Background: Audit of adverse outcome might allow identification of substand
ard surgical results, To test this hypothesis statistical modelling was app
lied to two indicator vascular procedures (elective abdominal aortic aneury
sm repair and carotid endarterectomy) with accepted adverse event rates.
Methods: Binomial statistical models for varying adverse event rates were c
onstructed. A power calculation was used in an attempt to predict the case
numbers required to determine substandard results for individual surgeons a
nd vascular units. Two scenarios were considered: first a base adverse even
t rate of 6 per cent and surgical practice with 9, 12 and 24 per cent morbi
dity rates, and second a base adverse event rate of 3 per cent and surgical
practice with 6, 9 and 12 per cent morbidity rates.
Results: A mean of 57 elective abdominal aortic aneurysm repairs and 70 car
otid endarterectomies were performed per annum. The adverse evens rate for
both operations was 4 per cent. Power calculations revealed that 130 patien
ts would need to be studied to detect a surgeon with an adverse event rate
twice 6 per cent and over 280 patients would be required with an adverse ev
ent rate twice 3 per cent. To gather this number of patients 2 years of uni
t data and between 3 and 22 years of individual data would need to be studi
ed for a base adverse event rate of 6 per cent. A base rate of 3 per cent r
equires 7-47 years for an individual and 4-65 years for the unit. With a ba
se adverse event rate of 6 per cent, detection of widely variant surgical p
ractice (four times the morbidity rate as base) requires only 21 procedures
.
Conclusion: Statistical modelling demands assumptions about accepted advers
e event races, confidence criteria and what constitutes substandard results
, Data from large numbers of patients are required even for common operatio
ns with accepted adverse event rates. These data raise serious questions as
to the feasibility of performing clinical governance on the basis of morbi
dity and mortality event rates alone.