MIC Drift Early-Warning Model for Ghana, Africa and Global AMR Surveillance

Antimicrobial resistance surveillance usually detects threats after clinical breakpoints are crossed. This project will develop a MIC Drift Early-Warning Model that uses the full minimum inhibitory concentration (MIC) distribution to identify emerging resistance signals earlier, before conventional susceptible/intermediate/resistant summaries show clear deterioration. We will request and harmonize all Vivli AMR Register datasets eligible for the 2026 Data Challenge, excluding Merck datasets, and build a reusable Ghana-Africa-global modelling pipeline. The analysis will focus on repeated organism-antimicrobial-country-year combinations, stratified where possible by infection type, specimen type, age group and sex. Ghana will serve as the national demonstration case; other African countries will provide regional validation; the full global dataset will be used to benchmark generalizability and detect cross-regional emergence patterns. Our primary scientific question is: can longitudinal shifts in MIC distributions predict future resistance burden and identify high-risk organism-drug-country combinations earlier than standard breakpoint-based surveillance? We will model MICs using censored/interval regression, hierarchical Bayesian models, quantile regression, and change-point detection. For each organism-antimicrobial-country-year stratum, we will estimate an Early Warning Score based on upward MIC drift, increasing upper-tail MICs, proximity to CLSI/EUCAST breakpoints, acceleration over time, and uncertainty due to sample size. Model performance will be assessed by temporal validation: earlier years will be used to predict later resistance classifications and rising non-susceptibility. The project will generate an open, reproducible dashboard and codebase showing early-warning maps for Ghana, Africa and global comparisons. Outputs will include ranked priority pathogen-antimicrobial combinations, uncertainty-aware visualizations, and a policy-facing framework for when surveillance systems should intensify testing, stewardship, laboratory review, or guideline reassessment. The expected impact is a transferable early-warning tool that converts historical MIC surveillance into actionable intelligence for AMR preparedness, antibiotic stewardship, and national and international AMR policy. If feasible, Global AMR R&D Hub data will contextualize early-warning threats against investment gaps.