New York City – Colorectal cancer kills 52,000+ Americans every year. Early detection could substantially improve mortality, but nearly half of eligible adults skip their recommended screenings. New research from Columbia Business School reveals a breakthrough: an AI model that identifies high-risk patients and recommends reaching out to them doesn't just dramatically boost colonoscopy rates by over 200%, it saves lives by reducing deaths by 6.2 percentage points within two years, which is a 43% decrease overall.
In the new research, "Cancer Screening Outreach Guided by Machine Learning: The Benefits of Proactive Care," a paper that was recently accepted for publication in the journal, Manufacturing & Service Operations Management, Professor Carri W. Chan, the Cain Brothers and Company Professor of Healthcare Management at Columbia Business School, and her co-authors Minje Park from The University of Hong Kong, Keith Boell, Elliot G. Mitchell, and David K. Vawdrey from Geisinger Health System, and Abdul A. Tariq from Children's Hospital of Philadelphia, studied an AI screening model used at Pennsylvania's Geisinger Health System starting in 2019. The AI model routinely analyzed blood test results from patients aged 51-75 who were overdue for colorectal cancer screening, scoring each patient's cancer risk based on blood markers, age, and gender. The program processed approximately 450 risk scores each week, analyzing over 62,000 total risk assessments during the study period. When patients scored above 0.150 on the risk scale, the system automatically flagged them for outreach, and nurse coordinators called these high-risk patients to explain their elevated risk and help schedule colonoscopy appointments. The researchers compared outcomes for patients who were flagged by the AI versus similar patients who scored just below the 0.150 threshold and didn't receive calls. The results were striking: flagged patients were 214% more likely to get a colonoscopy within three months and 117% more likely within six months. Most importantly, they were 43% less likely to die within two years.
This research proves AI can move beyond prediction to actually save lives, offering health systems nationwide a blueprint to dramatically improve cancer outcomes while making smarter use of limited resources. Similar AI-guided programs could transform screening for breast cancer, lung cancer, and other deadly diseases where early detection saves lives.
To explore research on the use of AI in healthcare at Columbia Business School visit our AI+Healthcare page.
