Despite considerable technological advances and the increased epidemiologic knowledge of the last decades, our health systems have been operating under service models too costly and too difficult to manage. Beyond diagnosis and treatment of conditions, both health professionals and federated learning researchers should focus more on the prevention of health problems. In our current epidemiological paradigm, this operational shift is particularly relevant as 48% of the global burden of disease is associated with modifiable risk factors. These risk factors are behaviours and habits, such as smoking and physical inactivity, amenable to change and with potential to prevent certain conditions and their consequences. Similarly to vaccination and cancer screening initiatives, acting on these behaviours can result in considerable health and economic gains. However, ongoing scientific research has not yet adequately addressed role federated learning can have in contributing to disease prevention. In this chapter, we present a novel medical and public health perspective, focusing on this issue. First, it is discussed how federated learning approaches can leverage AI and medical devices applications to act on modifiable risk factors. Secondly, we review the potential of high quality individualized health data for privacy-preserving population-level knowledge generation. Additionally, some hypothetical use cases and opportunities are presented, as well as estimations of their health and economic impacts. Overall, this chapter introduces the federated learning community to the untapped potential of addressing modifiable risk factors, considering both the expected impacts and the scale of this global problem.