Time-origin misalignment in registry-derived baseline tables: analysis code and locked result tables

Analysis code and locked result tables for the article “Time-origin misalignment in registry-derived baseline tables inflates the apparent advantage of tree ensembles over regression” (Fang YW, Huang YC, Wang JT, Tsai MH). Contents. result_tables/: the locked result files from which every number in the manuscript and its Supplementary Material is derived. scripts/: the analysis and figure scripts, with English documentation. kidit_survival/: the prediction-pipeline package (cohort derivation, temporal folds, candidate grid, preprocessing, metrics). environment/: pip lock files for the two Python environments. MANIFEST.txt: the SHA-256 of every file. README.txt describes each script, the environment it runs in and whether it runs from this package alone. Not included. The source registry files and three result files with one row per patient are withheld. The records are deidentified but remain re-identifiable: the cohort is from a single centre, and follow-up time and dialysis vintage are recorded precisely enough to reconstruct calendar dates. The per-patient anchor intervals are replaced by an aggregated table. scripts/18_simulation.py needs no source data and runs independently. Licences. Code (scripts/, kidit_survival/) is released under the MIT License; the result tables and documentation under CC BY 4.0.