For the first time, researchers have used machine learning – a type of artificial intelligence (AI) – to identify the most important drivers of cancer survival in nearly all the countries in the world.
The study, which is published in the leading cancer journal Annals of Oncology [1] today (Wednesday), provides information on which improvements or policy changes can be made in each country that would have the greatest impact on improving cancer survival. By going to the online tool created by the researchers, anyone can find their country and scroll down to see which factors, such as national wealth, access to radiotherapy and universal health coverage, are most associated with cancer outcomes.
Dr Edward Christopher Dee, a resident physician in radiation oncology at Memorial Sloan Kettering (MSK) Cancer Center, New York, USA, who co-led the research, said: "Global cancer outcomes vary greatly, largely due to differences in national health systems. We wanted to create an actionable, data-driven framework that helps countries identify their most impactful policy levers to reduce cancer mortality and close equity gaps.
"We found that access to radiotherapy, universal health coverage and economic strength were often important levers being associated with better national cancer outcomes. However, other key factors were relevant as well."
Dr Dee and his colleagues used machine learning to analyse data on cancer incidence and deaths from the Global Cancer Observatory (GLOBOCAN 2022) for 185 countries. They also collected information on health systems from the World Health Organization, the World Bank, United Nations agencies and the Directory of Radiotherapy Centres. This included health spending as a percentage of GDP, GDP per capita, the numbers of physicians, nurses, midwives and surgical workforce per 1000 of the population, universal health coverage, the availability of pathology service, an index of human development, the number of radiotherapy centres per 1000 of the population, a gender inequality index, and the percentage of out-of-pocket expenditure.
Mr Milit Patel created the machine-learning model based on these global health systems data. He is the first author of the study and a researcher in biochemistry, statistics and data science, healthcare reform and innovation at the University of Texas at Austin, USA, and at MSK.
Mr Patel said: "We chose to use machine learning models because they allow us to generate estimates – and related predictions – specific to each country. We are, of course, aware of the limitations of population level data but hope these findings can guide cancer system planning globally."
The model generates mortality-to-incidence ratios (MIR), which reflects the proportion of cancer cases that result in death and which serves as a proxy for the effectiveness of cancer care. It uses a method of explaining individual predictions by quantifying each factor's contribution to the prediction; this is called SHAP (Shapley Additive exPlanations).
Mr Patel said: "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. As the global cancer burden grows, these insights can help nations prioritise resources and close survival gaps in the most equitable and effective way possible. International organisations, healthcare providers, and advocates may also use the web-based tool to highlight areas for investment, especially in resource-limited settings."
To take Brazil as an example, the model shows that the factor that has the greatest positive impact on the mortality-to-incidence ratio is universal health coverage (UHC), while pathology services, and nurses and midwives per 1000 of the population may not have as substantial an effect on outcomes. The researchers say this suggests that Brazil should make UHC a priority.
As another example, the density of radiotherapy services, GDP per capita, and UHC index demonstrated the greatest impact in Poland on cancer outcomes, among other key factors. This finding suggests that current efforts to strengthen health insurance and service access has led to more pronounced gains than factors such as general health spending, which shows a smaller positive effect.
In Japan, the USA and the UK, the data show that all health system factors are linked to improved cancer outcomes, with the density of radiotherapy centres in Japan, and GDP per capita in the USA and the UK having the greatest impact. This suggests that these are the factors on which policymakers should focus.
The picture is more mixed in China. GDP per capita, UHC and density of radiotherapy centres are the factors that contribute the most to better cancer outcomes. Out-of-pocket expenditure, surgical workforce per 1000 of the population and health spending as a percentage of GDP are factors that are less likely to explain the differences in cancer outcomes at present.
In their paper, the researchers write about China: "High direct costs for patients remain a critical barrier to optimal cancer outcomes, even amidst national improvements in health financing and access. These findings underscore that while China's rapid health system development is yielding important gains in cancer control, disparities in financial protection and coverage persist, warranting intensified policy focus on reducing out-of-pocket expenditures and further strengthening UHC implementation to maximise health system impact."
Mr Patel explained the significance of the green and red bars that show in the graphs for each country. "The green bars represent factors that currently appear most strongly and positively associated with improved cancer outcomes in a given country. These are areas where continued or increased investment is most likely to result in meaningful impact. However, the red bars do not indicate that these areas are unimportant or should be neglected. Rather, they reflect domains that, according to the model and current data, are less likely to explain the largest differences in outcomes right now. This may be due to already strong performance in these aspects, limitations of the available data, or other context-specific factors.
"Importantly, seeing a 'red' bar should never be interpreted as a reason to stop efforts to strengthen that pillar of cancer care – improvement in those areas can still be valuable for a country's overall health system. Our results simply suggest that, if the goal is to maximise improvement in cancer outcomes as defined by the model, focusing first on the strongest positive (green) drivers may be the most impactful strategy."
Strengths of the study include: coverage of nearly all countries, use of up-to-date global health data; provision of actionable, country-specific policy recommendations (not just global averages), and more explainable AI models. Limitations include: reliance on aggregated, national-level data, not individual patient records; variability in registry and data quality, especially in many low-income countries; national trends may overlook within-country disparities, which merit further investigation; and the study is not able to prove that focusing on particular area will cause an improvement in cancer outcomes, only that such efforts may be associated with cancer outcomes. The findings will help policymakers prioritise, but further studies that intervene in particular areas are needed.
Dr Dee concluded: "As the global cancer burden grows, this model helps countries maximise impact with limited resources. It turns complex data into understandable, actionable advice for policymakers, making precision public health possible."
