Reproducibility of machine learning applications in clinical informatics heavily relies on data preparation. However, preprocessing pipelines are not systematically reported in a consistent, standardized way, limiting adaptability and reproducibility when code and datasets are shared. To address this gap, we propose a set of best practices that define minimum, actionable principles for reporting preprocessing in clinical informatics, aligned with the FAIR principles, which stand for Findable, Accessible, Interoperable, and Reusable. Our framework encourages transparent, consistent, FAIR reporting of data preparation steps, making pipelines easier to understand, reuse, and compare across studies. We present FAIR4prep, a JSON-LD schema that operationalizes the machine-readable implementation of these best practices for FAIR-alignment preprocessing. By establishing a shared baseline for preprocessing documentation, this work aims to improve reproducibility, facilitate collaboration, and support more reliable translation into clinical settings.
Cobo et al. (2026) studied this question.