The EFA resulted in a new model with only minor alterations in regard to the original factor structure of CAQ [4], in the removal of three items. The original model showed questionable fit in the CFA, while the models with reduced amount of items showed a better fit to the data. Adding Reassurance Seeking as a fourth factor did not improve the fit of the model, nor was this structure supported by exploratory analysis. Additionally, the CAQ showed acceptable psychometric properties in a Swedish population of post-MI patients.
Psychometric validation
The full questionnaire exhibited excellent internal consistency suggesting that the Swedish translation of CAQ still measures a singular coherent structure. The new subscales also demonstrated a good internal consistency, except for Attention which was just below the desired cut-off. Test-retest reliability was good for both the full scale as well as the individual subscales which also indicates that the questionnaire is stable over time. The convergent correlations with PCL-C and HADS-anxiety, and the fact that the correlations with depressive indexes were lower, suggest that the CAQ still measures symptoms of anxiety. These findings point to that the CAQ works psychometrically well in a Swedish post-MI population.
Exploratory factor analysis
In the EFA, four items (items 4, 10, 13 and 18) were initially cross-loaded between the factors Fear/Worry and Attention. Three of them were removed and one lost its cross-loading (item 18) during this process, and was thus retained in the model.
Item 4 was originally part of the factor Attention. However this item rather seem to tap into sleep disturbance than that of hypervigilance or monitoring. This could be a reflection of the common occurrence of sleep disturbance within anxiety disorders [29]. However, it could also be that this item represents some other aspect of anxiety. It also describes a sudden onset of chest pain/discomfort. A sudden onset of discomfort is also a common symptom of Panic Disorder, which an early study suggested CA to be a variation of [30]. Possibly this item describes a shared trait with Panic Disorder. Furthermore, this items has been found in varying factors or been deleted in previous analyses [5, 8, 10]. Another item that demonstrates a similar quality (item 3) is included in the model generated in the EFA. When rotating the 4-factor solution these two items loaded on a fourth factor, but as only two items in a factor is unacceptable [26] this model was discarded. As such, it is unclear whether these two items are part of the Attention aspect of CA or if they describe something else. It would be interesting for future studies to investigate the role of sleep disturbance and symptoms of panic in relation to CA.
Item 10 describes a tendency to worry, even when evidence against the need for worry has been demonstrated. This item demonstrated the strongest cross-loading in the EFA and was removed first. This suggests that its content relates both to Fear/Worry and to Attention. Both worrying and focused attention are cognitive processes, and while fear and worry are closely related concepts, so is hypervigilance and worry. As some of the items in the factor Fear/Worry describes a more emotional aspect of anxiety, rather than cognitive, this could be an explanation of the shared relation with the more cognitive factor Attention. However, in all previous studies of EFA on the CAQ this item has had a single salient loading on the factor describing Fear/Worry [4,5,6,7,8, 10]. It could be further theorised whether the questionnaire would benefit from a clearer distinction between cognitive and emotional aspects.
Item 13 concerns both worry and chest pain. However, the worry does not focus on whether the chest pain is dangerous or not, but rather concerns if others could be trusted. As such, it should be reconsidered whether this is a mark of CA or something else, and if it should be permanently removed from the questionnaire.
It is worth considering that item 10 and 13 were both removed in this study and in the study by Dragioti et al. [8]. This could be an indication that these items suffer from issues with translation. However, this idea is not supported by the study by Sardinha et al. [10], where neither of the four removed items were 10 or 13.
The factor Fear/Worry was reduced by two items from the original model. However, it is still the largest subscale and this modification should only have minor implications. This notion is supported by the correlation with the total measure of CAQ, suggesting that it continues to be a valid and central part of the concept CA. Fear/Worry also correlated more strongly with other measures of anxiety, which could indicate that it is also closer to the general concept of anxiety than the other factors.
The factor Avoidance has included the same items in every study of CAQ to date, making it the most robust of all factors. Still, it could be criticised for its lack of specificity. The items don’t specify the reason for avoidance of physical activity. An individual who avoids exercise may get high scores even if they do not avoid it for anxiety-related reasons.
Similar to Fear/Worry, the factor Attention also has fewer items but remains otherwise unchanged. Whether or not waking up at night is a part of attention, the remaining items seem to be a valid part of CAQ. Still, much like the factor Avoidance, these items could benefit from being more specific in regard to what drives the increased attention. Additionally, this factor demonstrated some problems with internal consistency being below the desired cut-off [26]. This is possibly the result of it being the smallest factor, as fewer items makes internal consistency go down. However, in this case, the advantages of adhering to theory and not underfactoring outweighs the disadvantages of a slightly lower internal consistency (.67 < .70).
Confirmatory factor analysis
In summary, the model with best fit was the 10-item version by Dragioti et al. [8], followed by the 3-factor solution generated in the EFA. The models that showed inferior fit to the data were the 1-factor solution, the three different 4-factor solutions [5,6,7], the reduced model by Sardinha et al. [10] and – interestingly – the original model by Eifert et al. [4].
The 3-factor solution by Dragioti et al. has eight removed items and its superior fit to the data could be a reasonable suggestion for a short-form of the CAQ. The superior fit of the 3-factor solution over that of the original version suggests that the validity of the questionnaire could benefit from a removal or change of some of the items. The three four-factor solutions did not demonstrate an adequate fit to the data, adding doubt to the suggestion of the addition of Reassurance Seeking to the questionnaire, although it is a theoretically sound suggestion.
Strengths and limitations
A limitation is that the study population included patients with a recent MI. As such, the factor solution generated in the EFA may not be generalizable to other populations. However, in regard to the psychometric properties, the instrument has been found valid and reliable in groups with non-cardiac chest pain as well [5, 7,8,9].
As demonstrated in Table 1, the methodologies of previous studies are varying. Some of them use principal components analysis or principal axis factoring to extract the factors while others perform both EFA and CFA in the same sample. Additionally, they rarely report if the analyses have been modelled for ordinal data. As such, this is the first study of the CAQ that has combined both EFA and CFA and followed proper statistical considerations.