Search for: All records

Abstract Using data from New York City from January 2020 to April 2020, we found an estimated 28-day lag between the onset of reduced subway use and the end of the exponential growth period of severe acute respiratory syndrome coronavirus 2 within New York City boroughs. We also conducted a cross-sectional analysis of the associations between human mobility (i.e., subway ridership) on the week of April 11, 2020, sociodemographic factors, and coronavirus disease 2019 (COVID-19) incidence as of April 26, 2020. Areas with lower median income, a greater percentage of individuals who identify as non-White and/or Hispanic/Latino, a greater percentage of essential workers, and a greater percentage of health-care essential workers had more mobility during the pandemic. When adjusted for the percentage of essential workers, these associations did not remain, suggesting essential work drives human movement in these areas. Increased mobility and all sociodemographic variables (except percentage of people older than 75 years old and percentage of health-care essential workers) were associated with a higher rate of COVID-19 cases per 100,000 people, when adjusted for testing effort. Our study demonstrates that the most socially disadvantaged not only are at an increased risk for COVID-19 infection, they lack the privilege to fully engage in social distancing interventions. 

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.