Using Biomedical Researcher Judgments to Predict Clinical Trial Outcomes
Description
Human patients should only be assigned to experimental medical treatments when investigators are truly uncertain about the novel treatment’s clinical utility. As such, the outcomes of clinical trials are difficult to predict by design. The goal of this project is to work toward building a predictive model of clinical trials. The first step is to categorize treatments based on their history and diseases based on their treatability using FDA records among other data sources. In collaboration with the Biomedical Ethics Unit at McGill University, we have collected many probability predictions about scientific and operational outcomes of newly registered clinical trials. When pre-processing is completed, we will begin building a model to predict the judgments of medical experts based on several trial and researcher characteristics. This model can be used to assess whether medical researchers are biased in their judgments about their own trials. Finally, we aim to assemble these components to develop a model to predict the outcomes of the clinical trials by accounting for the history of the treatment, treatability of the disease, and judgments of medical research accounting for revealed biases.
Advisors
Skills Required by the team
- Classification
- Predictive modeling
- Data Collection