Geisinger AI Outreach Program Associated With 6.2% Absolute Drop in Two-Year Mortality for Colorectal Cancer Screening

The observational evaluation examined Geisinger’s 2019-initiated outreach program, which applied gradient-boosted trees to complete blood count values, age, and sex to rank patients overdue for colorectal screening. Nurse coordinators then contacted high-risk individuals for education and scheduling. Compared with similar non-contacted patients, the intervention arm showed higher screening uptake at both three and six months, with the mortality benefit emerging within the two-year follow-up window.

Colorectal cancer remains the second-leading cause of U.S. cancer deaths, yet screening rates plateau below targets. The Geisinger results demonstrate that AI can function as operational infrastructure rather than isolated prediction, converting risk scores into scheduled procedures at scale. This shifts the value proposition from diagnostic accuracy alone to measurable population-level survival gains.

The study design was quasi-experimental and single-site; residual confounding from unmeasured patient engagement differences cannot be excluded. Capacity constraints on endoscopy suites and variable outreach fidelity across systems remain open questions. Multi-center cluster-randomized trials with longer follow-up are required to confirm transportability and cost-effectiveness.

Next steps include testing the same framework for lung and breast cancer outreach and embedding automated capacity forecasting to prevent screening backlogs when programs expand.