GPs Ability Unaided and Help Question Alone Appears As Good As 3 Questions
I have great admimiration for the new study by Bruce Arroll and colleagues, in part due to the excellent sample size and in part due to the authors attempt to compare different test methods to GP’s detection rate alone. It is surprising how few studies of putative diagnostic tests include a comparison with “diagnosis-as-usual” or in other words the clinicians routine ability.
However, I would like to raise a few points that are either unclear or deserve comment. The authors state that GP’s sensitivity of 79% is an improvement on that previously reported. The figure quoted is an improvement, but I suggest that the comparison with older studies is unfair for the following reason. In this study “The patient showed the general practitioner his or her written responses to the screening and help questions. The general practitioners could ask any questions.”. The GP arm of the study was therefore screening tool plus clinical ability rather than unaided clinical ability alone. This also may explain why the overall performance of the GP arm (by PPV to NPV differential or by likelihood ratio) appears comparable or even better than the 3-question test. Indeed, if the authors maintain that GPs were not influenced by the availability of the 3-question test, then the possible value of the test appears mathematically negligible.
Can the authors also explain why the help question alone was quite successful? The authors state that the help question was phrased “is this something with which you would like help?” Surely the offer of help question was conducted only after a patient had read question 1 and 2 in which case the “diagnostic test” is not really the help question alone but a summation of Q1 2 and 3 with post-hoc separation of items. Taken at face value, in performance terms the help question ALONE does as well as any other combination suggesting that a single item may suffice after all. But herein lies a conundrum because in order to extract maximum diagnostic accuracy from a single item question, such a question may need to be phrased “Do you want help today or in the future for often feeling low, down or hopeless during the last month or in the same period have been bothered by little interest or pleasure in doing things (footnote state yes if any condition applies).”
Clearly, there are problems in clinical practice summing various stems into a long tortuous question. It will be interesting to see if a meta-analysis currently being conducted reveals the benefits of the ultra-short (1, 2 or 3 question) approach to diagnosing depression.
Competing interests: No competing interests