Case fatality rates after admission to hospital with stroke: linked database study

BMJ 2003; 326 doi: (Published 25 January 2003) Cite this as: BMJ 2003;326:193
  1. Stephen E Roberts, statistician,
  2. Michael J Goldacre (michael.goldacre{at}, director
  1. Unit of Health-Care Epidemiology, Department of Public Health, Institute of Health Sciences, University of Oxford, Oxford OX3 7LF
  1. Correspondence to: M J Goldacre
  • Accepted 24 June 2002

Population based mortality for stroke has declined in most Western countries during the past few decades.1 This is probably because of a decrease in both the incidence of and case fatality from stroke over this period.24 Routine statistics generally do not provide long term trends in case fatality, and few studies have looked at differences in case fatality rates between hospitals.5 We used routine hospital data, which had been linked to mortality data in the former Oxford health region of England, to study case fatality rates after admission for stroke and to compare rates between different periods and different hospitals.

Standardised case fatality rates for patients admitted for stroke, by hospital of admission during 1978–87 and 1988-97*

View this table:

Methods and results

We calculated case fatality rates by dividing the number of deaths from all causes after admission by the number of admissions for stroke and multiplying by 100. We compared case fatality rates for deaths anywhere at 30, 90, and 365 days after admission and for deaths in the hospital admission for acute stroke. Following convention, we termed the latter “in-hospital” deaths. We analysed only emergency admissions for which stroke was the principal diagnosis. We used ICD-9 (international classification of diseases, 9th revision) codes 431–434 and 436 (ICD-10 codes I61-I64). Data had been collected by the region from 10 acute hospitals for 1978–87 and 12 for 1988–97 (with follow up during 1988 and 1998, respectively). Some hospital trusts had merged by 1998.

During the 20 years of data collection, 34 080 people were admitted to hospital with strokes; 18 126 (53.2%) were women. The mean age was 73.7 (SD 12.3) years.

In 1978-87, case fatality rates for all hospitals combined were 39.7% at 30 days and 56.9% at one year (table). In 1988-97, the corresponding figures were 32.9% and 48.9%. The table shows that case fatality rates were lower for the region's teaching hospital (I) than for all other hospitals combined. Significant differences were also seen between individual non-teaching hospitals. Variation between the teaching hospital and individual non-teaching hospitals reduced over time.

Differences in case fatality rates during the first 30 days accounted for the differences between hospitals and most of the difference over time (table). For 1988-97, the low case fatality rate for in-hospital deaths in hospital III within 30 days, calculated without data linked to death certificates, was as high as that in other hospitals when linked data were used. Otherwise, case fatality rates for in-hospital deaths were good predictors of hospitals' relative rankings when rates were calculated with data linked to death certificates (Spearman's rank correlation coefficients between in-hospital deaths and deaths anywhere at 30, 90, and 365 days were 0.83, 0.86, and 0.88 in 1987–98 and 0.93, 0.81, and 0.71 in 1988-97).

Age and sex standardised case fatality rates at 30 days in 1978–87 and 1988–97 were 33.2 and 24.7 in people <75 years and 27.5 and 19.8 in those <65. Differences between hospitals for case fatality rates in patients <75 and <65 years were similar to those found for patients of all ages, although with diminishing statistical power not all comparisons reached significance (see tables A-C on


Case fatality rates after hospital admission for stroke were high: about half of all patients died within one year. Differences in case fatality rates over time, and between hospitals, might be explained by differences in the case mix and particularly by differences in the severity of stroke and the extent to which patients were managed at home rather than in hospital. For these reasons, differences are hard to interpret. If the differences can be attributed to standards of care, rather than case mix, their impact is greatest in the acute phase of care: the reductions over time, and the differences between hospitals, were predominantly seen in fatality rates within 30 days of admission.


We thank Ruth M Ripley for advice about statistical methods.

Contributors: SR designed the study, analysed the data, and wrote the manuscript. MG designed the study and wrote the manuscript. MG and SR are guarantors for the paper.


  • Funding SR receives funding from the Department of Health as part of its funding for the National Centre for Health Outcomes Development (the views expressed in this paper are those of the authors and not necessarily those of the Department of Health). The Unit of Health-Care Epidemiology is funded by the South East Regional Office of the NHS Executive.

  • Competing interests None declared.

  • Embedded Image Extra tables appear on


  1. 1.
  2. 2.
  3. 3.
  4. 4.
  5. 5.
View Abstract