[epicureanideal]

But if the cost becomes so high that people simply go without the devices, is that really a good tradeoff? Maybe a device with 1 in 10,000 failures is better than no device at all, for some subset of illnesses?

Also, I suspect this is FUD by some entrenched medical devices companies who know how to play the game of complying with standards. Anyone here ever gone through compliance testing with software? I've seen systems pass that are worse than what I would have just thought up in 30 minutes and coded up for fun. It's just a warm fuzzy feeling.

[candiodari]

That's one thing. But there's other problems:

1) certification that takes 10 years (more like 3, but ...) means you, at the very best, have 10 year old technology

2) Certified at a higher level. That can for various reasons be very different from reliable at a different level. For instance, circumstances change or knowledge advances.

For instance certification tends to take "proof" that something works. Yet the most reliable robots are pretty bad hardware, with the ability for multiple components to do the same job "most" of the time. Such a device, despite being much more reliable, is disqualified a priori in nearly all certification processes I've seen.

(needless to say, every time the things they take as proof tend to be ... less than proof)

Plus we've all been in a company having this discussion. "For the price of this one 'reliable' server we could have 20 normal ones, and they would be a hell of a lot more reliable together. Hell, just give me 3 of the cheap ones and I'll make it more reliable". And we all know what the boss's answer and the resulting reliability was.

3) You assume no regulatory capture (or outright dishonesty on the part of government employees and/or lawmakers)