Intended for healthcare professionals


Regional differences in outcome from subarachnoid haemorrhage: comparative audit

BMJ 2004; 328 doi: (Published 20 May 2004) Cite this as: BMJ 2004;328:1234
  1. P Mitchell, senior lecturer (Patrick.Mitchell{at},
  2. T Hope, consultant2,
  3. B A Gregson, principle research associate1,
  4. A David Mendelow, professor1
  1. 1 Department of Neurosurgery, University of Newcastle upon Tyne, Newcastle General Hospital, Newcastle upon Tyne NE4 6BE
  2. 2 Department of Neurosurgery, Queen's Medical Centre, Nottingham NG7 2UH
  1. Correspondence to: P Mitchell
  • Accepted 15 January 2004


Subarachnoid haemorrhage affects 10 per 100 000 UK residents a year. More than half the cases are fatal, and serious disability is common among the survivors. Modern management has reduced death and disability by about 30% compared with 30 years ago.1 We conducted a prospective collaborative audit. No other units were involved.

Participants, methods, and results

The Newcastle neurosurgery unit serves a population of 2.4 million2 and has 78 adult beds. It is one of five units in the British Isles that are deemed by Safe Neurosurgery 2000 to have enough beds for their populations.2 The Nottingham unit serves a population of three million2 and has 36 beds. It is one of the three most under-resourced units in the British Isles.

We audited all patients presenting with a subarachnoid haemorrhage confirmed on computed tomography or lumbar puncture between 1992 and 1998. Patients' demographic and presenting clinical data were recorded during their admission. Outcome was recorded at clinic follow up, by postal questionnaire, or telephone and was obtained for 1822 of the 1851 cases in the study. The shortest interval between presentation and follow up was 6 months, and the average 12 months; these were similar for both units.

Full time research assistants were employed in each unit to collect the data. After careful and in-depth work, important errors were found and corrected in a quarter of cases. Funding was not available after 1998.

Good recovery and moderate disability (according to the Glasgow outcome score3) were classed as favourable outcomes; severe disability, vegetative state, or death were unfavourable.

We used the χ2 test to compare the unfavourable outcome rates of the two units and the time periods (up to or after 1995), and we used logistic regression to include age and presenting condition (according to the World Federation of Neurological Surgeons' (WFNS) grading4).

The rate of an unfavourable outcome was 35% in Newcastle and 19% in Nottingham. This difference was significant (P < 0.0001). The results in Newcastle worsened over time. These differences disappeared when the effects of age and presenting condition were included. Newcastle operated a less selective admissions policy than Nottingham because it did not have the deficiency of beds that Nottingham had. Between 1992 and 1998 Newcastle became progressively less selective, admitting more patients with a poor WFNS grading and more older patients. The table shows the independent effects of age, WFNS grade, and neurosurgery unit.

Logistic regression of outcomes

View this table:


The observed difference in outcomes between the units does not necessarily reflect the quality of care given, but rather it can be explained by case mix and the impact of the availability of resources on admission criteria. This only became evident through careful and specifically funded audit. The use of the crude results to guide clinical governance and policy making would have been highly pernicious.

It is easy to apply methods of performance analysis to medicine. The problems are not a lack of such methods but rather a lack of appropriate processes for collecting data and a poor understanding of likely confounding factors and how to measure them. Political motivation leads to pressure to produce easily accessible results. This approach is considerably worse than doing nothing and should be resisted. Collection of data on factors that may influence outcome is a prerequisite of the statistical comparison of results between units. These include, but are not restricted to, the quality of care given. Had Newcastle come under pressure from clinical governance to improve results without this being appreciated, the service it offers would have been compromised.


  • This article was posted on on 22 April 2004:

  • We thank R P Sengupta, C J Gerber, A Jenkins, A Gholkar, N V Todd, B White, P Byrne, I Robertson, R Ashpole, J Firth, J Stevens, and P Moody for their care of some of the patients in the study.

  • Contributors PM conceived the study and wrote the paper. BAG maintained the database and did the statistical analysis. TH and ADM conducted the study in Nottingham and Newcastle respectively. PM is guarantor for the paper.

  • Funding The study was funded by grants from the Northern Brainwave Trust and the Newcastle Neurosurgery Foundation.

  • Competing interests None declared

  • Ethical approval At the time of our study, audits did not require approval from ethics committees but did require registration with the Data Protection Agency, a requirement we fulfilled.


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