Letters

Midwife led debriefing to reduce maternal depression

BMJ 2000; 321 doi: https://doi.org/10.1136/bmj.321.7274.1470 (Published 09 December 2000) Cite this as: BMJ 2000;321:1470

Effectiveness of individual midwives is unclear

  1. J M Bland, professor of medical statistics (mbland{at}sghms.ac.uk)
  1. St George's Hospital Medical School, London SW17 0RE
  2. Centre for the Study of Mothers' and Children's Health, School of Public Health, La Trobe University, Carlton, Victoria 3053, Australia

    EDITOR—I should like to comment on the randomised controlled trial of midwife led debriefing to reduce maternal depression by Small et al.1 The first randomised clinical trial was of a drug treatment, streptomycin for pulmonary tuberculosis,2 which has provided the model for clinical trials ever since. In a drug trial we are not usually concerned with who is giving the drug because the effect of the drug itself is being measured. The treatment is impersonal, and we should be justified in assuming that the effect of a drug given by one person will be the same as it would be given by another.

    When we carry out trials of more personal treatments, however, as in the trial by Small et al, we should be aware that the treatment given by one operator might not be the same as that given by another. Surgeons are not all equally skilful, for example. So if we were to carry out a trial comparing two surgical techniques the surgeons would be a non-random sample from the wider population of surgeons. This might not be too bad if each surgeon carried out both techniques because there would be some sort of balance. If different groups of surgeons carried out each technique this would not be so. We could randomise surgeons to treatments, but this would probably be difficult to achieve. We could, and I think should, take surgeon variation into account—for example, by multilevel modelling.3 The inevitable result would be to make confidence intervals wider and P values bigger, as happens when cluster randomised trials are analysed correctly.4

    In the trial by Small et al the situation is even more complicated. The intervention is debriefing by a midwife—a very personal intervention. It is easy to believe that the individual skills of midwives in this complex task vary greatly. Clearly, the mothers in this trial are a sample from which we want to draw some conclusions about mothers in general. But surely the midwives are a sample too. We are asking whether debriefing by midwives is helpful. The two midwives here have somehow to represent the effectiveness of midwives everywhere. It may be that these particular midwives are not very good at debriefing rather than that debriefing is ineffective. Half of us are below average, after all.

    It is difficult to see how we could analyse the trial to take the midwife variation into account as such variation exists only in one arm. Although stratification by midwife is mentioned, the midwife's intervention is received by women in only one arm of the trial. Stratification therefore cannot allow for variation between midwives.

    I cannot criticise researchers for not applying a statistical technique yet to be invented, or at least to be noticed by myself. Medical research is in its infancy. There are many unanswered questions and, I suspect, many that are yet to be asked. We do not really know how to do it yet. It will be an interesting challenge to find out.

    References

    1. 1.
    2. 2.
    3. 3.
    4. 4.

    Authors' reply

    1. Judith Lumley, professor,
    2. Rhonda Small, research fellow (R.Small{at}latrobe.edu.au)
    1. St George's Hospital Medical School, London SW17 0RE
    2. Centre for the Study of Mothers' and Children's Health, School of Public Health, La Trobe University, Carlton, Victoria 3053, Australia

      EDITOR—Bland's point about practitioner variables in implementing interventions that are not drugs is important. Our approach to implementing the trial of debriefing after operative birth was to identify the key elements, summarised by Wessely et al as identifying emotional responses, encouraging their expression, and legitimising them.1 We also sought to define the necessary skills—active listening; reflection; encouraging the expression of women's experiences; accepting distress, anger, and pain; being able to name and normalise the experience; and being able to avoid offering solutions. We then selected two midwives who had these skills to a high degree so that our trial of debriefing would give the intervention the best possible chance of showing whether it was effective in reducing depression. During the run-in period all debriefing sessions were taped (with the women's written consent) to assess the quality of the intervention against the key elements. We saw the trial as a phase III trial, in the language of the recent Medical Research Council paper,2 and had foreshadowed in the grant application subsequent work to develop a manual and training programme for midwives if the trial were effective, to be followed by well-designed cluster randomised dissemination trials (phase IV).

      Bland is misleading when he writes, “We are asking whether debriefing by midwives is helpful.” We were indeed asking that, and women responded overwhelmingly that it had been helpful or very helpful. This is one of the principal intentions of debriefing, to reduce the immediate psychological distress after a traumatic experience, so women's responses were reassuring. The trial was not designed to answer the helpfulness question but to see whether debriefing could prevent the subsequent development of depression. It did not—despite the more than average skills of the two midwives.

      Bland's critique (the practitioners weren't up to the task) has often been used to explain away trial findings in perinatal work (antenatal cardiotocography, routine antenatal ultrasonography) in which interventions in widespread use have performed poorly within trials. One contribution would be for journals to require (and publish) enough detail about the implementation of complex interventions for readers to make informed judgments. As for practitioner and institutional variation, there seems to be no alternative but a wider use of cluster randomisation, despite the difficulties and challenges.

      References

      1. 1.
      2. 2.
      View Abstract

      Sign in

      Log in through your institution

      Subscribe