Journal of Abnormal Child Psychology

, Volume 40, Issue 1, pp 133–143

A Structural Equation Analysis of Family Accommodation in Pediatric Obsessive-Compulsive Disorder


  • Nicole E. Caporino
    • Department of PsychologyUniversity of South Florida
    • Department of PediatricsUniversity of South Florida
  • Jessica Morgan
    • Department of PediatricsUniversity of South Florida
  • Jason Beckstead
    • College of NursingUniversity of South Florida
  • Vicky Phares
    • Department of PsychologyUniversity of South Florida
  • Tanya K. Murphy
    • Department of PediatricsUniversity of South Florida
    • Department of PsychiatryUniversity of South Florida
    • Department of PsychologyUniversity of South Florida
    • Department of PediatricsUniversity of South Florida
    • Department of PsychiatryUniversity of South Florida
    • Department of PediatricsUniversity of South Florida

DOI: 10.1007/s10802-011-9549-8

Cite this article as:
Caporino, N.E., Morgan, J., Beckstead, J. et al. J Abnorm Child Psychol (2012) 40: 133. doi:10.1007/s10802-011-9549-8


Family accommodation of symptoms is counter to the primary goals of cognitive-behavioral therapy for pediatric obsessive-compulsive disorder (OCD) and can pose an obstacle to positive treatment outcomes. Although increased attention has been given to family accommodation in pediatric OCD, relatively little is known about associated child and parent characteristics, and their mediating/moderating effects. This study examined a structural equation model of parent and child variables related to parent reports of family accommodation. Sixty-one children with OCD (ages 6–17 years, 39% female) and their parents were recruited from a university-based clinic. They were administered clinician- and parent-rated measures of child OCD symptom severity, OCD-specific impairment, internalizing problems, and externalizing problems as well as parent anxiety, depression, empathy, consideration of future consequences, and accommodation. Results generally supported the hypothesized model. Family accommodation mediated the relationship between OCD symptom severity and parent-rated functional impairment; child internalizing problems mediated the relationship between parent anxiety and family accommodation; and parent empathy and consideration of future consequences interacted to predict family accommodation. Child externalizing problems were significantly associated with family accommodation but neither of these two variables was associated with parent depression. Findings suggest that reductions in family accommodation might be maximized by routinely screening for comorbid psychopathology in children with OCD and their parents, and using prescriptive or modular approaches to intervention. Directions for future research are discussed.


Obsessive-compulsive disorderChildrenTreatmentAssessmentFamily accommodation

Obsessive-compulsive disorder (OCD) is characterized by recurrent obsessions (i.e., intrusive thoughts or images) and/or compulsions (i.e., ritualized behaviors or avoidance) that cause significant distress and impairment. Approximately 1% of children and adolescents are affected by OCD (e.g., Rapoport et al. 2000), which runs a chronic course in the absence of treatment (Rufer et al. 2005) and is associated with significant impairment across multiple domains (e.g., Piacentini et al. 2007).

Cognitive-behavioral therapy (CBT) with exposure and response prevention (E/RP), delivered alone or in conjunction with a serotonin reuptake inhibitor medication, is considered the frontline intervention for youth with OCD (Geller in press; Pediatric OCD Treatment Study Team 2004). However, as many as 35% of treated youth do not demonstrate improvement and many treatment responders have residual symptoms (e.g., Pediatric OCD Treatment Study [POTS] 2004). Since many youth do not respond optimally or at all to extant treatments (e.g., CBT, antidepressant medication), increasing attention has been given to identifying variables that predict treatment response. Given its clinical relevance and association with treatment outcome (Merlo et al. 2009), family accommodation has emerged as a variable of interest. Family accommodation refers to actions taken by family members to facilitate rituals (e.g., providing necessary objects), yield to the child’s demands (e.g., following a certain routine in order to minimize anxiety), provide reassurance to the child (e.g., answering questions repeatedly), assist with or complete tasks (e.g., homework) for the child, or decrease the child’s responsibility (e.g., limiting attempts at discipline) because obsessive-compulsive symptoms interfere with his/her ability to meet expectations (Storch et al. 2007a).

Family accommodation reinforces OCD-related rituals and avoidance. It conflicts with CBT goals by reducing obsessional anxiety in a fashion similar to ritual engagement, preventing the habituation of anxiety and limiting the child’s opportunities to learn that feared consequences are not likely to occur (Storch et al. 2007a). Also, accommodating a child’s symptoms diminishes the aversive consequences of obsessive-compulsive behavior (e.g., interference with preferred activities), perhaps reducing motivation for change. Although treatments have been modified to address family involvement in OCD symptoms (Barrett et al. 2004; Storch et al. 2007b), continued research on accommodation is necessary so that treatment response can be maximized by tailoring the content of interventions to families.

Several studies in the adult literature have investigated correlates of family accommodation (e.g., Calvocoressi et al. 1995; Stewart et al. 2008). To date, only two such studies have been published in the pediatric literature (Storch et al. 2007a; Peris et al. 2008). Storch et al. (2007a) examined family accommodation in relation to obsessive-compulsive symptom severity, functional impairment, and internalizing and externalizing problems in 57 clinic-referred youth and their parents. Family accommodation was positively related to each of these variables and mediated the relationship between obsessive-compulsive symptom severity and parent-rated functional impairment. In contrast, Peris et al. (2008) found that family accommodation was not significantly associated with obsessive-compulsive symptom severity, externalizing problems, or internalizing problems. However, symptom severity was related to parents’ involvement in rituals, while externalizing symptomology was associated with parent report of worse consequences when refraining from accommodation. Given that these studies yielded mixed findings, neither corrected for measurement error, and several clinically and theoretically relevant variables were not included in the study battery, further research is necessary to clarify the relationship between family accommodation and child variables that could be the target of intervention.

Regarding parent-level correlates of family accommodation, Peris et al. (2008) reported that overall psychopathology was associated with total scores on a measure of family accommodation and that parental anxiety was related to involvement in children’s rituals. There are several ways to understand these findings. First, parents who experience anxiety might be especially inclined to accommodate symptoms because they share their children’s fears or relate to the distress that their children experience in fear-provoking situations, increasing their motivation to “rescue” their children. Second, anxious parents might fear escalating expressions of distress from their child (e.g., panic-like symptoms), especially if they experience their own arousal-related body sensations as scary. Given that anxious parents often struggle to manage their own symptoms, they may lack the energy or resources to cope actively with their children’s expressions of distress. Finally, anxious parents have a learning history that involves repeated reinforcement of the use of avoidance strategies, which could be expected to generalize to novel anxiety-producing situations involving their children. Accommodating their children’s obsessive-compulsive symptoms is a way to avoid or escape any anxiety that they may experience when their children are distressed or confronted with situations that the parents also fear. Given these possibilities, the relationship between parental anxiety and accommodation might be explained, at least in part, through the effect of child internalizing problems on accommodation.

In addition to parental anxiety, parental depressive symptoms might impact levels of accommodation, although this has yet to be investigated. Depressed parents are less likely than non-depressed parents to expect that they will be competent and effective parents (e.g., Bor and Sanders 2004; Haslam et al. 2006). Further, depression appears to influence parents’ ability to be firm and consistent in their discipline of children and to avoid “giving in” to tantrums (e.g., Lovejoy et al. 2000). Thus, depressed parents might experience decreased motivation to refrain from accommodation because they expect that efforts to resist their child’s OCD-related demands will be unsuccessful, particularly when their child exhibits externalizing problems. If this were the case, one would expect child externalizing problems to mediate the relationship between parental depression and family accommodation.

In addition to mental health problems, individual differences in traits such as empathy and tendency to focus on the present might contribute to family accommodation. Empathy involves taking another person’s perspective and experiencing a visceral, emotional reaction that is congruent with the perceived welfare of that person (Batson and Coke 1981; Davis 1983). Parents with relatively high levels of trait empathy might be especially likely to accommodate symptoms because they have a strong emotional reaction to the anxiety that their children exhibit upon experiencing intrusive thoughts and because they are able to recognize the feared consequences of not engaging in a ritual. Accommodating behaviors could be intended to reduce the child’s distress and improve his/her functioning; however, they might also serve to alleviate the parent’s personal distress.

The extent to which empathy motivates accommodation may be affected by whether or not parents consider the long-term consequences of accommodating their children’s symptoms at the same time that they attend to their children’s distress. Clinical observations and preliminary data suggest that at least some parents recognize that family accommodation exacerbates symptoms in the long run but feel pressured to utilize a short-term solution (Merlo et al. 2009; Caporino et al. 2010). Whether or not parents submit to such pressure might be determined, in part at least, by individual differences at the trait level. Consideration of future consequences (CFC; Strathman et al. 1994) represents a relatively stable cognitive mindset that influences decisions about how to behave when there is a conflict between immediate and long-term outcomes. It refers to the extent to which individuals consider the potential distant outcomes of their current behaviors and are influenced by these potential outcomes. Parents who have low levels of CFC might be especially likely to accommodate their children’s symptoms because in the moment, they have difficulty appreciating the long-term benefits of refraining from accommodation.

This study tested a model of child and parent variables hypothesized to be related to parent reports of family accommodation. Specifically, four hypotheses were examined: (1) Family accommodation would mediate the relationship between OCD severity and functional impairment. (2) Parental anxiety symptoms would be positively associated with family accommodation and this relationship would be mediated by child internalizing symptoms. (3) Parental depressive symptoms would be positively associated with family accommodation and this relationship would be mediated by child externalizing symptoms. (4) Parental consideration of future consequences would moderate the relationship between parental empathy and family accommodation such that empathy would be more strongly associated with accommodation when parents give relatively little consideration to future consequences.



Sixty-one children and adolescents with OCD (n = 24 female) and their parents were recruited from a university-based specialty clinic in the southeastern United States. The participation rate was approximately 96%. Exclusion criteria were the presence of mental retardation or psychotic symptoms in the child. Youth ranged in age from 6 to 17 years (M = 11.61, SD = 3.18). The majority of youth were identified by their parents as non-Hispanic White (87%; n = 53), with the remaining identified as Hispanic/Latino (n = 4), Asian (n = 2), Native Hawaiian/Other Pacific Islander (n = 1), and biracial (n = 1). Eighty-three percent of caregivers who participated in the study were the child’s biological mother and 15% were the child’s biological father; one caregiver (2%) was the child’s stepmother. The majority of parent respondents (89%) were married or cohabitating. Approximately 18% of participants reported a combined household income of less than $50,000, 38% of families had an income between $50,000 and $100,000, 41% had an income over $100,000, and 3% chose not to report their household income.

Approximately 61% of youth participants met criteria for at least one comorbid diagnosis (34% met criteria for two or more additional diagnoses). Thirty percent of youth had a comorbid anxiety disorder, 23% had a comorbid disruptive behavior disorder, 20% had a comorbid obsessive-compulsive spectrum disorder (e.g., trichotillomania), and 12% had a comorbid mood disorder. Few youth in this sample had Asperger’s disorder (n = 2), an elimination disorder (n = 2), or an eating disorder (n = 1). Approximately 61% of youth were taking psychotropic medication and 54% had seen a mental health professional for psychotherapy prior to participating in the study.

Diagnoses were determined by a licensed clinical psychologist with expertise in the treatment of OCD, based on a clinical interview and according to criteria outlined in the Diagnostic and Statistical Manual of Mental Disorders—Fourth Edition Text Revision (American Psychiatric Association 2000). In 72% (n = 47) of the cases, diagnoses were confirmed using either the Anxiety Disorders Interview Schedule for Children—Child and Parent Version (ADIS-C/P; Silverman and Albano 1996) or the Schedule for Affective Disorders and Schizophrenia for School Age Children—Present and Lifetime Version (K-SADS; Kaufman et al. 1997). For the other 28% (n = 18) of cases, diagnoses were confirmed by the first author using case notes generated by the interviewing psychologist during the intake evaluation (kappa = 1.00).


Children’s Yale-Brown Obsessive-Compulsive Scale (CY-BOCS; Scahill et al. 1997)

The CY-BOCS is a clinician-rated, semi-structured inventory that assesses the presence/absence of 62 obsessive-compulsive symptoms along with their associated severity (i.e., distress, frequency, interference, resistance, control) across ten items. The CY-BOCS severity scale has good internal consistency, inter-rater, and test-retest reliability (Scahill et al. 1997; Storch et al. 2004), construct validity (Scahill et al. 1997; Storch et al. 2004), and treatment sensitivity (POTS 2004). For the current sample, the intraclass correlation coefficient between two raters was 0.97. CY-BOCS scores ranged from 12 to 33, with a possible range of 0 to 40.

Child Obsessive-Compulsive Impact Scale—Revised (COIS-R/P; Piacentini et al. 2007)

The parent-report version of the COIS-R consists of 33 items that assess the extent to which pediatric OCD has caused impairment in specific areas of psychosocial functioning over the past month. In support of the validity of the COIS-R/P, total and subscale scores were significantly associated with clinician severity ratings and comorbid internalizing and externalizing symptomology (Piacentini et al. 2007). The COIS-R/P has good internal consistency and 2-week test-retest reliability for the total and subscale scores (ICC = 80–0.88). The possible range of scores on the COIS is 0 to 99; in this study, scores ranged from 1 to 81.

Child Behavior Checklist/6-18 (CBCL; Achenbach and Rescorla 2001)

The CBCL is a parent-report measure of childhood internalizing and externalizing symptoms over the past 6 months. It consists of 118 items that yield a Total Score, an Internalizing Problems Composite Score, and an Externalizing Problems Composite Score. The psychometric properties of the CBCL have been demonstrated across a variety of clinical and nonclinical populations (Achenbach and Rescorla 2001). The possible range of CBCL Internalizing raw scores is 0 to 60; scores in this study ranged from 1 to 38. The range of possible CBCL Externalizing scores is 0 to 66; scores in this study ranged from 0 to 41.

Brief Symptom Inventory (BSI; Derogatis 1993)

Parents completed the BSI, which is a 53-item self-report questionnaire that measures current psychopathology along nine dimensions. For this study, symptoms from only two dimensions were assessed: Anxiety and Depression. Internal consistency for each dimension is good (Derogatis 1993). Test-retest reliability has been demonstrated for 2-week intervals (Derogatis 1993). The possible range of BSI Anxiety scores is 0 to 24; scores in this study ranged from 0 to 20. The possible range of BSI Depression scores is 0 to 24; scores in this study ranged from 0 to 16.

Interpersonal Reactivity Index (IRI; Davis 1983)

Parental empathy was measured using the Perspective-Taking (PT) and Empathic Concern (EC) scales of the IRI. Each of these scales requires participants to respond to seven statements using a 5-point Likert-type scale. The PT scale reflects the cognitive component of empathy and consists of items that assess the tendency to spontaneously adopt the point of view of others (e.g., “When I’m upset at someone, I usually try to ‘put myself in his shoes’ for awhile”). The EC scale reflects the affective component of empathy and consists of items that assess the degree to which the respondent experiences feelings of warmth, compassion, and concern for others (e.g., “I am quite often touched by things that I see happen”). Internal consistency and test-retest reliability for the PT and EC scales are adequate (Davis 1980), and convergent and divergent validity have been established (Davis 1983). As in previous studies (e.g., Galinsky et al. 2008), items from these scales were combined to form a single measure of empathy. IRI scores ranged from 7 to 56, with a possible range of 0 to 56.

Consideration of Future Consequences Scale (CFC Scale; Strathman et al. 1994)

The CFC scale is a 12-item measure of the extent to which individuals consider the potential distant outcomes of their current behaviors and the extent to which they are influenced by these potential outcomes. Good internal consistency was reported in the original validation study and test-retest reliability coefficients ranged from 0.76 to 0.72 for 2- and 5-week intervals, respectively (Strathman et al. 1994). Convergent and divergent validity were demonstrated using measures of deferment of gratification, locus of control, future orientation, and attitudes toward social causes. Incremental validity was established by demonstrating that scores on the CFC scale predicted a significant amount of unique variance in measures of environmentalism and concern for health over and above related individual-difference measures such as conscientiousness. CFC scores ranged from 19 to 56, with a possible range of 12 to 60.

Family Accommodation Items

Family accommodation was measured using nine items from Calvocoressi et al. (1995) that tap into the behavioral involvement of family members in the child’s OCD (e.g., participation in rituals, modification of daily routines). These items have demonstrated good internal consistency (Cronbach’s alpha = 0.88; Peris et al. 2008). Summative scores have been negatively associated with family organization and functioning, and positively associated with familial stress, OCD symptom severity, and impairment (Calvocoressi et al. 1995; Peris et al. 2008; Storch et al. 2007a; 2010). The possible range of scores on these family accommodation items is 0 to 36; the range obtained in this study was 1 to 32.


All study procedures were approved by the local Institutional Review Board. Participants were recruited following an intake evaluation with a licensed psychologist experienced in the assessment and treatment of OCD. When more than one parent presented to the initial interview, the individual who spends the most time with the affected child was invited to participate in the study. Prior to the start of data collection, research assistants were trained to a reliable standard on the ADIS-P, the K-SADS, and the CY-BOCS through didactics, joint interviews, and supervision discussion. Diagnostic interviews (i.e., the ADIS-P or the K-SADS) were administered to each parent and child jointly or alone (depending on clinical appropriateness) as part of their participation in unrelated treatment studies; such assessments were within 10 days after administration of the CY-BOCS and other study measures. The CY-BOCS was then administered to each parent–child dyad and clinical judgment was used to make final ratings based on information from both respondents. Parents completed self-report measures in the following order: background questionnaire, COIS-R/P, CBCL, BSI, CFC, and IRI. The clinician-rated family accommodation items were administered to parents last.

Analytic Plan

Structural Equation Modeling (SEM) was used to examine hypotheses. Advantages of SEM over traditional regression models include the ability to model constructs as latent variables, accounting for measurement error in observed variables. With SEM, the minimum sample size depends on levels of communality (i.e., the portion of the variance in an observed variable that is accounted for by the latent variables) and overdetermination (i.e., the degree to which each factor is clearly represented by a sufficient number of variables). Monte Carlo studies (e.g., MacCallum et al. 1999) have demonstrated that when communalities are high (values of 0.6, 0.7, or 0.8), the roles of sample size and overdetermination are less important than when communalities are low (values of 0.2, 0.3, or 0.4) or wide ranging (values of 0.2 to 0.8). Because single indicators were used in the current study and communalities were high (i.e., internal consistency of the measures was strong), a sample size of 60 (versus 100, 200, or 400) was estimated to be sufficient to produce an admissible solution. In other words, Cronbach’s alphas were used to fix rather than estimate reliability, minimizing the ratio of parameters to be estimated to the number of observations.

Hypothesized pathways were tested with LISREL 8.80 (Jöreskog and Sörbom 2006) using the maximum likelihood (ML) method of parameter estimation. Models were evaluated for convergence using absolute, incremental, residual-based, and population-based fit indices (Kline 2005). The goodness of fit index (GFI) is an absolute fit index that represents the proportion of variability in the sample covariance matrix explained by the model. The GFI ranges from 0 to 1, with values of 0.9 or greater indicating good fit (Jöreskog and Sörbom 2006). The comparative fit index (CFI) is an incremental fit index that involves assessing fit relative to an independence or null model using a noncentrality parameter, which can be estimated as χ2df (Bentler 1988). The CFI ranges from 0 to 1, with values greater than or equal to 0.9 indicating good fit. The standardized root mean square residual (SRMR) is a residual-based fit index that represents the average of the differences between the observed correlations and the estimated correlations. It ranges from 0 to 1, with values less than or equal to 0.09 indicating good fit (Hu and Bentler 1999). The root mean square error of approximation (RMSEA; Steiger 1990) is a population-based index that involves analysis of residuals and reflects the lack of fit due to reliability and model misspecification (Browne and Cudeck 1993). The RMSEA indicates fit per degree of freedom of the model; a value of less than 0.08 is considered acceptable, with values of 0.05 or lower indicating very good fit. LISREL provides a p-value for a ‘test of close fit,’ which examines the null hypothesis that the RMSEA is no greater than 0.05.

To improve the hypothesized model, standardized residuals, modification indices, and expected change values were inspected. Trivial and non-significant relationships among exogenous variables, which were allowed to correlate freely in the original model, were fixed (i.e., set to 0) in the final model, increasing the degrees of freedom for hypothesis testing.

Hypotheses 1 through 3 were examined using tests of significance for indirect effects (Sobel 1982), which refer to the influence of an independent variable (e.g., parent anxiety symptoms) on a dependent variable (e.g., family accommodation) as mediated by one or more intervening variables (e.g., child internalizing problems). Hypothesis 4, which involves latent variable interaction, was tested according to the procedure outlined by Ping (1996) that involves estimating error variances for the indicators of latent variables (i.e., parent empathy and CFC), creating an interaction latent variable by forming a product of the indicators of latent variables (Kenny and Judd 1984), using error variances for these indicators to calculate the error variance of the interaction latent variable, and specifying the resulting error variance as a constant in the structural model containing the interaction variable.


Preliminary Analyses

Missing Data

All study measures were completed for each participant. There were very few missing data. For each measure, a total of 0 to 4 observations was missing across all participants included in the study. Measures were scored by averaging responses across the items that make up each scale.

Descriptive Statistics

Descriptive statistics and correlations among indicator variables in the structural equation model are presented in Table 1.1 Given that latent variables were represented by single indicators with high levels of reliability (alphas > 0.70), the measurement model is considered adequate.
Table 1

Zero order correlations, item means, standard deviations, and reliability estimates for indicators











1. BSI Anxiety



2. BSI Depression




3. IRI (PT+EC)





4. CFC













6. CBCL Internalizing








7. CBCL Externalizing



















9. FA






























n = 61 parent–child dyads. Cronbach’s alphas are listed on the diagonal. BSI Brief Symptom Inventory (0–4), IRI (PT+EC) Interpersonal Reactivity Index, Perspective Taking + Empathic Concern scales (0–4), CFC Consideration of Future Consequences Scale (1–5), CY-BOCS Children’s Yale-Brown Obsessive Compulsive Inventory (0–4), CBCL Child Behavior Checklist (0–2), COIS-R/P Child OC Impact Scale-Revised, Parent Report (0–3), FA family accommodation items(0–4)

*p < 0.05

Model Fit

The minimum fit function chi-square for the hypothesized model was significant, \( \chi_{\text{M}}^2\left( {{21}} \right) = {35}.{84} \), p = 0.02, reflecting less than adequate fit. With the exception of the RMSEA, other fit indices suggest that the model fit the data reasonably well (see Table 2). The second model that was tested was identical to the first model except that trivial and non-significant associations among exogenous variables were specified. The second model yielded a chi-square that was not significant, \( \chi_{\text{M}}^2\left( {{3}0} \right) = {39}.{845} \), p = 0.11. The RMSEA was much improved, with the test of close fit (p = 0.108) indicating that this index did not differ significantly from the 0.05 standard of very good fit. The CFI was also good; the GFI and SRMR were reasonable. The second model did not differ significantly from the first model in terms of fit [\( \chi_{{_{\text{difference}}}}^2\left( {9} \right) = {3}.{9}0{1} \), p = 0.92] but was retained based on the consideration of parsimony. An examination of the 55 standardized residuals revealed that they were normally distributed and with the exception of one, they were all smaller than ±2.58. Standardized parameter estimates for the final model are displayed in Fig. 1. Included are estimates of measurement error and structural coefficients corrected for attenuation.
Table 2

Summary of model fit statistics








Hypothesized model







Parsimonious model







n = 61 parent–child dyads. RMSEA root mean square error of approximation, SRMR standardized root mean square residual, CFI comparative fit index, GFI goodness of fit index
Fig. 1

Final structural equation model (n = 61). Standardized path coefficients are reported. BSI = Brief Symptom Inventory; IRI = Interpersonal Reactivity Index; CFC = Consideration of Future Consequences Scale; CY-BOCS = Children’s Yale-Brown Obsessive Compulsive Inventory; CBCL = Child Behavior Checklist; COIS-R = Child OC Impact Scale-Revised, Parent Report; FA = family accommodation items. *p < 0.05

Hypothesis Testing

Standardized total effects of the latent variables on each other are presented in Table 3. The final solution explained over 80% of the variance (1-0.188) in family accommodation.
Table 3

Standardized total effects among latent variables in final model


Exogenous on Endogenous

Endogenous on Endogenous

































ANX parent anxiety, DEP parent depression, SEV child OCD symptom severity, EMP parent empathy, CFC parent consideration of future consequences, INT child internalizing problems, EXT child externalizing problems, ACC family accommodation, IMP child OC-specific impairment

*p < 0.05

Relationships Among Latent Variables

Child obsessive-compulsive symptom severity had a significant indirect effect on child OCD-specific impairment via family accommodation, t(60) = 4.94, p < 0.05. Modification indices did not suggest the addition of a direct path from child OCD symptom severity to OCD-specific impairment (MI = 0.30), supporting full mediation.

Parent self-reported anxiety symptoms were significantly associated with child internalizing problems, which in turn, were significantly associated with parent reports of family accommodation, t(60) = 2.20, p < 0.05. Modification indices did not suggest the addition of a direct path from parent anxiety symptoms to family accommodation (MI = 1.93), supporting full mediation. Parent anxiety symptoms were also significantly associated with child functional impairment, t(60) = 2.17, p < 0.05, suggesting three-path mediation.

Although child externalizing problems were significantly associated with family accommodation, t(60) = 2.35, p < 0.05, the path from parent depressive symptoms to child externalizing problems was not significant. Modification indices suggest that the fit of the model would be improved substantially by allowing a direct path from parent depressive symptoms to family accommodation (MI = 4.23). However, the expected change value (SEPC = −0.219) indicates that the sign of the path coefficient would be negative even though the zero-order correlation between parent depressive symptoms and family accommodation was positive. This situation occurs when there are suppression effects (Conger 1974; MacKinnon et al. 2000); thus, a direct path was not added. Had this path been included, it would have inflated the influence of other variables.

Finally, parent self-reported empathy and CFC did not have significant direct effects on parent reports of family accommodation; however, as predicted, their interaction was significant, t(60) = −2.20, p < 0.05. Follow-up tests of simple slopes (Aiken and West 1991) revealed that at low levels of CFC (i.e., one standard deviation below the mean), the relationship between empathy and accommodation was positive and significant, β = 0.25, t = 2.44, p < 0.05. That is, among parents who are present-oriented, highly empathic parents accommodate more than parents who score low on measures of empathy. At average levels of CFC (i.e., at the mean), the slopes were not significant, β = 0.01, t = 0.14, p = 0.89, nor were they significant at high levels of CFC (i.e., one standard deviation above the mean), β = −0.23, t = −1.80, p = 0.08. Figure 2 illustrates these relationships.
Fig. 2

The interaction between parent empathy and consideration of future consequences in the prediction of family accommodation


This study extends the literature on family accommodation by examining its relationship to a number of theoretically relevant child and parent variables that could be the target of intervention. SEM was used to examine multiple relationships simultaneously while accounting for measurement error, which provided a more powerful test of hypotheses than multiple regression. As expected and consistent with Storch et al. (2007a), parent reports of family accommodation mediated the relationship between child obsessive-compulsive symptom severity and parent-rated functional impairment. One interpretation of this finding is that symptom severity might influence the extent of family accommodation, which in turn, contributes to the level of functional impairment in children, perhaps by limiting their opportunities to develop problem solving skills and putting a strain on relationships with family members (e.g., Steketee and Van Noppen 2003). Some data indicate that parents accommodate obsessive-compulsive symptoms to minimize impairment and a subset of parents believe that accommodation is helpful to their child in the present and the long run (Caporino et al. 2010). Taken together, these findings suggest that educating parents about the negative consequences of accommodating behaviors (i.e., their relationship to impairment) might be a necessary first step to decreasing their frequency.

As hypothesized, child internalizing and externalizing symptoms were each significantly associated with accommodation, and internalizing symptoms mediated the relationship between parental anxiety symptoms and family accommodation. These findings are generally consistent with others (Storch et al. 2007a; Peris et al. 2008) and suggest that the treatment of pediatric OCD could be enhanced by addressing certain comorbid conditions in the affected child and his/her parents. Specifically, addressing parental anxiety may reduce family accommodation by increasing parents’ ability to tolerate exposing their children to feared situations as well as their resources for coping with resulting expressions of distress. The intensity of treatment may be a function of the level of parental distress; more anxious parents may benefit from more intensive and possibly multimodal intervention to reduce their anxiety. Also, training in behavior management techniques (e.g., Kazdin 2005) may assist parents in refraining from accommodating OCD symptoms in the face of tantrums or other angry responses from their children. At the child level, treating children’s internalizing problems may improve their ability to manage the distress that they experience when accommodation is refused or when they are exposed to an anxious trigger. Intervention studies should be conducted to more directly examine these possibilities.

The hypothesis that child externalizing problems would mediate the relationship between parental depressive symptoms and family accommodation was not supported. Parental depressive symptoms had neither a direct nor an indirect effect on family accommodation. It may be that parental depressive symptoms do not contribute to understanding family accommodation. Alternatively, it’s possible that significant effects would have emerged if there had been more variability in parental depressive symptoms. Or, child externalizing problems may have a moderating versus mediating effect on the relationship between parental depressive symptoms and family accommodation. For depressed parents whose children exhibit defiant or aggressive behavior, it might be easier (i.e., require less energy or cause less disruption) to accommodate obsessive-compulsive symptoms than to refrain from accommodation. However, for depressed parents whose children are not oppositional, doing nothing in response to obsessive-compulsive symptoms is likely easier than accommodating them.

There was a significant interaction between parent empathy and CFC such that at low levels of CFC, empathy was positively associated with family accommodation. Parents who were highly empathic and less likely to consider future consequences (e.g., in the face of daily hassles or expressions of distress from their children) exhibited relatively high levels of accommodation. Given that these constructs had not been examined in the parenting literature prior to this study, findings should be replicated. However, the current results suggest that it might be useful to consider CFC and empathy in exploring parent-focused strategies for targeting accommodating behaviors. Motivational interviewing, for example, focuses on resolving ambivalence about behavior change that has short-term negative consequences but long-term positive consequences. Low levels of CFC have predicted the success of motivational interviewing interventions in other populations (Carey et al. 2007). These interventions might be effective for parents who continue to accommodate obsessive-compulsive symptoms after being educated about the long-term benefits of refraining from accommodation. Also, given that most parents accommodate in an effort to reduce their child’s distress (Caporino et al. 2010), which is an immediate outcome, cognitive distraction techniques (e.g., Linehan 1993) might be helpful to present-oriented and highly empathic parents who have difficulty disengaging from their child’s rituals. This possibility should be further explored with additional research.

Although this study advances the literature on family accommodation, there were several limitations. First, not all OCD diagnoses were confirmed using a structured clinical interview (28%). However, there was 100% agreement between two clinicians with expertise in OCD for these cases and CY-BOCS scores were all above 12 (an accepted cutoff for clinically significant symptoms). Second, the sample was relatively homogeneous demographically, limiting the extent to which findings generalize to the larger population of youth with OCD. Third, because this study was cross-sectional, causal effects could not be inferred. Prospective longitudinal studies are needed to confirm the direction of the relationships found significant. Fourth, the magnitude of some relationships might have been inflated by common method variance resulting from the use of parents as central informants. It will be necessary to replicate findings using multiple raters to ensure that observed relationships can be attributed to the traits of interest and not same-source bias. Finally, there may be other variables not assessed in this model that account for variance in family accommodation; however, the fact that the model explained over 80% of the variance suggests the clinical relevance of the included variables.


This study examined complex relationships among child and parent variables that further our understanding of family accommodation in pediatric OCD and have important clinical implications. First, screening for comorbid psychopathology in children and parents as part of routine clinical assessment may be necessary to maximize reductions in family accommodation, which has been associated with positive treatment response (Merlo et al. 2009). Assessing levels of accommodation without understanding its function (e.g., to alleviate parental anxiety) within the family context may not provide sufficient information for optimal treatment planning. Second, prescriptive or modular approaches to intervention with families of children who have OCD might facilitate large reductions in family accommodation. Modular treatment protocols allow CBT procedures to be applied in a highly individualized manner (Chorpita et al. 2004) and address issues that may be specific to the child. Future studies should explore the utility of adding to the standard CBT protocol brief modules that focus on psychoeducation about the impact of accommodation, cognitive and behavioral strategies for reducing parental anxiety, contingency management strategies, distraction techniques for tolerating empathic responses to child distress, and motivational interviewing.


Non-normality was addressed using square-root transformations. Results did not vary substantively when analyses were conducted with transformed variables. Thus, results of analyses with non-transformed variables are reported.



The contributions of Jeannette Reid and Adam Lewin, Ph.D. are acknowledged.

Copyright information

© Springer Science+Business Media, LLC 2011