thanks for this editorial. I agree with you that much of the talk on HIT successes may indeed be marketing pitch, and I completely agree that "it's been tried before. [...] 8 years ago [and didn't work]. What has changed since then? Why do we expect RHIO's will work now?". Indeed the nagging interoperability problems are still a big practical concern. And I completely agree that if RHIOs turn into mini-standard organizations it is a lot of effort wasted.

However, I disagree with how you move on from there. You are saying that (a) HL7 doesn't cut it and that (b) the only thing that works is VA's Vista and Kaiser's Epic implementation.

Let me take it one by one:

(a) HL7 is our only realistic chance, in fact, without a single comprehensive interoperability standard that specifies all relevant data conceptualizations and representations including data-structures and terminology as well as remote transactions the idea of RHIOs will never materialize. Without such standards you will never create ubiquitous HIT infrastructure, since no one has the power to decree a single system (and IMO noone should try that, because it would be stifling innovation and would ultimately harm HIT cost & quality). So, to connect the islands, a standard is the only way. And HL7 is the only standard that begins to have adoption basis, relevant domain content, and addresses the broad spectrum of issues.

(b) VA's Vista and Kaiser's Epic installation are not the only things that work. For one, both VA's Vista and Kaiser's Epic are simply a single large institution's EMR system, a large island of data to use your terms. In addition, Kaiser's Epic is operational only since very few years, so that's not a great example.

When it comes to RHIOs at the very least you need to mention the Regenstrief Medical Record System (RMRS) which connect almost the entire city of Indianapolis and is rapidly growing to cover the whole state of Indiana. As far as I know Salt Lake City is also pretty well connected and so might be Boston. Those regional networks are real and operational for a long time.

I agree that the shallow integration using web-portals are not the same as a deep data integration, and I agree that this can be a big usability drawback (but then the RMRS is not just a surface integration). But this only goes to show that there is still a lot to be done and no silver bullet has yet been found. I think that at the present time anyone stepping up and touting their system as the only credible solution or even the most credible solution is at least premature.