I have just come off a rotation on our hospital's arrest team*, and at the last arrest I attended the orthopaedic registrar (a hulking great lump of bloke), had managed to break the patient's ribs and puncture a lung before we got there
*basically a group of acute medics and nursing staff who run to "arrests" in the hospital - though 9 times out of 10 it involves running to the other end of the hospital to try and revive a patient who died overnight and the nurses only realised when they did their morning observations, panicked, and put out an arrest call lol
Yeah I was told it was something like 1 in 40000, but the figures give you an idea that CPR alone won't really save their lives, all your doing is keeping them viable until someone with a defibrilator turns up...
I think the figure is around 5-10% for out-of-hospital arrests, provided effective CPR is started asap. For in-hospital arrests, it's closer to 3% (due to co-existing illness, and increased liklihood of non-shockable rhythms).
Which is why gatting as many people comfortable with performing CPR as possible will save lives, as well as getting more AEDs into public places.
It's not true to say don't start unless you are prepared to continue CPR indefinately - most people doing CPR will become less and less effective as the tire, and the majority cannot maintain effective CPR beyond 3/4 cycles (ie 10-12mins). You should always start CPR if warranted, and continue as long as you can manage, but don't be put off starting by thinking you may not be able to last.
BTW, the old recommendations for the UK were 15:2 compressions:breaths (or 5:1 with two people doing CPR). Currently it's 30:2, though I would not be surprised if it shifts to 60-100:2 with the next set of guidelines in line with recent evidence