Clinical Environmental Reasoning Benchmark: RC-01 Reproducibility Archive

Abstract

This RC-01 reproducibility archive is publicly available through Zenodo at https://doi.org/10.5281/zenodo.22698036. The corresponding GitHub development repository remains private at the time of this release. This archive contains the formal N4-v1.0 specification, deterministic benchmark generator and fixtures, reference implementation, automated tests, verification workflow, and frozen reproducibility evidence supporting “Development and Computational Verification of a Theory-Informed Clinical Environmental Reasoning Framework for Addiction Treatment”. There are 490 deterministic formal rule-testing fixtures: 370 primary + 120 counterfactual. All 11/11 verification families passed with 0 failures; seven executable tests passed. The fixtures are not synthetic patients. Results establish implementation fidelity/internal computational coherence and reproducibility only. No human participants or patient data were used. Clinical validation remains future work. The 490 cases are deterministic formal rule-testing fixtures, not synthetic patients or a simulated patient population. Results support formal specification, deterministic implementation fidelity, internal computational coherence, exact reproducibility, and the specified uncertainty-preserving, compensation-aware, comparative-driver and counterfactual-accountability behaviors only. They do not establish clinical validation, content validity, inter-rater reliability, construct validity, predictive validity, sensitivity/specificity, calibration, clinical utility or safety, causal identification in patients, ASAM level-of-care validity, synthetic-patient realism, or population fidelity. The addiction-specific computational study operationalizes selected constructs of the Theory of Clinical Environmental Reasoning (TCER). The separate theory manuscript, “Theory of Clinical Environmental Reasoning: A Middle-Range Nursing Theory of Consequence, Driver Attribution, and Reassessment”, is submitted to Journal of Advanced Nursing and under editorial consideration, manuscript ID 5971775 (author-reported status). Computational verification does not empirically validate TCER. No PHI was used. Prospective human validation remains future work. TCER is proposed and unvalidated. The JAM study operationalizes selected TCER constructs in an addiction-specific methodological/computational context; it does not empirically validate TCER. Mixed licensing: Software source code (*.py), tests, scripts, build/configuration files (pyproject.toml, workflow YAML, .gitignore and .gitattributes) are licensed under Apache-2.0. Documentation, benchmark descriptions, fixture CSVs, the N4 JSON specification, result/verification JSON, citation and archival metadata, and other author-owned non-code scholarly materials are licensed under CC-BY-4.0. These are separate file scopes, not an option to use either license for every file. Official license texts and third-party dependencies retain their own terms. Published version DOI: 10.5281/zenodo.22698036. Concept DOI: 10.5281/zenodo.22698035. Record 22698036, version 1.0.0-rc01, is published and the archive is publicly available.