Often lost amid the hurrahs over the promise and potential of big data to transform healthcare is how unwieldy it can be to apply in real-world settings—and how wary clinical decision-makers can be when they ’re presented with it.
Such was the gist of a news report that aired on National Public Radio on Jan. 5.
In the broadcast, NPR correspondent Amy Standen introduced Jennifer Frankovich, MD, of the pediatric intensive care unit at Stanford’s Lucile Packard Children’s Hospital in Palo Alto, Calif. Frankovich told what happened after a gravely ill girl from Reno, Nev., was flown by helicopter to the unit when her kidneys began to fail.
Tests showed the girl had lupus, the chronic immune-system disease that can damage any part of the body yet is manageable if diagnosed early and treated properly. Frankovich had previously treated children with the disease and recalled that some of them had developed potentially fatal blood clots. Should the hospital administer anticoagulants, or would that introduce risk of stroke?
“There wasn’t enough published literature to guide this decision, and [so] really the best route was to not do anything,” Frankovich told NPR. Meanwhile, she knew she had the girl’s charts, all electronic, in a searchable database. “I brought the data back to the big team of doctors gathered around her bed, and I said, ‘Hey, this is the number of lupus patients we’ve had, and this is the number that had a clot. What do you think?’”
The team agreed that, based on the numbers, it would be best to head off a potential clot. The anticoagulant decision proved wise. No clots formed and, over time, the girl’s lupus got better. “As far as I know, she’s doing well,” said Frankovich.
Then came the rub. “We’re actually not doing this anymore,” said Frankovich, referring to the use of data to guide decisions in this way.
“The system just isn’t ready, the hospital decided,” reported NPR’s Standen. “What if Frankovich had used the wrong search terms or the engine itself had bugs? What if the records had been mis-transcribed? Even Frankovich agrees that it’s just too risky.”
“The data is there,” said Frankovich. “Now we have to develop a system to use it in a thoughtful, safe way.”
Atul Butte, MD, PhD, a Stanford professor of medical informatics and a pediatric endocrinologist, noted that, despite the one-off nature of the lupus case at Stanford, the experience stands as an example of “a seismic shift” coming to medicine.
“The scientific method—which we learned in elementary school—is growing obsolete,” he said. “You come up with a question or a hypothesis and go make the measurements to address and answer that question or hypothesis.” But today the answers already exist, he added. “We already have the measurements and the data. The struggle is to figure out what do we want to ask of all that data.”
NPR concluded: “Getting that system in place, Frankovich and others hope, will lead to better, faster, cheaper medicine. But it’s still many years away.”
The posting of the broadcast on NPR’s website spurred an interesting discussion among some thoughtful, albeit anonymous and non-expert, commenters.
One said that, in the early ’90s, he administered an engineering computer network and “pondered the windfall for medicine” when healthcare providers all adopted electronic medical records. “Over the coming decades I watched all other industries make this progression and am amazed that healthcare has avoided it,” he wrote. “Doctors [who] say it is a burden remind me of the engineers [who] resisted CAD and computer aided engineering. Some engineers never did get the hang of it, but no company in their right mind would ever go back to the drafting board.”
This commenter added that, 15 years ago, his wife fell ill. Today, he said, “most of her medical records may as well be in a black hole. We have noticed that the most competent doctors were already keeping electronic records and never have a problem providing them. Meanwhile, most other doctors have little interest her medical history and frequently discount the history I have compiled.”
Another anonymous poster referenced the recent Ebola crisis. “I'm reminded of the triage nurse in Dallas recording that the patient with the fever had recently returned from West Africa, but the treating physician not hearing it and sending the patient home,” he posted. “What the hell are all those questions and electronic records for, then?”
A third stated: “My basic problem with big data in medicine