Eating disorders have a profoundly negative impact on patients' lives, yet to date there is no satisfactory measure of this impairment. Such a measure would be of great value in the clinical assessment of patients and their response to treatment. It would also help inform epidemiological studies of the burden imposed by eating disorder psychopathology.
The aim of this research was to develop a clinically useful measure of the psychosocial impairment secondary to eating disorder features and to test its psychometric properties. To this end the 16-item, self-report CIA was developed and its reliability, validity, sensitivity to change and ability to predict eating disorder case status were evaluated using data collected in the context of a treatment trial.
It was found that the internal consistency of the preliminary 22-item instrument was high and that all items correlated positively with the global score. However, IRT analysis showed that the questionnaire was significantly non-unidimensional and that six items either had a notable floor effect, significantly misfit or were inconsistent with the underlying construct. These items were therefore excluded. The resulting 16-item instrument (the CIA) was found to be unidimensional and showed high internal consistency. Principal component analysis suggested the existence of three factors (named ‘personal’, ‘social’, and ‘cognitive’), explaining 77% of the variance. All three domains were also found to be unidimensional, making it possible to use both global and domain-specific scores. Overall the findings suggest that the CIA measures one construct. Given the content of the instrument and the fact that its scores correlate closely with clinicians' ratings of secondary psychosocial impairment, it is highly likely that it is measuring psychosocial impairment secondary to the presence of eating disorder features.
The test–retest reliability of the CIA was examined by administering the CIA twice within three days to participants with varying degrees of eating disorder psychopathology. It was found that there was a statistically significant positive correlation between the CIA scores at the two time points. The scores were on average slightly lower on the second occasion but the difference was small. Overall, the findings suggest that the CIA has satisfactory test–retest reliability.
Two tests of the construct validity of the CIA were conducted. The first was an indirect one and it involved comparing scores on the EDE-Q and the CIA. A statistically significant positive correlation was found indicating that higher levels of eating disorder psychopathology were associated with higher levels of secondary psychosocial impairment. The second test involved comparing scores on the CIA with ratings of secondary psychosocial impairment made by expert clinicians. Again there was a strong positive correlation between the two.
The discriminant validity of the CIA was tested by comparing the global and domain-specific CIA scores of patients with an eating disorder with those of patients who were judged no longer to have one. It was found that the scores of the two groups differed significantly. A ROC analysis revealed that a cut-point of 16 on the CIA best predicted eating disorder case status, with a sensitivity of 76% and a specificity of 86%.
Lastly, sensitivity to change was tested. It was found that there was a significant decrease in patients' CIA global and domain scores following cognitive behaviour therapy, and that there were significant positive correlations between change in clinician-rated secondary impairment and change in the global CIA score.
Certain points about this work are of note. First, this new instrument, the CIA, addresses impairment due to all the main elements of eating disorder psychopathology including concerns about shape. None of the other measures explicitly assesses the effect of such concerns despite the fact that they are central to the “core psychopathology” of the eating disorders (Fairburn, 2008
). In this study, patients are likely to have had all the main features of their eating disorder at the forefront of their mind when completing the CIA, as they were asked to complete the EDE-Q, a measure of all aspects of eating disorder psychopathology, immediately before the CIA. A second strength of this research is that many aspects of the performance of the CIA were tested including sensitivity to change and ability to predict case status. Equivalent data are not available for existing measures of eating disorder-specific health-related quality of life. Third, the sample used is likely to have been representative of many other outpatient samples of adults with an eating disorder, given its transdiagnostic composition and the two catchment area-based sampling frames. Fourth, the participants were assessed both before and after treatment and over an extended period of follow-up. Therefore CIA data were available on people with the full range of severity of eating disorder psychopathology. This was important since we wanted to test the CIA's performance across its entire scoring range. Lastly, the ability of the CIA to assess secondary psychosocial impairment was validated against a simultaneous but independent assessment of impairment made by expert clinicians, the best available validator.
The main limitation of the study was the relatively small sample size, especially given the need to employ subsets of the data for the various analyses. Despite this, significant associations were found, all of which support the validity and utility of the CIA. Another limitation is that the sample contained few patients with anorexia nervosa. More data are needed on this subgroup of patients. It might be thought that a limitation of the study of discriminant validity was the fact that no CIA data on healthy controls were used: instead, discriminant validity was tested by comparing patients with a clinical eating disorder with patients who were judged to no longer to have one. In fact this was a particularly stringent test since the latter participants were likely to have residual levels of eating disorder psychopathology making discrimination more difficult. Nevertheless it would be of value to have normative data on the CIA to ascertain the degree of impairment that is associated with the level of eating disorder features found in the general population. As always, it is important to stress that the findings require replication, preferably using larger samples and ones differing in case composition, age and ethnicity. Finally, it should be recalled that the CIA does not attempt to assess secondary physical impairment. This is because we do not think it is possible for patients to reliably identify those aspects of their physical functioning that are being impaired exclusively as a result of their eating disorder. We recommend that any assessment of secondary physical impairment should be made by a physician alongside the administration of the CIA.
In conclusion, this paper has described the development and performance of a brief self-report measure of the nature and severity of the psychosocial impairment that arises from eating disorder psychopathology. The findings support the validity and utility of the instrument.