How useful are non-random comparisons of outcomes and quality of care in purchasing hospital stroke services?

We performed a prospective observational study using a stroke register, case-note review and survey of carers with 6 months of follow-up in two adjacent health districts in East London. District 1 was a teaching district and had no special stroke service; District 2 had a comprehensive stroke service comprising stroke unit, review of all stroke admissions and community follow-up. Three hundred and sixty-one consecutive patients with stroke admitted to hospital and 103 carers were surveyed at 6 months from admission using the Royal College of Physicians (London) Stroke Audit standards. We also assessed mortality, disability, perceived health, mood, and satisfaction with services 6 months after stroke, carer mood, perceived health and satisfaction with services. The standard of care was below that set by the Royal College of Physicians of London in both districts and there were no significant differences between the districts in age-standardized mortality at 1 and 6 months, Barthel score, extended ADL score, Geriatric Depression score, Nottingham Health Profile score and patient satisfaction with services at 6 months. Carer outcomes did not differ between districts. Service costs, particularly costs of rehabilitation services, were much lower in District 2. A comprehensive district stroke service was not associated with major differences in patient outcomes or standards of care. This may have been because the non-random nature of the comparison meant that the patients differed in other ways than in the nature of treatment. Caution is needed when using these techniques in making purchasing decisions.