2021 Grantee: Wansu Chen, PhD, MS
Kaiser Permanente Southern California
Research Project: Extracting and Validating Symptoms Prior to Pancreatic Cancer Diagnosis from Electronic Health Records (EHR)
Award: 2021 Pancreatic Cancer Action Network Research Project
Award Period: July 1, 2021 – June 30, 2022
Dr. Wansu Chen began her education in China, receiving a BS degree in Computer Science from Fudan University, and then an MS in Medical Statistics at Shanghai University, II. From there, she went on to earn another MS in Biostatistics from the University of Oklahoma and a PhD in Biostatistics from University of Southern California. She is currently a research scientist II in the Department of Research & Evaluation at Kaiser Permanente Southern California (KPSC).
With both statistics and computer science as her background, Dr. Chen is interested in predictive modeling using machine learning and statistical methods. Early detection of pancreatic cancer is the area she has devoted the most time and energy over the past three to four years. The efforts have resulted in more than 10 publications and submitted manuscripts. Dr. Chen works closely with Bechien Wu, MD, gastroenterologist at KPSC and co-principal investigator of PanCAN’s Early Detection Initiative. Under the leadership of Drs. Wu and Chen, the pancreas research team at KPSC developed, validated and externally tested a machine learning-based prediction model aiming to identify high risk patients for early detection of pancreatic cancer.
Patients who are diagnosed with pancreatic cancer at an earlier stage have better survival and outcomes compared to patients diagnosed with late-stage disease. Currently, only about 20% of pancreatic cancer patients are diagnosed early enough to be surgical candidates. In the absence of a standard early detection test for the disease, one of the hurdles is the lack of distinct symptoms to alert the patient and their medical team that they should be evaluated for pancreatic cancer.
PanCAN’s Early Detection Initiative aims to develop a screening strategy for patients with a lesser-known symptom of pancreatic cancer – a recent diagnosis of diabetes. Based on the person’s age and changes to their weight and blood sugar, a subset of participants in the Early Detection Initiative study will undergo imaging tests to determine whether their diabetes was caused by a not-yet-diagnosed pancreatic tumor.
The first site that will enroll participants in the Early Detection Initiative study will be KPSC. Eligible participants will be identified through their electronic health records (EHR) based on their age (over 50 years) and a diagnosis of new onset diabetes and/or hyperglycemia (high blood sugar). The main goal is to use the information gathered from the EHR to identify onset of hyperglycemia/diabetes at its earliest discovery and evaluate if imaging at the time of new onset results in earlier detection of pancreatic cancer, ideally at a stage eligible for surgery.
In parallel to the objectives of the Early Detection Initiative, Dr. Chen will dive deeper into health records, including notes stored within the EHR system, to identify and validate other symptoms that may signify the presence of pancreatic cancer before an official diagnosis. In this case-control study, the research team will extract both common and rare symptoms of pancreatic cancer based on KPSC EHR data between 2005 and 2019 for cancer cases diagnosed between 2010 and 2019. They will compare the results to records of healthy individuals over the same time period. Symptoms found in pancreatic cancer patients’ medical records will be classified as early or late, based on when they arose in relation to the pancreatic cancer diagnosis.
The information gathered from this study may be incorporated into PanCAN’s Early Detection Initiative to further refine the criteria to determine which participants are at highest risk of having pancreatic cancer. There is also a potential for broader applicability to other medical institutions to incorporate into their EHR systems to flag symptoms that may allow more pancreatic cancer cases to be diagnosed in their earlier, more treatable stages.