Current Research and Scholarly Interests
In addition to my clinical research in head and neck and lung cancer, I work on the application of computer science and machine learning to cancer research. I develop tools for analyzing large datasets to improve outcomes and safety of cancer treatment. I developed a machine learning prognostic model using data from around 13,000 patients with metastatic cancer which performs better than traditional models and physicians [PubMed ID 33313792]. We recently completed a prospective randomized study in thousands of patients in which the model was used to help improve advance care planning conversations.
I also work on the methods underpinning observational and predictive modeling research. My open source nnet-survival software that allows use of neural networks for survival modeling has been used by researchers internationally. In collaboration with the Stanford Research Informatics Center, I examined how electronic medical record (EMR) survival outcome data compares to gold-standard data from a cancer registry [PubMed ID 35802836]. The EMR data captured less than 50% of deaths, a finding that affects many studies being published that use EMR outcomes data.