What we need is long term surveillance studies of cohorts of patients who have been screened and diagnosed, including random samples of those who screened negative and were diagnosed negative as well as the positives. This is currently infeasible in the US. It's infeasible because following patients over time is too expensive, because the US lacks a coherent national electronic health record system. Things are, unfortunately, not too much better in most Canadian provinces. What should be happening is that these data ought to be accumulated as an automatic by-product of good clinical care. What's needed for that is a rational health information technology infrastructure — see here. For better and too often for worse, all of our problems are connected.