Answers are grounded in this paper and cite their sources.
Real-world epidemics emerge from interactions between viral transmission and human behavior shaped by socioeconomic conditions. Observed case data provide an indirect and delayed representation of these processes, limiting the ability to resolve transmission dynamics across time and scale, particularly in data-sparse settings. We develop a multiscale modeling framework for recovering structured temporal signals from reported incidence and evaluating their utility for short-term epidemic prediction. The approach integrates wavelet-based identification of dominant temporal scales with B-spline modeling of smooth baseline dynamics within a parsimonious SEIR formulation, where scale-dependent features modulate the transmission rate, β(t). Applied retrospectively to county-level COVID-19 data from Wyoming, the model captures both long-term trajectories and short-term fluctuations in daily incidence. Peak timing is recovered with subweek accuracy across counties, whereas magnitude estimates show greater variability, indicating that temporal alignment is recovered more consistently than absolute case magnitude under sparse and noisy observations. These results suggest that short-term epidemic prediction is strongly influenced by how temporal structure is represented in the data. The proposed framework provides a low-dimensional and interpretable characterization of multiscale transmission dynamics that preserves predictive performance while supporting identification of epidemic turning points, particularly in rural and low-density settings.
No takes yet. Share an insight, caveat, or question.
Xu et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: