Predictive Analytics and Machine Learning in Medicine

Clinical prediction models have existed for decades, 1, 2 but they are just now attracting a wider interest in medicine.3, 4 Given the complexity and abundance of clinical information, the limited ability of clinicians to accurately prognosticate outcomes, and the plethora of prediction problems that clinicians face daily, it is not surprising that there is a growing interest in the field of prediction research. Unfortunately, while many models have been developed, few are appropriately validated, and almost none are routinely used in clinical practice.5, 6 This paradigm will likely change drastically in the next few years given the intersection of four major forces: (a) the growing complexity and cost of patient care, (b) the advent of Big Data in medicine, (c) the adoption of machine learning methodology in biomedical research, and (d) the development of the digital infrastructure in healthcare.