Clinical Study

BJC Open article

British Journal of Cancer (2008) 98, 1934–1943. doi:10.1038/sj.bjc.6604396 www.bjcancer.com
Published online 27 May 2008

Are one or two simple questions sufficient to detect depression in cancer and palliative care? A Bayesian meta-analysis

A J Mitchell1,2

  1. 1Department of Cancer & Molecular Medicine, Leicester Royal Infirmary, Leicester LE1 5WW, UK
  2. 2Leicester General Hospital, Leicester LE5 4PW, UK

Correspondence: Dr AJ Mitchell, Department of Cancer & Molecular Medicine, Leicester Royal Infirmary, Leicester LE1 5WW, UK. E-mail: alex.mitchell@leicspart.nhs.uk

Received 4 March 2008; Accepted 31 March 2008
Advance online publication 27 May 2008

Top

Abstract

The purpose of this study is to examine the value of one or two simple verbal questions in the detection of depression in cancer settings. This study is a systematic literature search of abstract and full text databases to January 2008. Key authors were contacted for unpublished studies. Seventeen analyses were found. Of these, 13 were conducted in late stage palliative settings. (1) Single depression question: across nine studies, the prevalence of depression was 16%. A single ‘depression’ question enabled the detection of depression in 160 out of 223 true cases, a sensitivity of 72%, and correctly reassured 964 out of 1166 non-depressed cancer sufferers, a specificity of 83%. The positive predictive value (PPV) was 44% and the negative predictive value (NPV) 94%. (2) Single interest question: there were only three studies examining the ‘loss-of-interest’ question, with a combined prevalence of 14%. This question allowed the detection of 60 out of 72 cases (sensitivity 83%) and excluded 394 from 459 non-depressed cases (specificity of 86%). The PPV was 48% and the NPV 97%. (3) Two questions (low mood and low interest): five studies examined two questions with a combined prevalence of 17%. The two-question combination facilitated a diagnosis of depression in 138 of 151 true cases (sensitivity 91%) and gave correct reassurance to 645 of 749 non-cases (specificity 86%). The PPV was 57% and the NPV 98%. Simple verbal methods perform well at excluding depression in the non-depressed but perform poorly at confirming depression. The ‘two question’ method is significantly more accurate than either single question but clinicians should not rely on these simple questions alone and should be prepared to assess the patient more thoroughly.

Keywords:

depression; diagnostic validity; meta-analysis; sensitivity