Your Emergency Contact Information Could Be Used in a Medical Study

Will you be my emergency contact?

When you’re dating, the question is a sign that you’ve made it to the this-is-really-serious category. When you’re friends, it’s a sign that you’re truly beloved or truly responsible. And if you’re related, it may mean that you will now be entered into a medical study together so scientists can figure out if sinus infections or anxiety run in your family.

What? That’s right. Researchers have begun experimenting with using emergency contacts gathered from medical records to build family trees that can be used to study the heritability of hundreds of different attributes, and possibly advance research into diseases and responses to medications.

“It’s a way of looking at genetics but without having any genetic data,” said Nicholas Tatonetti, a data scientist at Columbia University Medical Center and one of the researchers who came up with the novel approach, which was outlined in a paper published Thursday in the journal Cell.

The data set, which they are now making available to other institutions, contains anonymized information from about 2 million patients at Columbia University, Weill Cornell Medicine, NewYork-Presbyterian Hospital and Icahn School of Medicine at Mount Sinai. The researchers are also sharing their algorithm so other hospitals can conduct their own studies using emergency contacts.

“If we have more data, we’re going to be able to advance medicine much more over the next 50, 60 years by observing patients instead of just guessing,” he said.

But this is not what patients signed up for when they filled out their emergency contact forms, said one bioethicist, who feels so outraged that he’s considering withholding that information on future doctors’ visits.

“Ethically this is really problematic,” said Craig Klugman, a professor at DePaul University and an editor at bioethics.net.

Tatonetti came up with the idea of using emergency contacts while trying to figure out how to effectively study adverse reactions to medications. Genetics, he believes, frequently plays a role in why some people respond badly to a medication.

One known example is the cancer medication mercaptopurine; while highly effective for many people, it is toxic to patients with a genetic variant.

“If we can figure out other examples like that we can design safer drugs for patients,” he said, adding that would help create specialized medications for smaller groups.

What has hindered work in this area, he said, is the lack of a large data set containing multiple generations of detailed medical histories. Most heritability research, so far, has relied on studies of twins and specialized research looking at families that share a particular disease, like schizophrenia. Genetic research has also been limited regarding race.

“The majority of research on disease heritability has been done in Caucasians of mostly northern European descent,” said Fernanda Polubriaginof, a graduate student in biomedical informatics at Columbia University and the study’s lead author. The electronic medical records from the New York hospitals were far more diverse.

She and her colleagues didn’t know if emergency contacts could fill these gaps, but they decided to try by building an algorithm and then comparing their estimates of how likely it was that traits were inherited to existing research.

First they whittled down the records, looking for cases where the people named as emergency contacts had been patients at the same institution. Then they grabbed diagnostic codes and information such as height, body mass index and blood serum levels from lab reports and bills; the research consent form most of us sign at the doctor’s office allows this.

They knew there were numerous reasons that connections in their data set might not be biologically related. To create a statistical model, they compared a subset against 1,500 patients that had also consented to providing DNA.

The team was pleased to find that height, obesity, diabetes, acne and sickle cell anemia all came out within a few percentage points of more traditional research studies. The data set also offered potential heritability data for about 400 items that had not previously been examined this way, including sinus infections, tooth decay, irregular menstruation and thyroid disorders. And Tatonetti was surprised to learn during the research that runny noses are highly heritable — a trait confirmed by his data set.

What does this change for the patients whose data was used? Actually nothing. Rather, it is expected to primarily affect the direction of future studies. And even if a major breakthrough were to emerge from those studies, patients whose records helped get scientists there and could benefit from that information would not be notified because the privacy agreement does not permit the researchers to contact physicians or patients directly.

All of this is part of why Klugman, the bioethicist, objects to this study and others like it.

The idea that “we’re going to find information to help your health but we’re not going to give you that information,” undermines the reason for building the data set in the first place, he said. It also highlights what he sees as a broader problem with generalized consent forms: they provide cover for uses of medical records that patients are unlikely to expect.

He compared researchers’ use of emergency contacts with the now-defunct Cambridge Analytica’s use of personality quiz data. But he said it’s more egregious.

“On Facebook, there is no expectation of privacy. It is clear that we pay for this free service by giving away our information,” he said. “In medicine, there is a strong expectation of privacy going all the way to the Hippocratic oath.”

Other efforts to anonymize family health data have provoked criticism that stripping identifying information may not adequately protect one’s privacy.

Tatonetti said that they went through a thorough vetting process with the institutions’ review boards. He agrees that it’s not ideal that they don’t have a better way to communicate their findings. In the future, he’d like to see participating institutions provide more specific consent forms. The authors also acknowledged other limitations. One, that just because someone didn’t have a diagnosis in their paperwork, it doesn’t mean that person didn’t have that condition. Second, environmental factors, not genes, might be the reason for a shared condition.

Still other researchers shared the authors’ excitement about the novel approach. Already a number of scientists have begun using the data.

“We need to have more inquiry into how the public feels about such efforts,” said Laura Koehly, a researcher at the National Human Genome Research Institute at the National Institutes of Health, who was not involved in the project. But given how difficult it is to conduct family studies, from her perspective, “This is a valuable endeavor.”