Overview
- Using nationwide Swedish registries, the study tracked 6,036,186 adults over five years and found 38,582 melanoma cases.
- The best machine‑learning model correctly separated future melanoma cases from non‑cases about 73% of the time, compared with about 64% using only age and sex.
- By combining diagnoses, medications and sociodemographic data, the models isolated very small subgroups with roughly a 33% chance of melanoma within five years.
- The authors say selective screening of these high‑risk groups could catch more cancers sooner and make better use of clinic time, though they call for more studies and policy guidance before rollout.
- The work, led by Sam Polesie with key analysis by Martin Gillstedt, was published in Acta Dermato‑Venereologica and involved the University of Gothenburg and Chalmers University of Technology.