Oxygen prescription as diagnosis: arithmetic limits of the SpO₂/FiO₂ ratio and a randomised demonstration of target-dependent misclassification — protocol, analysis code and result records

This deposit contains the pre-specified analysis protocol, the analysis code and the machine-readable result records for a study of how the SpO2/FiO2 (S/F) ratio behaves arithmetically when it is used as a substitute for PaO2/FiO2 (P/F) in respiratory-failure severity criteria. The protocol fixes the estimand, the P/F-to-S/F conversion rule, the trial eligibility and extraction rules, and the inference procedure together with its stopping rules, in advance of any confirmatory computation. The analysis code reproduces every reported quantity from arm-level summary statistics extracted from published trial reports and their supplements. Contents: the registration protocol (version 7) with its full version history; the Python analysis scripts; the result records those scripts emit; and a SHA-256 manifest of every file, so that the state of the protocol at registration can be checked against the files used in the analysis. The manuscript reporting the results is not part of this version and will be deposited or linked separately. No individual patient data and no identifiable information are contained in this deposit. All inputs are published aggregate data. Version 2 (15 September 2026) adds one script and one result record: analysis/sofa_cutoff_vacuity.py and data/calculations/C-NIRS-SFPF-14/result.json, a post-hoc corollary not in the protocol, applying the boundary formula of protocol section 4 to the S/F cut-offs of the respiratory SOFA component (Pandharipande et al. 2009). Arithmetic only, no inference. Every file carried over from version 1 is byte-identical; the regenerated manifest marks each file unchanged or new, and CHANGES-v2.md lists the additions.