May 23, 2018
Biostatistics doctoral student Stephen Lauer is the lead author of a paper published in a recent issue of the Proceedings of the National Academy of Sciences (PNAS). Titled “Prospective forecasts of annual dengue hemorrhagic fever incidence in Thailand, 2010-2014,” the article presents findings on dengue hemorrhagic fever prediction forecasting models tested over the course of a five-year period.
Dengue hemorrhagic fever (DHF), a severe manifestation of dengue viral infection that can cause severe bleeding, organ impairment, and even death, affects between 15,000 and 105,000 people each year in Thailand. The number and location of cases vary dramatically from year-to-year, which makes planning prevention and treatment activities prior to the dengue season difficult. Accurately forecasting where DHF outbreaks occur prior to the dengue season could help public health officials prioritize public health activities.
Their paper, Lauer explains, was conceived as a response to a question posed to them by a Thai public health official: “How many dengue hemorrhagic fever cases will there be this year?”
Says Lauer, “This simple query is surprisingly unexplored in an area of research that primarily focuses on short-term forecasts. In answering this question, we sought to build a forecasting model that would make predictions in a timely manner using the data that would be available to officials in ‘real time,’ that is at the moment policy-relevant forecasts would be made.”
Lauer, with faculty advisor Nicholas Reich and postdoctoral research associate Krzysztof Sakrejda in the University of Massachusetts Amherst Department of Biostatistics and Epidemiology, along with colleagues from Mount Holyoke College, Johns Hopkins University, the University of Florida, and the Thai Ministry of Public Health, developed statistical models with biologically motivated covariates, such as climate, demographics, and other incidence-related drivers, observed by April each year, to make forecasts for each Thai province every year.
The group’s methodology allows for flexible model selection and complex relationships between the covariates and their forecasted outcome. As a result, their model makes accurate DHF forecasts for a long timeframe earlier in the year than any other published model to date.
“The annual forecasts produced by our model consistently outperform a baseline model based on median seasonal incidence, correctly order provinces by outbreak risk, and have well-calibrated uncertainty (e.g., the 80% prediction interval covers exactly 80% actual annual incidences),” says Lauer.
These early, accurate forecasts of DHF incidence could help public health officials determine where to allocate their resources in the future in a policy-relevant time-frame.
“Accurate forecasts could play an important role in implementing targeted interventions designed to reduce transmission, such as in helping to determine the location and timing of vector control activities and the mobilization of additional resources, as well as for reporting risk of infection to the public,” the authors write. “Additionally, they could play a critical role in a systematic study of how well different interventions prevent or reduce the size of disease outbreaks. Collaborative efforts between public health agencies and academic- or industry-based teams with predictive modeling expertise are critical to helping propel this field forward. With the rapid growth and maturation of disease surveillance systems worldwide, developing our understanding of the best methods for creating and evaluating forecasts of infectious disease should continue to be a global health priority.”
This project was funded by NIH NIAID grant 1R01AI102939 and NIGMS grant R35GM119582.