7 September 2021 in npj Digital Medicine
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.
23 October 2019 in Nature Communications
The recent trend of rapid urbanization makes it imperative to understand urban characteristics such as infrastructure, population distribution, jobs, and services that play a key role in urban livability and sustainability. A healthy debate exists on what constitutes optimal structure regarding livability in cities, interpolating, for instance, between mono- and poly-centric organization. Here anonymous and aggregated flows generated from three hundred million users, opted-in to Location History, are used to extract global Intra-urban trips. We develop a metric that allows us to classify cities and to establish a connection between mobility organization and key urban indicators. We demonstrate that cities with strong hierarchical mobility structure display an extensive use of public transport, higher levels of walkability, lower pollutant emissions per capita and better health indicators. Our framework outperforms previous metrics, is highly scalable and can be deployed with little cost, even in areas without resources for traditional data collection.
18 September 2020 in Science
Better relaxing lockdown together: Even during a pandemic, all countries—even islands—are dependent in one way or another on their neighbors. Without coordinated relaxation of nonpharmaceutical interventions (NPIs) among the most closely connected countries, it is difficult to envisage maintaining control of infectious viruses such as severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2). Ruktanonchai et al. used mobility data from smartphones to estimate movements between administrative units across Europe before and after the implementation of NPIs for coronavirus disease 2019 (COVID-19). Modeling disease dynamics under alternative scenarios of countries releasing NPIs, in particular stay-at-home orders, showed that if countries do not coordinate their NPIs when they relax lockdown, resurgence of disease occurs sooner. Coordination of on-off NPIs would significantly increase their effectiveness at reducing transmission across Europe. Science , this issue p. 1465
18 May 2020 in Nature Human Behaviour