This study from the US aims to improve the so called ‘grand round’ in hospitals.[1] The intervention, co-designed with patients, aims to make the patient a more active participant in the ward round. I loved ward rounds, especially when I was a medical student in Johannesburg. I enjoyed the learning experience and, depending on the specialist in charge, the theatricality of the occasion. The specialists seldom talked as though the patient was not there and I think patients also enjoyed the rounds. But it would be wrong to say that the patient was centre stage. The intervention developed in this paper tries to change that. For instance the ward round starts with the patient’s voice – the patient is asked to comment at regular intervals. Patients receive a summary of the conversation. The intervention consisted of training sessions that included role-play. The intervention was implemented over seven sites and each site made specific adaptations to the intervention.

The intervention was studied in a before and after design. It was associated with small, but statistically and (I would say) clinically, significant improvements in both patient and carer satisfaction. Satisfaction with doctor and nurse communication improved. The investigators did not only measure satisfaction and communication. They also measured error rates and adverse events. Overall errors were unchanged but errors resulting in harm were reduced following the intervention. But then unavoidable adverse events also declined. Co-design and implementation of a family friendly way of communicating with families of children in hospitals.

So, what to make of the above American educational intervention? I have four concerns. First, as the authors themselves state, it is hard to draw cause and effect conclusions from a before and after study. However, I am not impressed with the authors attempt to sweep this concern aside by pointing out that RCTs may be difficult in improvement research. Researchers can deploy stronger alternatives to before and after studies in addition to standard cluster RCTs. Time series, non-randomised designs with contemporaneous controls [2] and step-wedge cluster RCTs,[3] for instance. Second, while the effect on satisfaction seems robust, the fact that non-avoidable adverse events declined in parallel with adverse events caused by error is evidence in support of the hypothesis that the changes seen in adverse event are part of a temporal trend rather than a result of the intervention. Third, the theory that this intervention should reduce adverse events due to error is not strong and has to rely on some fairly nebulous reasoning about ‘accident chains’. However, the effect on communication has high face validity; in Bayesian terms one might expect an enthusiastic prior for an effect on satisfaction. There must be greater uncertainty regarding the finding of a reduction in adverse events. Fourth, the measurement of errors and adverse events was not entirely third party, but included data collected by participants meaning that it could have been ‘reactive’ (there could have been in interaction between the phase of the trial and the measurement obtained).

Would I have recommended publication had I been a reviewer? Certainly, but I would have urged greater caution regarding the putative causal effect on errors.

Richard Lilford, ARC WM Director

References:

1. Khan A, Spector ND, Baird JD, et al. Patient safety after implementation of a coproduced family centered communication programme: multicenter before and after intervention study. BMJ. 2018; 363: k4764.
2. Benning A, Dixon-Woods M, Nwulu U, et al. Multiple component patient safety intervention in English hospitals: controlled evaluation of second phase. BMJ. 2011; 342: d199.
3. Hemming K, Haines TP, Chilton PJ, et al. The stepped wedge cluster randomised trial: rationale, design, analysis, and reporting. BMJ. 2015; 350: h391.