No further research needed : Abandoning the Hospital and Anxiety Depression Scale (HADS)

Cosco and colleagues [this issue] provide a well done and transparently reported systematic review of the Hospital and Anxiety Depression Scale (HADS) literature of the past decade. They conclude that the underlying structure of the HADS is inconsistent across samples and highly dependent on the statistical methods used to establish that structure. The implication is that the HADS is not a dependable means of differentiating anxiety and depression for the purposes of assessing the absolute or relative levels of these variables. These results can also go far in explaining the confusing difficulties that have arisen in research concerning use of the HADS as the first stage of two-stage screening procedures for depression and anxiety disorders or case identification purposes. Comparisons between results obtained with the HADS in a particular sample and the available literature can prove bewildering. It is quite uncommon to validate the presumed structure of cleanly separated anxiety and depression subscales, but there is little consistency whether the best fit is a unitary single factor solution, a two-factor solution without a clean separation of anxiety and depression, or a three factor solution. And then there is the anomalous finding in a large Danish study [1] that breast cancer patients had lower anxiety and depression scores on the HADS than women drawn from the general population. Do we accept that finding, dispute it on the basis of flaws we can identify in the study's methodology, or doubt the basic validity of the HADS? Fundamental problems with the HADS are unintended consequences of deliberate decisions made in its construction. Cosco and colleagues provide only an easily missed clue to a basic problem by noting in passing, “Item 7 was found to anomalously load in 20 studies, indicating that it is a particularly poor item.” The item, “I can sit at ease and feel relaxed,” is one of six positively valence items, with greater endorsement representing less anxiety, but the response key is both reversed (1 equals definitely; 4 equals not at all) and has different anchors than the response key (1 equals not at all; 4 equals most of the time) for the item just prior to it. While such reversals of wording and varying response keys were intended to avoid effects of a response style, they are disorienting, and unless patients are particularly vigilant, they will miss the changes in direction of the items and scoring.