Saturday, May 14, 2016

Image Sharing: Are we there yet? It seems not.

Short version: Why are we still using CDs? Its not the lack of standards or commercial solutions, it seems to be the lack of will, aka. incentives.

Long version.

In Joe Biden's Full Interview With Robin Roberts on the Cancer Moonshot he rightly bemoans (at 08:20 minutes in) the inability of two prestigious organizations, Walter Reed Hospital in Washington, D.C., and MD Anderson Cancer Center in Houston, TX, to share his son's medical imaging data electronically, without resorting to flying discs across the country (and even that apparently required the intervention of his son-in-law, who is a surgeon). Unfortunately, he attributes this to an absence of a "common language", which for this particular case is not true (since we have DICOM, which is the lingua franca of images). Earlier in the interview, the issue of incentives is discussed though.

This experience mirrors my own, dealing with family attending Memorial Sloan Kettering Cancer Center (MSKCC)  in New York, NY. The only mechanism I have to obtain images from there is again via CD. Speaking to one of the radiologists at Memorial, I was told that the inbound problem is just as bad; they employ 10 (!) FTEs whose only function is to stuff CDs received into drives to import them. Apparently they do have one of the commercial network image sharing alternatives installed, but are planning on ditching it and going with another vendor, not sure why. "Continuing bandwidth issues" were cited as a concern. MSKCC has a limited patient portal, which does have radiology results available through it (plain text of course, nothing structured to download), but apparently making images available (whether to View, Download or Transmit) through the portal is not a priority. It does make paying the bills easier though (I guess that is important for them).

Now, it is great that CDs work at all, and work relatively well. And of course they are thoroughly standardized (using the DICOM PS3.10 files that are specified by IHE PDI), as long as they don't come from older Stentor/Philips crap. But surely, well into the 21st Century, we can do better than "sneaker net", especially between major medical centers.

Yesterday, on a call with the HIMSS-SIIM Enterprise Imaging Joint Workgroup Best Practice Image Exchange and Sharing (Team 3) (which I have belatedly joined), there was a discussion about reorganizing the work groups and starting a new one on Standards and Interoperability. I was keen to emphasize that I don't think the interoperability problem is one of a lack of standards or implementation of them, but rather a lack of incentives, funding, prioritization or indeed a clearly articulated value proposition for deploying solutions, using the standards that we already have (or even using a non-standard solution, if it works).

When the UK folks were facing the problem of image sharing, and the NHS failed to deliver a suitable central solution, an ad hoc network of push-driven sharing evolved, the Image Exchange Portal (IEP), which has been bought and expanded by Sectra. They claim that:

"100% of NHS Acute Trusts in England plus private hospitals are connected to one another via the IEP network".

As I understand it, these guys were no more incentivized to develop, join or use the IEP sharing than are their counterparts in the US, nor were there any disincentives for not bothering to share images. Perhaps there were just no funds available to employ an army of CD-stuffers to work around the problem, so the pain was being felt by the decision makers. Or perhaps the resources for repeat imaging were more tightly controlled (as opposed to being a potential source of more revenue in the US), so the shared images were the only images available. I am just guessing, but I doubt it was because the Brits are any more altruistic or sensible than their Cousins (I can say that, since I am nominally a Brit, even though I have lived and worked in the US for decades).

The Canadians have their much vaunted, centrally funded, regional Diagnostic Image Repositories (DI-r's), but am I told that, in some provinces at least, you are lucky if you can get out what you put in, and there is little if any useful access to images submitted by other sites. Some provinces have apparently been able to do better though.

Regardless, all of us who work in medical imaging IT know that the technology is there, and is affordable, and the workflow is manageable despite having to deal with stupid things like the lack of a single national patient identifier. It doesn't really matter for the sharing use case which standard or combination of standards you choose for the transfer, as long as the payload is DICOM. Whether you push them or pull them, use traditional DICOM protocols or DICOMweb or XDS-I RAD-69 or XDR-I or some proprietary mechanism, or follow IHE Import Reconciliation Workflow (IRWF) to deal with the identifiers or do it your own way, with a little configuration, the images are going to get where they need to be. It is really just a question of motivating sites to get off their collective asses.

In the "collective" probably lies part of the problem, since on a large scale, what motivates competitors to share?

For once though, the problem can hardly be laid at the door of the evil vendors who might be accused of "data blocking". For image sharing, there is an army of vendors willing to help solve your sharing problem, as well as open source components to assemble your own, there are no format issues, the problem is way simpler than that of general EHR interoperability, and there is no debate over documents versus APIs (all of the radiology and cardiology images, at least, are already in DICOM format and document-like in that respect).

When I discussed this in late 2012 with Farzad Mostashari, after expressing my disappointment that the MU2 didn't insist on image sharing, he wrote that:

"My hope is that the business case for this is so clear that it will happen regardless (perhaps with some help from convening, best practices, etc) and we can point to the on-the-ground reality in two years as the ultimate refutation of the concerns."
Now here we are three and a half years later, not two, with a plethora of commercial solutions as well as multitude of standards for image sharing, but the "business case" is apparently not so clear after all, if the Vice President of the United States still needs to arrange to fly CDs around.

Shame on us all for failing him and his family.


PS. As far as I have been able to ascertain, the MACRA proposed rule doesn't provide any incentives or requirements for imaging sharing either. This may be as much because nobody has submitted sharing related performance measures as the lack of central recognition that this is important or a priority. Maybe the VP should submit comments on it!

PPS. In the same interview, Joe Biden also takes a shot at the much reviled editor of the NEJM, Jeffrey Drazen,  over his ill-considered "data parasites" comments (actually "research parasites", in the editorial co-authored with Deputy Editor Dan Longo). While Drazen may be well on his way to becoming the most hated man in America (perhaps overshadowing Martin Shkreli, the AIDS drug robber baron) the issues raised in Drazen's editorial are about a different kind of "sharing" than the subject of this post.

No doubt Drazen's comments reflect the opinion of many in the "elite healthcare research establishment", who seem to regard the right to solely exploit their taxpayer-funded research and data in order to exclude success by their funding competitors (not to mention their unwillingness to have their own data and analysis scrutinized for integrity and repeatability) as something akin to the divine right of kings. Again, this all seems to be a matter of incentives, this time the perverse incentives of the research funding infrastructure that encourage data hoarding rather than sharing due to the competitive nature of the process. NIH, perhaps crippled by the Bayh–Dole Act, doesn't seem to have any teeth in its data sharing policy when it comes to reviewing and approving grant applications or monitoring their performance, so there is no "level playing field" of mandatory and immediate sharing. Since most of what is published is probably false anyway, perhaps it doesn't matter:(

There is something for everyone in the interview, and the lack of open access to research publications comes in for its share of criticism too. Hear, hear!

I wish the VP every success in his crusade.


colnedwardrhodes said...

David, this is a great summary of the current situation in image sharing.

One point that you hit on is the lack of bandwidth to send CDs. This is of course a reality for large institutions who must pay as a chargeback from IT. In such cases Sneaker-net appears cheaper.

There ARE lots of great solutions to this problem and I think you've covered them ALL. It's time to advocate for adoption and insist on timely transfer of data for continuity of care!

May 16, 2016 at 1:54 PM

Anunaya Jain said...

We solved this issue way back in 2012 in western NY to create a large image sharing network and successfully trusting patients to save costs.

Anonymous said...

I can't believe none of the big pacs vendors have figured out that giving away a free Image Exchange service would make them the default choice for pacs. Or maybe somebody already did and I didn't notice.

David Clunie said...

Judging by Mark Bronkalla's recent post (, it seems we still aren't there yet, and some places lack the operational support to make even CDs work effectively.