AMR Sentinel: a competing-risks reframing of AMR burden and a burden–funding–surveillance mismatch analysis (Vivli 2026 AMR Surveillance Open Data Re-Use Data Challenge)

Pre-registered, fully reproducible analysis pipeline for the 2026 Vivli AMR Surveillance Open Data Re-Use Data Challenge. Anchored on the SPIDAAR sub-Saharan-African hospital cohort (with Pfizer ATLAS and the Global AMR R&D Hub as cross-domain datasets), it reframes the resistance→mortality question as resistance-attributable excess bed-days under competing risks, formalises the adjusted SSA/LMIC mortality evidence in a Bayesian synthesis, quantifies the empiric-adequacy stewardship lever with a g-formula (shipped as a Streamlit tool), and builds a three-axis Cross-Domain mismatch — disease burden vs R&D funding vs disease surveillance — showing all three neglect the same sub-Saharan-African community Gram-negatives. Estimators are implemented directly in NumPy/pandas and unit-tested on synthetic data (155 tests); a self-checking confirmatory harness reproduces every reported headline number. Every deviation from the OSF pre-registration is logged.