Depression and anxiety frequently occur in RA. This study compared two commonly used screening scales, HADS and DASS, in relation to their measurement range and cut points to consider if prevalence rates may be due to scale-specific case definition.
The individual subscales of depression and anxiety (HADS and DASS) and stress (DASS only) showed a good fit to Rasch model expectations. After some adjustments to the ordering of thresholds and for local dependency for the DASS Anxiety and Depression subscales, interval scale transformation was achieved for all subscales. When pooled together, DASS Anxiety and HADS Anxiety subscales required some modifications to create a unidimensional scale whereas only minor changes were necessary to merge DASS Depression and HADS Depression. This suggests that the anxiety subscales across the two scales may measure different aspects of this construct. This finding has implications for interpretation of anxiety rates across studies and the selection of anxiety scales. It is important to further investigate the clinical relevance of those scales.
When, however, all subscales were mapped together as testlets, they were found to measure a common underlying trait. This is consistent with the testlet design, when used in this fashion, being equivalent to the bi-factor model, where all sub-domains load onto one common domain, while unique components load on to secondary domains . In this instance it was found that the unique variance was low, and that the majority of variance was shared by the five subscales. This finding may be due to the use of testlets, as in removing local dependency those testlets may also remove some discrete uniqueness within each subscale (which when individually tested all met Rasch analysis expectations). It may also be due to an overlap between the two constructs, depression and anxiety, either as a result of their measures assessing general distress  to a degree, or that they may in fact be demonstrating their share common causal factors . Nonetheless, the two constructs are seen as distinctly different, anxiety being motivated by fear while depression by sadness . Finally, the commonality detected may also be due to a high level of comorbidity of depression and anxiety in this study's sample. This comorbidity is particularly significant given its occurrence in the context of a physical comorbidity (RA). It highlights the clinical complexity and associated implications for the treatment of both depression and anxiety in the presence of the physical manifestations of RA. It is also important that further research explores the structure of depression and anxiety constructs in relation to their common underlying construct of psychological distress.
Irrespective of the derivation of the cut points, it was possible through the common metric provided by Rasch analysis, to compare DASS cut points with those of the HADS. Depending upon which cut points were chosen, prevalence estimates could be similar. For example, the HADS 'probable' and DASS 'extremely severe' cut points for depression were found to be closely matched. For anxiety, however, comparable cut points were only found at a lower end. That is, the DASS-Anxiety 'mild' cut point and above matched the HADS-Anxiety 'possible' and above cut point. The DASS-Anxiety 'extremely severe' cut point was higher on the construct than the HADS-Anxiety 'probable' cut point. This may be due to the difference in focus of the anxiety items, with the DASS reflecting feelings of panic and physical symptoms (e.g. dryness of mouth) while the HADS subscale was developed to exclude somatic symptoms and is closely aligned with the generalised anxiety disorder . This is an important finding suggesting that these anxiety scales need further clinical verification to determine which aspects of anxiety they measure, and their concordance with clinical diagnostic assessments.
As such, these findings have important implications for screening and interpretation of research findings across studies. First, this study supports previous findings that elevated levels of both depression and anxiety occur in RA [i.e. ]. Second, it permits a direct comparison of the ability of these two measures to distinguish between depression and anxiety, and their derived estimates of prevalence. The operational ranges of measurement between the scales were found to vary and the cut points needed combined calibration in order to determine appropriate comparability.
The DASS cut points were originally determined by distributional difference of their norming sample, later confirmed in clinical populations [34, 49, 50]. However, in the current study an interesting observation is that the severity-specific DASS cut points for the anxiety and depression subscales all appeared to measure the same point on the underlying metric, irrespective of the specific construct they represent, giving almost identical prevalence for depression and anxiety. In contrast, this was not the case for the HADS. This similarity for DASS cut points requires further investigation.
The study has a number of limitations. Although both scales have previously been evaluated against clinical diagnostic assessment [34, 51] it is not possible to determine which one may do so with more specificity/sensitivity in this study. The study was also restricted to RA population, and it would be useful to see if the same results were to be found in other rheumatic diseases given that it is not clear to what degree depression and anxiety presentation and severity may vary across other rheumatic conditions. While the study was not designed to look at responsiveness of these scales, it has shown that their structures are invariant across countries (UK and Australia) and time (examined in UK only). This is an essential requirement for future studies to assess responsiveness of those scales.
In conclusion, the current study highlights the comorbidity of anxiety and depression with RA. It also highlights the challenges of measuring depression and anxiety given the limitations of assessment measures, and the complexity of those conditions. However, this study shows that at least some confusion can be addressed with a co-calibration of depression and anxiety screening scales to enable the comparison of the estimates across scales and studies. That is, the results for the two scales compared in this study suggest that for depression a comparison between the 'probable' category on the HADS and 'extremely severe' on the DASS may be made, but that for anxiety, a comparison of DASS 'mild' and HADS 'possible' is better aligned. This study also highlights the importance of developing a common metric upon which all scales, and their cut points, may be calibrated.