A system for independent evaluation of clinical prediction models while preserving intellectual property and data privacy

Background Evaluating clinical prediction models (CPMs) in new patients and settings is essential for building the evidence needed to support their adoption in practice. Independent evaluation by researchers distinct from the CPM developers is particularly valued, yet remains underutilised due to barriers of data access, governance, and technical complexity. Methods Here we propose a novel two-component system enabling any researcher to conduct a fully independent evaluation of a static CPM, with no requirement to share patient data, no specialist infrastructure, and minimal code. We introduce evaluatr, a free R package that implements the workflow. Results The system lowers the barrier to evaluation while providing developers with confidence that CPMs are evaluated exactly as intended. The system also resolves a growing tension between open science and commercial viability: model parameters remain under the developer's control and are protected throughout, preserving intellectual property while enabling evaluation to proceed. Conclusions The proposed system provides a viable pathway toward widespread evaluation and ultimately impact for promising CPMs, where few routes currently exist.