Drivers of sylvatic yellow fever transmission in southeast Brazil: a joint epidemiological and phylodynamic inference.
Abstract
Background
Yellow fever virus (YFV) has caused substantial human disease in Brazil, driven by spillover from outbreaks in nearby non-human primates (NHPs). As disease surveillance in NHPs is challenging, the impact of environmental conditions on YFV outbreaks in NHPs remains poorly understood. In this study, we aimed to use joint inference modelling to overcome these obstacles by integrating diverse data streams and to estimate YFV transmission dynamics in NHPs to enhance understanding of yellow fever disease ecology.
Methods
We applied a multistage analysis using epidemiological, phylogenetic, demographic, and meteorological data in conjunction with mechanistic and statistical modelling. We used EpiFusion joint inference models to infer YFV infection dynamics in NHPs from 2015 to 2019, and statistical generalised additive models to test hypothesised associations between NHP infections and environmental variables and human cases.
Findings
The effective reproduction number (Rt) of YFV in NHPs was approximately 1·0, indicating that small changes in transmissibility can enable explosive sylvatic outbreaks. We also found the possibility of large, entirely unobserved outbreaks in NHPs. El Niño sea surface temperature anomalies at medium-term and long-term lags were important predictors of YFV infection dynamics in NHPs. Finally, we found that the risk of YFV spillover to humans, measured in terms of the effect of NHP infections on human cases, was correlated with the number of NHP infections, with a 7-day to 21-day lag.
Interpretation
YFV might be establishing recurrent transmission in NHP populations in southeast Brazil, indicating that more comprehensive and consistent surveillance is needed. Identifying key environmental drivers and their lags with human YFV cases can be used to develop early warning systems to guide reactive vaccination and public health campaigns to prevent future outbreaks.
Funding
None.