A self-report questionnaire for measuring separation anxiety in adulthood☆☆☆

Measures of separation anxiety 

Most existing measures of SA have focused on symptoms in the juvenile phase of development. With the inclusion of juvenile separation anxiety disorder in DSM-III,12 DSM-IV,4 and ICD-10,5 a number of interview-based instruments have been developed to elicit relevant symptoms directly from children, and indirectly from the observations of parents.13, 14, 15, 16 There are several other childhood measures that assess juvenile SA in a dimensional or psychodynamic manner, such as the Hansberg Separation Anxiety Test,17, 18 the Multidimensional Anxiety Scale for Children,19 and the Screen for Child Anxiety Related Emotional Disorder.20 A retrospective measure of juvenile SA, the Separation Anxiety Symptom Inventory (SASI),21 records adults' memories of SA experiences over the first 18 years of life. This 15-item self-report measure has a coherent factor structure, satisfactory internal consistency (Cronbach's alpha = .84 to .88) and sound test-retest reliability (intraclass correlations of .86 to .98).21

In adulthood, the focus of measurement has tended to be on attachment style, a construct derived from attachment theory, in which the focus is on a pervasive pattern of bonding. The Berkeley Adult Attachment Interview22 and the Attachment Style Questionnaire23 are examples of such measures. In contrast, relatively little attention has been given specifically to developing a phenomenological or symptom measure of SA for adulthood. The few existing instruments have assumed that SA is a dimensional construct representing an underlying personality trait. For example, the Interpersonal Sensitivity Measure (IPSM)24 consists of 36 items with SA forming one subscale. Gilbert et al.25 have developed a 10-item self-report measure with five items measuring SA and the other five, social anxiety in adulthood. It is unclear whether the SA subscale represents a construct equivalent to a putative adulthood category of separation anxiety disorder. A measure relevant only to maternal attachment to infants, the Maternal Separation Anxiety Scale (MSAC), was developed by Hock et al.26 Maternal SA was defined as the “unpleasant emotional state reflecting a mother's apprehensions about leaving her child.”26 The measure would not be relevant to assessing general symptoms of SA in adulthood.

Semistructured diagnostic interviews for adults such as the Diagnostic Interview Schedule,27 the Composite International Diagnostic Interview (CIDI),28 and the Structured Clinical Interview for DSM (SCID)29 do not include modules for current separation anxiety disorder. Moreover, there are no recent clinical or epidemiological studies that have included separation anxiety disorder as a possible category in adulthood.30, 31, 32, 33

Preliminary investigations into the measurement of adult separation anxiety 

Data outlining the development of the Adult Separation Anxiety Semistructured Interview (ASA-SI) have been provided previously.7 The interview was used in the present study as the “gold standard” to make a provisional diagnosis of separation anxiety disorder in adulthood. Items for the interview were derived from a content analysis of responses to a semistructured assessment of patients selected on clinical grounds to be suffering from high levels of SA. The qualitative study was terminated when the themes yielded by successive interviews became repetitive, suggesting that the constellation of symptoms documented was exhaustive.34 Items of the ASA-SI assessed for both current symptoms (in the last 3 months and lasting for several weeks) and past symptoms (prior to the last 3-month period and lasting for several weeks in adulthood).

On completion of the interview, a global clinical assignment to the putative diagnosis of adult separation anxiety disorder (ASAD) was made by trained clinicians.7 (The term “ASAD” will be used for convenience throughout the remainder of the paper although the designation is intended to be provisional given its uncertain nosological status). Training of interviewers involved listening to 10 audiotaped ASA-SI interviews, half with a predetermined diagnosis of ASAD. Interviewers were then required to rate a further 10 audiotapes. Optimal levels of inter-rater reliability (100%) have been achieved in previous studies.7, 9 Furthermore, a DSM-IV-based diagnosis based on specified criteria in the manual corresponded closely to the interviewer's global clinical assignments on the ASA-SI (kappa = .74).7

Development of the adult separation anxiety questionare 

The self-report questionnaire for adult SA (ASA-27), whose properties will be examined further in the present study, is based on items that are identical to those contained in the ASA-SI.7 The early development of the ASA-27 has been described in previous studies.7, 9 Each item on the questionnaire was rated on a four-point scale, where a rating of 0 indicated “this has never happened” and 3 indicated “this happens very often.” In a clinic-based study of anxiety patients, preliminary psychometric analyses were undertaken on a 16-item version of the ASA-27.9 It seemed necessary therefore to undertake a further investigation using the full item pool and including a larger, more heterogenous sample with a substantial portion of cases assigned to ASAD according to the ASA-SI.

The goals of this present study therefore were to examine more comprehensively the psychometric properties of the ASA-27, particularly focusing on its factorial structure and on deriving threshold scores by comparing the questionnaire with the ASA-SI. A broader aim was to test more comprehensively the capacity of novel measures to identify a coherent construct of SA in adulthood.

Method 

Results 

The whole sample (N = 352) included 111 males (31.5%) with a mean age of 43 (SD = 12.4) years, and 241 females (68.5%) with a mean age of 40 (SD = 12.6) years. The overall mean age of the combined sample was 41 (SD = 12.6; range, 19 to 80) years.

Factor structure of the ASA-27 

A principal components analysis without restriction on the number of factors was performed on the 27 items of the ASA-27 (n = 352). A factor loading of .35 was set, a priori, for item inclusion. The purpose of this analysis was to examine whether items would load highly on a principal factor, reflecting a cohesive underlying construct of SA.

The principal components analysis generated five factors with eigenvalues greater than 1.0. All items loaded positively on the first factor (Table 1) and their loadings were greater than or roughly equal to those on other factors.

Table 1. Factor Loadings From the Unrotated Principal Components Analysis of the ASA-27 (N = 352)

		Factor
	Question item	1	2	3	4	5
	1.Feels more secure at home with close attachments	.66	.40	.06	.12	−.12
	2. Experiences difficulty in staying away from home for several hours	.66	.22	−.24	.06	.03
	3. Carries around something in purse or wallet for security or comfort	.56	−.02	−.26	.29	.18
	4. Experiences extreme stress when leaving home to go on a long trip	.60	.15	−.42	.22	.12
	5. Suffers from nightmares or dreams about separation from close attachments	.70	−.06	.00	−.37	.28
	6. Experiences extreme stress before leaving someone close when going away on a trip	.73	−.08	−.18	−.16	.18
	7. Becomes very upset when usual routine is disrupted	.61	−.05	−.30	.16	.06
	8. Worries about the intensity of relationships with close attachments	.72	−.33	.01	−.14	.01
	9. Experiences physical symptoms before leaving to go to work or other regular activities	.58	.13	−.36	.34	.08
	10. Talks a lot in order to keep close attachments around	.55	−.30	.33	.45	.21
	11. Concerned where close attachments are going when separated from them	.75	−.09	−.01	−.03	−.22
	12. Experiences difficulty in sleeping alone at night	.69	.33	.25	.06	−.12
	13. Better able to sleep if he/she can hear the voices of close attachments or voices on the TV or radio	.61	.44	.27	.02	.17
	14. Becomes very distressed when thinking about being away from close attachments	.80	.15	.00	−.05	−.12
	15. Suffers from nightmares or dreams about separation from home	.59	.01	−.12	−.22	.51
	16. Worries about close attachments coming to serious harm	.68	−.04	.03	−.30	.02
	17. Becomes very upset with change to usual daily routine if it interferes with contact with close attachments	.71	−.11	−.01	−.02	−.02
	18. Worries a lot about close attachments leaving	.76	−.26	.11	−.23	−.10
	19. Sleeps better if the lights are on in the house or bedroom	.49	.42	.36	−.14	.39
	20. Tries to avoid being at home alone when close attachments are out	.64	.35	.33	.12	−.18
	21. Suffers from panic attacks when thinking about leaving close attachments or about them leaving	.76	.02	−.08	.02	−.22
	22. Anxiety about not speaking to close attachments on the telephone regularly	.78	−.07	.09	.00	−.27
	23. Afraid that he/she would not be able to cope if close attachments left	.77	−.15	.05	−.03	−.27
	24. Suffers from panic attacks when separated from close attachments	.76	.01	−.22	−.01	−.16
	25. Worries about possible events that may separate him/her from close attachments	.73	−.14	.03	−.04	−.16
	26. Close attachments have mentioned that he/she talks a lot	.38	−.46	.41	.44	.20
	27. Worries that relationships are so close it may cause others problems	.62	−.46	.06	−.10	.05

Abbreviation: ASA-27, Adult Separation Anxiety Questionnaire.

Loadings on this first factor ranged from .38 to .80, with most items loading in the .60 to .75 range. The first factor alone accounted for 45% of the variance, with each of the remaining factors contributing between 4% and 6% of the variance only. Orthogonal rotation of the five-factor solution failed to strengthen the loadings on the additional dimensions, with all items continuing to load strongly on the first rotated factor. Thus, the complete 27-item measure was retained for further psychometric analysis.

Sensitivity and specificity 

Participants (n = 178) who completed both the interview (ASA-SI) and questionnaire (ASA-27) were included in this analysis. Figure 1 shows the distributions of ASA-27 scores for those assigned to the ASAD category according to the interview (ASA-SI).

View Large Image Download to PowerPoint

Fig. 1. Frequency of ASA-27 total scores for ASAD and non-ASAD participants.

The mean ASA-27 score for participants with ASAD was almost three times higher than for those not so assigned (37.8 v 14.2; t = 12.1, df = 176, P < .001).

A receiver operation characteristic (ROC) analysis35 was conducted to assess more closely the relationship between the ASA-27 and case-assignment by the ASA-SI (Fig 2).

View Large Image Download to PowerPoint

Fig. 2. ROC for the ASA-27. T1, threshold 1; T2, threshold 2.

According to Swets,36 area under the curve (AUC) values between .5 and .7 are regarded as low, while values above .9 are judged to be high. The ROC AUC for the ASA-27 was .90. Sensitivity and specificity curves were computed for two possible cut-off scores (Fig 3).

View Large Image Download to PowerPoint

Fig. 3. Sensitivity and specificity estimates for the ASA-27 compared to the structured interview. T1, threshold 1 (97% sensitivity, 66% specificity); T2, threshold 2 (81% sensitivity, 84% specificity).

A score of 22 (see “Threshold 2” in Fig 3) on the ASA-27 represented the intersection point for sensitivity and specificity (sensitivity = 81%; specificity = 84%). This score yielded a misclassification rate of 17% (positive predictive power = 76%; negative predictive power = 88%) indicating that at this cutoff, participants were almost equally likely to be incorrectly assigned to either the “separation anxiety” or “non-separation anxiety” categories.

One other cut-off score was computed. This aimed to provide a cutoff with high sensitivity which might be applied in community surveys aiming to capture all cases of ASAD in the first screen, even if the subpopulation included some “non-cases.” A cut-off score of 16 (see “Threshold 1” in Fig 3) resulted in 97% sensitivity and a specificity of 66% (misclassification rate = 25%; positive predictive power = 62%; negative predictive power = 97%). Thus, application of that cutoff would yield a sample that included nearly all cases of ASAD but 39% would need to be excluded by a further more rigorous assessment, for example, by applying the ASA-SI.

Discussion 

The development of the ASA-SI represents an attempt to provide a clinically based measure that allows trained clinicians to make an accurate assignment to a putative category of separation anxiety disorder in adults. In our previous studies,7, 9 we found that clinicians could be instructed to make such diagnoses after a short training period, achieving high levels of inter-rater reliability (100%).

The principal components analysis of the ASA-27 in the present study provide further evidence that symptoms of SA in adulthood may form a coherent construct. Although the principal components analysis generated five factors, the first factor alone accounted for 45% of the variance, with subsequent factors making a negligible contribution. Cronbach's alpha for the whole scale was .95, indicating a high level of internal consistency for the total item pool. The finding of high test-retest reliability (.86) indicated that respondents maintained consistency in their ratings of SA over time.

Apart from supporting the construct of SA in adulthood, the ASA-27 was developed as a conveniently administered self-report measure that might substitute partially for the logistically more exacting interview-based method of assignment. Participants assigned to the ASAD category by the ASA-SI returned scores almost three times higher on the questionnaire measure than those not so assigned. The AUC for the ASA-27 in the ROC analysis was high (.9), indicating a satisfactory level of correspondence between the questionnaire and the ASA-SI. Two cut-off scores on the ASA-27 were derived using the ROC analysis. A score of 22 on the ASA-27 reflected the point of transection of the ROC curves providing sound levels of sensitivity (81%) and specificity (84%). To achieve 97% sensitivity, for example, as a first screen for ASAD in future community samples, a score of 16 emerged as most appropriate. Once screened, those that score above this threshold could be interviewed using the ASA-SI to derive a definite diagnosis of ASAD.

Sensitivity and specificity estimates tend to vary across samples, yielding lower sensitivities and higher specificity rates in general population samples than in patient samples.37 For this reason, a mixed sample comprising both community volunteers and anxiety patients was used for the present analyses. Even so, the samples were chosen partly for convenience and hence further studies on a wider range of populations (e.g., a random community sample) would be necessary to replicate the current findings and to test the robustness of the thresholds derived.

These findings represent a tentative step in examining the nosological status of separation anxiety disorder as a putative entity in adulthood. Further clinical and epidemiological research isrequired to determine whether the inclusion of a category of ASAD is warranted in future typologies of the anxiety disorders in adulthood.

Data from the present study add to evidence obtained from our earlier investigations6, 8, 9 in suggesting that a construct of SA may be identified in adults. Psychometric evaluation of the questionnaire revealed a strong principal component of SA, high levels of internal and test-retest reliability, and a robust ROC analysis yielding two strategic thresholds for case assignment. Such data add further evidence that builds to the argument that a category of separation anxiety is both identifiable and measurable in adulthood.