The availability of a vast amount of complex data coming from new technologies is revolutionizing healthcare research, providing a huge potential to both preventive and prognostic activities. To properly synthetize and manage the information deriving from different sources of clinical data (e.g., texts, medical imaging, omics), advanced methods are required, shifting the priority from predictive modeling to representation learning and fingerprint extraction. In this paper, we will discuss the challenges that this new perspective introduces in healthcare research, focusing on how health analytics is able to provide meaningful answers.