A University student developed a machine-learning model to identify factors associated with cancer outcomes across the globe, according to a Jan. 14 study. 

Biochemistry senior Milit Patel developed an interactive tool that combines global cancer data with countries’ health system information to analyze how strongly different factors correlate with cancer survival rates in different countries. 

The tool was the focus of a study, co-led by Patel, and published in the cancer research journal Annals of Oncology. The platform allows policymakers to compare countries and identify which metrics, like gross domestic product and universal health coverage, correlate with cancer survival. The goal of the tool is to understand how much each factor affects cancer survival rates at an individual nation’s level.

Traditional linear models typically used in global cancer research may not always accurately portray the relationship between metrics and cancer survival rate because they do not fully account for the nuanced country-and context-specific interactions between health care system factors and cancer outcomes, according to the study. A predictor improving outcomes in one country may have a different effect in another, making the results highly country-specific, Patel said.

 “We have a good amount of evidence that supports the association between health systems and cancer outcomes,” wrote Dr. Edward Christopher Dee, resident physician in radiation oncology at Memorial Sloan Kettering Cancer Center and co-corresponding author of the study, in an email. “We recognize that having more providers or having more resources overall is important, but how should countries prioritize different levers in the health system?”

The model Patel developed is supposed to make seeing associations in the data easier for policymakers. The model generated thousands of data points, which were then used to create simple interactive visualizations. 

“(The policymakers are) not going to … want to ingest all of the data themselves,” Patel said. “(It is) for someone who wanted to look at the data but look past the figures in the paper.”

According to the research paper, global cases and deaths are projected to rise sharply through 2050, particularly in “resource-constrained areas.” The study emphasizes policymakers should not take specific findings and factors as a cause of higher or lower cancer survival rates, the researchers said. 

“I think of it as a guide rather than a definitive answer,” wrote Dr. James Fan Wu, physician fellow at Medical College of Wisconsin and another co-author of the study, in an email. “It helps highlight which health system factors appear most closely linked to cancer outcomes in a specific country, which can be useful for setting priorities. But it’s not saying that fixing one variable will automatically improve outcomes, and it shouldn’t be used to rank countries or assign blame.”

Policymakers should consider the data but also examine work from other fields, such as health and economics, when making policy decisions that can affect patients’ lives, Dee said in an email. 

“Our hope is that this work can be used to inform cancer policy decisions throughout the world,” Dee wrote in an email.