AI maps the hidden forces shaping cancer survival worldwide

https://www.sciencedaily.com/releases/2026/01/260117053526.htm

Publish Date: 2026-01-17 09:59:00

Source Domain: www.sciencedaily.com

Scientists used machine learning, a form of AI, to identify factors linked to cancer survival in nearly every country globally.
Published in the Annals of Oncology, the study pinpoints specific policy changes or system improvements that could maximize impact on cancer survival in each nation.
The team created an online tool that lets users select a country and see the relation between cancer outcomes and factors such as national wealth, radiotherapy access, and universal health coverage.
Dr. Edward Christopher Dee emphasized the need for a data-driven framework to help countries tackle equity gaps in cancer mortality, highlighting access to radiotherapy, universal health coverage, and economic strength as significant levers.
The research combined cancer incidence and death data with health system information from over 185 countries, analyzing factors like health spending, number of healthcare workers, and levels of universal health coverage.
The machine learning model, developed by Mr. Milit Patel, provides estimates specific to each country and uses SHAP values to explain variable contributions.
The aim was to guide cancer system planning globally by providing actionable insights to prioritize health system investments.
Country-specific results show varying priorities: Brazil benefits most from universal health coverage, Poland from radiotherapy services, and Japan from radiotherapy centers, while the USA and UK focus on GDP per capita.
China faces mixed results; while national improvements yield gains in cancer control, out-of-pocket spending remains a significant barrier.
The study highlights that green bars represent key areas with the strongest positive association with improved outcomes, while red bars indicate areas less likely to show major differences currently.
Strengths of the study include comprehensive global coverage and transparent AI models; its limitations involve reliance on national-level data and variations in data quality across countries.