In terms of MRIs, you have some odd effects with pricing, particularly when the MRI is seated inside of a massive tertiary care center instead of a standalone facility. If you look at the pricing discrepancies, it is almost always related to getting the MRI done at an academic medical center versus one of the ambulatory care centers. The problem is actually pretty simple: hospitals are terrible at cost accounting and totally game it. Instead of taking the leasing costs over the expected uses of the machine, adding in time for the technician and a bit of a real estate or facility charge, they allocate hospital costs (from all departments/overhead) to services based on their expectations on what they can charge. Michael Porter and his staff at HBS are looking at this right now.
Further complicating MRIs (I'm not sure if this is included in the study's cost estimate) is that radiologists essentially operate in a cartel fashion. They are rarely, if ever, employed by the hospital (like most doctors), but band together and set outrageous prices for reading images. Radiology, despite being non-patient facing and limited liability (they render opinions to other doctors, not patients), is one of the most lucrative medical trades. Eventually, traditional radiology should give way - either through disruption (overseas or computers) or by other doctors simply saying why the heck should a radiologist get money for reading an image I can read myself and will then have to intervene on anyways?
Startups have emerged in price/transparency space (e.g. Castlight Health) and will hopefully start to put pressure on hospitals/physicians to actually compete with one another and bring down costs. Since they have so much local market power, there is only so far an insurer can go without owning an entire market.