Hospital Anxiety and Depression Scale (HADS): validation in a Greek general hospital sample

The study was performed by the Second Department of Psychiatry at the 'Attikon' General Hospital in Athens. Three groups participated; one group of elderly inpatients in the Internal Medicine and Surgical Departments, one group of outpatients waiting to be examined in the Internal Medicine Outpatients Department and one general population group which was assessed by mail (this group is referred to as 'controls'). A total of 521 participants completed the study; 246 patients and 275 controls.

The HADS is a self-report rating scale of 14 items on a 4-point Likert scale (range 0–3). It is designed to measure anxiety and depression (7 items for each subscale). The total score is the sum of the 14 items, and for each subscale the score is the sum of the respective seven items (ranging from 0–21). It is worth noting that items referring to depression symptoms that describe somatic aspects of depression (e.g. insomnia and weight loss) are not included in the scale. The Greek translation by 'nFer Nelson Publishing' (The Chiswick Centre, 414 Chiswick High Road, London, UK) was used with permission.

The Beck Depression Inventory (BDI) was used to measure depression. It is designed to examine both somatic and cognitive aspects of depression. The BDI is a 21-item self-reporting scale that has been used, apart from its original purpose (assessment of the severity of known depression), for screening purposes. The Greek version has been translated and validated previously [20] and has been widely used to date.

The State-Trait Anxiety Inventory (STAI)[21] developed by Spielberger is used to measure anxiety. It is a 40-item scale made up of two 20-item subscales (one state and one trait), and has been widely used to asses anxiety not only in clinical but in non-clinical samples. The STAI (Form X) has been translated and validated in Greek [22]. The BDI and STAI were administered to patients only.

All of the scales used are self-rated and were administered by five of the researchers. The aim was that the examiners would interfere as little as possible in the patient's completion of the scales. For homogeneity of the results, the scoring of the scales was performed by only one of the researchers.

The following tests were used for the statistical analysis of the data: the Pearson Chi-square test was used for comparison of percentages, and thee Student t test and one-way analysis of variance (ANOVA, with Bonferroni correction) for comparison of means of variables. Correlations were tested by the Pearson r or the Spearman r_s coefficients, depending on whether the variables were normally distributed or not. The psychometric properties of the HADS were evaluated by the following: construct validity was assessed by inter-item and inter-scale correlations and exploratory factor analysis (principal components with varimax rotation). The intraclass correlation coefficient was used to explore the test-retest reliability. The internal consistency of the scale was calculated with Cronbach's alpha coefficient (minimum acceptable value for alpha was 0.7). Concurrent validity was assessed by calculating correlations between the HADS and the BDI (the gold standard for depression) and the STAI (the gold standard for anxiety). Statistical analysis was carried out using SPSS (Version 11.0) for Windows (SPSS Inc., Chicago, IL, USA).

The demographic data for the participants are listed in Table 1. The patients group consisted of two subgroups: 150 inpatients (elderly inpatients of age > 65 years attending the Internal Medicine and Surgical Departments) and 96 outpatients of all ages waiting to be examined in the Internal Medicine Outpatients Department. The differences in age and sex among the three groups were of statistical significance.

The mean scores for HADS, BDI and STAI are listed in Table 2. Comparing patients as a whole to the controls (t test) showed that patients had greater values as assessed by HADS, HADS-D (depression) and HADS-A (anxiety) with a level of statistical significance p < 0.001. The same finding was generally observed when inpatients and outpatients were compared to controls separately and to each other (using ANOVA after Bonferroni correction); inpatients and outpatients showed higher scores than controls. It is worth noting that although inpatients and outpatients had similar scores on the HADS-D, outpatients showed higher scores on the HADS-A.

Zigmond and Snaith [1] have suggested two cut-off scores for detecting depression and anxiety that have generally been used in most studies; scores of 8 to 10 = doubtful cases, and scores of 11 and higher = valid cases. Bjelland et al., in their review, report that most studies conclude the cut-off score of 8 in general population and in somatic patients samples is correct [7]. The same score has been recently proposed by Olsson et al. for outpatients [23]. In our patient sample (inpatients and outpatients), the prevalence of doubtful cases was 14.2% for depression and 16.3% for anxiety. The prevalence of valid cases was 13.4% for depression and 15.1% for anxiety. These percentages for patients only were 22.3% for doubtful cases for depression and 17.4% for anxiety and 22.7% for valid cases for depression and 21.9% for anxiety.