Reliability testing of the hospital anxiety and depression (HAD) scale in the English, Maltese and back-translation versions

This paper discusses the translation of the hospital anxiety and depression (HAD) scale (Zigmond, Snaith, Acta Psychiatr. Scand. 67 (1983) 361) into the Maltese language. The HAD scale is a well-validated and reliable measure of anxiety and depression originating in the United Kingdom. To ensure accuracy in the translation of the tool, the translation process was based on the Maltese Translation Guidelines issued by Chetcuti (Tahrig ghall-ezamijiet tal-Malti. Biex taghmel traduzzjoni tajba, Veritas Press, Malta, 1975, pp. 9-10) and those of Sechrest et al. (J. Cross-Cultural Psychol., 3 (1) (1972) 41). The composition of the Maltese language, which is Semitic in nature, is described and examples from different languages are given. The Maltese version of the HAD scale will facilitate the investigation of mood states in future studies on patients. Reliability testing of the HAD scale is then performed on the English version, the Maltese version and on the back-translation. The test-retest reliability of the three versions is examined using cross-tabulations of each item (pre values with post values), all of which gave highly significant values of chi-squared (p = 0.0000). These cross-tabulations also yielded high values for the Kappa measure of reliability and for Spearman's coefficient of correlation (kappagreater than or equal to0.8 and rhogreater than or equal to0.9 for most items of anxiety and depression in all the three versions). The internal consistency of the three versions is also examined statistically using Cronbach's alpha and factor analysis. Both the anxiety and the depression subscales in the Maltese version can each be parsimoniously described by one factor, Each subscale therefore has a dimensionality of one. This explains why good levels of internal consistency are observed for the Maltese translation of the HAD scale (alpha = 0.79 for the anxiety subscale, 0.70 for depression, and 0.85 for both subscales together). This validates the Maltese version of the HAD scale, which can thus be used safely in future studies on Maltese patients. The anxiety subscale is also unidimensional in the original and in the back-translation, and also showed satisfactory values for Cronbach's alpha (0.73 and 0.74, respectively). Unfortunately, for the depression subscale, correlations between its items were rather low, thus leading to low alphas (about 0.45 and 0.51, respectively) for the internal consistency of this subscale in these versions. (C) 2002 Elsevier Science Ltd. All rights reserved.