In the present study, the HADS was tested on a sample of Greek general hospital patients (inpatients and outpatients), and controls from the community. The HADS appears to have high internal consistency; Cronbach's α value for the total HADS was 0.884.
The Greek version of the HADS seems to be bidimensional; thus, it could be considered that the two subscales of the HADS measure anxiety and depression independently. The HADS and its two subscales showed high correlations with the gold standards that were used to measure depression (BDI) and anxiety (STAI).
As expected, patients in general appeared to be more depressed and anxious than the subjects from the general population (controls). Outpatients seemed to be more affected than inpatients in presenting anxiety. This could be attributed to the outpatients' concern about hearing bad news as a result of their consultation. By contrast, inpatients are in a way settled in the 'safety' of the ward.
The psychometric properties of the G reek version of the HADS are similar with those of other languages [8, 9, 12–14]. The HADS generally appears to have a high internal consistency; Cronbach's α values ranged from 0.870 to 0.885 for all the items of the scale. The item-subscale correlations were moderate to high; from 0.608 to 0.762 for the anxiety items, and from 0.540 to 0.804 for the depression items. The HADS appears to be bidimensional as in the original study by Zigmond and Snaith . All items but one (item 14) loaded in the appropriate factor. Similar findings for one or two, but not always the same, items have been reported not loading to the appropriate factor in many other studies. Most studies report the two factor structure of the HADS [2, 7], though there are some exceptions that have reported one factor , or three [10, 17, 24], or even four factors . The study of Mykletun et al. with 51,930 participants, which is the largest of all in the literature, concluded a bidimensional structure for the HADS was correct .
The HADS property of consisting of two independent subscales can also be shown by its correlations with the scales that were used as gold standards for depression and anxiety. The HADS/anxiety correlated highly with the STAI and the HADS/depression correlated highly with the BDI. There are some studies where the total HADS showed greater correlations than its subscales with BDI and STAI correspondingly [8, 12, 16, 25], but in our study the total HADS correlated to almost the same levels as its corresponding subscales with BDI and STAI (a little higher than HADS/depression with BDI, and a little lower than HADS/anxiety with STAI). It is worth noting that the two HADS subscales had a moderate correlation (0.559) to each other. This could be expected, bearing in mind that depression and anxiety show great comorbidity, especially in general health care settings .
There are some limitations in our study; there were gender and age differences between the groups, and the test-retest reliability was carried out with the control group only.