Code and Analytic Output for: Frontal versus Kors-Derived Spatial QRS-T Angle and 27-Year Heart-Disease Mortality in NHANES III

This repository contains the complete, reproducible computational pipeline, survey-weighted analytical code, tabular results, and execution provenance for the study: "Frontal versus Kors-derived Spatial QRS-T Angle and 27-Year Heart-Disease Mortality: A nationally representative competing-risks and incremental-value analysis — NHANES III (1988–1994) linked to the National Death Index through 2019." Overview The study evaluates the long-term prognostic value of the frontal versus reconstructed spatial QRS-T angle over 27 years (126,945 person-years; 1,048 heart-disease deaths) among 6,325 participants aged 40+ free of baseline cardiovascular disease in NHANES III. Key methodological pillars include: Head-to-head comparison: Direct evaluation of whether the free, automated frontal QRS-T angle loses prognostic value relative to the Kors-reconstructed spatial angle. Competing-risks framework: Cause-specific cumulative incidence using Aalen-Johansen and Fine-Gray models to correct for non-cardiac death overstatement by Kaplan-Meier. Non-proportional hazards: Landmark and period-specific analyses capturing time-attenuation of repolarization risk over 27 years. Incremental discrimination & net benefit: Delta Harrell's C, time-dependent AUC, continuous/categorical NRI, IDI, calibration deciles, and decision-curve analysis (DCA). Biological mediation: Quantification of excess cardiac mortality mediated through ECG-estimated left ventricular mass index (LVMI). Contents of this Archive QRST_angle_NHANES3_high_impact_analysis.ipynb: Fully executed Python notebook containing data intake, automated sentinel cleaning, cohort derivation, multiple imputation (M=20), survival modeling, and visual generation. design_based_analysis.R: Standalone R script for survey-weighted Cox regression (survey::svycoxph) and Fine-Gray subdistribution modeling (cmprsk::crr). provenance/run_manifest.json: Machine-readable audit file logging runtime parameters, package versions, and complete analysis specifications. provenance/data_provenance.json: SHA-256 cryptographic digests, byte counts, and CDC source URLs for all raw input files. tables/: Full machine-readable CSV tables (baseline characteristics, cause-specific Cox models, period-specific effects, reclassification metrics, and sensitivity analyses). figures/: Publication-ready figures (STROBE flow, Aalen-Johansen cumulative incidence, adjusted hazard ratios, spline dose-response, and Bland-Altman agreement). Data Access and Compliance Notice In strict compliance with the National Center for Health Statistics (NCHS) Data Use Agreement and federal confidentiality statutes, raw participant-level microdata are not redistributed in this repository. All survey data (exam, laboratory, adult interview, and ECG data) and linked National Death Index (NDI) records are programmatically fetched from official CDC endpoints and authenticated via SHA-256 checksums embedded in the analysis code.