Parent–child concordance in reporting of child eating disorder pathology as assessed by the eating disorder examination


  • Paige Mariano BPsych,

    1. School of Psychology, Murdoch University, Perth, Australia
    Search for more papers by this author
  • Hunna J. Watson PhD, MPsych(Clin),

    Corresponding author
    1. Eating Disorders Program, Princess Margaret Hospital for Children, Perth, Australia
    2. Centre for Clinical Interventions, Department of Health in Western Australia, Perth, Australia
    3. School of Paediatrics and Child Health, The University of Western Australia, Perth, Australia
    4. School of Psychology and Speech Pathology, Curtin University, Perth, Australia
    • Correspondence to: Dr. Hunna Watson, Princess Margaret Hospital for Children, GPO Box D184, Perth, Western Australia, Australia, 6840. E-mail:

    Search for more papers by this author
  • David J. Leach MEPsych,

    1. School of Psychology, Murdoch University, Perth, Australia
    Search for more papers by this author
  • Julie McCormack MSc(Psych),

    1. Eating Disorders Program, Princess Margaret Hospital for Children, Perth, Australia
    Search for more papers by this author
  • David A. Forbes MBBS, FRACP

    1. Eating Disorders Program, Princess Margaret Hospital for Children, Perth, Australia
    2. School of Paediatrics and Child Health, The University of Western Australia, Perth, Australia
    Search for more papers by this author



The aim of this study was to examine parent–youth concordance in reporting of eating disorder pathology, as assessed by the Eating Disorder Examination (EDE) in a clinical pediatric sample.


The sample comprised 619 parent-youth dyads of youth (8–18 years) presenting for treatment at a specialist eating disorder clinic. A cross-sectional correlational design was used to examine the association between parent and youth symptom reports.


On the whole, parent–youth inter-rater agreement was poor to moderate. Agreement was acceptable for the presence of behavioral symptoms, with the exception of excessive exercise (PAK = 0.48–0.98). There was poor inter-rater agreement on frequency of behavioral symptoms, with parents providing lower estimates than youth (ICC = 0.07–0.52). Although we predicted that inter-rater agreement on cognitive symptoms would by higher with adolescents than children, both groups were discordant with parent reports. Younger children identified less severe eating disorder cognitions than parents and the opposite occurred for adolescents. An anorexia nervosa presentation and lower malnutrition were not associated with lower inter-rater agreement, as might have been expected through ego syntonicity. Youth with bulimia nervosa presentations reported significantly higher severity of cognitive symptoms and more frequent disordered eating behaviors compared with their parents.


Results support the utility of parent–youth assessment via the EDE to obtain a wider clinical picture of eating disorder psychopathology in children and adolescents, particularly for younger children. Clinical implications pertinent to administration of the EDE and parent literacy regarding eating disorder symptoms are discussed. © 2013 Wiley Periodicals, Inc. (Int J Eat Disord 2013; 46:617–625)


The Eating Disorder Examination (EDE)[1] is a structured clinical interview widely held as the gold standard for assessing eating disorder psychopathology. The EDE has been extensively validated among adults,[2-5] yet only a small body of research explores its application among youth.[6, 7] Multiple informant assessment is widely recognized as achieving a more comprehensive understanding of youth psychopathology[8, 9] and may be particularly valuable when assessing eating disorders. While the introduction of a parent adaptation of the EDE has enhanced the measurement of eating disorder symptomatology,[7] literature assessing parent–youth concordance on eating disorder symptoms is scarce. Henceforth, we use the terms “child” (<13 years), “adolescent” (13–17 years), and “youth” (<18 years) to demarcate age groups.

Generally, multi-informant psychological assessment of youth yields low to moderate concordance.[10] Greater parent–youth concordance exists for externalizing symptoms (observable behaviors) than internalizing symptoms (cognitive, mood, and physiological elements).[11] In an Australian clinical sample of 379 youth (11–18 years) whereby parents and youth agreed on the presence of a psychiatric disorder, higher concordance appeared for externalizing (attention deficit hyperactivity disorder and conduct disorders) than internalizing (anxiety, depression, and somatoform disorders) disorders.[12] Eating disorder behaviors, such as binge eating and purging, are generally concealed and carried out privately, making it unclear whether concordance for observable behaviors generalizes in the same manner. Disordered eating research in nonclinical samples suggests nonconcordance of observations, for example, parents of average and overweight youth (∼6–14 years) detected disordered eating cognitions (e.g., weight and shape concern) more accurately than overeating and binge eating behaviors.[13-15] Parent–youth (10–18 years) reports of binge eating found that parents were able to identify if their child did not have binge eating disorder (82% agreement) but were less likely to recognize a subclinical (15% agreement) or clinical binge eating disorder (25% agreement).[16]

Youth characteristics may influence concordance on diagnostic assessments.[17] Developmentally, children have lower verbal ability, capacity for abstract thought, insight toward their emotions, and behavioral constraint.[18] Children can be influenced by impulsivity in responding, difficulty attending, and fatigue during assessment. Inclination to provide socially desirable responses that conform with the question direction or what the child believes the assessor wants to hear can occur.[19] Children may experience difficulty, more so than adolescents, recognizing cognitive antecedents as opposed to environmental antecedents to certain behaviors, due to the required ability for introspection.[20] For these reasons, the age of the informant is a critical variable influencing the reliability of clinical interview. Edelbrock and colleagues[21] found that reliability of youth (6–18 years) reports on the Diagnostic Interview Schedule for Children improved with age, and that younger children, aged 6–9 years old, were not particularly reliable in disclosing their anxiety symptoms.

Adolescents as a cohort score lower on the EDE relative to adult counterparts.[22] Adolescents have been found to have greater levels of denial and lower levels of desire for help than adults diagnosed with eating disorders,[23] potentially leading to poor parent–youth concordance on eating disorder cognitions and behaviors. The amplified nature of egosyntonicity and the impact of minimization may yield lower scores among youth-informants, as might clinical presentations with complete absence of weight and shape-based disturbance, whereas the parent may superimpose these concerns as part of their explanatory model for the symptoms. Inhibited personal disclosure of behavioral and physical manifestations (binge eating and purging) may occur because of shame and guilt related to perceived stigma.[24, 25] The severe broader health implications of anorexia nervosa (AN), with potential adverse impacts on growth and brain development,[26] may in very severe cases impair affected individuals' capacity to report on their cognitive symptoms. During an active phase of AN, 50% of individuals suffer from mild cognitive impairment.[27] Factors that influence youth reporting of eating disorders present a substantial concern due to the resulting barriers for early identification and treatment.

Parent-related characteristics can influence responding on parent-informant measures. Some parents may underreport to suppress their child's behavioral problems, to avoid their child being stigmatized, or to conceal family problems and relationship conflicts and present well to assessment agencies.[28, 29] Parental insight may be limited to the domains in which they observe their child, reducing awareness of problematic behavior outside the home, such as at school or with peers, particularly for adolescents. A diagnostic concordance investigation in a community adolescent (14–18 years) sample demonstrated substantial parent–adolescent agreement for a key AN indicator (refusal to maintain body weight at or above 85% for age and height) but only moderate agreement for a key indicator of bulimia nervosa (BN), recurrent episodes of binge eating.[8] In a nonclinical sample of 319 female youth (12–15 years), mothers' reports of daughters' eating disorder symptoms were significantly lower than those reported by their daughters and concordance for bulimic behaviors was near zero.[30] Mother–daughter concordance for cognitive symptoms was greater but on the whole modest. By contrast, in another study, parents reported significantly more binge eating episodes and significantly fewer compensatory episodes in 6–12-year olds relative to children's self-report.[13] Parents may have a better conceptual understanding of the content domain being assessed compared to children.

Recently, Couturier and researchers[7] undertook the first investigation examining parent–youth concordance on child eating disorder symptom ratings in a clinical population. Youth (N = 117) aged 9–18 years and their parents were administered parallel parent and child versions of the EDE. For youth with AN or restricting symptoms, self-reported scores were significantly lower on the subscales of restraint and weight concern compared with parental reports. Those with BN or binge-purge symptoms reported higher restraint and shape concerns in contrast with parental reports. We propose to extend this line of research by investigating concordance on eating disordered behaviors, which was not examined in Couturier's study, and to examine explanatory factors associated with concordance.

The first aim of the present study was to describe and investigate parent–youth concordance on the EDE with respect to cognitions and eating disordered behaviors (Hyp 1-3a). No a priori predictions were made regarding the level of inter-rater agreement, given previous limited research and clinical complexities including egosyntonicity, which may lead to patient minimization of symptoms. The second aim was to examine how parent–youth concordance on the EDE varies with age, diagnosis, and malnutrition (Hyp 3b-7). Specific hypotheses follow:

  • Hyp 1: Given the concealed nature of eating disorder behaviors, there will be poor parent–youth inter-rater agreement on frequency of behavioral symptoms.
  • Hyp 2: There will be acceptable agreement on presence of behavioral symptoms.
  • Hyp 3a: Agreement for disordered eating behaviors will be higher for youth who reported the absence of these behaviors compared with youth who reported their presence.
  • Hyp 3b: Inter-rater agreement on scale scores will be higher for adolescents than children, due to cognitive developmental limitations[31] and increases in reliability of reporting that improve with age.[21]
  • Hyp 4: Due to the impact of egosyntonicity on symptom reporting within AN presentations,[25] inter-rater agreement on global eating disorder psychopathology will be lower for youth with AN or eating disorders not otherwise specified (EDNOS) AN, compared with BN or EDNOS-BN
  • Hyp 5: Youth with AN or EDNOS-AN will report less-elevated psychopathology relative to parents.
  • Hyp 6: Youth with BN or EDNOS-BN will identify a higher frequency of binge eating and purging relative to parents.
  • Hyp 7: Youth with higher levels of malnutrition will report less severe psychopathology relative to parents.



The initial population pool (N = 961) of this study was children (<13 years) and adolescents (13–18 years) consecutively referred to the Eating Disorders Program at Princess Margaret Hospital for Children (PMH EDP) in Perth, Western Australia, between 1996 (service inception) and 2012. PMH EDP operates the only public sector pediatric eating disorder program in the state. Referrals are accepted from general practitioners, psychiatrists, psychologists, and school nurses and counselors, and the general community. PMH EDP is multidisciplinary with staff representing many health professions including medicine, psychiatry, clinical psychology, nursing, dietetics, occupational therapy, physiotherapy and social work; and schoolteachers are provided through the Department of Education Hospital School Services. A continuum of care is offered ranging from outpatient to day program and inpatient, with consultation and liaison support, plus an education and research program. There are no service exclusion criteria, although mild to moderate eating disorders are generally redirected toward community services. We define children as <13 years because the average age of menarche onset is 12.8 years.[32]

This study is part of a larger project called the Helping to Outline Paediatric Eating Disorders (HOPE) Project, which is an ongoing clinical cohort registry study that has recruited nearly 1,000 families to date.[33] The sample for the present study comprised 619 parent–youth pairs whereby the child met the sole inclusion criterion of a Diagnostic and Statistical Manual (DSM-IV)[34] diagnosis of AN (restricting type or binge-purge subtype), BN (purging or non-purging subtype) or EDNOS. Diagnoses were yielded through information from a child informant version of the EDE1 and medical review of expected BMI. All clinicians administering the EDE undergo training which includes familiarization with the original manual, updates, adaptations for children, and relevant literature on norms, reliability, and validity, observation of experienced clinicians administering the instrument, co-scoring (i.e., training clinician in the room during EDE administration) with comparison of scores with the experienced clinician, and then administration of the EDE in the presence of the experienced clinician. At least four EDEs would be jointly scored, and more if good inter-rater reliability is not obtained.

The sole exclusion criterion was a medical record that did not include the parent and child EDE records, for which 65 parent–youth pairs were excluded. T-tests and chi-square analyses comparing included with excluded participants on demographic (age, sex, intact family, rural residence) and clinical features (diagnosis, duration of illness, BMI [body mass index] z score) evaluated the potential for selection bias. The alpha level was adjusted to control for inflation of the family-wise error rate using the Bonferroni procedure. There were no significant differences indicating no apparent selection bias.

Ethics approval for the study was obtained from Murdoch University Human Research Ethics Committee (HREC) and Princess Margaret Hospital for Children HREC.


Eating disorder psychopathology was assessed via a child informant version of the EDE, adapted from the adult-based EDE,1 which has sound discriminant validity and psychometrics[35]; minor changes in wording were made for age-appropriateness only. Behavioral symptoms of objective binge eating, self-induced vomiting, laxative misuse, and diuretic misuse are assessed by frequency of episodes over the previous 28 days, and driven exercise by number of days out of the previous 28 days. Twenty-two cognitive symptoms are graded by the clinician on a six-point scale of severity or incidence, organized into subscales; restraint, eating concern, weight concern, and shape concern, which together yield a global score. The reliability and validity of the adult-based measure is well established.[1] Although the proposed global and subscale structure is generally not confirmed in factor analysis, the original scores are used widespread clinically and as outcomes in clinical trials. An additional factor analytically derived subscale “shape and weight concern” with eight items identified in Wade and colleagues[36] (dissatisfaction with weight, prescribed weighing, importance of weight, dissatisfaction with shape, importance of shape, discomfort seeing body, avoidance of exposure, and feelings of fatness) that was stable across two independent samples[36] was also included.

The child-informant version used at PMH EDP is akin but not identical to the ChEDE,[6] as program and intake clinic planning and the inception of the service predated publication.6 The version has construct validity, among the broader HOPE Project cohort with DSM-IV eating disorders and since introduction of the Eating Disorder Inventory (EDI-3)[37] into routine intake assessment (N = 40), the correlations between the EDE global score and EDI-3 eating disorder risk scales are 0.86 (p < 0.001) for drive for thinness, 0.56 (p < 0.001) for bulimic symptoms, and 0.76 (p < 0.001) for body dissatisfaction.[38] Internal consistency of the scales in the broader HOPE Project cohort with DSM-IV eating disorders (N = 663) is acceptable (global α = 0.93, restraint α = 0.73, eating concern α = 0.78, shape concern α = 0.89, weight concern α = 0.80). Diagnosis based on EDE responses followed manualized standardized procedures. A trained doctoral clinical psychology student blindly diagnosed a subset of cases in the HOPE Project cohort (n = 19) and kappa agreement was established with diagnoses (categories: AN, BN, EDNOS, no eating disorder) assigned by author H.W. Inter-rater reliability of diagnoses yielded with the EDE (and medical review data) was acceptable (kappa = 0.81; percentage of exact agreement = 89%).

Malnutrition was defined as an age-adjusted BMI z score ≤ −1.65 (5th centile), consistent with World Health Organization literature on young people and recent BMI discussion about definition of overweight and obesity (i.e., obesity 95th centile). BMI z scores were calculated using Epi Info 7.[39] Z-scores are age-adjusted based on the Centers for Disease Control and Prevention[40] growth charts.


Following referral, routine intake assessment spans two half-days. Both parents are encouraged to attend, and the assessment includes a psychosocial interview with the parents and their child, a medical review, dietetic review, and psychometric evaluation. In some cases, patients were assessed following admission to hospital as a result of physiological instability or malnutrition. During the medical review, the physician measured height and weight. Youth were weighed wearing underwear and a hospital gown, using electronic scales accurate to 50 g. Height was measured with the patient standing without shoes and socks, standing on a hard surface and using a Hardpenden stadiometer, accurate to 1 mm. With respect to EDE administration, the youth was interviewed separately from parent/s. If co-parents presented they were interviewed by the clinician collaboratively. Assessors of each parent–youth dyad were blind to the other's assessment, as child and parent EDEs occurred concurrently with different assessors.

Statistical Analysis

Inter-rater agreement was calculated by methods of Cohen's kappa,[41] weighted kappa[42] using a matrix method,[43] and the intraclass correlation coefficient (ICC), depending on the measurement level of the variable. Kappa values (k) are most informative when the negative and positive ratings are balanced, otherwise kappa is very sensitive, for instance, when one category is high, kappa may be unexpectedly low.[44] A prevalence-adjusted method of calculating kappa is available.[45, 46] Because there is disagreement on the appropriate method, we report unadjusted and prevalence-adjusted kappa (PAK) for binary variables. According to Landis and Koch, [47] kappa values of <0 indicate less than chance agreement, 0.01–0.20 slight agreement, 0.21–0.40 fair agreement, 0.41–0.60 moderate agreement, 0.61–0.80 substantial agreement, and 0.81–0.99 almost perfect agreement. The interpretation of ICC is similar to kappa. In both instances, a score of >0.7 is considered evidence of an acceptable level of inter-rater agreement. Pearson product-moment correlations (r) between youth and parent scores were computed. Percentage of exact agreement (PEA), the percentage of parent-youth dyads reporting “yes” simultaneously and “no” simultaneously, was calculated for binary variables.

Hypotheses evaluating parent and youth score differences, including models with covariates where relevant, were tested using repeated measures general linear modeling (GLM). Repeated measures analyses are appropriate when data are dependent and within-group variance is restricted. Hypotheses were a priori so no adjustment to conventional alpha (0.05) was made.


Descriptive Statistics

Descriptive statistics are displayed in Table 1. The most common diagnosis was EDNOS, typical of youth eating disorder populations,[48-50] and as shown by BMI z score, participants mostly had restricting presentations.

Table 1. Sample demographic and clinical characteristics
 Sample (N = 619)
  1. Note: BMI, body mass index.

  2. a

    n =554.

  3. b

    n =585.

  4. c

    n =615.

Age (years), M (SD)14.63 (1.65)
Sex (female), n (%)578 (93.4)
Rural residence, n (%)102 (16.5)
Intact family, n (%)a358 (64.6)
Diagnostic category, n (%) 
Anorexia nervosa153 (24.7)
Restricting127 (20.5)
Binge eating/purging26 (4.2)
Bulimia nervosa46 (7.4)
Purging42 (6.8)
Non-purging4 (0.6)
Eating disorder not otherwise specified420 (67.9)
Anorexia nervosa-like231 (37.0)
Bulimia nervosa-like20 (3.0)
Binge eating disorder3 (0.5)
Unspecified166 (27.0)
Duration of illness (months), M (SD) rangeb11 (11.36) 1–120
BMI z score, M (SD) rangec−1.42 (1.43) −6.49–2.13

Inter-rater Agreement

Estimates of inter-rater agreement are shown in the online Supplementary Table, Supporting Information. In 64% of EDE items, youth-informant mean scores were higher than parent-informant mean scores. Kappa scores for the 22 EDE items and ICCs for the EDE subscales and global scale showed unacceptable agreement for all outcomes.

ICC values for frequency of disordered eating behaviors (see online Supplementary Table, Supporting Information, for presentation of these results) were unacceptable, supporting Hyp 1, namely that parent–youth agreement on frequency of disordered eating behaviors would be poor. PEAs for presence/absence of disordered eating behaviors were diuretic misuse (99%), laxative misuse (95%), self-induced vomiting (87%), binge eating (86%), and driven exercise (74%). Given the imbalance in prevalences and the supporting information provided by PEA estimates, the PAK was interpreted for inter-rater agreement on presence of disordered eating. There was acceptable agreement for objective binge eating, self-induced vomiting, laxative misuse, and diuretic misuse, but not driven exercise, thus generally supporting Hyp 2, namely that parent–youth agreement on presence/absence of disordered eating behaviors would be acceptable (see online Supplementary Table, Supporting Information).

Inter-rater agreement for disordered eating behaviors was investigated further by splitting patients into two groups of endorsed or did not endorse the behavior. Kappa and PAK values were computed; in some instances, a cell of 0 mean kappa was not available (NA). The kappa and PEA (i.e., parent/s also report symptom presence) values for the patient group who self-reported symptoms were diuretic misuse (PAK = 0.99, k = 0.24, PEA = 99%), laxative misuse (PAK = 0.91, k = 0.56, PEA = 91%), objective binge eating (PAK = 0.87, k = NA, PEA = 52%), self-induced vomiting (PAK = 0.75, k = 0.66, PEA = 84%), and driven exercise (PAK = 0.58, k = 0.49, PEA = 74%). For the patient group who did not self-report symptoms, kappa and PEA (i.e., parent also reports symptom absence) values were diuretic misuse (PAK = 0.99, k = 0.40, PEA = 99%), laxative misuse (PAK = 0.91, k = 0.29, PEA = 95%), objective binge eating (PAK = 0.87, k = NA, PEA = 94%), self-induced vomiting (PAK = 0.75, k = 0.60, PEA = 87%), and driven exercise (PAK = 0.45, k = 0.44, PEA = 72%). The inter-rater agreement estimates did not reveal a pattern of higher parent–youth inter-rater agreement for those who self-reported the absence of symptoms compared with those who self-reported the presence of symptoms, thereby failing to support Hyp 3a.

For cognitive symptoms, results show poor inter-rater agreement (see online Supplementary Table, Supporting Information), and in contrast to the behavioral symptoms, mean symptom scores on some indices were significantly different; on global EDE [F(1, 584) = 6.37, p = 0.01], restraint [F(1, 612) = 4.30, p = 0.04], eating concern [F(1, 608) = 10.08, p = 0.002], shape concern [F(1, 604) = 9.74, p = 0.002], and factor analytically derived shape/weight concern [F(1, 579) = 33.31, p = <0.001]. In all instances, the youth informant group reported more severe symptoms than the parent informant group.

Factors of Influence

Age, diagnosis, and level of malnutrition were hypothesized to impact on inter-rater agreement for cognitive and behavioral symptoms. Testing showed that there was a significant age × informant interaction on all EDE scale scores, partially confirming Hyp 3b, which proposed that inter-rater agreement on scale scores would be higher for adolescents than children due to developmental limitations and increased reliability of reporting with age. Consistent with predictions, children reported lower cognitive symptoms relative to parents, however, in contrast to predictions, adolescents reported higher cognitive symptoms relative to parent informants (see Figs. 1 and 2). Significant interactions were found on global EDE [F(1, 583) = 24.47, p < 0.001, ηp2 = 0.04], restraint [F(1, 610) = 19.38, p < 0.001, ηp2 = 0.03], eating concern [F(1, 607) = 14.39, p < 0.001, ηp2 = 0.02], shape concern [F(1, 602) = 27.60, p < 0.001, ηp2 = 0.04], and the factor analytically derived weight/shape concern scale, [F(1, 578) = 17.00, p < 0.001, ηp2 = 0.03], but not on weight concern [F(1, 593) = 5.48, p = 0.02, ηp2 = 0.009].

Figure 1.

Children (≤12 years) and parent mean scores (and standard error) on the scales on the Eating Disorder Examination (EDE).

Figure 2.

Adolescent (13–18 years) and parent mean scores (and standard error) on the scales of the Eating Disorder Examination (EDE).

Hyp 4 proposed that there would be lower inter-rater agreement for AN and EDNOS-AN presentations compared with BN and EDNOS-BN presentations among youth, as might be expected through egosyntonicity. Diagnostic concordance was compared, restricting analyses to AN, EDNOS-AN, BN, and EDNOS-BN patients. There was a significant diagnosis × informant interaction as predicted; however, marginal means showed that the pattern of predicted observations was not apparent; thus Hyp 4 was unsupported. Specifically, there were significant interactions on global EDE [F(1, 424) = 11.13, p = 0.001, ηp2 = 0.03], restraint [F(1, 443) = 15.26, p < 0.001, ηp2 = 0.03], eating concern [F(1, 441) = 9.89, p = 0.02], shape concern [F(1, 437) = 5.68, p = 0.02, ηp2 = 0.01], but not weight concern [F(1, 431) = 3.12, p = 0.08, ηp2 = 0.007] or the factor analytically derived weight/shape concern scale [F(1, 421) = 2.60, p = 0.11 ηp2 = 0.006]. The marginal means showed good concordance between parent and youth scores in the AN group, and poor concordance between parent and youth scores in the BN group; specifically, the parents of patient with BN reported lower cognitive symptoms than their children.

Tests on the AN and EDNOS-AN group alone showed in general no significant differences between youth and parent reporting on the EDE scales, thus failing to support Hyp 5, that youth with AN or AN-like presentations would report less-elevated cognitive symptoms relative to parents. The results were global EDE [F(1, 363) = 0.24, p = 0.63, ηp2 = 0.001], restraint [F(1, 379) = 0.56, p = 0.46, ηp2 = 0.004], eating concern [F(1, 378) = 1.58, p = 0.21, ηp2 = 0.004], shape concern [F(1, 437) = 5.68, p = 0.02, ηp2 = 0.003], weight concern, F(1, 431) = 3.12, p = 0.08, ηp2 = 0.006], and the factor analytically derived weight/shape concern scale [F(1, 421) = 2.60, p = 0.11 ηp2 = 0.03].

Testing of Hyp 6 revealed that, as predicted, youth with BN or EDNOS-BN reported a significantly higher frequency of binge eating episodes than parent informants [F(1, 51) = 6.60, p = 0.01, ηp2 = 0.11]. Against predictions, there were no significant differences by informant on frequency of self-induced vomiting [F(1, 56) = 0.09, p = 0.77, ηp2 = 0.002], laxative misuse [F(1, 57) = 0.02, p = 0.88, ηp2 < 0.001], or diuretic misuse [F(1, 55) =.66, p = 0.42, ηp2 = 0.012].

Testing showed only one significant malnutrition × informant interaction and not in the predicted direction, thus failing to support Hyp 7, that malnutrition would be associated with lower reporting of psychopathology relative to parents. Specifically, malnutrition was not associated with youth reporting of lower scores on global EDE [F(1, 579) = 1.82, p = 0.18, ηp2 = 0.003], restraint [F(1, 606) = 7.27, p = 0.007, ηp2 = 0.01], eating concern [F(1, 603) = 1.25, p = 0.26, ηp2 = 0.002], shape concern [F(1, 598) = 2.24, p = 0.13, ηp2 = 0.004], weight concern, F(1, 589) = 0.16, p = 0.69, ηp2 < 0.001], or the factor analytically derived weight/shape concern scale [F(1, 574) = 0.44, p = 0.51 ηp2 = 0.001]. For restraint, youth reported higher scores than parents.


This study examined parent–youth concordance in reporting of youth eating disorder pathology as assessed by the EDE. On the whole, agreement between youth- and parent-reported eating disorder symptoms was poor to moderate.

As expected, inter-rater agreement on frequency of behavioral symptoms, including vomiting, diuretic misuse, laxative misuse, excessive exercise, and objective binge eating, was poor. This finding contradicts the general consensus that there is greater parent–youth concordance for externalizing symptoms.[11, 51] Although there is a lack of concordance research in general on sensitive externalizing behaviors, such as drug use or risky sexual behaviors. Perhaps due to the secret and private nature of eating disorder behaviors, these might be regarded as more akin to other internalized conditions which also typically yield low parent–child agreement.[52]

Consistent with predictions, acceptable agreement for the presence of behavioral symptoms was found. The most unacceptable agreement was found for driven exercise, similar to another study using a different measure,[13] and the highest level of agreement was for self-induced vomiting. The finding that binge eating was detected by parents does not corroborate some previous work which showed that parents had trouble detecting binge eating when present in nonclinical children and adolescents.[13, 15, 16] This may be because the studies cited recruited nonclinical school children, yet our study involved a treatment-seeking sample where parents were sufficiently aware of and concerned about their children's eating attitudes and habits. Parents may have received or sought out education in the course of obtaining a health professional referral which is a common referral pathway to the program. The acceptable concordance on presence of binge eating may also reflect the timing of the administration of the EDE during the assessment process, in that clinical experience suggests that children frequently disclose the presence, if not the details of eating disorder behaviors to parents in the context of the psychosocial or medical interviews that are carried out prior to EDEs. Nevertheless, binge eating severity (frequency) presents a challenge for assessment, as it is typically a secret event and shameful, which may lead to underestimates in parent and possibly youth reports. Where binge eating is accompanied with a sense of detachment, young people may experience difficulty recalling the details of binge episodes impacting the validity of their reports.[53] Clinically, both informant reports on eating disorder behaviors generally are meaningful but for different reasons; the youth report is arguably more informative for measuring behaviors, whereas the parent report can be a clinically useful tool for the assessing clinician to judge needed areas for parent education.

When patients were classified into child and adolescent groups, children reported lower cognitive symptoms relative to their parents, and by contrast, adolescents reported greater symptoms. For the child group, it is possible that level of cognitive development or intrapsychic awareness precluded insight to eating disorder cognitions, leading to lower endorsements of eating disorder cognitions than their parents or that adolescents and parents are more likely to employ cognitive explanatory models of symptoms. Alternatively, some EDNOS unspecified cases may have had no actual psychic disturbance of weight and shape and may reflect presentations of the proposed DSM-5 eating disorder category of avoidant/restrictive food intake disorder. The finding that adolescents endorsed more cognitive symptoms than their parents is interesting, as while adolescents are generally more cognitively sophisticated than younger children, they have as a group been described as particularly prone to denial and minimization of their condition.[25] Adolescent's relationships with their parents undergo substantial changes during the pubertal period.[54] Time interacting, emotional intimacy, and defaulting to parents in decision-making decrease during this time[55] with adolescent friendships becoming closer, more intimate, and more disclosing.[53] Parents may consequently have less awareness and opportunity to observe their adolescent's problem. Thomson and colleagues[56] emphasized the role of parental avoidance and denial in the recognition of eating disorders, suggesting that while parents are able to detect signs of an eating problem they may be reluctant to label it as an eating disorder as acknowledgement of the problem may make it seem “real” and “scary”.

An unexpected finding was the good concordance between parent and youth scores for AN presentations, and no support for the idea that youth with AN presentations would identify less-elevated psychopathology as was the case in Couturier's[7] research and might be anticipated via egosyntonicity. This finding may be due to the context of EDE assessment in the study's clinical setting, in that considerable rapport is established by the time the EDE is administered, or alternatively because of the tertiary treatment setting, it may reflect processes that occur prior to assessment such as participation in primary care and other treatments. We had expected that there would be lower inter-rater agreement in the group of malnourished youth, on the basis of potential egosyntonicity, but this was not the case. Clinically our impression is that the majority of very unwell youth attending our service are aware that they are unwell and feel safe enough by administration of the EDE to report symptoms, but that a subgroup exists that endorse few symptoms, along with their parents. We propose exploring these in a later study, given that initial assessment can be misleading for these youth, including those within the wider study population pool that might not have received an eating disorder diagnosis, who go on to have a significant period of treatment.

There was poor concordance between parent–youth reports among youth with BN presentations. Parents reported lower eating disorder cognitions, replicating the finding of Couturier et al.[7] and supporting their suggestion that youth with BN are more reliable informants of their symptoms. Youth with BN presentations were anticipated to identify more elevated binge eating and purging relative to parents. They reported a significantly higher frequency of binge eating episodes; however, no significant differences were found between informants for vomiting, laxative misuse, and diuretic misuse.

Future Research

There were no mediating or moderating variables measured in this study. Future work might identify variables that predict increased parent–youth agreement. For example, children and parents from closer, more unified families may yield higher agreement compared with children and parents from more combative family systems. Potentially youth and families who have had previous treatment experiences may also yield higher agreement. Likewise, constraints on parent–child interactions (i.e., work schedule) and the cohabitation status of the parent may be relevant. As some connections between maternal depression or psychopathology and the ratings of behavioral concerns in children have been suggested,[57] family history of eating disorders and parental psychopathology may be useful to assess. Further development of normative data is required to increase the developmental sensitivity and utility of the EDE in youth. This would be particularly useful for adolescent males, who are a growing clinical population diagnosed with eating disorders.[58] The impact of procedural and contextual factors on disclosure and concordance could be investigated, for instance, the setting and timing of EDE administration.

Clinical Implications

The present results continue to support the utility of multi-informant assessment for youth with eating disorders, particularly children. The poor to moderate concordance between parents and their children raises the issue of how to integrate findings to enhance clinical decision-making in practice. Because DSM-IV prioritizes the youth informant in assessment and diagnosis of eating disorders, historically the youth EDE would be prioritized for diagnosis, but proposed DSM-5 goes wider to include parent report also. For clinical purposes, we recommend that clinicians take both EDEs into account equally, as well as other information and formulate an understanding of the differences, which may include developmental stage, motivational stage, communication patterns in the family, and then use this to formulate working diagnosis and treatment plan. If instrument choice is warranted for economic or other needs, then our informed view would be to prioritize the parent-informant EDE for youth aged 10 years and younger, and child-informant EDE otherwise. In our clinical service, despite resource constraints, our decision so far has been to retain a multi-informant assessment, due to our anecdotal impression of the clinical utility such as increasing the quality of our understanding, the value for engaging both child and parent early in the treatment process, and the important implication of increasing a parent's understanding of eating disorder symptoms through their participation in the EDE.

Anecdotally, we have found utility in parent education and skills training programs to help parents become informed about the symptoms of eating disorders and feel empowered. The public should be provided with easy access to mental health literacy material, especially that which decrease stigma and increase understanding of psychological issues, so that parents in the community can substantiate their intuition about a suspected problem; otherwise, they may resist taking necessary action and mistake potential symptoms of eating disorders as pressures of adolescence.[56] A recent Australian initiative, the website of the National Eating Disorders Collaboration, is one example that aims to improve community access to trustworthy information about eating disorders; visible community-based organizations are another important public resource.


Landis and Koch's method for interpreting kappa is widely used, although subsequent researchers have proposed alternate benchmarks.[43] Ultimately, the choice of benchmarks is arbitrary but can lead to differences in the interpretation of findings.[43] Multiple clinicians collected data, and intraclustering of ratings may occur, which could not be evaluated in the present research due to the degree of missing EDE administrator codes, but would be worthwhile investigating generally. The study population comprises more serious clinical cases referred to a specialist tertiary hospital-based setting, with diversion of mild to moderate cases at triage to community settings, and so may have limits to generalizablity.


Results support the utility of parent–youth assessment via the EDE to obtain a wider clinical picture of eating disorder psychopathology in children and adolescents, particularly for younger children. With the DSM-5 changing criterion to include behavioral observations by significant others, the potential of informants such as parents, to influence diagnosis and clinical formulation widens, and the importance of understanding how best to interpret and integrate findings from various assessment methods increases.

Earn CE credit for this article!

Visit: for additional information. There may be a delay in the posting of the article, so continue to check back and look for the section on Eating Disorders. Additional information about the programis available at