Creating strong predictive models in oncology.

Gensheimer, Michael F · Patterns (N Y) · 2026

review · Level V

Where this comes from

Abstract

Many oncology predictive models fail to improve care. Issues include risks of bias, underpowered radiomics studies, and limited clinical impact. A path forward involves an emphasis on clinically actionable questions, rigor, and generalizability.