Reproducible pipeline and verification ledgers for "Estimating Drug-Drug Interaction Alert Burden Before Deployment" (NHANES 2017-2018 + DDInter + RxNorm)

This archive contains the full reproducible pipeline, raw public inputs, run logs, and verification ledgers for the study "Estimating Drug-Drug Interaction Alert Burden Before Deployment: A Reproducible Framework Using Nationally Representative Public Data". The pipeline links NHANES 2017-2018 prescription files (14,434 named prescriptions from 4,045 medication users) to the DDInter interaction knowledge base (128,680 unique rule pairs) by normalized-name matching, yielding 6,001 potential DDI alerts (1.4836 per user; 37.65% of users with at least one alert; 41.58 per 100 prescriptions; 593 Major). It includes dual-caliber construction (all within-person pairs vs true co-prescription pairs), survey-weighted estimates with design-based confidence intervals (Taylor linearization, WTMEC2YR), a Major-only sensitivity variant (593 alerts; 9.79% of users), external triangulation against FAERS via openFDA (all 30 highest-volume pairs have co-reports) and TWOSIDES (92.3% of evaluable pairs carry significant signals), and a 71-item verification ledger. Two independent full runs (2026-10-26 and 2026-11-02) are key-by-key identical on every data field. See README.md for the reproduction guide.