Machine Learning Uncovers Global Cancer Outcome Drivers

NEW research shows that machine learning can pinpoint why cancer survival rates vary between countries. By analysing national health system data from 185 countries, researchers identified the strongest system-level drivers of cancer outcomes using interpretable models. Global Cancer Outcomes and Health System Inequities Despite advances in cancer prevention and treatment, large disparities persist in cancer mortality worldwide. Mortality-to-incidence ratios, commonly used as a proxy for cancer survival, remain substantially higher in many low- and middle-income countries. These differences reflect unequal access to diagnostics, treatment infrastructure, and financial protection, rather than biological variation alone. To address this challenge, researchers applied machine learning to understand how national health system characteristics shape cancer outcomes. Rather than relying on single indicators, the study integrated economic, infrastructure, workforce, and coverage measures to capture the complexity of cancer care delivery across diverse settings. The aim was not only prediction, but also policy relevance through transparent and interpretable modelling. Interpretable Machine Learning Reveals Key Drivers Using GLOBOCAN 2022 data, national mortality-to-incidence ratios were modelled for 185 countries with a CatBoost gradient-boosting algorithm. The machine learning framework incorporated indicators such as GDP per capita, universal health coverage index, radiotherapy centres per population, health spending as a percentage of GDP, workforce densities, and pathology availability. Model performance was strong, with an R2 of 0.852 (95% CI, 0.801-0.891), an RMSE of 0.057, and a correlation of r = 0.923 (P = 8.30 × 10−78). Globally, GDP per capita accounted for 22.5% of model contribution, followed by radiotherapy availability at 15.4% and universal health coverage at 12.9%. Country-specific analyses revealed marked heterogeneity, showing that dominant drivers varied substantially depending on national context. Policy Implications for Cancer Care Planning These findings demonstrate that machine learning can support evidence-informed cancer policy by identifying actionable system-level priorities. For clinicians and policymakers, the results reinforce that expanding radiotherapy capacity and strengthening universal health coverage may yield greater improvements in outcomes than increasing total health spending alone. The analysis provides a valuable framework for generating testable hypotheses and guiding prospective evaluations aimed at reducing global cancer mortality. Reference Patel MS et al. Machine learning reveals country-specific drivers of global cancer outcomes. Annals of Oncology. 2026;DOI:10.1016/j.annonc.2025.11.014.