Data Sharing: A Matter of Life and Death

Long a perennial demand, the call for improved data sharing at last week's Bio-IT World Expo in Boston was especially strong -- and emotional.

Credit: iStock/NiroDesign
Credit: iStock/NiroDesign

Heidi L. Rehm, Chief Laboratory Director for the Laboratory for Molecular Medicine, warned in her opening keynote that the consequences of failing to share data can lead to devastating results for patients. Rehm related the tragic tale of a doctor sending a fetal blood sample to her lab that showed a genetic mutation identified by researchers as likely to cause debilitating defects. Rehm's subsequent report to the doctor led to the mother terminating the pregnancy. Later research, however, demonstrated that this mutation was benign (and, in fact, common among certain ethnicities). This life-or-death discovery, said Rehm, could have been achieved sooner with better data sharing and consensus.

Now, Rehm devotes much of her work to getting genomics researchers to collaborate, and come to a consensus based upon their collective data.

The call for consistency
Rehm discovered that lack of consensus among genomicists does not always indicate true disagreement. Frequently, the problems are merely taxonomical."We've actually followed up with certain laboratories," said Rehm. "It turns out [one] lab doesn't use the 'likely benign' category."

This discovery and others like it, related Rehm, have led to greater resolution as she and her colleagues have begun to understand how inconsistent language and categorization have led to discordant scientific conclusions.

The call for standards
Standards -- or the lack of good, clear, consistently applied ones -- have also been problematic. "It would be great if we had better standards so we could all interpret in the same way," Rehm told her audience. "It is no easy task to try to professionally agree[.]"

Rehm described an effort she worked on standardizing terms related to how genomicists define the ability of an organism to cause modern diseases."Now the moment we published this guideline…we decided to do a bakeoff," said Rehm. "[W]e compared between labs using their old rules using the [new] rules." The initial results were "surprising;" the pre-standard and post-standard results between labs were mostly the same, demonstrating only 34-percent concordance.

The call for collaboration
The real problem? Human error. Rehm and her colleagues were only able to make this big-data discovery with small-data technology, picking up the phone. "[W]e walked through each of those pieces of evidence [the labs used]," Rehm said. "[W]e were able to agree as a group as to which rules actually did apply and which rules had been misapplied."

Improving the concordance of their data, Rehm explained, took more than the latest analytics technology. Rather, it took sharing both data and data methods, efforts requiring education on both the new standards and old ways of working.

"Some of them hadn't been able to work with these rules; some hadn't even read their own guideline," said Rehm. "So it became a really great framework for us to discuss evidence… Through this effort, we've worked to resolve either by phone or by email...exact interpretations by 70 percent, with a confidence level of 80 percent.

The moral here: Talk to colleagues to understand their thinking and get them to understand yours.

Or, as per Rehm, "if you're going to apply a ruleset, do it correctly."

Though conceding that individual professional judgment bars total consensus, Rehm insisted that the more researchers share with each other, the better work they will produce. "It's going to take…massive efforts in data sharing…to improve our knowledge of genome variation [and] improve our consistency," said Rehm. "The world is watching and patients' lives are at stake."

Joe Stanganelli, Attorney & Marketer

Joe Stanganelli is founder and principal of Beacon Hill Law, a Boston-based general practice law firm.  His expertise on legal topics has been sought for several major publications, including U.S. News and World Report and Personal Real Estate Investor Magazine. 

Joe is also a communications consultant.  He has been working with social media for many years -- even in the days of local BBSs (one of which he served as Co-System Operator for), well before the term "social media" was invented.

From 2003 to 2005, Joe ran Grandpa George Productions, a New England entertainment and media production company. He has also worked as a professional actor, director, and producer.  Additionally, Joe is a produced playwright.

When he's not lawyering, marketing, or social-media-ing, Joe writes scripts, songs, and stories.

He also finds time to lose at bridge a couple of times a month.

Follow Joe on Twitter: @JoeStanganelli

Also, check out his blog .

Clean Your Data with Visualization and Algorithmic Tests

Speakers at Bio-IT World explore techniques for biotech researchers and others working with big data to identify the accurate data in their data files.

Data Sharing: A Matter of Life and Death

Cooperation among medical researchers -- done right -- very simply can mean lives saved, but the research community needs education on how to execute on that collaboration.

Re: Consequences
  • 4/16/2016 12:16:58 PM

@ PredictableChaos.  I agree that is a very sad and traumatic experience for a patient to have.

As of 2015, surgeons still operate on the wrong body part or even the wrong person 1 out of 100,000 surgies in the U.S.  These are called never events.  Most are caused by communication errors between employees.

About 1 out of 4 all surgical errors are due technical or technology issues.

Re: Consequences
  • 4/15/2016 3:56:05 PM

"some hadn't even read their own guideline"

Yes, people have problems with "not reading the instructions" and this comes up repeatedly.

Computers have what might be the opposite problem: An inability to work around the instructions. What does that mean? One instruction out of place and the program might come to an unexpected halt.

Think about the rules for driving. And then imagine any number of cases where you might technically need to violate the letter of the law to make progress. One example - you're not supposed to cross the double yellow line. But what does traffic do if there is a stalled car on the side of the road that doesn't leave room to pass without crossing the double yello line? Everybody slows down and crosses the double yellow line briefly so traffic can continue to move forward.

In this situation, self-driving cars may be stuck or may just have to hand control back to the driver.

Re: Consequences
  • 4/15/2016 12:41:34 PM

Although "inconsistent language and categorization" among the data sharers is certainly the problem, I wonder if the folks who being used to make the choices when coding everything may lack the properr training to acheive consistent results among the collaborators, even when the "standards" have been laid out well. I would guess that some of the lower level employees making these judgments might be part of the problem? Or maybe the standards are way to complex that an average person can't make clear sense of what's to be done with the data.

  • 4/14/2016 6:18:16 PM

What a tremendously sad outcome for the patient who terminated a pregnancy based on data that turned out to be overly pessimistic.

Humans have been working to share data with each other for a long time. And we still don't get it right every time.

I think this challenge may be bigger than expected for self-driving cars and for certain robots. The Tay chatbot comes to mind.