Using the additional "help" screening question for depression; how can it improve the specificity without affecting the sensitivity?
Arroll et al  have assessed the usefulness of adding a “help question” to the previously tested  two screening questions for depression. Patients were considered as positive if they answered “yes” to either one of the two screening questions and “yes” to the help question. Authors concluded that adding the help question increases the specificity without changing the sensitivity. Looking at their results more closely, we would like to understand how authors came up with what appears to us to be inconsistent.
Using two tests combined with the condition “AND” should increase the specificity and optimally conserve the lowest sensitivity of each test taken alone. Arroll et al. reported a sensitivity of 74.5% for the help question alone. This is inconsistent with their reported sensitivity for the combined three screening questions which is of 96%.
Looking more closely at the published table 1 , it is possible to calculate that only two false negative results for the two-screening questions was observed. From table 2 we also learned that 12 depressed patients answered negatively to the help question. This should mean that at least 10 and at most 12 depressed patients answered positively to the two screening questions without asking for help. Combining both the screening questions and the help question, authors should therefore find between 33 and 35 true positive patients. Table 1 however reports that all 45 patients responded positively. We therefore wonder what happened to the 10 patients who were supposed to have answered negatively to the help question but positively to the two screening questions? We are also surprised that depressed patients apparently never consider they do not need any help.
We therefore believe Arroll et al’s reporting to be potentially misleading. If the sensitivity of the help question alone was to be confirmed, this could have important clinical implication if practitioners were to falsely acknowledge this test to conserve a high sensitivity. By using these published results, GPs could underestimate the number of false negative and not investigate depression for patients needing care.
1. Arroll B, Khin N, Kerse N. Screening for depression in primary care with two verbally asked questions: cross sectional study. BMJ. 2003 Nov 15;327(7424):1144-6.
2. Arroll B, Goodyear-Smith F, Kerse N, Fishman T, Gunn J. Effect of the addition of a "help" question to two screening questions on specificity for diagnosis of depression in general practice: diagnostic validity study. BMJ. 2005 Oct 15;331(7521):884.
Competing interests: We have tested the validity of the "help" screening questions for depression in patients followed-up in general practice (SODA study). Our results, we are about to publish, are in contradiction with those of Arroll et al. We are therefore trying to understand why.
Competing interests: No competing interests