Research Abstract

I am an Assistant Professor of Medicine at Harvard Medical School and Physician at Dana-Farber Cancer Institute, where I conduct research in the Division of Population Sciences and provide clinical care for patients with lung cancer in the Thoracic Oncology Program. My primary interest is in improving outcomes for patients with lung cancer on a population basis by identifying ways to optimize the implementation of scientific and clinical innovation. My specific goal is to develop a research program around applying modern analytic techniques to electronic health records (EHR) data to more readily use such information to improve care. One key challenge in oncology is that most adults with cancer do not participate in clinical trials, such that understanding the impact of innovation on a population basis requires analysis of ‘real-world’ data sources that were not originally intended for research purposes. The EHR could constitute an important source of such data, but to date, applying the EHR for this purpose has been limited by the fact that key clinical outcomes, even including whether cancer improves or worsens after a given treatment, are generally recorded only in unstructured free text reports. Structuring these data has therefore traditionally required manual medical records review, which is a resource-intensive, error-prone process at scale. I am therefore developing machine learning techniques for abstracting outcomes data from the EHR. Additionally, my work includes the development of techniques for understanding population-level treatment patterns, comparative effectiveness, and the impact of policy shifts using insurance claims data at scale.