Overtreated?

This week on Book TV, Shannon Brownlee discusses her book Overtreated: Why Too Much Medicine is Making Us Sicker and Poorer. She describes unnecessary care as "treatment that doesn't have a lot of evidence behind it." For example, many men receive a PSA test but she says that there's not really a lot of evidence about whether or not this will reduce mortality.

Her thesis is that valuable resources are being wasted by the medical profession on these treatments that aren't supported by evidence.

I don't find it very surprising that there are regions of the country in which people are receiving more "care" than they need, although Brownlee seems to be. Hot spots for unnecessary care include LA, Miami, New York, New Jersey, and Boston.

She says the central problem is that "we pay doctors to do more, not to do better." That is, we pay for visits or treatments, not for outcomes. There's an incentive for doctors and hospitals to "sell" us the goods and services they produce. Maybe we need a No Patient Left Behind to measure the effectiveness of doctors and hospitals. If they don't improve, they lose their accreditation.

She claims that overtreatment is a leading cause of death: 30,000 people are dieing each year as a result of unnecessary treatment.

But she admits that the greatest amount of unnecessary treatment occurs in areas of medicine where knowledge of good treatment levels is lacking, where we don't really know what's right and what isn't.

My questions:
1) How do we know that a treatment is unnecessary unless is has been used enough to provide the evidence that she seems to want. Brownlee wants treatments that have been shown to be effective, but how does that happen without different treatments being tried?
2) What are her criteria for unnecessary? She says that almost all of the over-treatment occurs in areas where there is no consensus so doctors are left to use their discretion. How does she know that a treatment is unnecessary when doctors don't?
3) How does she propose that we collect evidence without trying treatments that some might consider unnecessary?

If over-treatment occurs in areas where knowledge is lacking, how do we gain the knowledge we need to avoid the over-treatment without studying the populations who receive the treatment as compared with those who don't?