Hi, welcome to this course on predictive analytics. I am Ewout Steyerberg. I'm a researcher appointed at Leiden University Medical Center where I lead a Department of Biomedical Data Sciences. Together with my colleague David van Klaveren, I will guide you through the key issues in prediction modeling. David is a senior researcher at Erasmus MC, Rotterdam, The Netherlands. Predictive analytics has a long tradition. Discussions on prognosis was central to medical practice in ancient Greece. Hippocrates has been called the Father of Prognosis. He defined prognosis as foreseeing and foretelling, by the side of the sick, the present, the past, and the future. According to Hippocrates, the physician will then be more readily believed to be acquainted with the circumstances of the sick, so that men do have confidence to interest themselves to such a physician. This broad definition makes clear that Hippocrates was not only interested in predicting the future, but also appreciated the past and present situation of a patient. The past, present, and future need to be pieced together into a comprehensive story of the patient's health. In modern times, we rely on technology to make diagnostic and prognostic predictions. Here, we see an example from 1972. The London Hospital Survivor Predictor indicated where this patients in coma would survive. This gray box of electronics has a single indicator dial. If the needle swings to the right, it points to the letter S for survive. It can also point to letters, IBD which stands for irreversible brain damage or brain death. This device intended to determine whether a patient was in the IBD state or brain death. These enabled decisions as to whether organ donation could be carried out. The input for this device was based on 13 clinical characteristics, an ominous number by the way. This specific device was never actually used to determine whether a patient had already died or should have life-support his room. The quality of its predictions are unknown at least to be. In this course, you will learn how to make accurate prediction tools and how to assess their validity. We can use prediction models to better target preventive activities and deliver better health care. How can we learn about the key issues in predictive analytics? We have various sources at our disposal. There are also quite excellent textbooks that relate to prediction. Some are more mathematical or statistically oriented, such as this classical texts on statistical learning. There's a more practical book on regression modeling strategies by my colleague friend Carol. I myself try to write a book that is more pliant and pragmatic in nature called Clinical Prediction Models. It discusses a development, validation, and updating of prediction models. These topics are also addressed in this MOOC. Fortunately, there are also many scientific papers or medical prediction, including some reviews and reporting guidelines. One is called TRIPOD. This course extends on concepts that were discussed in the course: Population Health Study Design and in the course: Population Health Responsible Data Analysis. Concepts like sample size calculation, p-values, regression analysis, they all should be familiar to you. If not, I highly recommend following these courses first. This course consists of four modules. Through lectures, hands-on exercises in R, and self-study such as readings, you will gain knowledge and experience in basic and more advanced topics relevant to predictive analytics. In the first module, David van Klaveren introduces the role of predictive analytics for prevention, diagnosis, and effectiveness. In the second module, you will learn about key concepts in prediction modeling, such as the importance of sample size to prevent over-fitting, and methods to study the validity of a model. The third module will be devoted to model development, while the last module discusses model validation and updating. This course is intended for anyone interested in prediction in medical research. For example, master and PhD students, physicians or healthcare managers. Some topics are rather technical, but I trust anyone with basic knowledge about epidemiology and regression analysis can master the presented challenges. We hope you enjoy the course. Lets get started.