AI Models Identify High-Risk Melanoma Patients Using Routine Healthcare Data
Researchers at the University of Gothenburg developed AI models that identify people at elevated risk of melanoma by analyzing existing healthcare registry data. The study, based on information from Sweden's entire adult population, found that advanced machine learning models can distinguish future melanoma cases with 73% accuracy.
The analysis included 6 million adults tracked over five years. During that period, 38,582 people (0.64%) developed melanoma. Researchers examined age, sex, diagnoses, medication use, and socioeconomic status to train their models.
Better accuracy than age and sex alone
When researchers compared different approaches, the results were clear. A basic model using only age and sex correctly identified melanoma cases 64% of the time. The advanced model, incorporating diagnoses, medications, and sociodemographic data, reached 73% accuracy.
More importantly, the advanced model identified small population subgroups with substantially elevated risk. For these high-risk groups, the probability of developing melanoma within five years reached approximately 33%.
Potential for selective screening
The findings suggest that healthcare systems could use these risk assessments to target screening more efficiently. Rather than screening everyone equally, clinicians could focus resources on identified high-risk populations.
"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," said Sam Polesie, Associate Professor of Dermatology and Venereology at the University of Gothenburg.
The approach represents a shift toward precision medicine in dermatology. It combines population-level data with clinical judgment rather than replacing one with the other.
Next steps remain unclear
The researchers emphasized that implementation requires additional research and policy decisions. The current study demonstrates proof of concept using historical registry data, but healthcare systems would need to establish protocols, validate findings in other populations, and address implementation challenges before adopting the method.
Registry data already collected by healthcare systems contains patterns that can inform risk stratification. The study shows one application, but similar approaches could apply to other conditions where early identification affects outcomes.
For professionals in healthcare working with AI for Healthcare or AI Data Analysis, the research illustrates how machine learning models can extract actionable insights from routine administrative data without requiring new data collection.
Your membership also unlocks:
