Kolokotrones Professor of Biostatistics and Epidemiology
- Harvard-MIT Division of Health Sciences & Technology Faculty
- Associate Director, HSPH Program on Causal Inference
- Member, Biostatistics and Computational Biology Program, Dana Farber/Harvard Cancer Center
M.D., 1995, Universidad Autónoma de Madrid, Spain
M.P.H., 1996, Harvard University
Sc.M. (Biostatistics), 1999, Harvard University
Dr.P.H. (Epidemiology), 1999, Harvard University
My research and teaching are focused on methodology for causal inference, including comparative effectiveness research to guide policy and clinical decisions.
In an ideal world, most decisions would be based on randomized experiments. For example, public health recommendations to avoid saturated fat or medical prescription of a particular painkiller would be supported by long-term studies that compared the effects of interventions randomly assigned to large groups of people from the target population who adhered to the study interventions.
When randomized trials are available, we use cutting-edge statistical methods to complement the usual intention-to-treat estimates with appropriate estimates of the per-protocol effect that would have been observed under full adherence to the protocol of the study. Unfortunately, randomized experiments are often unethical, impractical, or simply too lengthy for timely decisions.
When randomized experiments are not available, my collaborators and I combine observational data, mostly untestable assumptions, and statistical methods to emulate the target experiment that would have answered the question of interest. We emphasize the need to formulate well-defined causal questions, and use analytic approaches whose validity does not require assumptions that conflict with subject-matter knowledge. For example, when experts suspect the presence of time-varying confounders affected by prior treatment, we do not use conventional adjustment methods that require the absence of such confounders. While causal inferences from observational data are risky, an observational analysis is often the only available evidence for decision-making.
Because there is no alternative to observational studies, we need to keep improving them. At the very least, sound observational studies may guide the design of future randomized experiments