From Evidence-Based Medicine to Computational Precision Health

This presentation will describe the evolution of paradigms for deciding what works in clinical medicine. The first paradigm – evidence-based medicine (EBM) – arose in the early 1990s and propounded the randomized controlled trial (RCT) as the “gold standard” for answering narrowly drawn hypotheses about the population-level effectiveness of clinical interventions. RCTs, and systematic reviews of RCTs, sit atop EBM’s hierarchy of evidence in which observational studies occupy lower levels of evidentiary strength. Much of the current medical literature still follows the EBM paradigm. In the 2000s, “big data” exploded onto the scene with datasets that were too big in volume, variety, and velocity (i.e., growth) for traditional data analytic methods. Instead, new “data-driven” methods that took advantage of cloud computing power created a new paradigm for assessing effectiveness, a paradigm that draws inferences about what works based on data alone without reference to specific hypotheses. More recently, quantum leaps in causal inference, artificial intelligence, and machine learning methods are increasingly upending how we “know” what works. A synthesis of the hypothesis-driven EBM and computational data-driven paradigms is now underway, with the potential to leverage the availability of huge amounts of electronic health record (EHR), genomic, biomedical literature, sensor, and other real-world data to go beyond population-level estimates to provide “precision” evidence for individualized decision-making. Thus, an era of computational precision health is dawning that will likely supplant EBM for assessing what works for populations and for individuals.