Climatic vs. demographic drivers of seasonality in historical childhood disease outbreaks
An understanding of the mechanisms responsible for seasonality and its spatial variation in the population dynamics of infectious diseases remains largely limited by the lack of long-term temporal data that are also spatially replicated. Here we take advantage of a recently digitized, spatially-resolved, historical dataset that spans the 20th century in the United States for four-childhood diseases: polio, typhoid fever, scarlet fever, and diphtheria. For each study system we first classify the peak timing of outbreaks relative to both time and space. We then ask whether environmental or demographic factors are primarily responsible for the observed seasonality in disease incidence. We specifically examine and compare the role of temperature, precipitation, humidity, birth rates, and a general seasonal forcing (or null) as different hypotheses in disease-specific transmission models. These models are implemented as Partially Observed Markov Processes and fitted to disease incidence data using Maximum likelihood by Iterated particle Filtering.
Across all four study systems, two unique spatial patterns can be observed in the timing of the outbreak-peaks, namely (I) a south-to-north gradient for the summer outbreaks of polio and typhoid fever, and (II) a geographical gradient that begins in the southeast and then spreads west and north for the winter outbreaks of diphtheria and scarlet fever. Comparison of the polio models that include real-world demography and either temperature or rainfall covariates, demonstrates that climate variables act as dominant seasonal drivers. Based on Maximum Likelihood Estimates (MLEs), temperature is a strong candidate as the primary driver of seasonal polio transmission. Models that include temperature as the seasonal driver are better able to capture the shape of seasonal outbreaks as well as their timing, although models that include rainfall better represent the deep winter troughs. This research begins to unravel the factors driving disease seasonality and demonstrates the relevance of climatic factors in childhood infections. The disease-specific transmission models can be extended to contemporary settings for forecasting future outbreaks and to gain a better understanding of why these pathogens continue to circulate today.