I know very little about medical research, so it is almost off-the-cuff that I say this. But I do want to get it written down, anyway.
It seems to me that in some ways the traditional methods of biology research and the more modern data-rich analysis are diverging incredibly. The medical community gives a nod towards 'informatics'...but it is still the case that diagnosis is a black art, and even drug interactions and recovery time expectations are in fact data-rich problems with very little data collection.
What we've done in the computer world is to take a small number of signals (e.g., how much CPU is your server using? What's the ambient temperature in the colo? How slow is your website?) and analyze this small number of items to death. We poll them every second, not every week, and we look for changes and trends. We don't for the most part have complex models, because the things we measure are both extremely complex and give off very quantitative signals.
I suppose I believe that a similar number of signals from the human body (and I don't know which ones will actually work) would yield a similar insight into diagnosis, recovery, and other problems, with a cost structure radically different than anything in the medical field today.
Today, physicians basically have no data or they have expensive data. For instance, they might send you in for bloodwork, which is usually supervised by at least a couple of people, at a relatively large cost "per bit of data obtained". This parallels the traditional research methods I mention above - a double-blind study over ten years with 2000 participants might glean a couple bits of clean data.
But I believe there are basically-free tests that could give lower-level signals about a person's health. Imagine if you could measure body mass by instrumenting a person's bed? Report it back to your PC and have a daily result for your intake/water retention/etc. for the day. What if you could instrument the chemistry/conductivity of the stuff you leave in your toilet? What if your toothbrush could do a pH analysis of your mouth?
These are extremely simplistic ideas, and a single sample of any of them is useless. But what we've learned with understanding how machines work is that the changes in them, and the trends over time, yield a huge understanding of a system.
I believe that the PC could be used to do this securely, and if aggregate data were needed, anonymously with the right software.
Likely there are a zillion people more informed on these topics than I am, but my impression is that this gap in techniques and approaches is widening, and there is a large opportunity to combine some of these techniques in a useful way.