A Reminiscence on the 10th anniversary of Direct.

This year’s DirctTrust Summit was held virtually on June 9th and 10th. You can follow participation and conversation on #DirectTrustSummit.

This was an interesting Twitter thread of the history on DirectTrust for an original participant – Arien Malec, SVP of product R&D at Change Healthcare, former ONC leader, and current HITAC member.

2/ (Back then, when you said “Hi, I’m from @ONC_HealthIT and we want you to join this new project” the success rate was “100!”
– I think the vendors have a little more initiative fatigue since then).

3/ The roots went back to a 2009 Standards Committee Implementation Workgroup public hearing where the committee heard about successful and failed standardization efforts, and heard from providers that simple things, like referrals, were really hard.

4/ “I’m trying to get paper out of my practice but the paper keeps pouring in” or “I have EHR A, & my colleague has EHR A & I can export a summary & she can import a summary but I can’t get the summary from here to there.”

5/ Out of that, the HIT Policy Committee started the “NHIN Workgroup” whose mission was to expand the NHIN(*) to meet the needs of small practices. That NHIN Workgroup was full of what became heavy hitters. @Farzad_MD (then “special advisor” to @DavidBlumenthal Todd Park etc.

6/ Oh, and lil’ ol’ me. The emerging thesis of the group was: address identity, trust & transport in a way that scaled down to individual practices.
I was my usual me, Sir Talksalot. Had I known that I was arguing with the future National Coordinator & US CTO…

7/ (I’d have probably done the same thing ?) During one of those meetings Farzad & Todd tried to twist my arm to join ONC to take this stuff on, and I resisted their Jedi Mind Tricks. Then came the Markle Foundation Consumer Workgroup meeting – the famous one…

8/ That was the meeting where, among other things, Adam Bosworth coined the term “Blue Button” as an analogy to the RSS orange button (which isn’t a thing anymore — this was The Good Old Days of Google Reader). Todd signed up CMS & VHA for Blue Button.

9/ I demoed a working production patient download.  Todd & @Fridsma pressganged me again, but this time with a special offer.
Take a leave of absence, take 9 months, focus on this problem, go get it done. & I said yes after sleeping on it.

10/ @Fridsma brought in @brianbehlendorf who had been working on Open Source issues with NHIN Connect (which also isn’t a thing? anymore) & we got a wiki & some basic open content rules, a mission, etc.

11/ (all this seemed like forever in real time, but I think all happened between October ’09 & March 1, ’10)

12/13/ All caught up to March 1st, 2010, Atlanta. I had my ONC badge & I walked around signing up EHR vendors.
My strategy:
1) Vendors first, not because I was vendor specific, but because we heard that implementors of the technology users use need to, um, implement.
2) I wanted to *require* commitment to deploy standards in production because I’d seen lots of “concern trolling” or theoretical arguments from folks who never had skin in the game
3) I wanted to keep it small – I was hoping to get 10 highly motivated organizations.

14/ We got more than 100. It was a handful. We setup up all the usual committees and workgroups, spent every meeting introducing people & defining what matters. We defined some clear community guidelines: http://wiki.directproject.org/Overall_Process

15/ The goal was to engineer good behavior through a parsimonious set of guardrails. You can’t “concern troll” if you have to define *exactly* what need to get done to fix

16/ But there was a ton of culture building. We tried really hard to make sure we heard from everyone. Every meeting had an “around the room” where we polled every participant to make sure they were heard.

17/ We tried really hard to implement “disagree and commit” & “patches welcome” as foundational principles. We wanted to hear from everyone, but we counted the voices of people who fixed stuff (wrote up specs, wrote code etc.) more than people who just talked.

18/ “I have concerns” — “OK, go fix it” This… surprised people, I think. Some people dropped out because it was hard work. But the ones who stayed got shit done.

19/ We also defined a timeline and stuck to it. The notion was that, as Parkinson’s Law states, “Work expands so as to fill the time available for its completion”, so we were going to march to some key milestones: requirements, spec, reference implementation, pilot, production.

20/ & we were going to set reasonable but aggressive timelines. Basically quarterly milestones. & we got it done! It was 11 months from that first day of HIMSS to the first implementation, & announced by @aneeshchopra, Todd, & @Farzad_MD in the HHS briefing room.

21/ There were some speedbumps along the way… We actually produced four specs, SOAP (remember that? no? it was a thing), REST, SMTP+S/MIME &; XMPP.

I still really like the REST spec. http://wiki.directproject.org/REST_Specification

22/ Since we had opinions about the best way to implement, forcing multiple specs & proofs of concept helped us separate opinions from reality & (un)fortunately everyone with an opinion made it work in practice.

23/ There was a meeting hosted by @Microsoft in Redmond where we had a split decision between SOAP, SMTP+S/MIME & REST. It nearly split the community. We spent a few weeks figuring out whether to force the issue, or let things play out.

24/ Finally, we agreed to SMTP+S/MIME with a SOAP (XDR) gateway option to split the difference. I got lectured to for an hour by @cshirky
on how there was this thing called REST that in my ignorant health care ways I might not know about & I was all like “dude, I know but…”

25/ That was the low point (not getting lectured to by @cshirky though that was pretty ironic) but feeling like the whole thing was so close and yet so far. The compromise worked, though. We got it done.

26/ Along the way, we made two other close call controversial decisions. One worked, the other didn’t. The one that worked was to move directory support to a second phase. Since we still don’t have widescale directories, I think that was a good call…

27/ (you can argue that the other way tho — this was a unique moment to get directories in place). I suspect that since we barely got Direct done, directories would have killed us.

28/ The second that didn’t work was to treat Direct as transport, & leave content up the stack.

Direct is often criticized for being “45 pages of meaningless information” and that turns out to be a C-CDA rendering problem “up the stack”.