An AI tool that uses detailed information from nearly all countries in the world has identified bespoke, national policy areas that could offer the greatest benefits in improving cancer survival.
The Global Health Predictor Insights Tool has been made freely available online and could help governments plan their cancer healthcare systems more effectively.
Machine learning highlighted national improvements in areas such as economic strength, radiotherapy accessibility and universal health coverage that could have the greatest impact on cancer outcomes.
The novel approach, published in the in the Annals of Oncology, could help prioritize resources to close survival gaps in the most equitable and effective way.
“Beyond simply describing disparities, our approach provides actionable, data-driven roadmaps for policymakers, showing precisely which health system investments are associated with the greatest impact for each country,” explained senior researcher Edward Dee, MD, from Memorial Sloan Kettering Cancer Center in New York.
He added: “As the global cancer burden grows, this model helps countries maximize impact with limited resources. It turns complex data into understandable, actionable advice for policymakers, making precision public health possible.”
The team studied the occurrence of cancer and deaths for 185 countries, spanning all income levels, recorded on the Global Cancer Observatory (GLOBOCAN 2022).
This was combined with demographic, economic and healthcare data from the World Health Organization, the World Bank, United Nations agencies and the Directory of Radiotherapy Centres.
These resources logged health spending as a proportion of gross domestic product (GDP), as well as GDP per capita, numbers of physicians, nurses, midwives and surgical workforce densities, and an index of universal health coverage.
They also included pathology availability, as well as an indices for human development and gender equality, radiotherapy centers per population, and the percentage of out-of-pocket expenditure.
The resulting machine-learning model created mortality-to-incidence ratios (MIR) that reflected the proportion of cancer cases ending in death and provided an indicator for the effectiveness of cancer care.
A Shapley Additive Explanations (SHAP) analysis revealed the magnitude of importance of each factor for each country, identifying those that might have most impact in specific national contexts.
In Brazil, universal health coverage had the greatest MIR benefits, indicating this is should be a focus for improvement. By contrast, there seemed to be less effect from pathology services, and the numbers of nurses and midwives per head of the population.
For Poland, the density of radiotherapy services, GDP per capita, and UHC index had the most importance, with general health spending having less impact.
While the density of radiotherapy centers had most impact in Japan, GDP per capita were most important in the U.S. and U.K.
GDP per capita was also important in China, as was its universal health coverage and density of radiotherapy centers. Less important here were out-of-pocket expenditure, surgical workforce per 1000 of the population and health spending as a percentage of GDP.
“The interpretability of country-specific SHAP profiles generated by our machine learning models offers new potential for directly informing national cancer control strategies,” the researchers maintained.
“Such tools enable policymakers to identify and prioritize interventions tailored to their unique health system bottlenecks and strengths, moving beyond generic recommendations toward precision public health approaches.
“Ultimately, adoption of explainable artificial intelligence (AI) and country-level modeling can help focus investments in resource-constrained settings, ensuring that expansions in infrastructure, work force, and coverage yield tangible improvements in cancer survival.”
