The practice of using outcomes data to indicate the quality of hospital stroke services is based on the assumption that, provided adjustment is made for differences in the characteristics of patients admitted and the play of chance, then most of the residual variation is likely to reflect differences in the quality of care. The paper by Lingsma1 and colleagues, published in this issue of J Neurol Neurosurg Psychiatry (see page 888), helps to show the extent to which this thinking is wrong and ultimately unhelpful in driving forward improvements in stroke care.

The study tested an idealised system of comparing the outcomes of patients admitted to 10 Dutch hospitals and …