Lancet. 2012 Apr 14;379(9824):1389-90; author reply 1390. No abstract available.
The results of the Quality in Acute Stroke (QASC) study (Nov 12, p 1699)1 are encouraging to those of us who have long fought to improve standards in basic interdisciplinary acute stroke care, yet we should be cautious about attributing the improvements in outcome to the team-building intervention, however well delivered. As pointed out in the accompanying Comment,2 there is little evidence from previous studies that physiological interventions can account for anything like the effect size claimed.
The purpose of randomisation in a therapeutic trial is to produce treatment groups that differ only by chance, but in a single-blind cluster trial, differences in the process of patient selection within the intervention and control clusters could unbalance the final comparison groups. In the QASC trial, 32% of patients initially assessed in the acute stroke units randomised to receive the intervention were enrolled, compared with 22% in the control units. In intervention centres, 20% of patients were labelled as “not stroke” versus 35% in control units. If there were no systematic differences in selection, a discrepancy as big as this would occur by chance once in a million times, so we cannot assume that the final patient groups were balanced. Indeed about 350 more patients who would have been rejected in the control stroke units might have been recruited in the intervention centres. If their outcome was better than average—as might be expected for patients with “stroke mimics”—this would totally disrupt the trial results.
Perhaps the proportion of non-strokes assessed in the intervention centres was the same in the pre-intervention phase, but without knowing this we might speculate that staff were reluctant to label patients as “non-stroke” once nurse-led care protocols had been started.