	1	9

Over 3 years of service use.

Parent level

Among the 70 families, 120 parents were involved at intake. Just over half of the parents were mothers (n = 68; 56.7%). Information on ethnicity was absent from many family files. Where such information was missing and there were no indications of an origin other than Dutch (e.g., an atypical family name), the code “probably of Dutch origin” was used. Otherwise, ethnicity was coded as missing. The largest proportion of parents (38.3%) was of Dutch origin, 40 parents (33.3%) were coded as probably of Dutch origin, and 14 parents (11.7%) had a migration background. The professional status of most parents was “not employed” (64.2%); 11 parents were “full‐time employed” (9.2%), and 5 parents held a part‐time job (4.2%). The professional status of 17 parents (14.2%) was coded as “other;” these parents were in, for example, an internship program, volunteered, or had sheltered employment. Table 3 presents the most often reported problem areas at the parent level.

Table 3

Reported problem areas at the parent level

N = 120	N	%
Physical problems	36	30.0
Substance abuse (not specified as addiction)	41	34.2
Addiction	20	16.7
Drug addiction	15	12.5
Alcohol addiction	5	6.0
Detained (in the past)	17	14.2
Problematic attitude toward social workers	69	57.5
Traumatic event in adult life	59	49.2
Intellectual abilities
Average	20	16.7
Mild	46	38.3
Suspected mild	14	11.7
Missing	40	33.3
DSM classification reports
No reports	31	25.8
DSM classification “unclear”a	70	58.3
Personality disorder	26	21.7
PTSS	25	20.8
Behavioral disorder	22	18.3
Autism spectrum disorder	10	8.3
Other	33	27.5
DSM classification “clear”b	19	15.8
Personality disorder	11	9.2
PTSS	5	4.2
Behavioral disorder	4	3.3
Other	5	4.2
Psychological problems	116	96.7
Negative feelings	77	64.2
Emotion regulation, impulse control	76	63.3
Lack of insight (in problems)	59	49.2
Complying with agreements made	30	25.0
Problem areas in childhood	98	81.7
Becoming a teen parent	27	22.5
Psychological problems	27	22.5
Out‐of‐home placement	26	21.7
Under supervision of the state	19	15.8
Adverse childhood events	72	60.0

Note. The 70 families included a total of 120 parents. DSM = Diagnostic and Statistical Manual of Mental Disorders; PTSS = posttraumatic stress syndrome.

A DSM classification was reported without specific information on the diagnostic assessment procedure.

Specific diagnostic assessment procedure information was available and judged as sufficient (recently employed by a certified professional following a comprehensive assessment).

Child level

The 70 families had, in total, 90 children who had been referred to the EC, of which 15 were unborn at the time of the referral. The number of boys (n = 40) and girls (n = 41) was almost equal among the referred children (missing values for unborn children). Most of the children were of Dutch origin (n = 63; 70%), although 22 children had a migration background (24.4%), and the ethnicity of 5 children was coded as probably of Dutch origin. Table 4 displays characteristics of child protection measures, placements, and reported problem areas at the child level. For 18.9% of the children, no problems were reported.

Table 4

Reported characteristics and problem areas upon referral at the child level

N	n	%
Supervision order	54	73.0
Out‐of‐home placement	45	60.0
Placement at intake
With birth family	20	26.7
With parents under supervision	6	8.0
Foster care	29	38.7
Kinship care	10	13.3
Other	10	13.3
No. of placements
0	36	48.0
1	24	32.0
2–5	10	13.3
Missing	5	6.7
Reported problem area
Emotion	25	33.3
Inconsolable crying	16	21.3
Behavior	26	34.7
Physicala	48	53.3
Physical complaints	21	28.0
Toxin exposure during pregnancya	16	17.8
Feeding	13	17.3
Sleeping	12	16.0
Physical development	10	13.3
Muscle tone	9	12.0
Motor development	4	5.3
Other physical problems	19	25.3
Adverse eventsa	62	68.9
Emotional neglect	30	40.0
Witness of domestic violence	28	37.3
Physical neglect	28	37.3
Physical abuse	13	17.3
Prenatal experiences of domestic violencea	12	13.3
Suspected child maltreatment (not specified)	7	9.3
Emotional abuse	5	6.7
Child maltreatment (not specified)	2	2.7
Other adverse events	18	24.0

Note. The 70 families included 75 children and 15 unborn children.

Variables were calculated with N = 90, as the unborn children were included.

Summary

Our analysis of the family files revealed a great number of reported issues at the family, parent, and child levels, varying from practical issues (e.g., related to housing and finances) to problems in parent functioning and between parents (e.g., in the partner relationship) and the environment (e.g., problems in the social network and in connection with social workers). Although the children involved were under age 3 years, most had already experienced child‐level problems, often physical, and profound adverse events in their young lives. These children proved to be highly vulnerable. In addition, it became clear that the problems the families had experienced, as documented in the files, were long‐lasting. For instance, almost all families had a long history of social service use, and for 81.7% of the parents, problems in their own childhood were reported (e.g., a history of out‐of‐home placements).

Descriptive statistics of atypical behavior during intervention

Frequency scores

Table 5 shows the frequency and rating scores of atypical behaviors for each AMBIANCE dimension during the EC intervention. The mean and maximum frequency scores for all except one dimension (role/boundary confusion), declined over time (T1 vs. T3). Role/boundary confusion behaviors increased between T1 and T2 and declined between T2 and T3, but still resulted in a higher mean frequency score at T3 as compared to T1. Most of the observed atypical behaviors were categorized under the dimensions of affective communication errors and intrusiveness/negativity. At the start of the intervention (T1), the fewest observed atypical behaviors were categorized under the dimensions of role/boundary confusion and withdrawal. At the end of the intervention, the fewest observed atypical behaviors fell under the dimensions of withdrawal and fearful/disoriented behavior.

Table 5

Descriptive statistics of frequency and rating scores for subscales of atypical maternal behavior during intervention

	Time 1			Time 2			Time 3
	(n = 30)			(n = 21)			(n = 19)
	M (SD)	max	% n>4a	M (SD)	max	% n>4a	M (SD)	max	% n>4a
ACE
frequency	13.1 (8.8)	40	53.3	11.1 (8.6)	32	52.4	7.3 (4.9)	17	26.3
rating	4.5 (2.0)	7		4.4 (1.8)	7		3.2 (1.7)	6
RBC
frequency	4.2 (4.1)	14	13.3	6.8 (7.1)	28	23.8	5.7 (6.8)	26	19.0
rating	2.5 (1.5)	6		2.8 (1.7)	7		2.6 (1.7)	7
FDB
frequency	7.9 (7.7)	31	26.7	6.5 (7.3)	22	19.0	3.9 (4.9)	21	0
rating	3.2 (1.8)	7		2.6 (1.6)	6		2.1 (1.1)	4
IN
frequency	10.7 (10.1)	37	40.0	9.1 (8.4)	26	42.9	7.1 (6.7)	20	31.6
rating	3.5 (2.1)	7		3.4 (2.2)	7		2.7 (1.8)	6
WIT
frequency	4.0 (4.2)	15	26.7	5.4 (5.2)	18	19.0	2.4 (1.8)	6	0
rating	2.9 (1.8)	6		2.6 (1.6)	6		2.0 (1.0)	4
OLD
rating	4.4 (1.8)	7	63.3	4.0 (1.9)	7	42.9	3.3 (1.9)	7	36.8

Note. On all dimensions, the minimum frequency score was 0, and the minimum rating score was 1. ACE = affective communication errors; RBC = role/boundary confusion; FDB = fearful/disoriented behavior; IN = intrusiveness/negativity; WIT = withdrawal; OLD = overall level of disruption.

Percentage with a rating score in the disrupted range.

Rating scores

At all times (T1, T2, and T3), atypical parental behaviors scored under the dimension of affective communication errors were rated as most severe, as compared to the other dimensions. The mean scores on this scale at T1 and T2 were greater than 4, signifying disrupted behavior. The mean rating scores for all dimensions, except role/boundary confusion, declined over the course of the intervention. At the start of the intervention (T1), a proportion of the sample was rated 5 or above on every dimension. At the end of the intervention, all parents scored under 5 on the dimensions of fearful/disoriented behavior and withdrawal. The mean score for overall level of disruption as well as the percentage of parents rated with a disrupted score (5, 6, or 7) declined over time. The mean of the overall level of disrupted behavior fell into the nondisrupted range. At the start of the intervention, 63.3% of the parents were classified as disrupted; at T2, this percentage was 42.9%, and at the end of the intervention, 36.8% of the parents were classified as disrupted.

Testing differences over time

For the group of parents with three measurements (n = 19), a Friedman test was conducted comparing the different measures for each AMBIANCE scale. The analysis rendered a significant chi‐squared value, P < .10 (see Table 6) for three scales: affective communication errors, fearful/disoriented behavior, and intrusiveness/negativity. A Wilcoxon signed‐rank test indicated a significant difference, P < .10, on the dimension of affective communication errors between T1 and T3, Z = −2.457, P = .014, and this was also the case for the difference on the dimension of fearful/disoriented behavior, Z = −2.277, P = .023, after Bonferroni correction.

Table 6

Mean rank scores and outcomes of the Friedman test of the frequency and rating scores

	M rank
n = 19	Time 1	Time 2	Time 3	χ2	P‐value
ACE
frequency	2.4	2.0	1.7	5.38	.068a
rating	2.4	1.8	1.8	6.33	.042a
RBC
frequency	1.8	2.1	2.1	1.06	.589
rating	1.8	2.2	2.0	2.18	.337
FDB
frequency	2.3	2.0	1.7	4.76	.093a
rating	2.2	2.0	1.8	3.04	.219
IN
frequency	2.3	2.0	1.6	4.95	.084a
rating	2.2	2.2	1.7	5.64	.059a
WIT
frequency	1.8	2.4	1.8	4.55	.103
rating	2.1	2.1	1.7	3.57	.168
OLD
Rating	2.3	2.0	1.7	4.54	.113

Note. ACE = affective communication errors; RBC = role/boundary confusion; FDB = fearful/disoriented behavior; IN = intrusiveness/negativity; WIT = withdrawal; OLD = overall level of disruption.

P <.10.

A Friedman test indicated differences in rating scores over time for affective communication errors and intrusiveness/negativity, P < .10 (see Table 6). A Wilcoxon signed‐rank test showed a significant difference, P < .10, on the dimension of affective communication errors between T1 and T3, Z = −2.431, P = .015, after Bonferroni correction.

Type of change

Our analysis of the type of changes observed in the group with at least two measurements available (n = 21) revealed that on all five dimensions, most parents could be assigned to the no change, nondisrupted group and the fewest parents assigned to the negative change group (see Figure 3). With respect to the total potential positive change toward good‐enough parenting (the sum of the groups of no change, disrupted and positive change), the greatest change occurred on the dimensions of fearful/disoriented behavior (5 of 7 possibilities) and withdrawal (4 of 5 possibilities). With respect to the overall level of disruption, the parents were spread quite evenly over the three groups of no change, nondisrupted (n = 7); no change, disrupted (n = 8); and positive change (n = 6).

Figure 3

Distributions of parents by type of change, in percentages; ACE = affective communication errors; RBC = role/boundary confusion; FDB = fearful/disoriented behavior; IN = intrusiveness/negativity; WIT = withdrawal; OLD = overall level of disruption

Eleven of the 21 parents fell into the positive change group on at least one of the dimensions. Examination of the cases revealed two patterns and four cases. First, 4 parents were assigned to the positive change group on the dimensions of affective communication errors AND intrusiveness/negativity AND overall level of disruptive behavior. For 3 parents, this was the case for fearful/disoriented AND withdrawal. One parent fell into the positive change group on role/boundary confusion, fearful/disoriented behavior, intrusiveness/negativity, AND overall level of disruption. The remaining 3 parents were assigned to the positive change group on a single dimension (see Table 7).

Table 7

Positive change per case for all dimensions including overall level of disruption

Case No.	1–4	5–7	8	9	10	11
Affective communication errors	x			x
Role/boundary confusion			x
Fearful/disoriented behavior		x	x		x
Intrusiveness/negativity	x		x
Withdrawal		x				X
Overall level disruptive behavior	x		x

Most of the atypical behaviors observed among parents in the EC program studied related to the dimensions of affective communication errors and intrusiveness/negativity. The fewest atypical behaviors observed fell into the dimensions of role/boundary confusion and withdrawal. Significant differences were found over time in relation to mean frequencies and mean rating scores with reference to the scales of affective communication errors, intrusiveness/negativity, and fearful/disoriented behavior.

At the end of the intervention (T3), all parents scored in the nondisrupted range regarding the scales of fearful/disoriented behavior and withdrawal. On the other scales, between 19.0 and 31.6% of parents scored in the disrupted range. Although no statistically significant decline in overall level of disruption was found, the proportion of parents with a disrupted score dropped during the intervention from 63.3 to 36.8%.

The greatest potential to change from a score in the disrupted range at the start of the intervention to a good‐enough score at T2 or T3 was found on the scale of overall level of disruption (n = 14), and for the dimensions of affective communication errors (n = 11) and intrusiveness/negativity (n = 9). Respectively, 42.8, 45.0, and 55.5% of these parents could be assigned to the positive change group. A higher percentage of positive change was found on the withdrawal scale (4 of 5 possibilities) and the fearful/disoriented scale (5 of 7 possibilities). Analysis of the type of changes observed suggests a connection between the three scales: All parents, except 1 parent who fell into the positive change group on the affective communication errors scale, went through a similar positive change on the intrusiveness/negativity scale and the overall level of disruption scale. The same pattern was found for three parents regarding the scales of fearful/disoriented behavior and withdrawal.

DISCUSSION

Conclusions and implications for practice

Target population

As outlined previously, we expected our sample to fit the descriptions found in the literature on families in multiproblem situations, including mental health issues. Our findings support this hypothesis, indicating that the EC reaches the intended target population: families in multiproblem situations with young children (0–2 years of age), of whom at least one parent has a mental illness, seeking family preservation. Such confirmation is essential in evaluation research to understand to whom exactly the study outcomes apply.

We found that the family files often lacked explicit information on the nature and severity of the psychiatric problems experienced, as indicated by the variable “unclear DSM classification” (Diagnostic and Statistical Manual of Mental Disorders; American Psychiatric Association, 2013) In these cases, referral to the EC seems highly relevant to gain a better understanding of the parenting situation, although it also implies that upon referral it is not always clear whether the families fit the inclusion criteria set by the EC. If the nature and severity of the problems are ambiguous, it is recommended that parental mental health be assessed in the referral and intake phase.

In addition, it became clear that intellectual disabilities, sometimes mild, were reported or suspected regarding a substantial proportion of the parents (50%). Since the combination of mental health issues and intellectual disability is known to be a risk factor in child maltreatment (Wilkins & Farmer, 2015), it is essential to identify these parents at the start of the intervention and to clarify during the program in what ways and to what extent the mental health issues and intellectual disability (or their combination) impact the capacity to parent, to safeguard the children and decide on treatment emphasis.

Further, our analysis found that problems in the partner relationship, such as intimate partner violence, were common among the target population. Since exposure to intimate partner violence (witnessed by 37% of the children) may well lead to trauma symptoms among very young children (Bogat, DeJonghe, Levendosky, Davidson, & Von Eye, 2006; Graham‐Bergmann & Levendosky, 1998), much attention should be given to this issue, and it needs to be targeted during treatment by, for example, providing relationship therapy.

Furthermore, the family files contained evidence that the children involved were highly vulnerable, not only due to their age but also because of the potentially traumatizing events that they had undergone and the problems a substantial proportion of them had experienced, as documented in the files. Therefore, these children should be closely monitored, and their well‐being must remain the primary consideration in the decision‐making process during the entire trajectory. There may be a risk of an overly narrow focus on the parents during the trajectory (also see Tausendfreund et al., 2016) because parenting is the main object of the assessment, and a trusting working alliance with parents needs to be established.

Parenting study

We also aimed to understand in what ways and to what extent the ability to parent was impaired among the target population. Our analysis provided detailed insight into the various aspects of atypical parental behaviors within our sample, which will be useful to further clarify the nature and severity of parenting problems within the target population and the treatment focus of FP services. Further, we sought to shed light on the outcomes of the intervention regarding the ability to parent. We assumed that we would find a decline in atypical behavior during the program. Our hypothesis was confirmed in relation to the scales of affective communication errors, intrusiveness/negativity, and fearful/disoriented behavior. These results are consistent with outcomes of two other studies (Benoit et al., 2001; Tereno et al., 2017) that have found a decline in atypical parental behavior following interventions targeting sensitive parenting.

Nevertheless, in our study, a proportion of parents scored in the disrupted range on three of the five scales and on overall level of disruption at the end of the intervention. This indicates that certain aspects of the ability to parent were still compromised among some of the parents after the clinical phase. If these were the parents who received a negative recommendation on FP, the outcomes on the AMBIANCE scales can be considered as a first indication of the validity of this decision. However, if these were parents who received a positive FP recommendation, ongoing support for these parents is recommended, targeting these issues in the home situation after the clinical phase. For instance, intensive aftercare or referral might be considered, alongside assurance of adequate transition to other services.

The duration of FP services is an important issue under discussion. FP interventions are designed to be intensive and of limited duration. However, there is evidence that the effects of FP services diminish after 12 months (Kirk & Griffith, 2004). Lindsey, Martin, and Doh (2002) argued that the problems experienced by families referred to FP services are too severe and complex to be resolved in the short‐term. A study of the outcomes of the EC in terms of the ability to parent in the long‐term is therefore recommended.

The decline in atypical behaviors that we found on three scales might indicate the target population's capacity to change toward sensitive parenting through intensive support. Another explanation for the changes registered in parental behavior is valid decision‐making by the EC in the three evaluations, as the families who scored lowest on the parenting scales were terminated or dropped out of the treatment. Further research is needed to explore the dynamics underlying these outcomes.

We found no significant difference over time concerning the scales of role/boundary confusion and withdrawal. This might be due to the fact that there was less potential for improvement in these behaviors, as most parents did not score high on these scales at the start of the intervention. However, it might also indicate that the treatment provided by the EC had no, or limited, impact on these aspects of parental behavior or that parents recognized these behaviors as inappropriate and made adjustments themselves while being directly observed.

Change was also examined more qualitatively by categorizing parents into four groups according to the type of change found. Our analysis identified two patterns and provided some indication that the AMBIANCE scales may be related. Further research is needed to examine these relationships and links between the scales.

Our study also yielded relevant lessons on the applicability of the AMBIANCE for the study of core parenting situations (feeding the child, physical care of the child, and putting the child to bed) using video data from naturalistic observation in a family psychiatric context aimed at FP. Since we identified evolution in atypical behaviors within our sample during the intervention, the AMBIANCE did appear to be a useful instrument for mapping changes in disruptive parental behavior within this target population.

Parenting, and the ability to parent, encompasses more than the atypical behaviors coded with this instrument. However, the quality of maternal behavior has proven to be a stronger predictor of long‐term outcomes over time than has infant attachment, indicating the importance of parental behavior (e.g., Dutra, Bureau, Holmes, Lyubchik, & Lyons‐Ruth, 2009; Shi, Bureau, Easterbrooks, Zhao, & Lyons‐Ruth, 2012). In addition, there is a sound theoretical and empirical foundation for the pathway of improving outcomes for children by targeting atypical behaviors of their parents and, by doing so, potentially affecting attachment quality as well. Therefore, the AMBIANCE seems a very valuable instrument for clinical assessment of parenting, for both treatment and decision‐making purposes, because specific, individual, and complex needs can be outlined based on the interaction between parent and child. Moreover, AMBIANCE codes relate to concrete observable behaviors, making them a very useful basis for dialogue with parents about the abstract concept of parenting and suitable for video‐feedback techniques. In addition, the AMBIANCE has proven to be highly informative in evaluation research on interventions targeting the quality of parenting.

Limitations and strengths

Target population description

Concerning the reliability of the target population analysis, note that by using file analysis we examined the reported characteristics of the families. We assumed that the reported characteristics of the family situations were considered relevant and significant in the context of a possible out‐of‐home placement by the professionals involved and therefore appropriate to describe the target population. However, we also noticed that the files contained reports of low quality. That is, information contained in the reports was often ambiguous and incomplete regarding, for example, the overall family situation, former service use, and the outcomes of services provided. Again, a lack of clarity on the overall family situation is often the reason for referral to the EC, and obtaining a good understanding of family functioning is one of the main objectives of treatment. We believe that the quality of reporting within child protection services can and should be improved. Reports overall need to be more accurate and comprehensive because they are often the basis for decision‐making and treatment emphasis. One strength of our study was the comprehensiveness of our target‐population description, which went well beyond the report of merely general background characteristics, which is common in evaluation research.

Our lack of a comparison group, for practical reasons, was a limitation of this study. The greatest challenge in finding a control group was the fact that our target group consisted mainly of families with a child who had been placed out of the home, meaning that it was not possible to assess parenting among families in a similar situation because parents typically do not care for their children following out‐of‐home placement. No alternative interventions with a similar target population were available that could be used to assemble a control group.

Furthermore, our sample size was relatively small, although data collection at the clinic lasted 2 years. Therefore, it is uncertain whether the sample is representative of the target population. In addition, the small sample size could lead to problems concerning the statistical power of the study. We did not employ multiple testing correction due to the explorative character of the research. However, we did find a significant and meaningful decline over time, suggesting that the sample was large enough to obtain relevant results. Finally, the reliability of the rating scores for the fearful/disoriented scale was marginally acceptable. For these reasons, the results of our study should be considered indicative rather than conclusive.

Use of the AMBIANCE in a repeated measures design with multiple parents constituted a unique and informative advance regarding clinical use of the instrument. Our sample included both mothers and fathers. While there has been a substantial shift in the involvement and role of fathers in child rearing and caregiving over the past decades (Cabrera, Tamis‐LeMonda, Bradley, Hofferth, & Lamb, 2003), there is a lack of research on the father–child relationship. Only one other published study using the AMBIANCE has engaged fathers and has aimed to explore the link between paternal behavior in the development of disorganized infant–father attachment (Madigan, Benoit, & Boucher, 2011). Similarly to De Wolff and van IJzendoorn (1997), who found that paternal sensitive responsiveness is a weak predictor of secure infant–father attachment (in contrast to maternal sensitive responsiveness and secure infant–mother attachment), prior studies have found that paternal atypical behavior does not predict infant–father disorganized attachment. Nevertheless, we believe that improvement of parental sensitive behavior is a desirable outcome and worth examining.

Conclusion

This evaluation study of EC intervention contributes to the evolving evidence on interventions targeting improvement of the parenting of young children in the context of permanency planning to increase child safety and prevent maltreatment of infants and toddlers. Furthermore, it is clear that the program provided by the EC has great potential to fulfill a very complex task in child protection: supporting a vulnerable target population in pursuing FP, preventing unnecessary caregiver changes, and providing permanency for infants and toddlers.

CONFLICT OF INTEREST

The authors declare that there is no conflict of interest.