It is desirable to predict outcomes for patients in advance of severe symptoms manifesting themselves, based on their vital parameters. For example, given the history of a cohort of patients, including their gender, age, weight, blood pressure, LDL and HDL levels, glucose tolerance, and various other routinely-measured parameters, we may like to assess their risk level for heart disease, and intervene early if required. Several machine-learning methods have been applied to this task, but the field is in its infancy, and a significant problem is getting sufficient quantities of "clean" data to train a machine-learning algorithm. We give an overview of the existing literature, and briefly outline some recently commenced research in collaboration with medical professionals in Chennai [joint work with Gautam Menon, Deepika Choubey (IMSc), Paul Ramesh, Jayashree Gopal (Apollo hospitals), Madhulika Dixit (IIT Madras) and others]