Issues of Reliability/Validity in Depression Diagnosis/BDI

524 individuals from inpatient, outpatient and community settings at 5 different sites (inter-rater reliability) were interviewed using DSm criteria then interviewed again 6 months later (test-retest reliability). Results showed that inter-rater reliability across different sites was 'fair to good' but 6 month test-retest reliability was 'fair' for dysthemia and 'poor to fair' for MDD.

Zanarini et al (2000) - inter-rater of correlation 0.8 for MDD and test-retest of .61 with one week between diagnosis sessions.

They suggest that, due to the 5 out of 9 symptoms having to be present, when the severity of a disorder is such that it is at the diagnostic threshold, a one item disagreement could make the difference between MDD and a less severe illness being diagnosed.

That 83.8% of patients admitted with depression had the same diagnosis when they left hospital.

McCullough et al (2003) compared 681 outpatients with various types of depression and found considerable overlap in symptoms, responses to treatment and other variables, thus it could be argued that it is difficult to justify distinct forms of depression.

Van Weel-Baumgarten - because GPs may focus on the patient's background as opposed to the presenting symptoms.

Sanchez-Villegas et al administered the standard interview from DSM-IV-TR to 62 participants with depression and 42 non-depressed participants. They found that 42/62 (68%) were diagnosed with depression and 34/42 (81%) were diagnosed without depression. This study needs careful interpretation because 19% could have been missed by earlier screening or the previously depressed patients could have been having a 'good day'.

Beck et al (1996) tested 26 outpatients at 2 therapy sessions one week apart from responses. There was a correlation of .93 - a significant level of test-retest reliability.

Visser et al (2006) assessed 92 patients with Parkinson's Disease (PD) for depression using the DSM diagnostic criteria and the BDI. In part 2 of the study, 60 PD patients completed the BDI again as part of a postal survey, producing a correlation of .88. This suggests the BDI is a reliable measure.

Content validity refers to whether the items in a test are representative of that which is being measured. BDI is considered to have high content validity as it was contructed as a result of a consensus among mental health clinicians concerning symptoms found among psychiatric patients.

Construct validity asses the degree to which a test measures an internal construct or variable. Research (eg. Caldwell and Redeker, 2008) has shown that BDI scores are positively correlated to related symptoms such as anxiety, loneliness, disturbed sleep patterns and alcoholism.

Concurrent validity is a measure of the ectent to which a test agrees with already existing standard ways of assessing the characteristic in question. Research (eg. Beck et al, 1988) has consistently demonstrated concurrent validity between the BDI and other measures of depression, such as the Hamilton Depression Scale.

Karasz gave a scenario describing depressive symptoms to diverse cultural groups in New York - 36 South Asian immigrants and 37 European Americans. The former identified the 'problem' in social and moral terms, with treatment suggestions emphasising self-management and non-professional help (STIGMA). The latter emphasised biological explanations.