Research in Focus shines a light on the innovative studies and discoveries taking shape across the UBC Department of Anesthesiology, Pharmacology & Therapeutics. Through each feature, we celebrate the minds driving meaningful change in research, education, and clinical practice across anesthesiology, pharmacology, and therapeutics.
Dr. Janny Ke and her team have developed predictive models that could transform how doctors monitor patients after elective colectomy. Using data from over 130,000 patients in the National Surgical Quality Improvement Program between 2014 and 2019, they created and validated time-to-event models to forecast major medical complications within 30 days of surgery.
Their analysis revealed striking differences in when complications typically occur: while bleeding requiring transfusion tended to happen almost immediately (within the first day after surgery), conditions like venous thromboembolism appeared much later, around 12 days postoperatively. Complication rates ranged from as low as 0.3% for cardiac arrest and acute renal failure to 5.3% for bleeding requiring transfusion, with an overall readmission rate of 8.6%.
The models showed strong predictive power for mortality, myocardial infarction, pneumonia, and renal failure, but were less accurate for readmission, venous thromboembolism, and sepsis. Once further validated, these tools could help clinicians identify high-risk patients and tailor monitoring to the periods when they are most vulnerable, potentially improving outcomes after surgery.
“Accurate risk prediction that responds to a patient’s perioperative course can help create monitoring plans tailored to individual risks and periods of vulnerability, optimizing both healthcare resources and outcomes.”
— Janny Ke, Clinical Assistant Professor, UBC
Meet Janny Ke!
Site: Providence Healthcare
Rank: Clinical Assistant Professor
Dr. Ke has always been passionate about research. Curious about the integration of artificial intelligence in medicine, she completed a Master of Science in Epidemiology, focusing on Big Data, prediction modeling, and population studies through her thesis and coursework. She credits numerous research mentors for their guidance throughout her career, particularly during her residency at Dalhousie, her Master’s program, and her work at UBC.
Responses have been edited for flow, clarity, and style.
What’s one thing you hope people will take away from this study?
This was a proof-of-concept set of models designed to predict not only who may experience complications, but also when those complications might occur. While some models performed well, others did not. Additional high-quality clinical data are needed to improve our ability to develop and evaluate prediction models.
How does this project fit into your broader research interests or goals?
Accurate risk prediction that responds to a patient’s perioperative course can help create monitoring plans tailored to individual risks and periods of vulnerability, optimizing both healthcare resources and outcomes. This capability is likely to improve with the use of artificial intelligence. However, prediction models require high-quality clinical data—which is often lacking—as well as rigorous validation and assessment to ensure accuracy and equity. As artificial intelligence becomes more integrated into risk prediction, clinician involvement in model development is essential. Education and standardization are also needed to enable clinicians to interpret and use these models effectively, while remaining aware of their limitations and potential biases.
Outside of work, what do you enjoy the most?
I love hiking, climbing, painting, and playing the piano, though these days I’m mostly bossed around by toddlers.
Read the publication: Development and internal validation of time-to-event risk prediction models for major medical complications within 30 days after elective colectomy