United to Solve SIDS
Binghamton alumni couple's loss inspires Microsoft data scientists to seek answers
As chief data analytics officer at Microsoft, John Kahan ’85 works with data that can be measured in petabytes and exabytes, amounts unfathomable to most people. As a father, the number he can’t fathom is one: One child, lost to Sudden Infant Death Syndrome (SIDS).
In 2003, his wife, Heather Kahan ’86, gave birth to their fourth child, a healthy boy named Aaron Matthew. Eight hours later the newborn stopped breathing. The cause of death was labeled SIDS.
“This diagnosis fundamentally means that the death had no explanation,” Heather says. “It’s a catch-all when all other known possibilities are eliminated. For the grieving parents, it is salt in the wound — there’s no chance for closure.”
In the 16 years since Aaron’s death, SIDS rates have not budged. Data science, in the meantime, has been racing forward, fueled by increased processing power and cloud computing.
Today, because of a discovery by a team of Microsoft data scientists, a world of statistics that had remained largely inaccessible is now revealing correlations that SIDS researchers hadn’t been able to see. “It turns out that the best way to find answers is through data, which, of course, is right up the alley of two computer science majors,” Heather says.
College, careers, kids and challenges
John and Heather met at Binghamton in 1984; they lived in Hinman’s Smith Hall, where John was a resident assistant. After graduating with bachelor’s degrees in computer science, they went to work for IBM and married in 1989. Fourteen years and three daughters later, John was recruited by Microsoft and the family moved to Seattle.
Aaron was born, and died, three months after that.
“I work with one of the most skilled, world-class data science teams that any company has,” John says, but no one has been able to answer the question he and Heather most want to know.
Risk factors for SIDS — poor prenatal care, smoking, putting the baby to sleep on his stomach — didn’t apply to them.
“None of this helped,” Heather says. “Life moved forward.”
The couple had another child, a daughter. Computing power continued to increase. And John became singularly focused on a mountain.
People don’t just live in the Pacific Northwest, Heather says, they live to be outdoors there, taking advantage of the hospitable year-round climate and natural beauty. The Kahan family shed their Northeast sensibilities and began enjoying casual hikes on local trails. But when a daughter spent a semester studying and hiking in New Zealand, John got the bug to do more.
That’s when the self-described “city boy” began training to climb Mount Kilimanjaro, the highest mountain in Africa. Not only would he use the climb to mark what would have been Aaron’s 13th birthday and bar mitzvah, he would raise money for SIDS research.
“John is the most driven person I know,” Heather says. “Once he sets his sights on something, he doesn’t stop until he’s achieved (or even over-achieved!) his goal.”
The whole experience was the first time that John had found a meaningful way to pay tribute to Aaron, Heather says. “The need to ‘give back’ and make a difference was always there, but the answer hadn’t come to him until this point.”
Heather honored Aaron in her own way, as a board member of a support group for parents who had lost babies to miscarriage, stillbirth and unexplained death. She did outreach, wrote articles and organized remembrance walks. “We participated in these events as a family, and it was one of the ways we kept Aaron’s memory alive among our other children,” she says.
John’s climb went according to plan, and he reached the summit of Kilimanjaro on June 29, 2016.
The following year, the Kahans established the Aaron Matthew SIDS Research Guild to fund research at Seattle Children’s Research Institute’s Center for Integrative Brain Research. At the close of 2018, the guild had raised more than $1.5 million through fundraising events, sales of John’s two wildlife photography books, corporate sponsorships and matching funds. In addition, Cribs for Kids, a 20-year-old national SIDS organization, announced the creation of a new research component called the Aaron Matthew Research Foundation, which will provide support and exposure to the guild.
Hidden in plain sight
Juan Lavista Ferres, senior director of data science at Microsoft, does not climb mountains.
He and John both chuckle as they make that point in separate interviews, because it turns out that what Lavista and some of his team did at their desks to support John took them on an unexpected journey.
Lavista says he remembers showing John a picture of his newborn daughter, swaddled in a hospital blanket, and remarking how similar it looked to the photo of the baby on John’s desk. He assumed it was one of the Kahan daughters. It was Aaron.
“That’s when he told me the story,” Lavista says. “I freaked out about SIDS. I thought, ‘If that happened to someone like John, it could happen to anyone.’”
So, while John prepared for his expedition, Lavista and about eight of his colleagues started a volunteer project, looking for data about SIDS.
What they found was the Centers for Disease Control and Prevention’s (CDC) Cohort Linked Birth/Infant Death Dataset, a publicly available but under-used source of statistics that includes tens of thousands of SIDS cases in the United States.
“It was amazing,” Lavista says. “Even though, by our standards, the data source isn’t that big, you still have 4 million rows, for 4 million births a year. So being able to process 10 years of data in some of these algorithms, it was just not possible 10 or 15 years ago because there wasn’t the processing power to do it.”
But first, the data scientists needed someone to tell them what was in the data: what to look for, how to sort it, what the terminology meant.
They turned to researchers at Seattle Children’s, where doctors had tried to save Aaron and where research was being done about a possible link between inner-ear damage and SIDS.
Solving the mystery of SIDS
About 3,500 infants in the United States die each year from Sudden Unexpected Infant Death (SUID), a category that includes three scenarios: SIDS, “unknown cause” and accidental suffocation and strangulation in bed, according to the CDC.
While SIDS rates in the United States have declined since the mid-1990s, most likely because of the “Back to Sleep” campaign, which taught parents to put babies to bed on their backs instead of their tummies, the overall rate for SUID between 1995 and 2016 remains relatively flat.
Research done around the world has looked at many factors, including air pollution, sleeparousal signals, gene mutations and inner-ear defects, as possible triggers for SIDS. But the datasets have tended to be small, which makes it harder to tease out information.
The original CDC dataset came with a 26- page document explaining the 250 columns, Kahan says. “And trust me, 99 percent of all medical researchers couldn’t understand what it is because it wasn’t where they ‘grew up.’ They grew up in a world looking at mice and small data sets.
“Our guys, in a couple hundred hours of [voluntary] work, downloaded it and put it up in the cloud. We ran machine learning models across it, visualized the data, and there’s a thousand years of research available.
“Seattle Children’s is one of the top pediatric hospitals, and they were blown away by our ability to manipulate data that they couldn’t even look at,” he says.
Dr. Tatiana Anderson, a neuroscientist in the Center for Integrated Brain Research at Seattle Children’s Research Institute, works closely on the collaboration with the Microsoft data scientists.
“They employ very sophisticated analyses and modeling techniques using a skillset that most research scientists do not have. On the other hand, scientists at Seattle Children’s are steeped in the literature and familiar with what is currently unknown in the field,” she says. “Therefore, we have formed an extremely productive collaboration wherein Seattle Children’s researchers and Microsoft data scientists jointly conceptualize and design each study, carefully interpret the results and co-author manuscripts that are submitted to medical journals.”
Nearly three years after they started, Lavista and some of his colleagues are still very much involved as volunteers. At a fall 2018 conference, they presented some new results to doctors. One inquiry they are pursuing is the difference between the average age of SIDS, which is 3 months, and early SIDS, which tends to happen in the first week of life.
“We did this because we wanted to contribute to the world, and it was rewarding. What I didn’t realize before was that it was an amazing experience learning not only about SIDS but about different data-science skills,” Lavista says.
Building a new database
One of the next steps is the building of a genome database for infants, not just for SIDS research but for future prenatal testing for families. The hospital asked the Kahans if it could start the database with Aaron.
“I said, ‘What are you talking about? Aaron died 15 years ago, and I don’t have parts of Aaron’s DNA anymore.’ And they said, ‘We do,’’’ John remembers.
Seattle Children’s had anonymously put away tissue samples for every child autopsied before 1994 (the procedure is now handled by the county), and, with the Kahans’ permission, it was able to track down Aaron’s. The database now has 250 children’s tissue samples from across the United States.
“Between Microsoft, which has donated the cloud resources, and the ability to build a genome database — which has been done for cancer research — we are now building the first infant database in the world, focused totally on infant mortality and ultimately solving this problem,” John says.
More at: givetostopsids.org