Activity Number:
|
128
- Modeling for the Masses: Tackling Infectious Disease for the Public Good
|
Type:
|
Topic Contributed
|
Date/Time:
|
Monday, August 3, 2020 : 1:00 PM to 2:50 PM
|
Sponsor:
|
Biometrics Section
|
Abstract #312518
|
|
Title:
|
Type-Specific Forecasts of Influenza-Like Illness Using Hierarchical Splines
|
Author(s):
|
Evan Ray* and Isabelle Beaudry and Graham C Gibson and Nicholas G Reich
|
Companies:
|
Mount Holyoke College and Pontificia Universidad Católica de Chile and University of Massachusetts, Amherst and University of Massachusetts at Amherst
|
Keywords:
|
forecasting;
Bayesian hierarchical modeling;
infectious disease;
influenza;
splines
|
Abstract:
|
Influenza-like illness (ILI) is responsible for substantial loss of human life and economic activity each year. Forecasts of ILI incidence could give public health officials advance warning and improve planning interventions designed to reduce or mitigate disease spread. We develop forecasts for incidence of ILI, which is defined as fever with temperature of 100°F [37.8°C] or greater and cough and/or sore throat. Our model uses latent splines with hierarchical structure across seasons and spatial units. We compare two variations on this model: one that models aggregated ILI in outpatient doctor visits, and a second that uses virologic testing data to decompose incidence of outpatient ILI into incidence attributable to influenza A and influenza B. Both variations on the hierarchical splines model have higher forecast skill and better calibration than a baseline seasonal ARIMA model for forecasts of short-term disease incidence. The decomposition by influenza type offers insights into the dynamics of influenza incidence that are not available from simpler single-pathogen models, particularly in seasons where influenza A and B peak at substantially different times.
|
Authors who are presenting talks have a * after their name.