The Clinician Administered Staging Instrument for Anorexia Nervosa: Development and Psychometric Properties

Abstract

Objective:

To develop and evaluate an instrument to assess severity in anorexia nervosa (AN), the Clinician Administered Staging Instrument for Anorexia Nervosa (CASIAN).

Method:

Candidate items for the CASIAN were developed in three phases (domain, content, and item generation) followed by a pilot study. The psychometric properties of the resultant 34-item questionnaire were investigated in cross-sectional and longitudinal samples (N = 171) with DSM-IV AN and subthreshold AN.

Results:

Item and factor analysis procedures resulted in a refined 23-item CASIAN comprising of six factors (“Motivation,” “Weight,” “Illness Duration,” “Obsessionality,” “Bulimic Behaviors,” and “Acute Issues”). The CASIAN had high internal consistency (.811), test–retest (.957), and interrater reliability (.973). Preliminary support for the convergent, discriminant, concurrent, and predictive validity of the CASIAN was found.

Discussion:

The CASIAN is a psychometrically sound instrument. Further studies are needed to confirm the factor structure and assess its clinical and research utility.

Introduction

Anorexia nervosa (AN) has been proposed to exist on a continuum of severity, and even the earliest descriptions of AN suggested stages of severity occur within the illness. Lasègue,¹ for example, described a gradual descent into the illness and distinguished between three distinct “phases” in its progression. Modern day diagnostic practices are predicated on the construct of severity and draw an arbitrary distinction somewhere along its continuum to divide illness from nonillness. Treatment regimens are often informed by and designed for differing levels of severity. Hence a definition of severity, a method for conceptualizing the continuum and a tool to assess it in its entirety, becomes an important endeavor in any disease, including AN.

There are numerous references to levels of illness severity in the eating disorders literature.^2–9 Indeed, it is widely accepted that the illness spectrum is larger than those cases eligible for full diagnosis according to DSM-IV.^10–15 Epidemiological studies also acknowledge this and include prevalence and incidence rates of both full-syndrome AN and partial AN.^{11,13,16–19} These studies define partial AN differently: some use an absence of amenorrhoea to designate a partial case,¹⁹ others define a subthreshold case as meeting all but the amenorrhoea or weight criteria^11,18 and still others include cases that meet the low-weight criteria plus at least one other of the three remaining DSM-IV diagnostic criteria.^13,16,17

Another approach to understanding the spectrum has been to examine the difference between full and partial cases. A recent latent class analysis attempted to identify clusters of illness within EDNOS and concluded that most cases resembled the existing specified categories of AN, BN, and BED.²⁰ In short, most EDNOS cases can be conceptualized as existing on a continuum with full-syndrome cases. In respect to AN, the majority of studies have demonstrated little difference on the axes assessed between full and partial syndrome.^21–24 A recent meta-analytic review concluded that full-syndrome AN represents the severe end of a continuum, with EDNOS “lying” closer to AN than the current diagnostic conceptualization suggests.²⁴

Many instruments have been designed to assess particular features of AN, several for the purposes of screening and diagnosis, and others to assess treatment outcome.²⁵ Because of the absence of a reliable and validated instrument to assess severity in AN, it has been common for these instruments, either singly or as part of a battery, to be used in research as proxy measures of severity.^26,27 Although these proxy measures are likely related to components of severity, these instruments are typically not validated as measures of severity. To our knowledge, only one (partly) validated instrument exists that explicitly attempts to assess overall illness severity. The Short Evaluation of Eating Disorders²⁸ is a relatively recently developed self-report questionnaire that assesses three of the four diagnostic criteria for AN rated 0–3 (0 = not present, 1 = mild, 2 = meets diagnosis, and 3 = extreme) with the most important symptom (i.e., weight loss) counted double. It derives a total severity index for anorexic symptoms and enables measurement of the diagnostic criteria for AN. However, as a brief instrument, it does not allow for a comprehensive evaluation of all of the clinical features purported to contribute to severity and relies heavily on the criteria that currently comprise DSM-IV diagnosis of AN. Studies of its reliability and discriminant validity are yet to be published.

In summary, despite what appears to be wide-spread assumptions about the importance of severity within the AN, there remains uncertainty about precise definitions and components involved in severity. Only limited attempts to develop empirically supported measures of this construct have been made. The purpose of the current study is to develop and test the psychometric properties of a clinician-administered multidimensional measure to assess severity. Here, we outline its development and describe a series of preliminary validation studies.

Method

Item Generation and Pilot Testing

Generation of items for the Clinician Administered Staging Instrument for Anorexia Nervosa (CASIAN) was conducted in three primary phases: domain generation, content generation, and item generation.

Domain generation: Five doctoral level eating disorder experts (DLG, HL, ST, LS, and PB) identified areas of illness symptomatology in AN believed to be pertinent to illness severity. The experts were specifically asked to consider the full spectrum of psychological, behavioral, and physical symptomatology of AN from mild through to severe. Once broad domains were generated, they were aggregated to remove redundancies and overlapping domains. This step produced seven general domain areas: weight/weight history, chronicity, dietary control, compensatory behaviors, psychological status, physical status, and egosyntonic features.

Content generation: Four experts (DLG, HL, ST, and LS) as well as the first author generated relevant symptoms to be assessed within each domain. Next, items designed to tap each of these content areas were generated. A broad range of instruments were reviewed for potentially applicable items to be adapted, including the Eating Disorder Inventory-3,²⁹ the Eating Attitudes Test,³⁰ the Morgan-Russell Outcome Assessment Schedule,^31,32 Rating of Eating Disorder Severity (REDS; Goldner, E. Rating of Eating Dsiorder Severity (REDS), Personal Communication), Anorexia Nervosa Stages of Change Questionnaire,³³ and the Eating Disorder Belief Questionnaire.³⁴ A final review by all five experts led to some items being eliminated, others revised, and some additions.

This initial version of the CASIAN was then reviewed by another independent expert panel of five clinicians to verify that the illness areas had been adequately addressed. Each member of this panel had a minimum of 10 years experience treating eating disorders. The panel included two dietitians, two nurses, and one general practitioner. As a result of this process, the CASIAN was revised with certain items eliminated, others added, and others reworded or redesigned.

As is a suggested practice,³⁵ this second draft of the CASIAN was then subjected to pilot testing. Sixteen pilot participants (females aged 15–44 years) with AN or EDNOS-AN subtype attending an eating disorders program in Sydney, Australia, were interviewed individually. The interview comprised the pilot version of the CASIAN followed by open-ended questions centering on patient experience of the questionnaire (e.g. “Can you gauge what the questionnaire is trying to achieve?,” “Do you think the range of questions asked address the severity of your illness?”). Interviews were digitally recorded.

Results of the pilot testing were analyzed by two of the four experts from the original panel. CASIAN items were further added, deleted, or modified through this process. This final version was distributed to the whole panel for final comment. This resulted in a 34-item clinician administered questionnaire assessing a broad range of AN symptomatology across seven illness domains of weight history, duration of illness, eating behaviors, compensatory behaviors, psychological status, physical status, and illness entrapment (full CASIAN available on request). As part of item generation and pilot testing, it was decided that current weight was the most important factor in the staging model and hence is given double weighting in item scores, a procedure followed by others.²⁸

Participant Recruitment

To obtain a sample with a broad range of illness severity, participants were recruited from five recognized specialist eating disorder treatment services and affiliated community service providers in three countries (Royal Prince Alfred Hospital, Wesley Private Hospital, and the University of Sydney Counseling Service, all Sydney, Australia; The University of Chicago, Chicago, USA; St Georges Hospital, London, UK; Leeds Eating Disorder Service, Leeds, UK).

DSM-IV diagnostic criteria were used to identify individuals, aged 16 years or above, with AN eligible for participation. To capture the full spectrum of illness severity, including those persons in partial recovery or in the early stages of illness not yet meeting full criteria, individuals with EDNOS were also included. Ricca et al.¹⁴ adjusted DSM-IV criteria for EDNOS were used to determine eligibility, that is, meeting all criteria for AN except criterion D [EDNOS-AN(m)] and except criterion A [EDNOS-AN(w)]. All participants were diagnosed by the primary clinicians at each site following routine interview and assessment. The mean age of the total sample (n = 171, 98% female) was 24.4 years (SD = 8.1; range = 16–58), with a mean body mass index (BMI = kg/m²) at baseline of 16.5 (SD = 2.3; range = 9.5–23.6). Mean illness duration was 7.9 years (SD = 7.6; range = 0–38), with 43.3% meeting full-criteria for AN, and 56.7% EDNOS-AN. Of the total sample, 23.4% did not meet criterion D and 33.3% did not meet criterion A. Half the sample (50.9%) were restricting AN and 49.1% were binge/purging type.³⁶ Information was not collected on the number or nature of persons who declined to participate in the study.

The psychometric properties of the CASIAN were evaluated in a series of four studies (see below) involving different subgroups of this sample.

Procedure

Participants were initially assessed within 2 weeks of being admitted to an intensive treatment program (inpatient, day-treatment, or outpatient family-based treatment) for their eating disorder, or at any time during usual care outpatient treatment. All interviewers were honors graduates in psychology or above and employed at the data collection site. All interviewers were trained in administration of the CASIAN by the first author (SM). Training involved a half-day workshop, followed by an initial trial administration of the CASIAN by each interviewer. A video-taped sample CASIAN interview was permanently available at each site and was watched by each interviewer following their first trial administration. The first test administrations of the CASIAN by each interviewer were either observed by SM and cross-scored or were taped and cross-scored. As the relevant period of examination for the CASIAN is the month before assessment, a maximum period of 2 weeks of intensive treatment within the last four was chosen to ensure that an assessment of the eating disorder was achieved without coercion.

Other Measures

Other measures and procedures were administered to particular cohorts of the total recruited sample, dependent on the validation study undertaken (see below).

Eating Attitudes Test (EAT^30,37–40;) is a 40-item self-report measure developed to evaluate a range of eating behaviors associated with AN.

Mizes Anorectic Cognitions Scale—Revised (MAC-R⁴¹) is a 24-item self-administered questionnaire examining eating disorder cognitions. The MAC-R has good psychometric properties and can discriminate between persons diagnosed with an AN-like illness and bulimia nervosa.

Morgan-Russell [Hayward] Outcome Assessment Schedule (MROAS^29,30) is a structured interview concerned with clinical factors central to the syndrome of AN. We used a modified version,⁴² because additional items provide a measure of binge/purge behaviors. The MROAS is the most widely used outcome assessment instrument in AN, and its graded outcome categories are often used as a proxy measure of severity in AN.⁷

Marlowe-Crowne Social Desirability Scale (MCS⁴³) is a 33-item instrument assessing the tendency to portray oneself in a favorable light. It has good divergent validity with measures of anxious and depressive psychopathology,⁹ suggesting that it may be an appropriate candidate against which to test the divergent validity of a measure of psychopathology in AN.

Patient Health Questionnaire (PHQ⁴⁵) is a self-report instrument that can reliably diagnose mental disorders among primary care patients.⁴⁴ Because of the already significant time burden placed on participants in this study, the PHQ was chosen to assess co-morbid diagnoses. An adolescent version, the PHQ-A, has been validated for use with individuals between 13 and 18 years of age,⁴⁵ and this was used for participants below the age of 18.

Obsessive Compulsive Inventory-Revised (OCI-R⁴⁶) is an 18-item self-administered questionnaire based on the earlier 84-item OCI⁴⁷ assessing both obsessions and compulsions commonly present in Obsessive Compulsive Disorder (OCD). The instrument has good convergent validity and discriminates OCD from other anxiety disorders.^48,49 Foa et al.⁴⁶ demonstrated good internal consistency and test–retest reliability for the OCI-R in clinical groups. A specific measure of this comorbid condition was included given high rates of OCD in AN.^50–52

Studies 1–4

Study 1.

(Longitudinal; n = 103). Participants in this cohort were young (M = 25.34 years; SD = 8.63), with a mean BMI at baseline of 16.08 (SD = 2.32). Mean illness duration was 8.69 years (SD = 7.95; range = 0–38), with 46.6% meeting full-criteria for AN, and 53.4% EDNOS-AN. Just over half (51.5%) were classified as restricting AN, with the remainder classified as binge/purging type. After being administered the CASIAN, weight, height, and medical status were confirmed using medical files or the primary attending physician. Participants completed a form assessing their treatment history and the following measures: MROAS, EAT, MAC-R, PHQ, and OCI-R. Their primary clinician was asked to rate their level of engagement with treatment. These measures were used to assess the convergent and concurrent validity of the CASIAN. These longitudinal participants were then reassessed 3 and 6 months later as part of the analysis of the CASIAN’s predictive validity.

Study 2.

(Cross-sectional; n = 68). Participants in this cohort were slightly younger (M = 22.9 years, SD = 6.9), with a mean BMI at baseline of 17.0 (SD = 2.5). Mean illness duration was 6.9 years (SD = 6.8. Just over a third (38.2%) met full-criteria for AN, with the remainder meeting EDNOS-AN criteria. Exactly half of the sample was classified as having restricting AN. As with Study 1, after being administered the CASIAN, weight, height, and medical status were confirmed using medical files, or the Primary Care Physician. The cohort then completed the MCS, a rating of socioeconomic status and a Subject Rating of Illness Severity. The participants Primary Clinician were asked to complete a Clinician Rating of Illness Severity. Data from this cohort were used to assess discriminant and concurrent validity.

Study 3.

Thirty consecutive consenting participants from Study 1 also had their Time 1 CASIAN interviews recorded on a digital voice recorder. The average age of participants in this cohort was 24.1 years (SD = 6.29), with a mean BMI of 17.06 (SD = 2.02) with an average duration of illness of 8.4 years (SD = 6.39; range 1–27). These taped interviews were reviewed and independently rated on the CASIAN by two health professionals; an acknowledged expert and a novice. Rater 1 was a PhD level clinical psychologist with over 20 years experience in the treatment of eating disorders, and rater 2 was a first year doctoral student in clinical psychology with no experience in the assessment or treatment of eating disorders. These data were used to assess interrater reliability.

Study 4.

Twenty-four consecutive consenting participants in Study 2 were readministered the CASIAN 1 week after initial assessment. The average age of this cohort was 22.79 years (SD = 6.29), average BMI 17.39 (SD = 2.3), and average duration of illness 6.69 years (SD = 5.84; range, 1–27). A short time period was selected to assess the stability because it was expected that constructs integral to the instrument (e.g., weight and food intake) could be expected to vary over the course of treatment. These data were used to assess test–retest reliability.

Results

Item Analysis and Selection

Item Analysis Procedures.

The item analyses were conducted on the entire pooled sample (n = 171). To determine the utility of each of the 34 items, a number of formal and staged data analyses steps were completed. First, frequencies and descriptive statistics were generated for all items to identify and remove those items with high rates of missing data or no variance. No items were removed as a result of this process. Next, correlations between CASIAN items and total scores on three other eating disorder psychopathology questionnaires (MROAS [Average and General Outcome Score], EAT, and MAC-R) were examined to reveal any items failing to show relationships with these measures. Next, corrected item-to-total correlations were analyzed to remove items that were not correlated or were negatively correlated with the total CASIAN score. An exploratory factor analysis was then performed, forcing all items into a single factor, to determine which items failed to load even at a moderate level (≥.30) on this single factor. Finally, preliminary test–retest correlations were conducted to determine any items showing particularly poor test–retest performance.

Results of Item Analysis Procedures

Based on combined results of each of the above procedures, two items were eliminated from further analysis. The first was “Time at Lowest Weight”: it showed poor associations with the other measures of eating disorder pathology, low item-to-total correlation, and it did not load even moderately on the single factor. Similarly, the item “Tolerating Directives” performed poorly; minimal variance, weak associations with measures of other eating disorder pathology, and, most importantly, a very low-test–retest correlation.

The objective bingeing item performed uniformly poorly across analyses; low item-to-total correlation, a low-single factor loading, and weak associations with the other measures of eating disorder pathology. The item assessing “Subjective Bulimic Binge” also performed poorly across tests; item-to-total correlation was low, as was single factor loading and correlations with other measures of eating disorder pathology were weak. However, as both of these items were deemed to be important clinical variables, attempts were made to preserve them within the analysis. These two items were combined and scores recoded assigning the maximum subjective and/or objective frequency for each case. This revised item called “Maximum Binge Frequency” underwent the same item analysis procedures described earlier and was included in further analyses. All subsequent analyses were conducted using the resultant 31-item CASIAN.

Factor Analysis

Exploratory Factor Analysis.

As an initial investigation of the construct validity of the CASIAN, an exploratory factor analysis was conducted. For this analysis, CASIAN item scores obtained for the total sample (n = 171) were used. The “Factor” procedure of SPSS Version 16⁵³ was used.

Extraction.

Maximum likelihood analysis was selected for factor extraction, because this approach provides a principal factor (as opposed to a principal component) solution and yields a statistical test of goodness-of-fit. Decisions about the retention of factors were guided by the Kaiser criterion,⁵⁴ with eigenvalues > 1 being retained. A six-factor solution accounting for 57.1% of the total variance was the best fit for the data. The goodness-of-fit test for this solution reached statistical significance (X² = 411.098, df = 294, p = .000) confirming the utility of the six-factor structure.

Rotation.

Promax was used as the method of rotation. Factors comprising “severity” in AN were expected to be correlated, and hence a nonorthogonal rotation (rather than orthogonal), which allows for correlated factors, was used.⁵⁵ A more conservative cut-off of .4 was used as the guideline for the minimum factor loading required for an item.⁵⁶ In addition, each variable was then required to load less than .4 on all other variables with a minimum difference between the highest loading (“on factor”) and next highest loading (“off factor”) for a variable of .2.

Following this initial factor analysis, “Interference Due to Preoccupations” was dropped as it did not load significantly on any factor. An iterative series of factor analyses were then conducted dropping one by one any item that did not meet the above requirements. In total, eight further factor analyses were run dropping sequentially the following items: “Average Daily Intake,” “Egosyntonic Attachment,” “Intimacy,” “Protection of Habits,” “Medical Complications,” “Menstruation,” and “Social Eating.” All these items failed to meet the factor loading requirements outlined earlier. This resulted in a factor solution for a 23-item CASIAN, which also comprised six factors.

Interfactor Correlations.

Interfactor correlations ranged from a very low, .035 to moderately high, .521 suggesting some evidence for a higher order factor representing severity but also separate components of illness that may work quite independently. A lower cut-off of .32 and an upper cut-off of .8 for intercorrelations between factors are recommended⁵⁷; with <.32 being a small and >.8 being large. It is evident that the majority of intercorrelations were small, none were large, and two fell in the moderate range. Overall, the results suggested that these are relatively independent factors.

Clinically Relevant Items Dropped from the Analysis

A number of clinically relevant variables were not supported psychometrically by the above analysis. The three items retained as clinically relevant, but would not contribute to the overall severity score are “Average Daily Intake” (3a), “Menstruation” (6a), and Medical Complications (6b). Nonetheless, these items may be of theoretical importance to the concept of severity in AN. As this validation study uses a relatively small sample and is the first upon which factor analyses for the CASIAN has been conducted, a conservative approach to the elimination of items from further analysis was adopted. Therefore, it was decided that these items should be part of the questionnaire, but not part of the scoring for the instrument. This solution allows for the gathering of clinical data thought important to an investigation of severity and will therefore be available for future analyses, without sacrificing the psychometric properties of the instrument. The abbreviated version of the 23 items to contribute to scoring and the three additional clinically relevant items retained (contact author for full version of the CASIAN) are given below.

Factor: Weight

• Current weight and height for BMI conversion (Current BMI)

• What was your stable resting weight prior to the commencement of weight losing behaviors? (Premorbid Drop)

• What is the lowest weight you have reached since the onset of weight loss? (Lowest BMI Ever)

• Over the past month what is the lowest weight you have wanted to be? (Lowest Ideal BMI)

Factor: Motivation

• As you may know your minimum healthy weight is ____. How do you feel about the idea that you should be at that weight? (Acceptance of Normal Weight)

• How much weight are you prepared to gain each week until you reach a healthy weight? How do you feel about 1 kg per week? (Acceptance of Weight Gain)

• Say that you did feel motivated to get well and you put your mind to it, do you feel like you could reach a BMI of 20/normal resting weight or have you doubts about your ability to do it? (Self-Efficacy to Reach Normal Weight)

• Say that you did feel motivated to get well and you put your mind to it, how much weight do you think you could gain per week until you reached a normal weight? How confident do you feel that you could gain a kilo a week? (Self-Efficacy to Gain Weight)

• Interviewer Rating: In your opinion is the person unmotivated to get well, somewhat motivated or very motivated to get or stay well? (Clinical Rating)

• How do you feel about your body when it is at its normal healthy resting weight? (Body Dissatisfaction)

• Do you think that you are better off continuing with your life just as it is at present? Would you like to permanently change the eating disorder? (Entrenchment)

Factor: Duration

• What age were you when you began to actively lose weight? At what age were you when your weight losing behaviors began to lead to weight loss? (Illness Duration)

• Since onset has there been any period in which you have returned to your normal weight range and been eating disorder free? (Net Length of Illness)

Factor: Bulimic Behaviors

• Over the past month have you experienced episodes of eating where once you began to eat you felt as if you could not stop or as if you had lost control? (Maximum Binge Frequency)

• Over the past month have you ever made yourself vomit or vomited spontaneously after eating? (Self-Induced Vomiting)

Factor: Obsessionality

• How much of your day do you currently spend thinking about or ruminating on food, weight, shape, fat and body? (Obsessional ED Thought)

• With regards the rules you follow around food and exercise, if over the past month you have broken one of those rules or circumstances prevent you from sticking to them, how have you felt? (Regimentation of Diet)

• Over the past month, has your eating and exercise pattern been for the most part stable? (Stability of Weight Reduction Regime)

• Can you describe a typical week’s activity? Is there any formal/structured exercise in the week? What about incidental activity? (Physical Activity)

Factor: Acute Issues

• Have you lost weight in the last 6 months? How much? Over what time frame? (Acute Weight Drop)

• How many times over the life of the illness have you lost 5 kg or more in less than 30 days? (Precipitous Weight Loss)

• Over the past month have you used any laxatives, either natural forms such as licorice and chewing gums or packet forms? (Laxative Use)

• How would you describe your mood over the past month? Have you felt sad, blue, down more days than not? (Depression)

Clinically Relevant Items Not Contributing to Scoring

• Record a typical days intake (all food and liquid) over the past month (Average Daily Intake).

• Have you ever menstruated/had a period? Do you get your periods at the moment? How regularly? (Menstruation)

• List all the medical complications the person has currently (Medical Complications)

Reliability of the CASIAN

The final 26-item CASIAN had high internal consistency (Cronbach’s α = .811), high test-retest reliability over a 1-week period (ICC = .957), and high interrater reliability (ICC = .973).

Validity of the CASIAN

The extent to which the CASIAN measures AN severity was assessed by investigating convergent, discriminant, concurrent, and predictive validity.

Convergent Validity.

The CASIAN total score correlated significantly with MROAS total score .319 (p = .000), along with the three MROAS subscales, food intake .502 (p = .000), psychosexual −.362 (p = .000), and socioeconomic −.318 (p <.001 as well as the EAT Total Score .443 (p = .000), and three subscales of the EAT, Dieting .347 (p = .000), Bulimia .305 (p = .000), and Oral Control .326 (p = .000). The total score of the MAC-R was also significantly correlated with the CASIAN total score .273 (p <.05). Additional support for the convergent validity of the CASIAN was evident in the significant correlations between the CASIAN total score and total number of inpatient admissions for an ED .186 (p <.05), and total number of hospital days (general and ED) .251 (p = .000).

Discriminant Validity.

There was no significant correlation between the CASIAN and the Marlowe-Crowne Social Desirability Measure .049 (p = .25) or combined family income −.309 (p = .39).

Concurrent Validity.

The total score of the CASIAN was positively and significantly correlated with both the Primary Clinician Rating of Severity .431 (p = .000) and the Subjects Rating of Illness Severity .372 (p = .001).

Predictive Validity.

Table 1 shows that scores obtained on the CASIAN at time 1 correlated significantly with both scores on the CASIAN and other measures of eating disorder symtomatology (EAT, MAC-R, and MROAS) and weight change at both 3 and 6-month follow-up

TABLE 1.

Pearsons correlations between Time 1 CASIAN Total Score and follow-up CASIAN and eating disorder symptomatology measures

Measure	CASIAN Total Score Time 1
CASIAN time 2	.668**
CASIAN time 3	.662**
MROAS average outcome time 2	−.456**
MROAS average outcome time 3	−.541**
EAT total time 2	.598**
EAT total time 3	.579**
MAC-R total time 2	.475**
MAC-R total time 3	.556**
Percent weight gain Time 1 to 2	.127
Percent weight gain Time 1 to 3	.208*

^*p <.05,

^**p <.01.

A series of hierarchical ordinary least squares regression analyses explored the ability of the CASIAN to account for variance on these measures administered at follow-up times one (3 months) and two (6 months). Results of these analyses are reported in Table 2. Co-morbid psychopathology was controlled for in these analyses; OCI-R Total Score and PHQ/PHQ-A subscale scores assessing the presence of a co-morbid depressive illness, panic, and/or anxiety disorder and alcohol abuse were entered into the first step of the regression, CASIAN Total Score at Time 1 was then entered at the second step. A consistent pattern emerged: total CASIAN severity score at initial assessment accounted for a significant amount of the variance in scores at Time 2 and 3 on all other measure of eating disorder symptomatology and psychopathology (MROAS, EAT, and MAC-R). Although the examination of the ability of the total severity score at initial assessment to account for observed weight gain over the follow-up period was not significant for either time points, it approached significance over the longer follow-up period.

TABLE 2.

Regression analyses (“Enter”) with CASIAN Time 1 total score as predictor of eating outcome measures

Measure	R Square Change	F Change	Significance of F Change
CASIAN time 2	.174	43.607	.000
CASIAN time 3	.310	40.186	.000
MROAS average outcome time 2	.153	16.097	.000
MROAS average outcome time 3	.205	20.070	.000
EAT total time 2	.135	17.996	.000
EAT total time 3	.134	16.359	.000
MAC-R total time 2	.091	9.468	.003
MAC-R total time 3	.146	16.647	.000
Percent weight gain Time 1 to 2	.018	1.455	.231
Percent weight gain Time 1 to 3	.050	3.744	.057

Discussion

The aim of this study was to develop and explore the psychometric properties of a multidimensional measure specifically designed to assess severity in AN. This is a first necessary step to enable exploration of what components or axes may be important in understanding illness severity. The study sought to investigate the full dimension of illness by including individuals along the AN spectrum, not just those meeting strict diagnostic criteria. The resultant CASIAN represents an attempt to describe the multiple features of AN that contribute to severity and to operationalize a method for assessing the extent of those features. Rigorous and thorough instrument development processes and item analysis procedures were followed yielding a 23-item instrument, which was subjected to psychometric evaluation, with an additional three items of clinical importance included in the administration of the CASIAN but not contributing to scoring.

It would appear that the construct of severity involves multiple dimensions with factor analytic studies here determining a six factor structure. Those factors comprising severity were motivation, weight, obsessionality, illness duration, bulimic behaviors, and acute illness issues. Some low-to-moderate intercorrelations were observed between factors that could be argued to be conceptually related (e.g., obsessionality and acute illness issues) but on the whole there appears to be evidence for both independence of the factors and some higher order factor accounting for variance across all six factors, which arguably could be severity.

The reliability of the CASIAN is supported by the high internal consistency of the CASIAN items, excellent 1-week stability, and very high interrater agreement. Given the nature of this instrument as clinician rated, and its likely use to assess illness severity and changes therein across different treatments and centers, high levels of interrater reliability are essential. The CASIAN is a distinct advancement in this regard to the MROAS, which, despite being interviewer rated, shows particularly poor interrater reliability.⁵⁸

Overall, the CASIAN observed a pattern of significant correlations with all the other instruments assessing eating disorder psychopathology. The CASIAN as a whole demonstrated significant positive relationships with both the clinician’s assessment of overall illness severity and the participant’s own rating of the same, suggesting that the instrument is providing an assessment of illness severity that matches clinical reality and indeed the individual sufferers own subjective reality. The CASIAN also demonstrated good discriminant validity showing no significant relationships with either of the measures used in this study. The correlation between combined family income and the CASIAN while not significant was substantial, this is likely the result of a recruitment bias, whereby the most severe cases in this study were recruited from public facilities and as such had a lower socioeconomic background. Finally, the CASIAN demonstrated an ability to predict the extent and intensity of eating disorder psychopathology at 3 and 6-month follow-up, suggesting that the instrument may hold promise as an indicator of prognosis in this illness, although longer-term follow-up studies are needed.

The current study had a number of limitations. The first of these was the absence of a standardized diagnostic instrument in the test battery. Although multiple endeavors were made to standardize diagnoses across sites and communication about eligibility for the study was maintained between all sites for the duration of the study, participants in this study were diagnosed according to DSM-IV AN criteria and adjusted EDNOS criteria by the primary clinicians at their site and hence there may have been site differences in the application of the diagnostic system. Second, while the sample size was large for studies on AN, in terms of the sample sizes required for instrument development, the present sample was relatively small. Hence all findings need to be regarded as preliminary. Further samples are needed to conduct factor analytic studies on a larger N to confirm the findings of this study. The strengths of this study include the sample method, which recruited from multiple treatment settings with differing clinical populations, providing for a large spread of illness severity across the sample.

Although still requiring further evaluation, the CASIAN is the first multidimensional measure of severity in AN to be validated in the literature. It offers an alternative to the MROAS as an instrument to provide a measure of ED status. The MROAS was devised over 20 years ago and has been the subject of some criticism concerning its psychometric properties.^43,58 The current instrument is arguably an improvement over the MROAS with demonstrated reliability and validity. Furthermore, the CASIAN takes account of the spectrum of AN symptomatology including psychological aspects such as motivation, body image disturbance, and depression.

The findings of this study require replication and extension in other samples. Further factor analytic studies are needed to confirm the factor structure of the CASIAN and contribute to further understanding of and decisions about items retained on the instrument. Importantly, studies of the long-term predictive utility of both the CASIAN as a measure of severity and the staging model derived from it are needed, before any conclusions can be drawn about the ability of the either to provide a real indicator of illness severity in AN.