Yep
Yep
It is good, aint it? Can only get better as technologies to deposit cells where we wish are developed, and also (this is one I can't wait for!) minature submarine robots are created to do surgery. Combine the two, and you have something that can be inejcted into the bloodstream to repair your damaged valve 100 cells at a time!
Anyway, back onto the topic. The stats should ideally be adjusted to show survival rates vs
prior predicted survival rate
For example, if a surgeon saved himself for the really tricky stuff - prior estimated chance of survival 50% - but had an 65% survival rate, then that surgeon is clearly top class
Better, in fact, than one who does easier ops which had 95% "prior to op survival probability" and who managed to get 90% of his patients to survive.
As with all stats, they need interpreting as a spread of data.
Patients should be divided into categories before op representing survivial probability - bands of 5%. So you'd get the 30-34%, 35-39%, 40-44% etc all the way up to 94-99%. This prior prediction should be made by several doctors - you don't want individuals "doing a scotty" and reducing survival estimates for a patient so as to look like a miracle worker!
For each band, it would be useful to then look at both relative numbers of mortality vs survival, and relative ratio.
The ratios are the most telling if there are sufficiently large numbers,
eg. if in the 60-64% band, 20 died but 200 survived, that averages out to 200/220= 91% sucess rate for a 60-64% band (such a surgeon would be a confirmed genius).
However, if the same surgeon had done 2 in this band, and both survived, that's a seemingly excellent 100% score - but with only 2 tries, it's early days.
Similarly, if a surgeon had done only 1 in the 30-34% band and that individual died,
you'd think it was a high failure rate, but 1 could very easily be a statistical blip.
If he had worked his way through 100 operations in the 25-29 band with 30% success, and another hundred in the 35-40% band with 45% success, you'd realise that it was indeed a statistical blip, and that his success with similarly risky bands indicated he was better than average!
So there we have it, folks. In order to properly evaluate a surgeon, you would need to look at the entire spread of data. Trouble is, I don't know if they record this prior estimate of success anywhere. If they don't, then their data is seriously lacking, and the public can only be confused.
Anyone for Maths? Innit great!