One of Howard et al’sReference Howard, Ryan, Trevillion, Anderson, Bick and Bye¹ conclusions in their paper is that their data confirm that the Whooley questionsReference Whooley, Avins, Miranda and Browner² ‘are a useful tool for case identification in early pregnancy’ (by midwives in routine clinical settings). This conclusion was principally based upon the obtained positive likelihood ratio in their study (5.8 for depression, anxiety and other related disorders) and high specificity (0.96), providing therefore a reasonable positive predictive value (0.66). Also, however, the authors explain that the Whooley questions had a low sensitivity of just 0.23. This means that they actually missed almost 80% of the women with these mental health disorders.

We feel that it is difficult to imagine a clinical service agreeing that an instrument that misses almost 80% of people with a condition could be considered ‘useful’, and is ‘a quick method for identifying that a mental disorder may be present’, despite the other receiver operating characteristic values reported for the questions.

We accept that the issue of what values, or combination of values, of a test's various screening metrics (for example positive likelihood ratio, sensitivity, specificity, positive predictive value) are indicative of a ‘good or clinically useful performance’ can be difficult to decide, is open to debate and will vary depending upon context. And we appreciate that Howard et al are clear in their reporting of their data, including the low sensitivity values and possible reasons for these, which they say include that the questions may not have been asked in a consistent and/or correct way by the midwives.

We would, however, question their main conclusions, these being that the obtained data ‘confirm … that (the Whooley questions) are a useful tool for case identification’ (p. 54) and that ‘(the two-item Whooley questions) can (therefore) be asked routinely by midwives when women attend for their routine antenatal booking appointment’ (p. 55). Rather, we would suggest that a different conclusion may be more appropriate, given their findings, this being along the lines of: screening positive on the Whooley questions, while being indicative of a reasonable likelihood of a woman having a mental health difficulty, needs to be tempered by the fact that most of the women with such disorders were not in fact detected by the questions in this study. These data therefore indicate that services would be unwise to implement these questions, in the way conducted in this study, if they consider that missing around 80% of women with a mental health difficulty is problematic.