Grant: Project Grant

In 2022, about 233,900 Canadians found out they had cancer. Thanks to better testing and treatments, more people are living longer after a cancer diagnosis. But this also means we need to pay attention to possible return of cancer, new cancer types, or side effects from treatments. Studying these in detail needs to first identify these cancer related information which usually requires long-term follow up and can be costly and time-consuming. Luckily, we have lots of health data like electronic medical records. By using artificial intelligence methods, like natural language processing and machine learning, we can better understand and use this data to learn more about cancer survivorship. Therefore, we aim to create artificial intelligence-based algorithms to identify cancer recurrence, pathologic response to treatment, cancer-specific death, chemotherapy toxicities, radiotherapy side effects, and surgical complications from population-based real-world data. We will leverage real-world data, which contains granular and timing-related clinical information, to extract relevant information on the outcomes and complications of interest, using state-of-the-art natural language processing techniques. Moreover, we will develop machine learning algorithms to identify changes in patterns of medical encounters (e.g., frequency of visits, physician specialty, and new rounds of cancer diagnosis or treatment) which indicate cancer-specific outcomes, such as recurrence. To validate the developed methods, we will conduct internal and external validation using chart review as a reference standard. This project would be the first to systematically develop algorithms for accurate and timely identification of a comprehensive set of cancer outcomes and treatment complications from real-world data. These algorithms would play an important role in developing tools for informed decision-making, which will improve cancer patient and survivor care.