What Google doesn't get

This is about the transformation and display of data.

Google Health screenshot from Google BlogoscopedHealth care computing isn't about search.

The winner of the electronic medical record (EMR) wars won't be the vendor who does the least evil. (Picture from an August story in Google Blogoscoped.)

This is about the transformation and display of data. The stuff is easy to find. Doctors and hospitals collect tons of data every day.

The question is, how do you translate from some squirrelly, proprietary scan format to something you can get on a screen somewhere else? And how do you display that data, on that screen, with as much of the original fidelity as possible?

The most promising aspect of the Google announcement yesterday is the involvement of the Cole Eye Institute, which is part of the Cleveland Clinic.

My own eye surgeon dropped his EMR solution years ago because it couldn't deal with new equipment he bought later. Tens of thousands of dollars down the drain and he's not anxious to go back there.

For now most of the publicity over Google and Microsoft involves issues of privacy. How do you give people control of their own data, outside the hospital, and deal with HIPAA concerns?

But I really think that's missing the point. The data patients themselves use is relatively simple and straightforward. The risks are in giving it back to them, not collecting it in the first place. Doctors and hospitals already do that.

Any solution to that problem isn't about privacy anyway. It's about identity. You have to prove yourself to the computer each time you interact with it, and that proof has to be foolproof, or else it's easy for me to become you, or a terrorist to be you.

Those questions generally lie outside the scope of these contracts but the simple answer is to chip everyone. Implant some silicon with enough capacity to read my record, separate the read only from the random access, read it through the skin.

No one is going there, so we're left with other identifiers, all of which can be faked. Although given that we're currently relying on a string of just 9 digits (a Social Security Number) anything we get will be better than what we have.

It's a long road getting from there to here. It's not something you can throw some storage and a few algorithms at. As I wrote yesterday I believe Microsoft has recognized that. Google will, too. But will it be in time?