Authors' reply

We are in agreement with Langford & Badenoch’s general comment on the need for holistic assessment in psychiatry. It is currently established that patients’ views, and especially quality of life (QoL) measures, should supplement the usual indicators of quality in healthcare.¹ However, we are doubtful about the relevance of their criticisms.

Langford & Badenoch denounced the following sentence: ‘Global satisfaction was significantly higher in the QoL feedback group […] compared with the standard psychiatric assessment […] and QoL assessment groups’.² This assumption was derived, however, directly from our results (i.e. the proportions of very satisfied patients were 73% in the QoL feedback group, 45% in the QoL assessment group and 68% in the standard group). The comparison performed using a chi-squared test was statistically significant (P = 0.025), allowing us to state that global satisfaction significantly differed between the three groups. As we have written in our Discussion, this finding did not prohibit us from suggesting that integrating QoL assessment and feedback with standard psychiatric assessment seemed relevant or that priority should be given to strategies that implement QoL measurements in routine practice.

Moreover, this assumption was in agreement with our study design (i.e. three arms) and the sample size calculation performed for this design. However, we recognise that multiple treatment arms in randomised controlled trials (RCTs) are sources of misunderstanding,³ especially because there are several possible comparisons.⁴ Langford & Badenoch re-wrote our primary outcome for a two-arm RCT as follows: ‘level of patient satisfaction in the QoL feedback group compared with standard psychiatric assessment’, implying pairwise chi-squared tests. However, our primary outcome and analysis were defined in accordance with the primary objective integrating the three-arm design. The objective was to globally determine the ‘positions’ of QoL feedback, QoL without feedback and the control group with respect to their relationships to satisfaction; we did not aim to question the relevance of using the QoL measure (which is already recognised in the literature) in 2×2 comparisons between the different arms. The primary criterion was thus analysed using a global chi-squared test, determined a priori; it was not analysed using pairwise chi-squared tests (as recommended by Langford & Badenoch), which were not planned and for which the alpha error risk was not controlled. It is also widely recognised that bias may be introduced if decisions regarding data analysis are driven by the data.³

Langford & Badenoch also claim that ‘The conclusions drawn by the authors, that their findings “provide strong support […]” and that “priority should be given to strategies to implement QoL” […] seem particularly unfounded’. We did not conclude with these two sentences, which were taken from the Discussion (the function of which differs from the conclusion⁵) without heed to what was written before and after. In fact, we stated that ‘Priority should be given to strategies to implement QoL measurements in routine practice’, especially because ‘clinicians did not optimally use the QoL feedback’ and ‘obtaining QoL data in an efficient, real-time manner is difficult and rare in clinical practice’.

Last, we were pleased to read that Langford & Badenoch felt that the existence of a nocebo effect in the QoL assessment group with feedback was the most salient finding, as this was an issue that we extensively discussed in our manuscript.

In conclusion, it is important to insist that any result reported in a study must be interpreted considering the objective and the design of the study and, more globally, in the context of current scientific knowledge. In agreement with Karl Popper, we believe that scientific objectivity is based on intersubjectivity and the ethics of discussion. We hope that our answer will close the gap between our scientific work and the understanding of Langford & Badenoch.