Analysis code accompanying the MOVER manuscript

These are the scripts used for the named cohort, numerical blood pressure processing, association models and reported audits. They require authorized MOVER data and prepared local intermediate tables. This archive is not a self-contained raw-data-to-publication runner. Participant-level intermediates and source credentials are deliberately not redistributed.

Run from a working directory with outputs/ and work/ subdirectories. Package versions used for statistical fits: pandas 3.0.6, numpy 2.5.3, statsmodels 0.15.0; scipy is required for ridge optimization. Some scripts prepend work/python_deps to sys.path; adapt that path to the approved local environment. Windows source archive paths must be adapted before executing scans. Inspect code before running scans, which may be resource-intensive.

Input contract and order
1. build_mover_formal_index_v02.py uses EPIC core/full information tables, events and the procedure-rule table. It defines first eligible patient-level operations and discharge chronology.
2. analyze_mover_epic_numeric_bp.py processes cleaned MAP blocks to case-level measurement summaries. Raw numeric scanning scripts document upstream extraction but depend on local archive organization.
3. extract_mover_bp_mortality_features.py joins work/mover_formal_analysis_v02.pkl, work/mover_epic_numeric_case_QC.pkl and work/epic_map_clean_blocks/*.pkl to produce the initial exposure table.
4. analyze_mover_bp_mortality_v01.py fits the initial models and fixes main-cohort IQR scales.
5. analyze_mover_bp_mortality_v02.py requires work/mover_BP_mortality_v02.pkl, which adds exact-60-second ARV and seven-day laboratory values.
6. scan_mover_mortality_pressors.py and audit_mover_severity_v03.py document preoperative support-record processing. analyze_mover_severity_v03.py requires the prepared severity table with documented-support indicators.
7. audit_mover_BP_QC_v04.py constructs strict-quality exposures; analyze_mover_BP_v04.py fits the quality, spline and ridge checks.

The ASA-5-excluded estimate uses the initial joint covariates plus log1p preoperative hospital days in the subset with ASA<5, with original main-cohort exposure scaling. It is documented in the manuscript and supplementary methods. This archive does not claim one-click reproduction of every extraction step. The exact procedure dictionary is supplied separately as Additional file 3; original linkage intermediates remain controlled by the data access conditions.

All analyses are observational. Scripts are research materials, not clinical decision-support software. No data usage rights are conveyed by this archive.
