Search for: All records

Abstract Arboviruses transmitted mainly byAedes(Stegomyia)aegyptiandAe. albopictus, including dengue, chikungunya, and Zika viruses, and yellow fever virus in urban settings, pose an escalating global threat. Existing risk maps, often hampered by surveillance biases, may underestimate or misrepresent the true distribution of these diseases and do not incorporate epidemiological similarities despite shared vector species. We address this by generating new global environmental suitability maps forAedes-borne arboviruses using a multi-disease ecological niche model with a nested surveillance model fit to a dataset of over 21,000 occurrence points. This reveals a convergence in suitability around a common global distribution with recent spread of chikungunya and Zika closely aligning with areas suitable for dengue. We estimate that 5.66 (95% confidence interval 5.64-5.68) billion people live in areas suitable for dengue, chikungunya and Zika and 1.54 (1.53-1.54) billion people for yellow fever. We find large national and subnational differences in surveillance capabilities with higher income more accessible areas more likely to detect, diagnose and report viral diseases, which may have led to overestimation of risk in the United States and Europe. When combined with estimates of uncertainty, these suitability maps can be used by ministries of health to target limited surveillance and intervention resources in new strategies against these emerging threats. 

Abstract Privacy protection is paramount in conducting health research. However, studies often rely on data stored in a centralized repository, where analysis is done with full access to the sensitive underlying content. Recent advances in federated learning enable building complex machine-learned models that are trained in a distributed fashion. These techniques facilitate the calculation of research study endpoints such that private data never leaves a given device or healthcare system. We show—on a diverse set of single and multi-site health studies—that federated models can achieve similar accuracy, precision, and generalizability, and lead to the same interpretation as standard centralized statistical models while achieving considerably stronger privacy protections and without significantly raising computational costs. This work is the first to apply modern and general federated learning methods that explicitly incorporate differential privacy to clinical and epidemiological research—across a spectrum of units of federation, model architectures, complexity of learning tasks and diseases. As a result, it enables health research participants to remain in control of their data and still contribute to advancing science—aspects that used to be at odds with each other.