We presented results of the psychometric properties of the German CAPS-5 in routine clinical settings. The descriptive statistics for the items of the German CAPS-5, the subscales of the German CAPS-5, and the overall CAPS-5 were investigated. In addition, the internal consistencies, test-retest reliability, interrater reliability, construct validity, and diagnostic accuracy of the German CAPS-5 and its subscales were also investigated. The German CAPS-5 was a structured diagnostic interview with good to excellent internal consistency, test-retest reliability, interrater reliability, construct validity, and diagnostic accuracy. Most of our hypotheses [see 23] were supported, and the results of this study support the value of the CAPS-5 for diagnosing PTSD in clinical and scientific settings.
The descriptive statistics of the items of the CAPS-5 showed relatively high mean scores. The participants showed a mean score of > 1 for most items. More than half of the participants fulfilled the criterion of a score ≥ 2 for 14 of 20 items of the CAPS-5. The lowest means were observed for items B3 (“dissociative reactions”), D1 (“inability to remember an important aspect”), and E2 (“reckless or self-destructive behavior”). These results align with the study of Weathers et al. [13], who also reported the most inferior means for these items. These items also showed relatively low mean scores in other studies and were excluded from confirmatory factor analyses [13, 15, 18, 67]. The mean sum scores of the German CAPS-5 and its subscales were similar to the results of other studies [13, 15, 67]. Most participants (≥ 74.8%) fulfilled the criteria of at least one subscale of the CAPS-5 according to the DSM-5 (Table 3). However, in the current sample, the diagnostic criteria of PTSD according to the DSM-5 were only fulfilled by 65.9% of the participants. These results showed that the current cohort had an overall high symptom severity and was highly mentally stressed, which are basic assumptions for investigating a diagnostic tool. However, these results represent an apparent variability and a relatively high percentage of TN cases, facilitating a differentiated analysis of the diagnostic accuracy of the German CAPS-5.
The internal consistencies of the German CAPS-5 were excellent and acceptable to good for the subscales of the German CAPS-5. Thus, the picture regarding the internal consistencies of the German CAPS-5 is consistent with studies regarding the original version [13] and other translations of the CAPS-5 [15, 16].
The other study on the German CAPS-5 showed comparable internal consistencies [19]. However, a closer look at the internal consistencies showed good coefficients for the subscales of cluster B (re-experiencing) and cluster D (cognitions and mood symptoms). In contrast, the internal consistencies for the subscales of cluster C (avoidance) and cluster E (hyperarousal) were acceptable. These differences in the internal consistencies for the subscales were also reported in other studies of the CAPS-5 [13, 15, 16, 19]. The cluster C subscale’s relatively lower and acceptable internal consistency is due to the number of items in this subscale. Scales with only two items tend to show a lower internal consistency [68]. The heterogeneity of the items of the cluster E subscale and the potential loading of these items on different factors [13, 67] could explain the relatively lower internal but acceptable consistency of the cluster E subscale.
The results of the CFA overall showed adequate to good model fit of the German CAPS-5 factorial structure according to the DSM-5 criteria of PTSD. Thus, the instrument can be used in clinical routines and research to diagnose PTSD. These results can also be seen as an indicator of sufficient construct validity [69].
The test-retest and interrater reliability of the German CAPS-5 and its subscales revealed excellent reliability coefficients. These coefficients were comparable to the results regarding the test-retest reliability and interrater reliability of the CAPS-5 [13, 15, 16] and slightly higher than the German CAPS-5 [19]. However, the number of participants who participated at both measurement points was relatively low (30. 2%). Only 39.2% of all interviews were independently rated for investigating interrater reliability. Thus, further studies regarding the test-retest and interrater reliability of the German CAPS-5 are needed to provide more robust results.
Regarding the construct validity of the German CAPS-5, the results confirmed the hypotheses regarding the convergent validity of the German CAPS-5 and its subscales. The German CAPS-5 and its subscales showed strong correlations with other measures of PTSD (PCL-5, IES-R) and cPTSD (SkPTBS), as well as measures of psychopathological constructs that are highly associated with PTSD (e.g., anxiety or depression). Measures of more divergent psychopathological constructs (e.g., worries or borderline personality disorder-associated thoughts and feelings) were also significant but less strongly associated with the CAPS-5 scores. Indeed, the hypotheses regarding the divergent validity of the German CAPS-5 and its subscales were only partially supported by the results. As expected, parts of the socially desirable responses (impression management) were not significantly associated with the German CAPS-5 sum score and its subscales. The self-deceptive enhancement aspect of the socially desirable responses revealed significant and negative associations with the German CAPS-5 sum score and its subscales. However, contrary to our hypotheses, all subscales of the SCL-90-R showed significant associations with the German CAPS-5 sum score and its subscales. These results underscore the high psychopathological distress of the current sample. In retrospect, the choice of instruments for investigating the divergent validity could have been better. For instance, other study results showed significant positive associations between PTSD and somatization [13, e.g., 70]. Future studies should focus more on measures that are less associated with psychopathology (e.g., personality measures).
The ROC analyses showed that the German CAPS-5 is a diagnostic instrument with good to excellent accuracy. Furthermore, these results showed that the German CAPS-5 has high sensitivity and specificity and ensures that persons with PTSD are diagnosed correctly; people without PTSD can also be adequately identified. Thus, the German CAPS-5 is an adequate diagnostic interview for the scientific, clinical, and expert assessment context.
Finally, the results investigating a cut-off score of the German CAPS-5 sum score showed an optimal score of ≥ 40 with both the PCL-5 and the IES-F as external criteria. To our knowledge, this is the first time a cut-off score was calculated for the CAPS-5. It can guide using the CAPS-5 in the scientific, clinical, and expert assessment context. However, this cut-off score should be used cautiously until it is replicated in further studies. Importantly, only self-rating questionnaires were used as external criteria, which could bias the resulting cut-off score. The mean PCL-5 score in this study was higher than the mean score of the CAPS-5 (PCL-5: M = 42.2 vs. CAPS-5: M = 31.36). Higher mean scores of the PCL-5 versus the CAPS-5 were also formerly reported [24]. This leads to the assumption that self-rating questionnaires of PTSD result in higher scores than structured diagnostic interviews. Thus, further investigations of the CAPS-5 cut-off score with another structured diagnostic interview as an external criterion could help answer this question. However, more detailed studies comparing the CAPS-5 and the PCL-5 are needed.
Challenges and limitations
This study does have some limitations. First, the current sample showed relatively high symptom severities in the CAPS-5 and its subscales. Accordingly, the results regarding diagnostic accuracy could be biased due to high scores. Furthermore, the sample consists of a relatively high percentage of male participants with a military background. Thus, the results should be generalized with caution. Also, the analyses regarding the divergent validity are limited. Therefore, this part of the construct validity still needs to be answered. Finally, we did not include data about comorbid diagnoses or control for other diagnoses that could interfere with PTSD symptoms (e.g., borderline personality disorder). Comorbidity should be considered as a potential covariate that could be relevant for the investigated associations. This should be noted in future studies.