Written by Siru Liu, Ph.D.
I was born and raised in Chengdu, China—a large city known for its pandas, a chill attitude towards life, spicy cuisine, tea culture, and folk music. In high school, I got involved in biomedical informatics research at West China Hospital, where I witnessed the transformative power of information technology in improving work efficiency. I also saw firsthand how immature tool design in electronic health records (EHR) could significantly impact healthcare professionals’ performance, especially considering the demanding workload in Chinese hospitals.
During my college years, I initially focused on studying statistics but gradually became captivated by the potential of computers and artificial intelligence (AI) to solve real-life problems. This led me to begin formal research in clinical decision support (CDS) at Harvard in 2016 under the guidance of Dr. Adam Wright, an esteemed expert in the field. I worked on a project detecting failures in CDS and gained valuable training in machine learning and natural language processing using healthcare data.
Driven by my passion for CDS, I pursued a Ph.D. in Biomedical Informatics at the University of Utah School of Medicine. Upon completing my doctoral work in 3.5 years, I recognized the importance of leveraging data-driven methods in CDS development and management. To further enhance my skills in CDS management, I returned to Dr. Adam Wright’s lab at Vanderbilt University Medical Center. I also actively engaged in career development activities such as ASPIRE on the Road, the Annual Career Symposium, the ASPIRE Networking Pacing Module, and Grant Pacing Workshops.
In 2022, I was honored to receive the NLM K99/R00 grant, which includes two years of mentored training and two years of independent research. It provides me with the remarkable opportunity to apply novel explainable AI (XAI) approaches to address the pressing issue of alert fatigue—a challenge encountered by many hospitals, including VUMC. My research aims to improve the logic of alerts and suppress unhelpful alerts. Through this work, I strive to develop a standards-based taxonomy of features affecting user response to alerts, a data-driven process for generating suggestions to enhance alerts, and expert-validated suggestions. Another direction of my research is focused on using large language models to improve CDS. In a recent study, I used ChatGPT to generate suggestions to improve alert logic, and the results were promising. Out of the 20 top-rated suggestions, nine were generated by ChatGPT.
Beyond my professional pursuits, I find joy in exploring nature, science, and art. Recently, I have been spending quality time with my 5-month-old Blenheim Cavapoo puppy named Culry. She is incredibly intelligent, friendly, and calm, bringing boundless happiness to my life. I also enjoy reading, cooking new recipes, making cocktails, and hiking. The most recent dishes I cooked were Volcano Ribs, street food from Bangkok, and Marc Forgione’s Bang Bang Shrimp in Lao Spicy & Sour Sauce with a glass of osmanthus-infused gin.
Overall, I am grateful for the opportunities that have shaped my journey and excited to improve the efficiency and quality of healthcare using AI.