Background. Research on the dimensional structure and reliability of the Hospital Anxiety and Depression Scale (HADS) and its relationship with age is scarce. Moreover, its efficacy in determining the presence of depression in different patient groups has been questioned.

Methods. Psychometric properties of the HADS were assessed in six different groups of Dutch subjects (N = 6165): (1) a random sample of younger adults (age 18–65 years) (N = 199); (2) a random sample of elderly subjects of 57 to 65 years of age (N = 1901); (3) a random sample of elderly subjects of 66 years or older (N = 3293); (4) a sample of consecutive general practice patients (N = 112); (5) a sample of consecutive general medical out-patients with unexplained somatic symptoms (N = 169); and (6) a sample of consecutive psychiatric out-patients (N = 491).

Results. Evidence for a two-factor solution corresponding to the original two subscales of the HADS was found, although anxiety and depression subscales were strongly correlated. Homogeneity and test–retest reliability of the total scale and the subscales were good. The dimensional structure and reliability of the HADS was stable across medical settings and age groups. The correlations between HADS scores and age were small. The total HADS scale showed a better balance between sensitivity and positive predictive value (PPV) in identifying cases of psychiatric disorder as defined by the Present State Examination than the depression subscale in identifying cases of unipolar depression as defined by ICD-8.

Conclusions. The moderate PPV suggests that the HADS is best used as a screening questionnaire and not as a ‘case-identifier’ for psychiatric disorder or depression.