Consent to treatment in adolescents with anorexia nervosa


  • Sheri L. Turrell PhD,

    Corresponding author
    1. Department of Human Development and Applied Psychology, Ontario Institute for Studies in Education, University of Toronto, Toronto, Ontario, Canada
    • Department of Human Development and Applied Psychology, Ontario Institute for Studies in Education, University of Toronto, Toronto, Ontario, Canada
    Search for more papers by this author
  • Michele Peterson-Badali PhD,

    1. Department of Human Development and Applied Psychology, Ontario Institute for Studies in Education, University of Toronto, Toronto, Ontario, Canada
    Search for more papers by this author
  • Debra K. Katzman MD, FRCP(C)

    1. Division of Adolescent Medicine, The Hospital for Sick Children, University of Toronto, Toronto, Ontario, Canada
    2. Department of Paediatrics, University of Toronto, Toronto, Ontario, Canada
    Search for more papers by this author



This study examined the ability of adolescents with anorexia nervosa (AN) to make treatment decisions.


The MacArthur Competence Assessment Tool-Treatment (MacCAT-T) was used to compare the decision making abilities of 35 adolescents with AN who were receiving inpatient treatment with that of 40 healthy, community-based adolescents. Vignettes of both a medical and psychiatric illness were provided, requiring participants to work through the process of making a hypothetical treatment decision. The MacCAT-T was also administered to participants with AN to examine decision-making about their own illness, which allowed for comparison of competencies across contexts.


Group differences were found, with the community group showing superior reasoning skills to the adolescents with AN.


The results provide evidence to suggest that adolescents with AN tend toward a thinking disposition that is concrete and lacking in abstract reasoning and reflection, which may negatively affect their ability to reason about treatment options. © 2010 by Wiley Periodicals, Inc. Int J Eat Disord 2010


In North America, patients can make their own decisions regarding medical treatment if they are considered capable to do so. The four criteria that define capacity to provide informed consent1, 2 include understanding the information provided regarding diagnosis, treatment options, and their risks and benefits; rational reasoning with the information provided; appreciating the diagnosis and the foreseeable consequences of a treatment decision in one's personal context; and the ability to make a treatment choice from available options.

Based upon the assumption that adolescents lack the necessary cognitive maturity to make their own decisions about treatment,3 parents have long acted as proxy decision makers. Existing evidence is now available to challenge this assumption. Research based on hypothetical vignettes2 and in vivo situations4–6 has generally yielded similar results: adolescents as young as 14 years of age are as able to make treatment decisions as adults.2, 4, 7 However, findings must be interpreted with caution due to limited ecological validity,2 the inclusion of outdated and paternalistic (e.g., reasonableness of choice) indices of capacity,2 failure to establish reliability of assessment tools,7 and inadequate description of the clinical status of participants in in vivo studies.7

Results from research with adults suggest that individuals who suffer from mental illness, although a heterogeneous group, demonstrate overall deficits in the ability to understand and reason about information that is relevant to their illness and treatment.1, 8, 9 However, to date, there have been no empirical comparisons of the ability of adolescents with and without mental illness to understand, appreciate, reason, and make treatment choices. Although the issue of capacity is a complex one for adolescents, it is even more daunting when the adolescent has anorexia nervosa (AN). Little is known about the long-term sequelae of the medical complications of adolescent onset AN, but timely treatment is believed essential to reduce the impact of long-term harm.10 Weight restoration is at the core of treatment designed to stabilize medically compromised adolescents with AN. However, intervention can be extremely problematic due to a combination of the tenacity and egosyntonic nature of the disorder and patient denial of illness, often resulting in reluctance to participate in recommended treatments.11

Using hypothetical vignettes describing medical and psychiatric illnesses requiring treatment decisions, we examined whether participants hospitalized with AN differed from a group of healthy participants in their capacity to consent to treatment. Although we examined all four components of competence typically used to assess capacity to consent to treatment—understanding, reasoning, making a choice (and for those with AN, appreciation)—we found no group differences in the ability to understand information or make a treatment choice; therefore, only the reasoning component of consent is reported in this paper. The use of hypothetical vignettes allows for experimental control over content and thus facilitates group comparisons but this methodology lacks ecological validity. To address this limitation, we also examined the ability of participants with AN to reason with regard to their own illness and compared this to their performance on the hypothetical vignettes. We predicted that participants with AN would have greater difficulty reasoning about their own illness compared to hypothetical illnesses because they may perceive their illness as being an integral part of their identity.12



Participants were 35 adolescents with AN, all of whom were inpatients in a tertiary care, university-affiliated hospital, and 40 healthy adolescents from the community. The inclusion criteria for participants with AN were as follows: female, age less than or equal to16 years 11 months, fluent in English, and a primary diagnosis of AN, either restricting or binge-purge subtype according to DSM-IV criteria.13 Diagnoses were made by a multidisciplinary team using standardized and structured clinical interviews and self-report questionnaires. The exclusion criteria were as follows: male, primary diagnosis other than AN (e.g., bulimia nervosa), administration of psychiatric medications, administration of steroids (which have been reported to affect cognitive functioning),14 a history of premorbid learning disability, head trauma with loss of consciousness greater than 30 min or significant neurological sequelae, and chronic illness unrelated to AN (e.g., systemic lupus erythematosus, irritable bowel syndrome, and diabetes).

Healthy participants were recruited via community postings. Inclusion criteria were the same as those for adolescents with AN, with the exception of a diagnosis of AN. The same exclusion criteria applied, with the addition of symptoms of an eating disorder, clinical levels of internalizing or externalizing symptoms (e.g., greater than 1 standard deviation above the mean on standardized questionnaires), or intellectual ability standard score below 85.

Procedure and Measures

Parents of potential participants were approached first and asked to give permission for the first author to approach their adolescent. Consent was obtained from adolescents 16 years and older and parental consent and participant assent was obtained for adolescents under 16. Of those who were approached to participate, all consented to do so.

All participants completed the same measures, outlined below. With the exception of the treatment decision-making interview, instruments were used as screening measures to ensure that the participants with AN were not compromised in terms of intellectual ability or memory skills, and that the healthy participants were not suffering from a DSM-IV Axis 1 disorder.

Body Mass Index (BMI)

Although not a definitive marker for healthy weight in adolescents, BMI was used as a broad measure of health. For the community group, participants' height and weight were measured at the time of interview. For adolescents with AN, weight on the day of the interview was taken from medical charts. Age-specific BMI was calculated using the website and a BMI of less than 17.5 kg/m2warranted further investigation.

Eating Disorder Symptoms

The 26-item Eating Attitudes Test (EAT-26)15 is a self-report questionnaire that asks respondents to rate the presence/severity of symptoms of disordered eating. Garner et al.15 reported internal consistency of the EAT-26 to be high (α = .90) as was criterion-related validity, in that the EAT-26 reliably predicted group membership divided as either healthy or having AN. Scores above 20 were considered suggestive of an eating disorder, and precluded inclusion for community participants.

Internalizing Symptoms

The Youth Self Report16 (YSR) is a self-report questionnaire for 11- to 18-year-olds covering a broad range of symptoms including both internalizing (e.g., depression and withdrawal) and externalizing (e.g., antisocial and oppositional) problems. It has good test–retest reliability (r = .82) as well as content and criterion-related validity.16 Only the internalizing scale was used in this study.

Intellectual Ability

The Wechsler Abbreviated Scale of Intelligence (WASI)17 was used to determine intellectual ability at the time of decision-making. It has demonstrated high internal consistency (α = .95–.97) and correlates at .87 with the full scale IQ of the WISC-III.18

Memory and Attention

The Immediate Verbal Memory subscale of the Children's Memory Scale19 (CMS) was administered to obtain a measure of memory for verbally presented material. To address the possibility of attention problems in patients with AN that may impact treatment decision-making, the Attention/Concentration subscale from the CMS was also administered. The CMS has demonstrated good psychometric properties overall.19

Treatment Decision-Making

The MacArthur Competence Assessment Tool-Treatment (MacCAT-T)20 is a semistructured interview that assesses the four components of capacity to consent to treatment in the context of real and/or hypothetical illnesses. It has demonstrated good interrater reliability (intraclass correlations = .87–.99) as well as construct and content validity.8

All participants were presented with two hypothetical vignettes (available upon request) depicting individuals with diagnosed illnesses: scoliosis (a medical illness) and depression (a psychiatric illness). In accordance with the MacCAT-T's standardized procedures, for each illness the interview began with the disclosure of information relevant to understanding the disorder and proceeded through two treatment options, their associated risks and benefits, and ended with participants reasoning about which treatment choice they would prefer. Information was provided in chunks, or elements (e.g., understanding the illness; understanding the treatment options), with participants repeating the information in their own words after each element was presented to ensure basic understanding prior to reasoning with the information. Responses for each of the four elements of reasoning were awarded a score of 0, 1, or 2, with higher scores corresponding to better responses. Summary scores were created with the overall reasoning score ranging from 0 to 8.

In addition to responding to the two vignettes, participants with AN completed the MacCAT-T in relation to information relevant to their own admission for AN, for which appreciation was also assessed. Although it was made clear to participants with AN that “no treatment” was not an actual option in their case, this was used as a second treatment option in the context of the MacCAT-T. Because there is some evidence to suggest that certain educational endeavors may unintentionally increase individuals' knowledge of, and participation in, eating disordered behavior,21 we did not administer a vignette depicting AN to the community group.

The order of presentation of the vignettes/MacCAT-Ts was counterbalanced within each group to ensure that group differences were not a result of the order in which the illnesses were presented. To assess inter-rater reliability, 20% of the interviews were audio taped.


The results section begins with descriptive analyses and between-group comparisons on demographic, cognitive, and screening measures. Primary analyses of group differences in decision-making consisted of repeated measures multivariate analysis of variance (MANOVA), with the subcomponents of reasoning serving as repeated measures.

Demographic and Screening Information

The average age of participants with AN was 14.9 years (SD = 1.44) and 14.4 years (SD = 1.25) for healthy participants. The majority of participants with AN and from the community were in high school (71% and 70%), from high occupational socioeconomic status (SES) families (e.g., professionals, high-level management, middle management) (80% and 88%),22 and self-identified their ethnicity as white/European (80% and 85%).

Participants with AN had significantly lower BMI and higher EAT-26 scores than those from the community. One participant from the community group was removed from the study due to symptoms of an eating disorder identified through this screening. YSR scores did not indicate significant symptamatology in the healthy participants. All participants achieved at least average scores with respect to IQ, verbal memory, and attention.

Decision-Making: MacCAT-T

Inter-rater reliability was conducted on the individual reasoning components of the MacCAT-T, using scores for each vignette from 10 participants in each group. κ Coefficients were .93 for consequential reasoning, .97 for comparative reasoning, .95 for generating consequences, and .93 for logical consistency.

Group Differences in Reasoning

A MANOVA was conducted to determine the effect of group membership on reasoning as a function of the context (scoliosis or depression) of the vignette. Each component of reasoning (consequential, comparative, generating consequences, and logical consistency) was entered into the MANOVA as a separate measure. Results revealed a significant main effect of group, F(4,70) = 5.42, p < .002, and of context, F(4,70) = 6.01, p < .001. The interaction between group and context was not significant, F(4,70) = .50, p = .736.

The associated follow-up univariate tests (ANOVA) indicated that the healthy participants outperformed the participants with AN group on every component but comparative reasoning (Table 1). The narrow range of the scale used in measuring reasoning should be taken into consideration when interpreting the effect sizes indicated in Table 1. ANOVAs conducted to follow up the context effect indicated that both groups obtained significantly higher consequential reasoning scores and overall reasoning scores for the scoliosis vignette compared with the depression vignette.

Table 1. Group differences in MacCAT-T reasoning ability
 Participants with ANHealthy ParticipantsGroup DifferencesEffect Size
Component of ReasoningMSEMSEF(1,73)Partial n2
  • *

    p < .05.

  • **

    p < .001.

Generating consequences.**.18
Logical consistency1.*.05
Overall reasoning4.**.18

Ability of AN Participants to Reason about their Own Illness versus Hypothetical Vignettes

A within-subject MANOVA was used to examine the ability of participants with AN to engage in reasoning as a function of context, with each component of reasoning used as a measure in the analysis. A significant multivariate effect of context, F(8,27) = 14.7, p < .001 was followed up by univariate tests (ANOVA), which indicated a significant effect of context on comparative reasoning. Follow-up t-tests indicated a significant advantage for comparative reasoning about scoliosis compared with AN, t(34) = −6.39, p < .001, and for comparative reasoning about depression compared with AN, t(34) = −5.31, p < .001 (Table2). There was no significant difference in reasoning about scoliosis versus depression, t(34) = .25, p = .81. A separate ANOVA was conducted to examine the effects of context on the overall reasoning score, F(2,33) = 11.39, p < .001. Here, participants performed better when reasoning about scoliosis compared to both AN, t(34) = −4.83, p < .001, and depression, t(34) = −2.53, p < .05.

Table 2. Effects of context on MacCAT-T components of reasoning for participants with AN
 Scoliosis VignetteDepression VignetteAnorexia NervosaContext DifferencesEffect Size
Component of ReasoningMSDMSDMSDF(2,68)Partial n2
  • Notes: Significance indicators (*) following superscript letters denote significance level of post hoc t-tests (*p < .001 and **p < .05).

  • a

    Significant difference between scoliosis and anorexia nervosa.

  • b

    Significant difference between depression and anorexia nervosa.

  • c

    Significant difference between depression and scoliosis.

Generating consequences.
Logical consistency1.80.531.71.621.57.611.42.04
Overall reasoning4.831.424.291.643.571.3111.39*a*c**.41


The aim of this study was to empirically examine the ability of adolescents with AN to make treatment decisions. Participants with AN were less able than community participants to engage in consequential reasoning, less apt to generate everyday consequences of their decisions, and less able to make logically consistent decisions with regard to the vignettes. In addition, they had greater difficulty with comparative reasoning and overall reasoning when it involved their own illness as opposed to hypothetical illnesses. The fact that many sufferers perceive AN as being an integral component of their identity contributes to the egosyntonic nature of the illness and may make it difficult for them to compare and contrast the pros and cons of accepting versus refusing treatment.

Despite a relatively small sample size, the pattern of results brings into question the overall reasoning skills in adolescents with AN, who at least in this sample, appeared to struggle across several treatment vignettes. This suggests that adolescents with AN may require support in the form of information and direct teaching of problem-solving skills in order to identify potential consequences of an option and to compare and contrast available options.

Limitations of our study should be noted. The illness status of participants was known to the examiner, which raises the possibility of bias in the administration and/or scoring of the measures. To protect against threat of bias, vignettes were read to both the healthy participants and participants with AN in order to standardize the information that was provided. Prompts for additional information were also standardized. Furthermore, MacCAT-T's were not scored until all of the interviews were complete and interrater reliability had been established.

Within the current sample of participants with AN, the ability to make treatment decisions proved to be a heterogeneous skill and supports the idea that patients can not be assumed to lack capacity as a function of diagnosis. It may be of benefit to enhance the knowledge base and assessment skills of professionals working with this population with respect to their potential decision-making skills, and in particular, the variety of ways in which reasoning ability can be elucidated.

The results of this study suggest directions for future research. The identification of clinical (e.g., restricting AN vs. AN-binge/purge) and demographic (e.g., number of admissions) factors that relate to indicators of capacity, as well as a better understanding of the impact of hospitalization on reasoning, may alert clinicians to individuals who require a more comprehensive assessment. In addition, studies with larger samples are necessary to expand on the current findings in order that we may truly act in the best interests of patients who suffer from AN and uphold their rights to informed consent regarding treatment.

The authors extend their sincere appreciation to the participants and their families for their time and energy. The authors also thank Drs. Judy Wiener and Paul Appelbaum for their contributions and support.