Microsoft data scientists and researchers at Seattle Children’s Hospital have teamed up to give a high-resolution picture of the link between unexplained death in newborns and smoking.
In a paper published today in the journal Pediatrics, the research team outlined how even small amounts of smoking can double the risk of Sudden Unexpected Infant Death Syndrome (SUIDS), a broad term that includes death from Sudden Infant Death Syndrome (SIDS), suffocation and undetermined causes.
The project was started by colleagues of John Kahan, chief data and analytics officer at Microsoft, who lost a child to SIDS more than a decade ago. In 2016, when Kahan was climbing Mount Kilimanjaro to raise money for SIDS research, his team hatched a plan: Using Microsoft’s Azure machine learning capabilities and data visualization program Power BI, they would analyze government data to find factors that increased the risk of SIDS.
The Microsoft team ultimately linked up with Dr. Tatiana Anderson, a research scientist in Seattle Children’s Center for Integrative Brain Research and lead author on the paper.
The teams worked together for a year and a half, using computational modeling to take data from The Centers for Disease Control and Prevention and turn it into something useful. They analyzed 20 million births over four years, which included 19,000 cases of SUIDS-related deaths. The work was funded by Microsoft’s AI for Good initiatives, NIH grants and donor contributions to the Aaron Matthew’s SIDS Research Guild, which was named for Kahan’s son.
The teams found that just one cigarette a day doubled the risk of SUIDS, and the odds increased with each additional cigarette. The researchers estimate that 20 percent — or 800 of the approximately 3,700 annual deaths from SUIDS in the U.S. — could be prevented if no women smoked.
“The correlation was known, but not to this high of a resolution,” Anderson said.
Mothers who smoked in the months before pregnancy but quit in the first trimester still had a 50 percent greater chance of SUIDS than non-smokers, Anderson said. Her main takeaway: “Plan to quit well before you start trying.”
Among smokers who become pregnant, roughly 20 percent quit and an additional 24 percent reduce smoking, the study showed.
Beyond the findings, which add yet more evidence that smoking during pregnancy is harmful, Anderson thinks the project provides a road map for using large datasets in scientific research.
“The system we’ve set up here could absolutely be duplicated for all kinds of different problems,” she said.