Forecasting Antimicrobial Resistance Trends Using Machine Learning on WHO GLASS Surveillance Data: A Retrieval-Augmented Generation Approach for Policy Decision Support

TL;DR

This study develops a machine learning framework using WHO GLASS data to forecast antimicrobial resistance trends and supports policy decisions with a retrieval-augmented generation system that provides source-attributed policy answers.

Contribution

It introduces a novel two-component framework combining AMR trend forecasting with an RAG-based policy support system, leveraging multiple ML models and a language model with document retrieval.

Findings

XGBoost achieved the best forecasting performance with MAE of 7.07% and R-squared of 0.854.

Prior-year resistance rate is the most important predictor for future resistance.

Regional MAE varies from 4.16% in Europe to 10.14% in South-East Asia.

Abstract

Antimicrobial resistance (AMR) is a growing global crisis projected to cause 10 million deaths per year by 2050. While the WHO Global Antimicrobial Resistance and Use Surveillance System (GLASS) provides standardized surveillance data across 44 countries, few studies have applied machine learning to forecast population-level resistance trends from this data. This paper presents a two-component framework for AMR trend forecasting and evidence-grounded policy decision support. We benchmark six models -- Naive, Linear Regression, Ridge Regression, XGBoost, LightGBM, and LSTM -- on 5,909 WHO GLASS observations across six WHO regions (2021-2023). XGBoost achieved the best performance with a test MAE of 7.07% and R-squared of 0.854, outperforming the naive baseline by 83.1%. Feature importance analysis identified the prior-year resistance rate as the dominant predictor (50.5% importance), while regional MAE ranged from 4.16% (European Region) to 10.14% (South-East Asia Region). We additionally implemented a Retrieval-Augmented Generation (RAG) pipeline combining a ChromaDB vector store of WHO policy documents with a locally deployed Phi-3 Mini language model, producing source-attributed, hallucination-constrained policy answers. Code and data are available at https://github.com/TanvirTurja