Researchers analyzed registry data covering the entire Swedish adult population to explore new ways to identify melanoma risk. The dataset includes information such as age, gender, medical diagnosis, medication use, and socio-economic status. A total of 6,036,186 people were included, and 38,582 (0.64%) developed melanoma during the 5-year study period.
Much of the analysis was done by Martin Gillstedt.
“Our study shows that data already available within the healthcare system can be used to identify individuals at high risk of melanoma,” says Martin Gilstedt, PhD student at Sahlgrenska Academy, University of Gothenburg and statistician at the Department of Dermatology and Venereal Diseases at Sahlgrenska University Hospital. “Although this is not a form of decision support currently available in routine healthcare, our results clearly demonstrate that registry data can be used more strategically in the future.”
AI models improve melanoma risk prediction accuracy
Researchers evaluated several artificial intelligence models and found clear differences in performance. The most advanced models accurately distinguished between those who later developed melanoma and those who did not in about 73% of cases. In contrast, accuracy using only age and gender was approximately 64%.
By incorporating a broader range of factors, including diagnosis, medications, and sociodemographic information, the model was able to accurately identify a small group of individuals at significantly higher risk. Within these groups, the chance of developing melanoma within five years was approximately 33%.
Targeted screening could improve detection and efficiency
The study was led by Sam Polesy, associate professor of dermatology and venereology at the University of Gothenburg and dermatologist at Sahlgrenska University Hospital.
“Our analysis suggests that selectively screening small, high-risk groups could lead to both more accurate monitoring and more efficient utilization of health care resources. This includes incorporating population data into precision medicine and complementing clinical assessments.”
Towards a personalized melanoma screening strategy
Although the study results are promising, the researchers note that additional research and policy decisions are needed before this approach can be used in routine medicine. Still, the results highlight the potential of AI trained on large-scale registry data to support more personalized risk assessments and guide future melanoma screening strategies.
The study was carried out in collaboration between the University of Gothenburg and Chalmers University of Technology.
