Lymelight, a machine learning system designed by researchers from Boston Children's Hospital and Google to analyze online search data, could be used to forecast the spread of Lyme disease in real time, a new study suggests.
The study was published in npj Digital Medicine and authored by a research team that included John Brownstein, PhD, chief innovation officer of Boston Children's. The team trained Lymelight to predict the spread of the tick-borne disease in geographical regions by analyzing online searches for Lyme disease and its symptoms — verified by clinical diagnostic criteria and physician assessments — in those regions over time.
When compared to CDC data for the same time period, Lymelight was found to be 92 percent accurate in forecasting the spread of the disease. Additionally, the researchers wrote, "Our results show that Lymelight can estimate real-world incidence of Lyme disease much earlier and more efficiently than the official Lyme disease tracking system, which often reports data with as much as a two-year delay."
They added, "In the light of the recent findings that Lyme disease incidence in the United States has been considerably underreported, our study offers practical ways to substantially improve Lyme disease monitoring in real time."
View the full study here.