Acta Derm Venereo
AI predicts high melanoma risk based on health data
Clinical takeaway: Registry-based risk modeling may help guide future melanoma outreach and screening.
Melanoma risk is difficult to target efficiently across large populations. Screening decisions typically rely on conventional clinical risk factors, rather than incorporating routinely collected health system data. A recent study set out to test if machine learning applied to commonly collected health data could be used to better understand who is most at risk.
Researchers analyzed Swedish registry data from more than six million adults, including age, sex, diagnoses, medications, and socioeconomic factors. They tested several computer models to see which best separated people who later developed melanoma from those who didn't, then used the strongest model to generate a five-year risk estimate for each person.
The most advanced model distinguished people who later developed melanoma from those who didn't in about 73% of cases, compared with about 64% when only age and sex were used. Adding diagnoses, medications, and sociodemographic data also identified small high-risk groups with about a 33% chance of developing melanoma within 5 years.
The findings suggest that routine registry data could help identify small groups at substantially higher melanoma risk for closer follow-up or more selective screening. Real-world use would depend on workflow, calibration, and whether targeted screening improves outcomes without adding too much overdiagnosis or burden.
“Our analyses suggest that selective screening of small, high-risk groups could lead to both more accurate monitoring and more efficient use of healthcare resources. This would involve bringing population data into precision medicine and supplementing clinical assessments,” said Sam Polesie, Associate Professor of Dermatology and Venereology at the University of Gothenburg and dermatologist at Sahlgrenska University Hospital.
The model now needs testing in other health systems and prospective calibration to determine whether identifying these high-risk groups improves screening efficiency or early detection without adding too much overdiagnosis, cost, or burden.
Source: Gillstedt M. Acta Derm Venereol. 2026 Apr 8. Predicting melanoma impact on the Swedish healthcare system from the adult population using machine learning on registry data