Do studies of the weekend effect really allow for differences in illness severity? An analysis of 14 years of stroke admissions

An increased mortality rate among patients admitted to hospital at weekends has been found in many studies, and this ‘weekend effect’ has been used to justify major health service reorganisation. Most studies have used routine administrative data to adjust for potential confounding factors, and are unable to measure differences in illness severity due to the tendency of patients with milder symptoms to stay home at weekends. We set out to estimate the importance of such unmeasured ‘confounding by severity’, using data from a hospital stroke register. All suspected acute stroke admissions in Gateshead were prospectively registered over a period of 14.3 years, including information on premorbid factors, measures of stroke severity and survival to hospital discharge. We examined whether each factor differed between weekday and weekend admissions, then used logistic regression to estimate the main contributors to variation in mortality rates. Stroke severity, measured by the Scandinavian Stroke Scale (SSS), was significantly greater among weekend admissions and strongly associated with mortality, and after adjustment for SSS score in logistic regression, the weekend effect completely disappeared. By contrast, most indicators of pre-stroke function, comorbidity or stroke type did not differ between weekday and weekend admissions, although some of them had prognostic significance. The sorts of factors measurable in administrative databases only account for a small part of the prognostically important differences in case-mix between weekday and weekend stroke admissions. While increasing weekend staffing might improve adherence to care standards, evidence that it could save lives is unreliable.