Scientific Method (Re: Hurry up and Wait)

[Timothy J Coats (SURG) 7728]

> For instance:  How does one PROVE, scientifically that helicopter based EMS 
> does or does not work, when one can't fly one in a lab, therefore simulating 
> in a realistic fashion, all the variables involved in the gathered data?
> 
> Or, does this merely become an exercise in semantics and theory?
> 
> Looking for serious answers please.

Kat,

The first stage is to do a sample size calculation. Assume that for 
severe injury there is a 25% decrease in death. Assume that the 
mortality in the control (non-helicopter) group is 20% (so the 
mortality in the helicopter group is estimated at 15%).

Assume a beta value of 0.8 (probability of finding a significant 
difference is it actually exists) and an alpha value of 0.05 (probability 
of the difference found being due to chance alone).

Feed these figures into your stats package, and the result is that 
950 controls and 950 subjects are required.

If the actual decrease in death is less than 25%, a correspondingly 
larger sample will be required.

In other words, any study of an EMS trauma helicopter that 
contains less than 2000 patients will not be able to detect a 25% 
decrease in mortality! No study has had this number of patients 
(and none has given an explicit sample size calculation) so all 
studies to date by definition are prone to a Type II error - that is 
failure to find a difference when one actually exists.

I would suggest that the minimum clinically significant effect of a 
helicopter system would be a 5% reduction in mortality. To detect 
this difference would entail a study with 50 000 cases. (Remember 
that it took international studies with tens of thousands of patients 
to prove that thrombolysis was of benefit following AMI).

Tim.





Timothy J Coats MD FRCS FFAEM
Senior Lecturer in Accident and Emergency / Pre-Hospital Care
Royal London Hospital, UK.