INTRODUCTION
Obsessive-compulsive disorder (OCD) is characterized by obsessions (intrusive thoughts, images, impulses experienced as unbearable distress) and compulsions (repetitive overt or mental rituals performed to reduce or prevent anxiety arising from obsessions) [1]. Exposure and response prevention (ERP) is currently considered as the most effective psychological treatment for OCD [2,3]. In recent years, acceptance and commitment therapy (ACT) has emerged as a primary therapy or an augmentation approach to enhance the efficacy of ERP [4-6].
With the use of ACT for the treatment of OCD, experiential avoidance (EA), one of the six core processes of psychopathology from the perspective of ACT [7], has also been conceptualized in the OCD context [6,8-10]. For example, a person with OCD who has an intrusive thought that he is likely to harm his daughter with a knife may take the knife away in front of him, not only because the knife is dangerous but also to get rid of his unwanted inner experiences [9]. Contrary to behavioral avoidance of cognitive-behavioral therapy, which focuses on the avoidance of external fear trigger, EA emphasizes the tendency to avoid inner experiences, such as unpleasant thoughts, emotions, memories, and physical sensations [11]. In addition, these time-consuming efforts to avoid invariably lead to unworkable actions inconsistent with one’s values and interfere life functioning [12]. In this vein, early studies in this filed demonstrated that targeting EA may have implications for OCD treatment [6,10,13]. Empirically, persons with higher levels of obsessive-compulsive (OC) symptoms engage in more EA than those with lower levels [14], and ACT for patients with OCD lessens EA [15]. Furthermore, in a comparison between EA and distress tolerance, unwillingness to endure (i.e., EA), rather than simple incapacity to tolerate unpleasant internal experiences, best predicts obsessional symptoms [13]. However, the contribution of EA to OC symptom prediction was found to be less significant. In fact, EA did not significantly contribute to the prediction of OC symptom dimensions over and above obsessive beliefs [14]. Researchers pointed out that the reason for these results is that the EA construct, measured by the Acceptance and Action Questionnaire-II (AAQ-II) [16], is too general (e.g., emotions cause problems in my life) to predict OC symptoms and is limited in capturing more specific and diverse obsessive beliefs [12].
This limitation and the need for domain-specific measures have been generally recognized across different psychiatric disorders. Diverse disorder-specific versions of the AAQ have been developed, for example, for chronic pain [17], psychotic symptoms [18], body image concerns [19], and hoarding [20]. Furthermore, Benoy et al. [21] observed greater treatment sensitivity of alternative versions with greater wording and context specificity compared with the AAQ-II. A recent review demonstrated that context-specific measures performed better than a generic measure of psychological flexibility in terms of incremental validity and treatment sensitivity [22].
In this context, Jacoby et al. [12] developed an AAQ version specific to OC symptoms, referred to as the Acceptance and Action Questionnaire for Obsessions and Compulsions (AAQ-OC). The AAQ-OC specifically measures unwanted intrusive thoughts and responses (i.e., EA) to them. Notably, a definition of intrusive thoughts and their examples are provided at the beginning of the measure. This measure comprises 13 items with two factors, namely, Valued Action and Willingness subscales, and generally exhibits good psychometric properties. In particular, the AAQ-OC demonstrates incremental validity in predicting two symptom dimensions of responsibility for harm and unacceptable thoughts. To the best of our knowledge, the Persian version of the AAQ-OC is the only published translation [23]. However, both studies reported the results based on the nonclinical university students. Despite the specificity of the questionnaire to OCD, the lack of study on how patients with OCD may respond to this questionnaire can be an obstacle to the increasing use of the AAQ-OC.
Therefore, this study aimed to evaluate the psychometric properties of the Korean version of the AAQ-OC in nonclinical (university medical students) and clinical (OCD patients) samples. In this study, the factor structure, reliability, as well as concurrent and criterion validities of the Korean version of the AAQ-OC were investigated.
METHODS
Participants
This study recruited undergraduate medical students at the Kyungpook National University School of Medicine as a nonclinical sample from 2021 to 2023. This research coincided with the biannual mental health evaluation conducted at the beginning of each semester across all grades. Before the evaluations, the procedure was explained, and consent was obtained from the students. All assessments were conducted using online self-reported questionnaires, and no compensation was provided to the participants. Data received from the school included demographic information such as sex, age, school grade, responses to medical and psychiatric history questions, and various psychological measures. Specifically, psychological measures such as the Beck Depression Inventory-II (BDI-II), Beck Anxiety Inventory (BAI), and Perceived Stress Scale (PSS) were obtained from regular mental health evaluation, whereas scales related to ACT were unique to this study. The collected data will be securely discarded one year after the completion of the research. Then, individuals with previous or current history of psychiatric or neurological diagnoses, severe medical illness, a BDI-II [24] score of 15 or above, or any significant inaccurate or incomplete data were excluded, leading to the removal of 53 individuals from the sample. The final nonclinical sample consisted of 561 medical students (414 male and 147 female; mean±standard deviation [SD] of age=21.2±2.4 years).
Patients with OCD aged 18-55 years were recruited through subway advertisements and psychiatric clinics at Kyungpook National University Hospital as a clinical sample from 2017 to 2023. Psychiatric interviews were conducted by two experienced psychiatrists for differential diagnoses according to the Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition (DSM-5) criteria. Individuals were excluded if they had any current comorbid psychiatric diagnosis (e.g., major depressive disorder, alcohol use disorder, schizophrenia spectrum and other psychotic disorders, and intellectual disability), neurological disorders, or a history of head injury or medical illness with cognitive sequelae. The final clinical sample consisted of 121 patients (71 male and 50 female; mean age±SD=27.2±7.6 years).
All the participants provided written informed consent after being briefed about the research protocol. This study was approved by the Institutional Review Board of Kyungpook National University Hospital (2021-04-032).
Psychological measures
Psychological measures were commonly administered to both nonclinical and clinical samples, except the PSS and Dimensional Obsessive-Compulsive Scale (DOCS), which were administered only to the students and patients, respectively.
AAQ-OC
The AAQ-OC [12] measures EA specific to OC symptoms. Individual items are scored on a 7-point Likert scale ranging from always true (7) to never true (1), with higher scores indicating greater EA (i.e., an elevated level of symptom severity). The AAQ-OC exhibits good psychometric properties and consists of 13 items with two factors: Factor 1 (Valued Action: items 1, 3, 4, 6, 8, 9, 10, and 12) indicates the extent to which the presence of intrusive thoughts impairs valued action, whereas Factor 2 (Willingness: items 2, 5, 7, 11, and 13) indicates the degree to which people with OCD are engaged in a struggle with (or accept less willingly) those intrusions. We translated the AAQ-OC with permission from the original authors. In the present sample, Cronbach’s α for all items was 0.93.
AAQ-II
The AAQ-II [25] is an 8-item questionnaire for evaluating the psychological inflexibility and EA emphasized in ACT. Each item is rated on a 7-point Likert scale, with higher scores indicating a higher degree of psychological inflexibility [26]. This study used the Korean version of the AAQ-II, which had good internal consistency (α=0.85).
Cognitive Fusion Questionnaire
The Cognitive Fusion Questionnaire (CFQ) [27] is a 7-item self-report measure for assessing excessive attachment to the literal content of thoughts. The items evaluate the tendency of thought entanglement and are rated on a scale from 1 (“never true”) to 7 (“always true”). Higher scores indicate higher levels of cognitive fusion. The CFQ has demonstrated excellent internal reliability (α=0.88 to 0.93) in both clinical and nonclinical samples [27,28].
DOCS
The DOCS [34] is a 20-item self-report measure that assesses the severity of the four most consistently replicated OC symptom dimensions: 1) contamination, 2) responsibility for harm and mistakes, 3) unacceptable thoughts, and 4) symmetry and ordering. Each item is rated on a 5-point Likert scale, yielding individual subscale scores ranging from 0 to 20 and a total score from 0 to 80. The DOCS exhibit excellent reliability in clinical samples, and the measure corresponds well with other measures of OC symptoms. The Korean version of the DOCS (α=0.91 to 0.95) was used in this study [35].
Statistical analysis
The Kolmogorov-Smirnov test was used as the goodnessof-fit index to evaluate the normal distribution of the AAQOC. The test results indicated that the total and subscale scores of the AAQ-OC were not normally distributed. Thus, nonparametric statistics were used whenever appropriate. Chi-squared tests were employed for model comparisons, and independent samples t-test was adopted for group comparisons. The internal consistency of the AAQ-OC and its subscales was estimated using Cronbach’s α. Regarding test-retest reliability, intraclass correlation coefficient (ICC) estimates were calculated based on a single measurement, absolute-agreement, and 2-way mixed-effects model using data from 37 patients who repeated the AAQ-OC at a 2-month interval. To assess the concurrent validity of the scale, Spearman’s correlations between AAQ-OC, AAQ-II, and CFQ were analyzed. To assess its discriminant validity, on the other hand, Spearman’s correlations between AAQ-OC, BDI-II, BAI, and PSS were analyzed. In addition, Spearman’s correlations between AAQOC and DOCS were analyzed in OCD patients to examine the criterion validity. Furthermore, we conducted receiver operating characteristic (ROC) analyses, in which the association between sensitivity and specificity is used to estimate the area under the curve (AUC) to indicate how well scores on a measure distinguish between positive (i.e., a diagnosis of OCD) and negative (i.e., nonclinical students) cases and to establish cutoff scores with optimal diagnostic accuracy for distinguishing between OCD and nonclinical groups. The IBM SPSS Statistics software for Windows, version 23 (IBM Corp., Armonk, NY, USA), was used to conduct all statistical analyses, and statistical significance was set at p<0.05.
Both explorative factor analysis (EFA) and confirmatory factor analysis (CFA) were conducted to examine the factor structure of the AAQ-OC. The nonclinical sample was divided into two subsamples using a random group generator with participant identifiers (1 to 561). According to the guideline by Anderson and Gerbing [36], subsample 1 was used to conduct an EFA, whereas subsample 2 was used to cross-validate those factors. CFA was conducted using the AMOS 20 software (IBM Corp., Armonk, NY, USA) in the IBM SPSS [37]. The maximum likelihood (ML) estimation approach was employed to estimate the parameters. To evaluate the goodness of fit of the CFA models, multiple indices were chosen, including the chi-squared statistic, comparative fit index (CFI) [38], standardized root mean square residual (SRMR) [39], and root mean square error of approximation (RMSEA) [40], as each of these indices provides different information. As a general rule, CFI >0.90, SRMR <0.10, and RMSEA <0.10 are considered adequate fit [41].
RESULTS
Descriptive statistics
The mean±SD of the AAQ-OC total, Valued Action subscale, and Willingness subscale scores in the nonclinical sample were 32.8±12.2, 17.1±7.9, and 14.6±5.7, respectively (Table 1). No significant sex difference was observed in the AAQ-OC total and Willingness subscale scores (t=1.26, df=559, p=0.209; t=-0.54, df=559, p=0.591, respectively), whereas significant sex differences were observed in the Valued Action score (male= 17.6±8.0, female=15.8±7.3; t=2.46, df=278.8, p=0.015). Age was not associated with the AAQ-OC total and its two subscales.
In the clinical sample, the mean±SD of the AAQ-OC total, Valued Action subscale, and Willingness subscale scores were 61.2±18.2, 38.3±12.1, and 22.9±7.0, respectively (Table 1). No significant sex difference and association with age were observed in any AAQ-OC scores.
The AAQ-OC total score showed positively and negatively skewed distributions in the nonclinical and OCD samples, respectively (Table 2). Compared with AAQ-II, AAQ-OC subscale was more negatively skewed in the OCD sample.
Factor structure
Both EFA and CFA were conducted to assess the factor structure of the AAQ-OC in the university student sample. Subsample 1 was used to conduct EFA (n=279; 211 male, 68 female; average±SD age=21.16±2.33 years), whereas subsample 2 was used to cross-validate those factors using CFA (n=282; 203 male, 79 female; average±SD age=21.23±2.55 years). No significant difference was observed between the two subsamples in terms of sex, age, scores of AAQ-OC and its two subscales, and all other psychological measures.
EFA was conducted on subsample 1 using ML estimation and oblique rotation (direct oblimin with delta=0). Inspection of the scree plot revealed a two-factor solution with eigenvalues of 5.84 and 2.25. The explained variances for these two factors were 44.9% and 17.3%, respectively. Factors 1 and 2 well matched the Valued Action and Willingness subscales, respectively (Table 3). However, item 3, originally loaded onto the Valued Action factor, was loaded onto the Willingness factor, whereas item 11, originally loaded onto the Willingness factor, was loaded onto the Valued Action subscale. The correlation between the two factors was 0.37. These results were consistent with those of the original study [12].
EFA was also conducted on the OCD sample. Inspection of the scree plot revealed a two-factor solution with eigenvalues of 7.68 and 1.39. The explained variances for these two factors were 59.0% and 10.7%, respectively. Item 3 was loaded onto the original Valued Action factor, whereas item 11 was still loaded onto the Valued Action subscale. Compared with the university student sample, the two-factor model seems to better explain the variances in the patients with OCD (Table 3).
CFA was conducted on subsample 2 to cross-validate the two-factor model. This confirmatory model was close to an acceptable fit with the data (χ2(64)=530.2, p<0.001; RMSEA=0.161; SRMR=0.139; CFI=0.742). To enhance goodness of fit, the CFA model was respecified with five correlated residuals based on the EFA. This revised CFA exhibited an improved fit with χ2(59)=242.2, p<0.001; RMSEA=0.105; SRMR=0.107; CFI=0.899 (Table 4).
On the other hand, based on the EFA in subsample 1 and the patient sample, the new model putting item 11 into factor 1 was also evaluated. The CFA results showed an acceptable fit with χ2(64)=265.6, p<0.001; RMSEA=0.106; SRMR=0.089; CFI=0.889. To enhance goodness of fit, the CFA model was respecified with three correlated residuals based on the EFA. This revised CFA exhibited an improved fit with χ2(61)=197.2, p<0.001; RMSEA=0.089; SRMR=0.088; CFI=0.925 (Table 4). However, consistent with the original study,12 the remaining analyses use the same items to compose the Valued Action subscale (items 1, 3, 4, 6, 8, 9, 10, and 12) and Willingness subscale (items 2, 5, 7, 11, and 13).
Reliability
The AAQ-OC demonstrated a high internal consistency in both samples. In the nonclinical sample, Cronbach’s α coefficients were 0.87, 0.86, and 0.72 for AAQ-OC total, Valued Action, and Willingness subscales, respectively. The item-total correlations varied from 0.46 (item 5 for AAQ-OC Willingness) to 0.66 (item 10 for AAQ-OC Valued Action).
For OCD patients, Cronbach’s α coefficients were 0.94, 0.93, and 0.84 for AAQ-OC total, Valued Action, and Willingness subscales, respectively. The item-total correlations varied from 0.52 (item 13 for AAQ-OC Willingness) to 0.85 (item 8 for AAQ-OC Valued Action). The ICC of the total scores of the AAQ-OC between test and retest with an 8-weeks interval, was 0.777 with a 95% confidence interval (CI) from 0.609 to 0.878 (F=7.9, p<0.001).
Concurrent validity
Regarding the convergent validity of the AAQ-OC factors, both the Valued Action and Willingness subscales were positively correlated with the AAQ-II and CFQ scores in both nonclinical and clinical samples (Table 5). As regards the discriminant validity of the AAQ-OC factors, although both factors were correlated with the BDI-II, BAI, and PSS scores, the associations were relatively weak (0.12<r<0.31) in the nonclinical sample. In the OCD group, although both factors were correlated with the BDI-II and BAI scores, the relationships were also weak (0.33<r<0.59), relative to their associations with AAQ-II and CFQ (0.57<r<0.76) (Table 5).
Criterion validity
The independent t-test indicated that the OCD patients had higher scores of AAQ-OC and its two subscales than the university students (Table 1). Correlational analyses were conducted in the OCD sample to assess the association between AAQ-OC and OCD symptoms. In the OCD patients, the Valued Action and Willingness subscale scores were significantly correlated with the DOCS total scores (r=0.44 p<0.001 and r=0.37 <0.001, respectively). Among the DOCS subscales, the two symptom dimensions of responsibility for harm and mistakes and unacceptable thoughts were consistently associated with both the Valued Action and Willingness subscales (Table 6). Finally, we determined whether the AAQ-OC exhibited incremental validity in predicting OC symptoms above and beyond the AAQ-II (Table 7). Hierarchical multiple regression analyses revealed that the amount of variance in the DOCS unacceptable thought scores significantly increased (∆R2=0.025 for Step 2, p<0.001) when the AAQ-OC was added in Step 2. However, no incremental validity was observed in the other three domains in the OCD sample.
Potential diagnostic value of the AAQ-OC
In distinguishing the OCD group from the nonclinical group, the ROC analysis of the AAQ-OC total score showed an AUC estimate of 0.90 (95% CI=0.86-0.94). Thus, the AAQ-OC total score appears to have great diagnostic accuracy. A cutoff score of 46 or higher provided the best balance between sensitivity and specificity, correctly classifying about 84% of OCD patients (sensitivity) and 84% of nonclinical participants (specificity).
DISCUSSION
This study examined the psychometric properties of the Korean version of the AAQ-OC, and the results showed its cross-cultural validity in Korean samples.
The findings of this study suggest that the results from our nonclinical AAQ-OC data were consistent with those of previous studies that reported 36.9±14.8 [12] and 44.3±13.5 [23] as the average of the AAQ-OC total scores. Notably, the AAQ-OC subscale was more negatively skewed than the AAQ-II particularly in the OCD sample, indicating that more individuals were distributed in the high-score band in the AAQ-OC than the AAQ-II.
In this study, no significant sex difference and relationship with age were observed in any AAQ-OC scores, but there was a sex difference in the AAQ-OC-VA in the nonclinical sample. A previous study also reported no gender differences in the AAQ-OC [12]. However, previous studies did not report a relationship with age, which has no significance in this study [12,23].
Factor analyses revealed that a two-factor structure of the Korean version of the AAQ-OC provided an acceptable fit to the data in the nonclinical sample. The items loaded on each factor well matched the Valued Action and Willingness subscales in the original scale;[12] however, items 3 and 11 were loaded onto the other factor in this study. On the other hand, the OCD group exhibited better match with the two original subscales. That is, item 3 was loaded onto the Valued Action subscale, where this item was originally assigned. Moreover, the explained variances of two factors showed an additional 10% increase in the OCD group. However, item 11 was still loaded onto Valued Action, not the original Willingness factor. We assumed that item 11 is less clear, unlike other items in the Valued Action subscale which have a more specific wording, i.e., including the word “avoid” or “control,” and may be misinterpreted as measuring affective intolerance or general distress. In fact, the loading of this item (0.5) was also low in the Persian translation[23].
The results of the present study indicated that the Korean version of the AAQ-OC and its factors had excellent internal consistencies in both the nonclinical (0.72 to 0.87 of αs) and clinical (0.84 to 0.94 of αs) samples in comparison with the original (0.82 to 0.93 of αs) and Persian (0.73 to 0.91 of αs) versions [12,23].
As regards convergent validity, the AAQ-OC total score was highly correlated with AAQ-II and CFQ in the nonclinical sample. In terms of the two subscales, Valued Action showed higher correlation with AAQ-II than Willingness in the nonclinical samples. These findings were consistent with those in the original study. Furthermore, Soltani et al. [23] reported that the Willingness subscale was not correlated with AAQ-II and other depression, anxiety, and stress scores. These tendencies were also observed in the OCD group, although the correlation coefficients generally tended to be higher in the OCD than the student group.
In terms of discriminant validity, the scores of AAQ-OC and its two subscales exhibited weak correlations (rs<0.31) with general depression anxiety and stress, although their relationships were statistically significant partially due to the large sample size in the nonclinical sample. Consistent with our findings, previous studies demonstrated that the AAQ-OC total score showed weak correlations with anxiety, depression, and stress in the nonclinical sample. In this study, the patients with OCD tended to represent moderate correlations between AAQ-OC and depression or anxiety. These findings suggest that AAQ-OC may be related not only to OC symptoms but to general negative affect (e.g., depression and anxiety).
The criterion validity of the AAQ-OC was examined by assessing the relationships between the AAQ-OC scores and OC symptoms in the OCD patients. First, the OCD patients had higher scores of AAQ-OC and its two subscales than the university students. Second, all AAQ-OC scores were significantly correlated with the DOCS total scores. This demonstrates the conceptual connection between AAQ and OC symptoms. Third, in terms of the specificity of the OC symptoms, the total and subscale scores of the AAQ-OC were correlated with only the two symptom dimensions in DOCS, namely, responsibility for harm and mistakes and unacceptable thoughts. Although the original study reported significant correlations between AAQ-OC and all four OC symptom dimensions, the correlation coefficients tended to be higher in the same two symptom dimensions (0.40 and 0.64 of αs) than the other contamination and symmetry dimensions (0.23 to 0.28 of αs) in the student sample. Wetterneck et al. [42] suggested that EA, as measured by the AAQ-II, is more related to autogenous obsessions (i.e., thoughts of being responsible for harm to others) than reactive obsessions (i.e., thoughts of contamination). This connection is thought to be more pronounced in the AAQ-OC, which focuses on intrusive thoughts and was rated by patients with OCD in this study.
This study has some limitations that need to be acknowledged. First, considering that our subjects were medical students, caution is needed when generalizing our findings. Specifically, it should be noted that medical students as non-clinical participants were included solely on the basis of students’ self-reports, without conducting direct interviews using diagnostic tools. It is also possible that medical students possess psychological characteristics themselves (e.g., higher level of OC symptoms) that differ from those of the general population [43]. Further research with a higher number of representative samples including various age groups is warranted. Another limitation is that the sample size of the OCD patients is insufficient to confirm the factor structures. Further studies with greater statistical power may be able to more affirmatively explore them. Finally, the translation process of the AAQ-OC has few limitations. It did not involve any back-translation or cross-verification. Although two experienced psychologists and one psychiatrist independently translated and arrived at a final version through consensus, there was no opportunity to compare preliminary test results or validate the readability of the translated Korean expression with a Korean language expert.
In conclusion, the AAQ-OC is considered a reliable and valid instrument for a Korean sample. Future studies are warranted to further investigate the efficacy of this measure in ACT.
PDF Links
PubReader
ePub Link
XML Download
Full text via DOI
Download Citation
Print
