The implantable cardioverter defibrillator (ICD) is the first line of treatment for life-threatening ventricular tachyarrhythmias (VTa’s) both as primary (patients who have an increased risk of experiencing VTa’s) and secondary prophylaxis (patients who have previously experienced VTa’s) [1
]. Although the ICD is well accepted by the majority of patients [4
], one in four patients experiences psychological distress post implantation, including anxiety and depression [5
]. Distress has not only been associated with the development of posttraumatic stress disorder post implantation [6
], but on its own or in combination with the distressed (Type D) personality profile also with increased risk for VTa’s and even mortality [9
In order to identify ICD patients who suffer from anxiety and depression, it is paramount that we have appropriate and well-validated measures available with appropriate cutoffs and sensitivity and specificity. Generally, a distinction is made between generic and disease-specific anxiety measures [11
]. Examples of generic measures include the Hospital Anxiety and Depression Scale (HADS) [12
], the Spielberger Trait-State Anxiety Inventory (STAI) [13
], the Beck Anxiety Inventory (BAI) [14
], and the Generalized Anxiety Disorder Scale (GAD-7) [15
]. Recently, disease-specific measures were also developed, such as the Florida Shock Anxiety Scale (FSAS) [16
] and the ICD Patient Concerns questionnaire (ICDC) [18
] that tap into anxiety related to ICD shocks.
Since ICD-related anxiety seems to be more prevalent than shock anxiety [18
], this study focused on two widely used generic anxiety scales, the (state) anxiety scales from the HADS and the STAI, referred to as the HADS-A and STAI-S, respectively. Both instruments are generic and commonly used for screening for anxiety in the general population, psychiatric and somatic patients, including cardiac patients and ICD patients. Although they are considered interchangeable, we know little about their equivalence in terms of identifying patients with probable clinical anxiety levels and whether they both reflect a common anxiety attribute. To our knowledge, no information is available whether the cutoffs used for the HADS-A and the STAI-S reflect comparable anxiety levels and thus produce equivalent prevalence rates. Should that not be the case, the choice for either the HADS-A or STAI-S may have consequences for treatment and care if patients are identified with anxiety on one scale but not the other and also for the results of epidemiologic research. Ideally, prevalence rates and treatment decisions should be independent of the specific anxiety measure used. Moreover, the HADS-A and STAI-S are supposed to reflect a common psychological attribute of anxiety
, but they may tap into other psychological subdomains as well, as their items are framed differently. Previous research has shown that in addition to anxiety, the HADS-A also taps into relaxed affect [19
]. The STAI-S on the other hand measures anxiety in terms of the presence versus absence of symptoms, which appear to be separate factors, as determined by factor analysis [20
]. Hence, subtle differences in the meaning of HADS-A and STAI-S scores may explain prevalence differences and the power to predict adverse treatment outcomes.
Hence, in the current study, we examined whether the HADS-A and STAI-S (i) are equivalent with respect to determining the prevalence of anxiety symptomatology as reflected in a crosswalk table for equivalent scores, (ii) reflect variation on a common anxiety attribute in ICD patients.
Ideally, cutoffs of different anxiety scales should reflect comparable levels of anxiety, but this study shows that the cutoffs of two widely scales, the HADS-A and STAI-S, do not (necessarily) yield similar screening criteria even though they are based on external criteria. In particular, the diagnostic cutoff on the STAI-S corresponding to the cutoff on the HADS-A is about six points higher than the conventionally used cutoff of 40 for the STAI-S. Likewise, the concordant cutoff of the STAI-S on the HADS-A is two points lower than the conventional cutoff. This means that using the ‘traditional’ cutoff scores for anxiety the STAI-S would classify patients as anxious who would not be identified as such by the HADS-A. Hence, the STAI-S appears to be less conservative and would thus produce a higher prevalence rate compared to HADS-A. Using the crosswalk table, equivalent cutoffs for different scales can be employed so as to obtain prevalence rates that are less sensitive to the scale that is used.
Our findings further suggest that HADS-A and STAI-S largely tap into the same generic anxiety attribute, although both scales also have their unique parts. Hence, results confirm that both questionnaires reflect somewhat different operationalization of state anxiety. However, if it is the common generic anxiety attribute that explains adverse health outcomes in ICD patients, both scales may be feasible, but results also showed that HADS-A measurements of the general anxiety trait were less precise than STAI-S. Future research should focus on the relationship between general and specific trait variation and the prognostic and screening properties of both questionnaires.
Both questionnaires are widely used as screening tools in research and clinical practice. They have shown to be predictive of adverse health outcomes in cardiac populations [23
] but have also received some critique over the past years. For example, it has been advocated that the HADS should be abandoned due to an unclear latent structure [38
]. However, others have proposed that the scale should at best be restructured [39
], as it is a strong predictor of morbidity and mortality in cardiac patients regardless of its structure [40
]. For the HADS depression scale, alternative cutoff scores have been proposed for cardiac populations. The cutoff scores that should be used for screening purposes seem to vary depending on the type of cardiac disease [42
]. Whether this is the case for the anxiety subscale is still unknown. As compared to the HADS, the STAI has more items and is thus more time consuming to administer and constitutes a greater burden to patients, which is a disadvantage when using it as a screening instrument in clinical practice.
In conclusion, both HADS-A and STAI-S showed good reliability and validity. However, the traditional cutoff scores used to indicate probable clinical levels of anxiety are not equivalent. The HADS-A appears to be more conservative as compared to the STAI-S and will thus produce lower prevalence rates of anxiety. The cutoff scores of the HADS-A seem to vary depending on the type of cardiac disease. Our findings illustrate that studies published to date using the HADS-A and STAI-S to assess anxiety with traditional cutoff scores are not comparable when it comes to prevalence rates. Thus, the prevalence rates should be interpreted in light of the assessment tool used.
For clinical practice, it is important to be aware of the fact that a substantial proportion of patients are classified differently using the HADS-A versus the STAI-S. As the traditionally used cutoff score for anxiety on the STAI-S is less conservative as compared to the HADS-A scale, using this scale will result in higher anxiety prevalence rates. This discrepancy has implications for both clinical practice and research. Hence, prior to implementing one of the scales as a screening tool, it is important to decide whether it is more important to prevent false positives or false negatives. To avoid a high number of false positives, the HADS-A scale should be used. By contrast, use of the STAI-S will reduce the number of false negatives. A crosswalk table allows converting total scores from the HADS-A scale to equivalent STAI-S total scores and vice versa, which may be beneficial when comparing patients who completed different questionnaires (e.g., for meta-analyses).
One of the limitations of this study is the fixed order in which the questionnaires were administered, which can be a confounding factor (e.g., carry over effects, fatigue, motivational problems). Second, the crosswalk table provides a useful tool for comparing scores from different scales, but caution should be exercised when applied to individuals in real-life screening settings. Pseudo-HADS-A scores should not be conceived as substitutes for the STAI-S (i.e., should not be seen as the score a person would have should he/she completed the other questionnaire). In addition, the use of generic anxiety measures in the current population might not have tapped sufficiently into the disease-specific anxieties that might be experienced by patients briefly after implantation. Hence, using disease-specific measures to assess anxiety might provide a more accurate reflection of anxiety symptomatology.
Future research should investigate whether it is feasible to divide the STAI-S into two scales (present and absent) to comply with the increasing demand for brief measures to reduce the burden to patients and in clinical practice while maintaining prognostic power. In addition, research on the predictive elements of the HADS scale should be considered. A bifactor model for the HADS scale has previously been proposed where anxiety and depression are considered as components of a hierarchical structure that includes a general distress factor [43
]. Total scores reflecting the full scope of general distress may be better predictors of poor health outcomes than subscale scores. Furthermore, examining whether different cutoff scores for anxiety should be employed depending on the type of cardiac disease and disease severity is warranted. Finally, results of the crosswalk table seem quite robust (i.e., precise and generalizable), but this needs to be confirmed in future research in other cohorts.
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.