What Works for Whom?
Moderators of the Effectiveness of a
Psychosocial Intervention for Children with
Chronic Illness and their Parents
Research Master Educational Sciences Thesis 2
Student: Elisa Napoleone (6266827/10048944)
Supervisors: Agnes Willemen (VU University Amsterdam) & Frans Oort Date: 27-06-2012
2 Acknowledgements
For the present thesis, I was very lucky to be able to participate in the analysis of data collected as part of the ―Samen Op Koers‖ research project at the VU University in Amsterdam. This has been a fantastic opportunity for me to deepen my understanding of a topic I find both fascinating and challenging. I would like to thank Agnes Willemen for her positive and always enthusiastic support, which has been a true inspiration for me. Thanks also to Frans Oort for his valuable guidance and advice, especially during my time in the wonderful world of statistics. Finally, a big thank you to Linde Scholten for giving me the opportunity to take part in her PhD project, and being always helpful along the way.
3 Abstract
Objective: To identify moderators of psychosocial outcomes of a cognitive-behavioural based
group intervention for children with chronic illness and their parents. Methods: Participants were 194 children (8-18 years old) in three randomised conditions: a child intervention, a child intervention with parallel parent group, and a wait-list control. Treatment outcomes were parent- and child-reported internalising and externalising problems six and twelve months after
intervention. Potential moderators were baseline illness severity and duration, age, coping style, attachment, parental psychopathology, and parenting stress. Data analyses were performed with multilevel modelling. Results: In the child intervention, children with a higher than average passive coping style at baseline had fewer in parent-reported internalising problems at post-test, but more self-reported externalising problems, compared to the wait-list. In the parent-child intervention, children with more secure attachment, and children of parents with more psychopathology, showed fewer externalising (parent report) and internalising (self-report) problems at post-test in comparison to the control group. Disease characteristics, age, active coping style, and parenting stress did not moderate the outcome in the treatment conditions.
Conclusions: Children who use more disengaged coping strategies, and parents with more
psychopathology, may have greater scope for change and gain more from interventions. Parent-child dyads that are less securely attached may need additional resources to improve. Future RCTs may consider stratifying the sample on the moderators that we found and replicating our results.
4 What Works for Whom? Moderators of the Effectiveness of a Psychosocial Intervention for
Children with Chronic Illness and their Parents
Chronic health conditions during childhood can have dramatic influences on the psychosocial adjustment of children and their families (LeBlanc, Goldsmith, & Patel, 2003). Psycho-educational interventions based on cognitive-behavioural principles, with the aim of promoting psychosocial adjustment in children with chronic illness, offer promising results (Barlow & Ellard, 2004; Beale, 2006; Brown, Daly, & Rickel, 2007; Last, Stam, Onland-van Nieuwenhuizen, & Grootenhuis, 2007; Thompson, Delaney, Flores, & Szigethy, 2011). However, research needs to move beyond effectiveness and investigate for whom and under what circumstances these interventions work best (Beale, 2006; Hinshaw, 2007; La Greca, Silverman, & Lochman, 2009; Wallander & Varni, 1998). In fact, treatments are rarely equally effective for all children (La Greca & Varni, 1993).
Once the effectiveness of an intervention has been established, exploring whether the effects differ depending on child and family characteristics is the crucial step towards the ―noble
but elusive goal‖ (Drotar, 1997, p. 599) of matching interventions to individual needs (Bauman, Drotar, Leventhal, Perrin, & Pless, 1997; La Greca et al., 2009). The scientific and clinical implications are manifold. Results can guide efforts at supplementing or modifying current treatment approaches; they can improve our understanding of treatment effectiveness; and they can direct clients to the most appropriate intervention (Rose, Holmbeck, Millstein Coakley, & Franks, 2004; Southam-Gerow, Kendall, & Weersing, 2001). Yet, despite the potential scientific and clinical gains, examples of moderator investigations are still rare in pediatric psychology intervention research (Beale, 2006; Drotar, 1997; La Greca & Varni, 1993).
5
Potential Moderating Variables
A moderator is a variable that influences the strength of the relationship between two other variables (Rose, Holmbeck, Millstein Coakley, & Franks, 2004). In an intervention context, participants’ characteristics measured at baseline act as moderating variables if they are found to
be differentially associated with the outcome for the treatment compared to the control group (Gardner, Hutchings, Bywater, & Whitaker, 2010). In other words, moderators help answer the question ―which children benefit most from interventions, and which may need a different strategy?‖ In order to identify relevant characteristics that could influence treatment outcome, an approach may be to consider a theoretical model of risk and resistance factors of children’s
psychosocial adjustment to chronic illness. In fact, factors that are posited to intervene in
psychosocial adaptation may also impact participation in treatment, and treatment gains (Kazdin, 1995).
Transactional models of child adaptation to chronic illness recognise the importance of numerous influences on children’s psychosocial adjustment (Brown et al., 2007; Thompson &
Gustafson, 1996; Wallander & Varni, 1998): disease parameters (e.g., severity, onset, duration), child characteristics (e.g., age, gender, coping style), and family variables (e.g., parental support, psychopathology, family functioning). Specifically, the stress and coping model identifies risk and resilience factors that could be targeted and modified in interventions (Wallander & Varni, 1998). According to this perspective, a chronic illness is an ongoing strain that requires
continual adjustment. Parameters such as severity and duration of the disease constitute risk factors that influence psychosocial stress: In turn, higher levels of stress put the individual at risk of psychosocial problems. On the other hand, resistance factors such as stress-processing skills and parental support are posited to intervene by buffering the relationship between stress and
6 adjustment. Baseline levels of risk and resistance factors may potentially moderate children’s response to cognitive-behavioural treatment. Below, we review the relevant available evidence on the impact that specific disease, child, and parent characteristics can have on the psychosocial adjustment of children with chronic illness, and on treatment response in different client groups.
Disease characteristics.
Illness severity and duration. Severity refers to objective measures of the illness, such as intensity and frequency of symptoms (Bleil, Ramesh, Miller, & Wood, 2000), number of
hospitalisations and surgeries, and use of medication. The more an illness is severe, the more it disrupts normal activities, including experiences that are important for development (Walters & Williamson, 1999). Both Lavigne and Faier-Routman (1993) and McQuaid, Kopel, and Nassau (2001) found that higher severity was significantly correlated with child maladjustment. In another study, limited capacity to carry out age-appropriate tasks was associated with
psychological adjustment problems (Stein & Jessop, 1984). Illness duration, on the other hand, refers to the length of time elapsed from the onset of the disease, or first diagnosis (Lavigne & Faier-Routman, 1993). In children and adolescents, longer illness duration is likely to coincide with onset during early childhood. Earlier onset is generally associated with better psychosocial adjustment (LeBlanc et al., 2003; Thompson et al., 2011). On the other hand, psychosocial adjustment is poorest at the early phases of the disease (Le Blanc et al., 2003).
Because of their widespread influence on the psychological resources of both child and family (MacLean, Perrin, Gortmaker, & Pierre, 1992), illness severity and duration may moderate the extent to which children profit from the interventions (La Greca & Varni, 1993). Kibby, Tyc, and Mulhern (1998), in a meta-analysis of interventions for children with chronic illness, did not find a moderating effect of illness severity or duration on treatment outcomes;
7 however, they acknowledged that several studies did not provide adequate information on these parameters. In another study, higher severity of illness was associated with lower medical regimen adherence in adults (DiMatteo, Haskard, & Williams, 2007).
Child characteristics.
Age. Adolescents tend to respond better than younger children to cognitively-based interventions (Kazdin & Crowley, 1997), possibly because of higher cognitive functioning (Weisz, Weiss, Han, Granger, & Morton, 1995). Both Kibby and colleagues (1998), and Weisz et al. (1995) in a meta-analysis of psychotherapy effects for various psychological conditions, found more favourable outcomes for adolescents.
Coping style. General coping style in children with chronic illness refers to their way of adapting to developmental tasks and to stressful events (Compas, Connor-Smith, Saltzman, Harding Thomsen, & Wadsworth, 2001; Meijer, Sinnema, Bijstra, Mellenbergh, & Wolters, 2002). Evidence suggests that making active attempts to engage with the stressor or with one’s own thoughts and feelings (such as by purposeful problem-solving, open communication, and seeking of social support; Compas et al., 2001) is associated with better psychological
adjustment (Lavigne & Faier-Routman, 1993; Meijer et al., 2002; Spirito, Stark, Gil, & Tyc, 1995). On the other hand, a more disengaged or passive coping style (e.g., wishful thinking, denial) appears to be related with more adjustment difficulties (Compas et al., 2001; Thompson & Gustafson, 1996).
There is preliminary evidence that coping style may moderate treatment outcome. The ―congruence hypothesis‖ (Christiano & Russ, 1998; Piira, Hayes, Goodenough, & von Baeyer,
2006) holds that interventions matching the child’s spontaneous coping style lead to better outcomes. Experimental studies in normative populations (Fanurik, Zeltzer, Roberts, & Blount,
8 1993; Piira et al., 2006) found that adolescents who spontaneously coped with painful stimuli by focusing away (passive coping) tolerated pain better after an ―imagery‖ intervention, while
adolescents who spontaneously coped by focusing on a stimulus (active coping) responded better to a ―sensory-focusing‖ intervention. One study with children with cancer reported findings in
the opposite direction (Smith, Ackerson, & Blotcky, 1989), but results have been put into question due to methodological issues (Fanurik et al., 1993).
Attachment. Secure parent-child attachment, which involves a sensitive and responsive caregiver to the child’s need for proximity, promotes optimal development (Wan & Green,
2009). On the other hand, insecure attachment, characterised by lower carer sensitivity, is associated with increased hopelessness (Wood, Klebba, & Miller, 2000) and more depressive symptoms (Bleil et al., 2000) in children with asthma. A secure attachment may act as a buffer in the relation between stressful life events, such as illness, and psychological processes in children (Willemen, Schuengel, & Koot, 2011; Wood et al., 2000), and could also moderate treatment effectiveness. For instance, preadolescents in group counselling were found to display more productive client behaviour and responsiveness to others if their attachment was secure (Shechtman & Dvir, 2006).
Parental psychopathology. Approximately 47% of parents of children with chronic illness report elevated levels of psychiatric symptoms (LeBlanc et al., 2003). Children of parents with psychopathology are at risk of suffering from psychopathology and poor functioning themselves (Wan & Green, 2009). Children with asthma are hospitalised more if their mother has mental health problems (Weil et al., 1999). In addition, parental psychopathology limits the extent to which psychological interventions can be beneficial (Drotar, 1997; Cobham et al., 1998; Dadds et al., 1999). For instance, Hutcheson et al. (1997) found that young children with a chronic
9 health condition (failure-to-thrive) showed more long-term improvements (in motor and
cognitive development, and interactive behaviour) following a home-based intervention if their mother did not suffer from depression. One reason may be that psychopathology reduces the extent to which parents take advantage of intervention resources (Drotar, 1997).
Parenting stress. Parenting stress (i.e., parents’ perception of their role as stressful; Levendosky & Graham-Bermann, 1998) is higher among parents of children with chronic illness (Deater-Deckard, 1998). Elevated pediatric parenting stress is correlated with poorer family functioning outcomes (Streisand, Kazak, & Tercyak, 2003) and more emotional and social problems in children (Colletti et al., 2008). High levels of parenting stress have been found to predict unfavourable outcomes in cognitive-behavioural therapy interventions for children with anxiety disorders (Crawford & Manassis, 2001) and antisocial behaviour (Kazdin, 1995).
The Current Study
On the whole the existing literature, although scarce, points at several characteristics that could influence the effectiveness of psychosocial interventions for children with chronic
illnesses. In an effort to advance knowledge in this area, the current study examined moderators of effectiveness of a multicentre randomised controlled trial (RCT) testing two formats of a cognitive-behavioural based group intervention for children with chronic illness and their parents: a child-focused intervention (Op Koers - ―On course‖) and the same intervention combined with a parallel parent education program (Samen Op Koers - ―Together on course‖; Scholten, Willemen, Grootenhuis, Maurice-Stam, Schuengel, et al., 2011).1 The child
intervention had the aim of teaching children disease-related active coping skills, in order to prevent or reduce their psychosocial problems and improve their resilience (Scholten et al.,
10 2011). The parallel parent group taught parents how to improve parent-child communication and motivate children to apply the learned skills to their everyday life. Within the stress and coping perspective, the child intervention can be viewed as influencing the stress-processing pathway, through the teaching of active coping skills; the parent-child intervention, by involving the parents in the intervention, also targets the social-ecological pathway, with the aim of improving parental support (see Figure 1).
--- INSERT FIGURE 1 ABOUT HERE --- The objective of the current investigation was to examine the extent to which baseline levels of the disease, child, and parent characteristics outlined above moderated the psychosocial outcomes of children following the two interventions compared to a wait-list control group. Given the scarcity of previous research, hypothesizing the direction of the moderating influences was tentative at best. Two alternative perspectives can be proposed. On the one hand, it may be expected that children who had baseline characteristics that put them at lower risk of
psychosocial stress, or equipped them with more resistance (e.g., milder illness severity, active coping style, low parental psychopathology), would gain the most following the interventions compared to the wait-list control group. This scenario would illustrate the so-called Matthew effect—that children who are better off optimally benefit from interventions (e.g., Bakermans-Kranenburg, van IJzendoorn, & Bradley, 2005). On the other hand, it may also be argued that at-risk children and families can gain more from interventions, because they have greater scope for change (e.g., Hautmann et al., 2009). Under this perspective, we may expect to find greatest improvements, in comparison to the control group, in the adjustment of children who were at higher risk of psychosocial problems or had lower resistance factors at baseline (e.g., passive coping style, higher parenting stress). It made sense to expect the same baseline variables to
11 moderate the outcome of both treatments, since they shared the same child-focused protocol. However, the stress and coping model acknowledges the important role that parental
characteristics and the family environment have in the psychosocial adjustment of children with chronic illness. Therefore, any effects may be more pronounced in the parent-child intervention than in the child intervention. To assess treatment outcome, we obtained the child’s self- and parent-reported internalising and externalising problems at baseline and at 6 and 12 months following intervention.
Method
This study is part of a larger treatment effectiveness investigation which sought to promote psychosocial adjustment in children with chronic illness. The research project was designed as a multicentre randomised controlled trial with three conditions: 1) the child
intervention ―Op Koers‖ (OK); 2) the parent-child intervention ―Samen Op Koers‖ (SOK); and
3) a wait-list control group (WL). Participants were assessed at four time-points: baseline (T0), 6 weeks (T1), 6 months (T2), and 12 months after baseline (T3). In the present study,
measurement points T0 (baseline), T2 (post-test), and T3 (follow-up) were used.
Participants
A total of 1134 families were contacted through doctors of participating centres and invited to take part in the study. At screening, 208 families met the inclusion criteria, and 203 signed the informed consent. At the baseline measurement after randomisation, the sample consisted of 194 children (97 boys and 97 girls) between the ages of 8 and 18 (M age = 12.03, SD = 2.68) spread in the three conditions (OK: n = 72, SOK: n = 48, WL: n = 74). Of the
participating families, 43.8% were recruited at academic centres, 42.8% at non-academic centres, and 13.4% at special needs schools. Mothers’ mean age was 43.18 years (SD = 5.74), and
12 fathers’ mean age was 45.71 years (SD = 6.13). Parents’ educational level (on a scale from 1: no
education, to 4: higher education) was fairly high (mother, M = 3.27, SD = .597; father, M =
3.22, SD = .72). The mean family income (on a scale from 1: less than the average income of € 30.500 per year, to 4: more than two times the average income) was 2.56 (SD = .96). The majority of families (76.4%) had a Dutch background. Single-parent families constituted 16.2% of the sample. Nine parents did not take part in the study. Of the participating members of the families, 82.7% were mothers, 13.5% were fathers, and 3.7% were other caregivers.
In order to meet the inclusion criteria, children had to have a chronic illness, defined as ―at least 6 months of continuous medical care, permanent lifestyle changes, and continuous
behavioural adaptation to the unpredictable course of the illness‖ (LeBlanc et al., 2003, p. 859). Exclusion criteria were severe learning difficulties, and the inability to complete questionnaires in Dutch.
Procedure
The study was advertised through information letters, posters, and leaflets in outpatient clinics of three academic hospitals, four non-academic hospitals, and two special needs schools in the Netherlands. Interested participants returned a stamped self-addressed application form and were then asked to sign informed consent letters. Inclusion criteria were established by telephone interview. Successively, randomisation to the conditions took place in each centre via a computerised block randomisation procedure (Altman & Bland, 1999a, 1999b, cited in
Scholten et al., 2011), and participating families were informed of the outcome by mail. The intervention started within a week of collecting the baseline assessments. Participants in the control condition were placed in a one-year wait-list, after which they were offered the
13 the primary caregiver, at intervals of 6 months between each measurement occasion. The
participants were rewarded financially for completing the assessments.
Intervention. Op Koers is a manualised cognitive-behavioural based group intervention
delivered by two trained child psychologists working in the participating hospitals. It consists of six weekly 90-minute sessions and one follow-up session after six months. Groups were formed with a minimum of four and a maximum of eight participants. The intervention was delivered in two slightly different formats depending on the age group of the participants (8 to 12, and 12 to 18). The parents’ group intervention was scheduled at the same time as the children’s one, but was located in a different room.
The learning goals of the child intervention were: information seeking and information giving about the disease, use of relaxation during stressful procedures, increased knowledge of self-management and compliance, enhancement of social competence, and positive thinking. The learning goals of the parent intervention were: understanding what the children learn, being sensitive to children’s cognitions and feelings, and stimulating children to apply learnt skills in
everyday life. Both interventions combined group discussions, role plays, information videos, and take-home assignments.
Measures
Outcome measures. To assess children’s psychosocial adjustment, we used both parent
and child reports:
Parent-reported adjustment. The primary caregiver was asked to fill in the Dutch version of the Child-Behaviour Checklist (CBCL, Achenbach, 1991a; Verhulst, van der Ende, & Koot, 1996). The CBCL measures parent-reported behavioural and emotional problems in children aged 4 to 18. It consists of 122 items rated on a 3-point Likert scale (from ―not true‖ to
14 ―often true‖). We computed two broadband syndrome scores (Internalising and Externalising),
which encompass eight narrowband syndromes (Withdrawn, Somatic Complaints,
Anxious/Depressed, Social Problems, Thought Problems, Attention Problems, Delinquent Behaviour, Aggressive Behaviour). The Somatic Complaints scale was not used because of the difficulties in interpreting physical symptoms in samples of children with chronic illness (Drotar, Stein, & Perrin, 1995). A higher score indicates more problem behaviour. The Dutch CBCL has good psychometric properties, with high internal consistency and test–retest reliability, and high external validity (Grietens, Onghena, Prinzie, Gadeyne, Van Assche, et al., 2004). In the present study, both the Internalising scale (Cronbach’s α at T0 = .88; T2: .85; T3: .86) and the
Externalising scale (α at T0 = .90; T2: .91; T3: .90) showed good internal consistency.
Child-reported adjustment. The Dutch version of the Youth Self-Report (YSR,
Achenbach, 1991b; Verhulst, van der Ende, & Koot, 1997) was administered to children older than 11 to obtain a measure of their perceived behavioural and emotional problems. It consists of 112 items rated on a 3-point Likert scale (from ―not true‖ to ―often true‖). The YSR is the self-report version of the CBCL and, as such, it assesses the same two broadband syndromes, Internalising and Externalising. As with the CBCL, the Somatic Complaints narrowband scale was not used. The Dutch YSR has good reliability and validity (Verhulst et al., 1997).
Cronbach’s α coefficients indicated good internal consistency of the Internalising (α at T0 = .86;
T2: .89; T3: .89) and Externalising scales (α at T0 = .84; T2: .89; T3: .91).
Moderator measures. The potential moderators considered in the present study were
assessed at the baseline measurement (T0):
Disease characteristics. Disease characteristics included illness severity and duration. Information regarding illness severity was obtained via a parent-reported questionnaire
15 developed for the study. Parents were asked to report whether thirteen ―yes/no‖ statements applied to their child. Statements covered aspects of the illness such as doctor visits and hospitalizations, school absences, dietary and visible consequences, medications and surgeries, use of appliances, and course of illness (improving and stable, or changing and deteriorating). For each item, a positive (―yes‖) response was scored as 1. The illness severity index was obtained by adding up individual items. In addition, parents reported the age of their child at first diagnosis of a chronic illness, and this information was used to calculate the duration of the illness (in years and months).
Child characteristics. Child characteristics included age, general coping style, and attachment. The short version of the Coping Strategies Inventory (CSI, Tobin, 1991) was used as a self-report measure of general coping style for children aged 7 and up. A backward-forward method was used to translate the 32-item questionnaire into Dutch, and was then adapted to a 12-item version. Items, rated on a 5-point Likert scale (from ―I never did this‖ to ―I always did this‖) assess coping thoughts and behaviours in response to a specific stressor. The child is
asked to describe a stressful event, and then to answer the questions. It consists of four subscales (problem engagement, emotion engagement, problem disengagement, and emotion
disengagement) which result in two primary scales representing engagement and disengagement (with six items each). The CSI is a well-established measure with strong psychometric
properties (Blount, Simons, Devine, Jaaniste, Cohen, et al., 2008). In the present study, levels of internal consistency were low (Engagement: α = .63; Disengagement: α = .58) but comparable to other studies that have used the short form (Addison et al., 2007).
The Security Scale (SS, Kerns, Aspelmeier, Gentzler, & Grabill, 2001) is a child self-report measure of perceived security of parent-child relationships. The questionnaire was
16 translated into Dutch by a native speaker. Children completed it twice; once for the mother, and once for the father. The SS consists of 15 items on a 4-point Likert scale (from ―fully agree‖ to ―fully disagree‖) measuring three domains: availability (perceived responsiveness and
accessibility of the attachment figure), reliability (tendency to rely on attachment figure during stressful times), and communication (perceived ease and interest in communicating with attachment figure). A higher total score represents more secure attachment. This measure has good psychometric properties (Kerns et al., 2001). In the present study, internal consistency was α = .75 and α = .81 for mother and father, respectively. The score of the scale relating to the
parent who was involved in the intervention (i.e., filled out the other questionnaires) was used in the present analyses. For children whose parents did not take part in the study, the score of the mother scale was used.
Parent characteristics. Parent characteristics included psychopathology and parenting stress. The Dutch short version of the Young Adult Self-Report (YASR, Achenbach, 1997; Wiznitzer, Verhulst, van den Brink, Koeter, van der Ende, et al., 1992, cited in Hofstra, van der Ende, & Verhulst, 2001) was used to measure parental psychopathology. It is the self-report measure of the CBCL for adults. It consists of 29 items that measure the same syndromes as the CBCL and the YSR. It has good reliability and validity (Hofstra et al., 2001). To assess the general mental health of the primary caregiver, only the Total problem scale was used in the present study (α = .89).
The Dutch short version of the Parenting Stress Index (Nijmegen Parenting Stress Index-Short – NOSIK, De Brock, Vermulst, Gerris, & Abidin, 1992) measured the amount of parenting stress that the parent experienced. The short version is a 25-item measure on a 6-point Likert scale (from ―strongly disagree‖ to ―strongly agree‖) divided into a parent stress domain and a
17 child stress domain. Only the parent domain was used in the present study, to prevent overlap between the child characteristics and the outcome measures. Fourteen items refer to parent characteristics (efficacy, depression) within the care-giving context. A higher score indicates more parenting stress. Internal consistency of this scale was adequate (α = .86).
Data Analysis
Analyses were performed on an intention-to-treat basis, i.e. participants were analysed according to the treatment condition assigned at randomisation. Multilevel modelling was used to account for dependency in the observations, due to the longitudinal design and the block randomisation procedure. Hence, measurement occasions (Level I) were treated as nested within individuals (Level II), and within groups (Level III) if appropriate. Data were analysed using a maximum-likelihood estimation procedure, which estimates all model parameters simultaneously to maximize the likelihood that the data were derived from the population (under the hypothesis that the model is true). The maximum-likelihood approach is the appropriate method for
comparing nested models, not only when they differ in their random but also in their fixed parts (Tabachnick & Fidell, 2007).
Model building was approached with a stepwise strategy. Given the large number of parameters to be estimated, the number of steps was kept to a minimum, and simpler, more parsimonious models (with fewer parameters) were preferred. The same procedure was repeated for each of the four dependent variables. Steps 1 and 2 were common to all potential
moderators, while Step 3 differed for each moderating variable. Preliminarily, the intra-class correlation was calculated on a three-level empty model (measurement occasion, subject, and group) for each dependent variable. Levels were rejected if the intra-class correlation indicating dependency in the observations was not significant (or below .05). In Step 1, an empty model
18 with the appropriate number of levels and an autoregressive within-subjects (co)variance
structure was fitted to the data.2 In Step 2, the fixed effects and interactions of measurement occasion and treatment condition were entered in the model. The final step (Step 3) tested moderation. In this step, the fixed effects and interactions of the potential moderator with measurement occasions and treatment conditions were estimated. If the introduction of these potential moderating effects significantly improved model fit, then significance of single effects was inspected. We were specifically interested in the three-way interactions between potential moderators, treatments (either OK or SOK), and post-tests (either T2 or T3), because these would indicate that the interventions had a different effect at post-test depending on the baseline level of the potential moderating variable, compared to the wait-list. At each step, differences in fit between nested models were evaluated using the χ2
difference test for deviance values (-2 Log-Likelihood). A significant χ2 value supported the alternative hypothesis that there were differences between the nested models being compared. Progression through the steps was only deemed appropriate based on the significance of the χ2 difference test. To assess the significance of single fixed effects, we used the F-statistic. An α level of .05 was used as significance
criterion for these tests.
Measurement occasions and treatment conditions were binary coded. All continuous variables (both dependent and explanatory) were standardised to aid interpretation and prevent multicollinearity in the interaction terms (Aiken & West, 1991). As a result, regression
2 A saturated empty model may have fitted the data better. However, for the sake of parsimony, we decided to fit the
simplest model appropriate for longitudinal data. Longitudinal data often have a proximally autoregressive variance-covariance structure, implying that waves of measurement closer in time correlate more highly than waves further apart (Campbell & Kenny, 1999, cited in Kwok et al., 2008). In addition, when focus is on fixed regression coefficients, misspecification of the within-subjects variance structure is not crucial (Hox, 2010).
19 coefficients of binary variables can be interpreted as Cohen’s d effect sizes, and coefficients of continuous variables as r effect sizes.
Results Preliminary Analyses
There was an 18% attrition rate during the study. Children who dropped out did not differ from the others in age, gender, ethnicity, setting, illness severity, or baseline problem levels. However, there were significant differences between conditions, with children in the parent-child intervention group withdrawing significantly less (8%; χ2 (2) = 6.55, p = .038) than children in the child intervention (21%; Exp B = .24; 95% CI [.07, .89]) and wait-list control group (20%; Exp B = .26, 95% CI [.07, .94]). Since multilevel modelling allows the inclusion of all participants for whom at least one data point in the dependent variable is available, analyses were carried out with a sample size of 187 for the parent-reported scales, and of 139 for the child-reported scales (completed by children who were, or turned, eleven years old during the study). A majority (74.3%) of children in the wait-list condition received some form of psychological help during the intervention period.
Preliminary analyses were run at the level of analysis appropriate for each variable: Hence, the dependent variables were analysed at Level I, and potential moderators at Level II (Tabachnick & Fidell, 2007). In the parent-reported scales, 4.6% of the 518 observations was missing. In the child report, 22.8% of the 451 observations of children old enough to have completed the questionnaire was missing. Missing values in the potential moderators ranged from 0 to 9.3%. The null hypothesis that data were missing completely at random could not be rejected based on Little’s MCAR test, χ2
(27) = 38.42, p = .07. To prevent different sample sizes in each moderator analysis and ensure comparability of nested models, missing values in these
20 variables were imputed using an expectation maximization technique. Through inspection of standardized scores, four univariate outliers were identified in the potential moderators, and nine in the dependent variables. To reduce the influence of the most extreme cases, raw scores were changed to one unit above the next highest extreme score in the distribution (Tabachnick & Fidell, 2007). After removal of univariate outliers, one multivariate outlier was found in the dependent variables: a case with extremely high scores on the child report, and extremely low scores on the parent report. Excluding this case from the analyses did not change results
appreciably; therefore, the decision was taken to keep it. Distribution tests indicated departures from normality; however, these tests are too sensitive in samples larger than 100, so violations are expected (Tabachnick & Fidell, 2007). Indeed, skewness and kurtosis values were within acceptable range; thus, variables were deemed normally distributed.
Table 1 shows the means and standard deviations of the dependent variables at each measurement occasion, and of the potential moderating variables at baseline, for each condition.
--- INSERT TABLE 1 ABOUT HERE --- Although participants were randomly assigned to the three conditions, there were significant differences between groups in age, severity, parental psychopathology, parenting stress, and parent-reported internalising scores at baseline, with participants in the parent-child intervention scoring higher than participants in the other groups. Furthermore, there were significant gender differences, with girls scoring higher than boys, in parent-reported internalising problems at follow-up, F (1,166) = 4.293, p = .040, and in self-reported internalising problems at all time points (baseline: F (1,109) = 7.425, p = .008; post-test: F (1,121) = 8.227, p = .005; follow-up: F (1,115) = 6.264, p = .014). In addition, there were significant differences in parent-reported externalizing problems at post-test, with parents of
21 children in special needs schools reporting more problems than in the academic and
non-academic settings, F (2, 166) = 4.106, p = .018. Thus, in the multilevel analyses, we controlled for gender in parent- and child-reported internalising problems, and for setting in parent-reported externalising problems.
Multilevel Moderation Analyses
Intra-class correlation coefficients ranged between 59 and 80% at Level II (subjects), and between 0 and 3% at Level III (groups) for all dependent variables: The third level was thus disregarded. Table 2 illustrates the model building steps, with the deviance values and significance of overall χ2
difference tests.
--- INSERT TABLE 2 ABOUT HERE ---
Main effects.3 The main effects and interactions of measurement occasions and
treatment conditions significantly improved the fit compared to the empty model for all outcome variables. Table 3 displays the coefficients and standard errors for the standardised parameters estimated in the second step of model building, for each outcome variable. Since the binary indicators of baseline measurement and control group were not entered in the equation, the intercept is to be interpreted as the average score on the dependent variable for the wait-list at baseline. The other parameter coefficients are thus difference scores to be added to the intercept estimate. The coefficients of the binary measurement occasion variables (T2 and T3) represent the differences from baseline for the wait-list group at post-test and follow-up. The estimates of the treatment groups (OK and SOK) are the average differences from wait-list for children in the intervention conditions at baseline. Finally, the coefficients of the interactions between
3
To investigate main effectiveness, different analyses were run by A. Willemen and L. Scholten. The results from the analyses hereby reported are for reference only, and will not be considered further.
22 measurement occasions and treatment groups indicate the differences from wait-list in average scores for children in the intervention conditions at post-test and follow-up.
--- INSERT TABLE 3 ABOUT HERE --- With regards to parent-reported problems in the internalising scale, there was a significant main effect of follow-up compared to baseline, with problems decreasing
significantly in all groups, F (1, 412.012) = 4.705, p = .031, 95% CI [-.572, -.028]. Baseline levels of problems were significantly higher for children in the parent-child intervention
compared to wait-list, F (1, 313.292) = 5.769, p = .017, 95% CI [-.080, -.801]. In addition, there was a significantly larger decrease in problems at post-test in the child intervention group
compared to the wait-list, F (1, 311.416) = 4.578, p = .033, 95% CI [-.611, -.025]. For parent-reported externalising problems, there was a significant decrease in parent-reported problems at post-test compared to baseline, F (2, 311.280) = 6.749, p = .01, 95% CI [-.323, -.045], but this effect did not differ according to treatment condition.
With regards to child-reported problems in the internalising scale, there were significant decreases in problem levels between baseline and post-test, F (1, 227.988) = 4.745, p = .03, 95% CI [.411, .021], and baseline and followup, F (1, 263.171) = 10.059, p = .002, 95% CI [.650, -.152], but no significant differences between conditions. For externalising problems there were no significant main effects of time or treatment condition, although coefficients indicated overall decreasing trends of problems over time.
Moderator effects. We inspected single moderator effects when the overall χ2 difference test indicated significant improvements in fit compared to the model in Step 2. Table 4 displays the standardised parameter estimates for Step 3 models in which we found significant single
23 moderation effects4.
--- INSERT TABLE 4 ABOUT HERE --- In this step, for each potential moderator, nine new parameters were added. The estimates of the moderator main effect and interactions with measurement occasions are to be interpreted as the effect that one standard deviation increase in the moderator has on the outcome (at baseline, post-test and follow-up, respectively) in the wait-list group. The coefficients of the interactions between moderator and treatment conditions are the effects of the moderator at baseline for the intervention groups. To investigate our hypotheses, we were interested in three-way interactions between moderator, treatment group, and measurement occasion (either T2 or T3). Significant interactions indicate that children’s psychosocial outcomes after the
intervention differ depending on the baseline level of the moderator. These effects are compared to the effect of the moderator on the wait-list at baseline, and controlled for the effect of the moderator at post-test in the wait-list. Since the single interaction effects cannot be interpreted by themselves (one needs to take into account the other relevant effects), the profiles of the significant moderators per treatment group are plotted in Figure 2.
--- INSERT FIGURE 2 ABOUT HERE --- Disengaged coping style significantly moderated parent-reported internalising problems at post-test for children in the child intervention compared to the wait-list, F (1, 315.937) = 9.997, p = .002, 95% CI [-.761; -.177]. As illustrated in Figure 2(a), children who were one standard deviation higher than average on the Disengagement scale at baseline decreased more at post-test than children who were lower on disengaged coping. The parameter estimate for the same effect in the parent-child intervention was in the same direction, but non-significant.
24 Another moderator effect of coping style was found in self-reports. Disengagement moderated self-reported externalising problems in the child intervention group at post-test, F (1, 218.109) = 6.659, p = .011, 95% CI [.070; .521]. Figure 2(b) shows that self-reported externalising
problems differed as a function of baseline coping level at post-test compared to baseline, for children in the child intervention. Specifically, children who were one standard deviation higher than average on the Disengagement scale at baseline increased in their reported externalising problems at post-test, while children who were one standard deviation lower decreased. Interestingly, parent-reported externalising problems were in the same direction for the child intervention, but not significantly so, r = .151, S.E. = .102, p = .142. It is important to note that the increase was only present at post-test. From the figure, it can be seen that externalising problems in the child intervention decreased by follow-up for children who were one standard deviation above average. Finally, there were no similar effects in the parent-child intervention, as indicated by the parameter estimates for self-reported problems in this group.
Attachment to the parent involved in the intervention significantly moderated the effect of the parent-child intervention at post-test, according to parent-reported externalising problems, F (1, 308.842) = 10.237, p = .002, 95% CI [-.548; -.131], and self-reported internalising problems, F (1, 218.070) = 10.005, p = .002, 95% CI [-.657; -.152]. As shown in Figure 2(c) and (d), the psychosocial adjustment problems of children who were one standard deviation more securely attached at baseline decreased significantly more than those of children who were less securely attached. Children in the child intervention appeared to follow the same trajectory for parent-reported externalising problem, but this effect was not significantly different from wait-list.
Parental psychopathology had a moderating effect on parent-reported internalising problems at post-test for children in the parent-child intervention, F (1, 298.563) = 5.625, p =
25 .018, 95% CI [-.611; -.057]. Figure 2(e) illustrates how children of parents with one standard deviation higher psychopathology at baseline significantly decreased in their internalising problems compared to children of parents with less mental health problems. Interestingly, with regards to self-reported internalising problems, it appeared that parental psychopathology acted as non-specific predictor of outcome (i.e., a variable that predicts the outcome irrespective of treatment condition; LaGreca et al., 2009). To further investigate this effect, we excluded the three-way interactions with treatment condition, and fitted a model with only the interactions between measurement occasions and the moderator. This model significantly improved the fit of the main effects model of Step 2, χ2
diff (3) = 13.157, p < .01. Children of parents with higher parental psychopathology reported significantly less internalising problems at post-test, F (1, 218.135) = 6.742, p = .01, r = -.145, S.E. = .055, 95% CI [-.255, -.035], and follow-up, F (1, 257.242) = 4.469, p = .035, r = -.156, S.E. = .074, 95% CI [-.301, -.011], compared to baseline and to children of parents with fewer mental health problems. This was not the case for self-reported externalising problems. Finally, all significant moderator effects were in the medium range, according to r effect size rules of thumb (Cohen, 1988).
Disease characteristics, age, and engaged coping style did not significantly moderate outcome for treatments on any of the dependent variables. Parenting stress had a significant moderating effect on self-reported internalising problems in the wait-list group at post-test and follow-up compared to baseline. Interestingly, there were significantly different effects in both the child and parent-child intervention, suggesting that parenting stress did not have moderating effects in the treatment conditions. Children in the wait-list who had parents with higher than average parenting stress at baseline reported significantly fewer problems at post-test, F (1, 238.993) = 7.131, p = .008, r = -.374, S.E. = .140, 95% CI [-.650, -.098], and follow-up, F (1,
26 266.748) = 5.564, p = .019, r = -.397, S.E. = .168, 95% CI [-.729, -.066].
Discussion
The present study sought to investigate whether specific disease, child, and parent characteristics associated with psychosocial adaptation in children with chronic illness would moderate the effectiveness of two formats of a cognitive-behavioural based group intervention that had the aim of promoting children’s adjustment and resilience. We proposed two alternative
perspectives: a) that children who had more advantageous characteristics at baseline would profit more from the interventions; and b) that children who started off with more risk factors would have more scope for change and thus benefit more. We also proposed that any effects would be stronger in the parent-child intervention, because of its additional parent component. In general, results did not lend uniform support to either perspective, and we found different moderators depending on the intervention format. Baseline risk and resistance factors that may have been most relevant to the specific treatment format moderated the outcome: Coping style moderated the effectiveness of the child intervention, while attachment and parental psychopathology had a moderating effect on the parent-child intervention. On the whole, our results provide interesting insights into the question ―which children benefit most from interventions, and which may need a different strategy?‖
A more disengaged coping style at baseline predicted a greater decrease in
parent-reported internalising problems at post-test for children in the child intervention, compared to the wait-list controls. This finding is in contrast with the ―congruence hypothesis‖ that interventions matching the child’s coping style are preferable (Piira et al., 2006). The empirical evidence available so far on this hypothesis is, however, based on children’s stress-processing and
27 respect, in that we considered general coping style as a person’s tendency to respond in a
particular way to a variety of stressful situations (Boekaerts & Röder, 1999), and targeted coping strategies for dealing with disease-related stressors, but not specifically medical procedures. From an alternative point of view, Blount, Davis, Powers, and Roberts (1991) propose that children should learn coping strategies that are inconsistent with their own, to increase adaptability to a wide range of situations. However, they also acknowledge that there is no evidence to suggest that children should be taught passive strategies. From a general
perspective, an active, engaged coping style is identified as an effective way of managing stress (LeBlanc et al., 2003; Meijer et al., 2002) while an avoidant or disengaged coping style is associated with more psychosocial problems (Compas et al., 2001; Thompson & Gustafson, 1996). In the present study, children who used a more disengaged coping style benefited more than others from an intervention aimed at increasing the use of active coping strategies. Thus, this specific subgroup of children particularly profits from an ―incongruent‖ approach.
Our results indicate that intervention efforts directed at teaching practical active coping skills work better with children with a lower stress-processing resistance factor (i.e., more disengaged coping style). Reporting a less engaged coping style could also be seen as a lower resistance factor. However, we did not find a moderating effect of this dimension. It may be that disengaged coping style is more relevant for identifying children who will profit more from the child intervention, because it is a dimension associated with maladjustment. Individuals who are most impaired may have more motivation to change and more room for improvement (De Ridder & Schreurs, 2001). This aspect, combined with the malleability of coping strategies in childhood and adolescence (Thompson et al., 2011), may have resulted in the child intervention improving the outcomes of those children who presumably needed the most support.
28 On the other hand, the results of the self-reported externalising problems appear to
contradict the findings discussed above. Children who started the intervention with a more disengaged coping style reported an increase in externalising problems at post-test, while children with a less disengaged style reported fewer problems. Since this is an unexpected finding, a few tentative interpretations can be advanced. One explanation may be that children with a more disengaged style at baseline were more reluctant to report their externalising
problems. Following the child intervention, they reflected more on their behaviours and actions, and their self-report became more truthful. However, there was a (non-significant) trend in the same direction in parent-reported externalising problems. Therefore, an alternative interpretation may be that children who reported a more disengaged coping style may need additional resources to reduce their externalising symptoms. The lack of a moderating effect of coping in the parent-child intervention seems to corroborate this interpretation. In fact, stress-processing skills may interact with resources offered by the social support of significant others (Boekaerts & Röder, 1999). Further research is needed to ascertain this possibility. On the whole, however, it is important to mention that self-reported externalising problem levels were below the norm threshold for the whole sample (Verhulst et al., 1997), therefore an increase may not necessarily be worrisome. In addition, the problems of children with higher disengaged coping tended to decrease at follow-up.
Another important finding of this study was the moderating effect of attachment in the adjustment of children following the parent-child intervention. We found that the intervention resulted in a larger decrease at post-test in parent-reported externalising problems and child-reported internalising problems for children who perceived a more secure relationship to the parent involved in the program. From the stress and coping perspective, securely-attached
29 children may be viewed as more ―resistant‖, because of the more supportive social-ecological
environment. The parent-child intervention in the present study involved training parents in observing and interacting with their children on feelings and cognitions. Our results suggest that these tasks may have found a more ―fertile‖ ground in parent-child relationships that were
characterised by security prior to the intervention, perhaps because these parent-child dyads found it easier to share their emotions and feelings.
On the other hand, children who perceived the relationship with their parent to be more insecure may need additional resources devoted at building a more secure bond. Perhaps, these children may benefit from more intervention sessions, or from an additional session in which the parent and the child work together on communication skills. It should be noted, however, that the results indicated a tendency for insecure children’s problems to decrease at the follow-up
measurement. Therefore, it is possible that these parent-child dyads simply needed more time to perceive the benefits of the intervention. In sum, attachment relationships have a widespread influence on psychosocial outcomes across the life-span (Waters, Merrick, Treboux, Crowell, & Albersheim, 2000). In addition, parental support is essential for children’s adjustment to the disease (El-Mallakh, Howard, & Inman, 2010). It is thus surprising that little attention has been paid to attachment as a moderator of treatment effectiveness. Particularly when parents are involved in an intervention, obtaining information about the parent-child relationship before beginning treatment may provide a basis for identifying parent-child dyads that may require additional resources devoted at improving their relationship.
A further noteworthy finding of this study was that, in the parent-child intervention, the parent-reported internalising problems decreased more for children of parents with higher psychopathology than for others. This result is in line with previous research with different
30 client groups. For instance, Cobham et al. (1998) found that a cognitive-behavioural intervention delivered with a parallel parent program was more effective (in terms of no longer meeting diagnostic criteria) for anxious children of parents with an anxiety disorder than for children of parents without psychopathology. Similarly, Gardner et al. (2010) found that the conduct problems of children after a parenting intervention were moderated by parental mental health, with a greater difference between treatment and wait-list for children of mothers with depression. The authors argued that this could be explained by the fact that the intervention equipped
mothers with goal-setting and problem-solving skills, important resources for improvement from depression. In the present study, the parent group may have helped parents with more mental health issues become more available and responsive to their child.
An alternative explanation may be that parental mental health improved during the intervention, perhaps through the increased social support that the groups offered. This aspect of the intervention may have led to parents feeling better and reporting fewer problems in their children. This possibility is promising, in that it suggests that the intervention may also have reduced the risk of children themselves suffering from psychopathology. One way to test the possibility that improvements in mental health mediate treatment outcome for children of parents with psychopathology (i.e., mediated moderation; Wu & Zumbo, 2008) would be to obtain parents’ mental health ratings not only at baseline, but also at further measurement occasions
following intervention. An additional point to consider is that the self-report of children of parents with higher psychopathology indicated significantly fewer internalising problems for all conditions at post-test. Since the majority of children in the wait-list received psychological help, this result may be interpreted as an indication that receiving any form of psychological support is perceived by the child as beneficial. On the other hand, another plausible explanation
31 is that these more at-risk children naturally regressed to the mean (Barnett, van der Pols, & Dobson, 2005).
Contrary to our expectations, we did not find a moderating effect of illness severity, duration, child’s age, or parenting stress. We examined disease parameters as moderators because previous research was inconclusive with regards to their impact on children’s
adjustment. The general impression in the literature is that, on the whole, disease parameters are not the most important influences on the adjustment of children with chronic illness (Wallander & Varni, 1998). However, it is of clear practical importance to establish whether psychosocial interventions on heterogeneous chronic conditions vary in their effectiveness along illness dimensions such as severity and duration. Thus, the present results may serve as a reassurance that the interventions can be applied following a non-categorical approach, which privileges the common life experiences and challenges that children with chronic illness face (Stein & Jessop, 1982). With regards to age effects, previous studies have not found age differences in emotional and behavioural problems in children with chronic illness (Thompson & Gustafson, 1996). Furthermore, since the interventions were adapted to the age group of the children, the lack of an age effect may be taken to indicate that the intervention protocols were successfully matched to the developmental requirements of the children. Finally, the fact that parenting stress did not influence the outcome in any of the treatment groups, but only in the wait-list, may be because it was not a stable baseline characteristic. In fact, it is plausible that levels of parenting stress changed as a result of the intervention, particularly in the parent-child format. Indeed, there is evidence that educating parents in child management techniques and parent-child communication is effective in reducing parenting stress (Deater-Deckard, 1998; Telleen, Herzog, & Kilbane, 1989). Under this light, parenting stress may be a mediator of treatment outcome, and may have
32 been, in retrospect, investigated as such.
The fact that we did not find the same moderators to operate in both interventions,
although in contrast with our expectations, is perhaps not surprising. Both interventions included the same child-focused protocol, but they differed, crucially, in parental involvement. The inclusion of the additional parent program in the parent-child intervention may not only have influenced the social-ecological environment, but also the processes set in motion by the child treatment protocol. Indeed, one of the aims of the parent program was to help children apply the learned skills to their everyday life (Scholten et al., 2011). Hence, for instance, parental support in the parent-child group may have reduced the differences that we observed between children with high or low disengaged coping in the child intervention, effectively attenuating the impact of baseline coping style on the treatment outcome of children in the parent-child intervention. Importantly, the absence, in one intervention, of a moderating effect that is present in the other intervention does not mean that the treatment was not effective, but simply that the main
effectiveness results hold for all children in that group, irrespective of the baseline characteristics under investigation.
Results should be interpreted in light of several limitations. First of all, findings of moderator effects on treatment outcome reported by the same rater of the moderating variable (either parent, or child) may be partly explained by shared method variance. This would be the case for the coping-externalising problems effect, and for the parental
psychopathology-internalising problems effect. With regards to attachment as a moderator, having found an effect in the same direction on both parent and child reported problems gives confidence in this result. Future research may consider including other informants, such as clinicians or the parent not involved in the intervention, and establishing whether the moderating effects that we found hold
33 across multiple informants.
Secondly, the scales we used to measure coping style in the present study had low internal validity, suggesting that the items possibly did not measure a single concept. This weakness might have arisen during the translation procedure, or the short-version adaptation, and may partially account for the contrasting effects on the parent and child’s reported problems.
However, similarly low psychometric properties of the short version of the Coping Strategies Inventory have also been reported in an African American sample (Addison et al., 2007). In the future, it would be recommended to replicate our results with an improved version of the
questionnaire, or to validate the self-report with observational instruments (Compas et al., 2001). Statistical limitations need to be taken into consideration as well. Moderator effect sizes are generally small (Chaplin, 1991, cited in Wu & Zumbo, 2008) and investigations are often underpowered to detect them (Hinshaw, 2007). The randomised controlled design, combined with multilevel modelling, increased the power of our analyses. However, the fact that most children in the wait-list received some form of treatment most certainly reduced the differences between groups, and, with that, the sizes of any moderator effects to be detected. Heterogeneous factors have also likely contributed to reducing any long-term effects, thus explaining why we only found evidence of moderation between baseline and post-test. Finally, with the many analyses performed, false positives are a possibility. Given the early stage of research in this field, taking precautions for multiple testing was not a priority. A key task for future research will be to validate our findings, possibly within a larger sample and with a more controlled comparison group. These improvements would also allow the investigation of multiple moderators at the same time, to examine their relationships and relative importance in moderating outcome.
34 In sum, the present study provides preliminary but valuable evidence that psychosocial interventions for children with chronic illness differ in their effects depending on baseline child and parent characteristics. Rather than endorsing one perspective (―Matthew effect‖) or the other (―more at-risk profit more‖), the pattern of findings suggests that it is important to consider the
type of characteristic under investigation. When we consider baseline characteristics that are ―intrinsic‖ to the child (coping style) or to the parent themselves (psychopathology), and that are
relevant to the treatment format (coping style in the child format, parental psychopathology in the parent-child format), the interventions work better for those who have greater scope for change. On the other hand, when we consider the characteristics of a mechanism through which the intervention may partly exert its effects (i.e., the parent-child relationship), those who are more advantaged at the start stand a greater chance of success. In other words, the interventions may be more effective for children and parents who need them the most; however, when
additional processes are involved, it is important to acknowledge that more favourable initial conditions of these processes may offer an advantage.
Empirical efforts should be devoted to continuing in this direction and expanding the evidence base of moderator effects in pediatric psychology interventions. Research of this kind is extremely important in guiding intervention development and informing clinical practice. With the information generated from this and similar studies, future RCTs can be stratified on factors that moderate treatment outcome, improving power and cost-effectiveness (Kraemer, Frank, & Kupfer, 2006). On a practical level, moderator investigations can provide clinicians with the information needed for customising interventions to clinically-relevant characteristics of pediatric populations (Drotar, 1997). With the results of the present study, we propose that the baseline level of coping style, attachment and, to a lesser extent, parental psychopathology may
35 be used to direct children to the most appropriate intervention format for their needs. For
instance, the child intervention may be prioritised for children who use more disengaged coping strategies and need to be taught alternative skills. Likewise, obtaining information regarding parent-child relationship security can advise clinicians on which dyads will benefit from a parent-child intervention and which may need additional resources. These and future efforts have the potential to accomplish an ultimate, fundamental outcome: delivering the all-round, evidence-based, and effective care that children with chronic illness deserve.
36 References
Achenbach, T. M. (1991a). Manual for the Child Behaviour Checklist/4-18 and 1991 Profile. Burlington, VT: University of Vermont, Department of Psychiatry.
Achenbach, T. M. (1991b). Manual for the Youth Self-Report and 1991 Profile. Burlington, VT: University of Vermont, Department of Psychiatry.
Achenbach, T. M. (1997). Manual for the Young Adult Self-Report and Young Adult Behavior Checklist. Burlington, VT: University of Vermont, Department of Psychiatry.
Addison, C. C., Campbell-Jenkins, B. W., Sarpong, D. F., Kibler, J., Singh, M., Dubbert, P., … Taylor, H. (2007). Psychometric evaluation of a Coping Strategies Inventory Short-Form (CSI-SF) in the Jackson Heart Study Cohort. International Journal of Environmental Research and Public Health, 4(4), 289–295. doi:10.3390/ijerph200704040004
Aiken, L. S., & West, S. G. (1991). Multiple regression: Testing and interpreting interactions. Newbury Park, CA: Sage.
Bakermans-Kranenburg, M. J., van IJzendoorn, M. H., & Bradley, R. H. (2005). Those who have, receive: The Matthew Effect in early childhood intervention in the home
environment. Review of Educational Research, 75(1), 1–26. doi:10.3102/00346543075001001
Barnett, A. G., van der Pols, J. C., & Dobson, A. J. (2005). Regression to the mean: what it is and how to deal with it. International Journal of Epidemiology, 34(1), 215–220.
doi:10.1093/ije/dyh299
Barrett, P. M., Dadds, M. R., & Rapee, R. M. (1996). Family treatment of childhood anxiety: A controlled trial. Journal of Consulting and Clinical Psychology, 64(2), 333–342.
37 Bauman, L. J., Drotar, D., Leventhal, J. M., Perrin, E. C., & Pless, I. B. (1997). A review of
psychosocial interventions for children with chronic health conditions. Pediatrics, 100(2), 244–251. doi:10.1542/peds.100.2.244
Beale, I. L. (2006). Scholarly literature review: Efficacy of psychological interventions for pediatric chronic illnesses. Journal of Pediatric Psychology, 31(5), 437–451.
doi:10.1093/jpepsy/jsj079
Bleil, M. E., Ramesh, S., Miller, B. D., & Wood, B. L. (2000). The influence of parent-child relatedness on depressive symptoms in children with asthma: Tests of moderator and mediator models. Journal of Pediatric Psychology, 25(7), 481–491.
doi:10.1093/jpepsy/25.7.481
Blount, R. L., Davis, N., Powers, S. W., & Roberts, M. C. (1991). The influence of environmental factors and coping style on children’s coping and distress. Clinical
Psychology Review, 11(1), 93–116. doi:10.1016/0272-7358(91)90139-L
Blount, R. L., Simons, L. E., Devine, K. A., Jaaniste, T., Cohen, L. L., Chambers, C. T., & Hayutin, L. G. (2008). Evidence-based assessment of coping and stress in pediatric psychology. Journal of Pediatric Psychology, 33(9), 1021–1045.
doi:10.1093/jpepsy/jsm071
Boekaerts, M., & Röder, I. (1999). Stress, coping, and adjustment in children with a chronic disease: A review of the literature. Disability & Rehabilitation, 21(7), 311–337. doi:10.1080/096382899297576
Brown, R. T., Daly, B. P., & Rickel, A. U. (2007). Chronic illness in children and adolescents. Advances in psychotherapy: Evidence-based practice (Vol. 9). Pennsylvania State
38 Christiano, B., & Russ, S. W. (1998). Matching preparatory intervention to coping style: The
effects on children’s distress in the dental setting. Journal of Pediatric Psychology, 23(1),
17–27. doi:10.1093/jpepsy/23.1.17
Cobham, V. E., Dadds, M. R., & Spence, S. H. (1998). The role of parental anxiety in the treatment of childhood anxiety. Journal of Consulting and Clinical Psychology, 66(6), 893–905. doi:10.1037/0022-006X.66.6.893
Cohen, J. (1988). Statistical power analysis for the behavioral sciences (2nd ed.). Hillsdale, NJ: Lawrence Erlbaum.
Colletti, C. J. M., Wolfe-Christensen, C., Carpentier, M. Y., Page, M. C., McNall-Knapp, R. Y., Meyer, W. H., … Mullins, L. L. (2008). The relationship of parental overprotection,
perceived vulnerability, and parenting stress to behavioral, emotional, and social adjustment in children with cancer. Pediatric Blood & Cancer, 51(2), 269–274. doi:10.1002/pbc.21577
Compas, B. E., Connor-Smith, J. K., Saltzman, H., Harding Thomsen, A., & Wadsworth, M. E. (2001). Coping with stress during childhood and adolescence: Problems, progress, and potential in theory and research. Psychological Bulletin, 127(1), 87–127.
doi:10.1037/0033-2909.127.1.87
Crawford, A. M., & Manassis, K. (2001). Familial predictors of treatment outcome in childhood anxiety disorders. Journal of the American Academy of Child & Adolescent Psychiatry, 40(10), 1182–1189. doi:10.1097/00004583-200110000-00012
Dadds, M. R., Holland, D. E., Laurens, K. R., Mullins, M., Barrett, P. M., & Spence, S. H.
(1999). Early intervention and prevention of anxiety disorders in children: Results at 2-year follow-up. Journal of Consulting and Clinical Psychology, 67(1), 145–150.
