Abstract
Routinely collected outcomes (case fatality by 30 days and discharge home by 56
days) have been used to indicate the quality of hospital stroke services in Scotland
since 1994. However, the validity of these data is in doubt. In particular, it is
difficult to know whether differences in outcome are due to differences in the quality
of care or to inadequate adjustment for casemix, biased measurement or the play of
chance. Also, because the majority of patients survive their stroke, the relevance and
sensitivity of the indicators is limited by the failure to report functional outcome.
It would be useful, therefore, to investigate whether a substantially improved system
(one which adjusts comparisons of outcome for important differences in casemix and
which measures functional outcome at a defined time after admission) would be
routinely feasible and provide valid measurements of the quality of stroke care.
We attempted to address these questions in the context of a study of 2724 patients
with an acute stroke admitted to five Scottish hospitals between 1995 and 1997. We
identified patients using routine hospital discharge information and then identified
cases of acute stroke and data describing casemix and the process of care from the
medical record. We collected case fatality by linkage to death certificate data and
functional outcome and institutionalisation by postal survey at six months. We
adjusted comparisons of outcome for casemix using a set of simple, externally
validated logistic regression models.
The study shows that a considerably improved routine system for measuring outcome
after stroke is a realistic possibility. Specifically, it suggests that the proposed
system for identifying hospital cohorts and collecting casemix data would be
reasonably accurate and that the proposed system for measuring functional outcome,
although compromised by non-response, would not be seriously biased.
Nonetheless, whether these improvements would result in valid measurements of the
quality of stroke care remains uncertain. At best, it appears that a system reporting
case fatality and death or dependency at six months might be sensitive to moderately
large differences in the quality of care. However, there may be alternative
explanations for this finding and the system would certainly fail to identify
opportunities to improve care at the majority of hospitals. The collection of data
describing simple but important aspects of the process of care in addition to outcome
might be preferable and should be investigated.