Predicting antimicrobial resistance in Africa using machine learning: a systematic review

Background

Antimicrobial resistance (AMR) is a global health emergency, posing a significant burden in Sub-Saharan Africa where surveillance capacity is limited. Artificial intelligence (AI) offers potential to enhance the prediction, monitoring, and control of AMR through routinely gathered data. This study aims to develop and validate state-of-the-art machine learning (ML) models using public surveillance data to predict resistance patterns across the region.

Methods

This study utilized country-level data from the WHO Global Antimicrobial Resistance Surveillance System (GLASS) combined with isolate-level susceptibility datasets. Following data harmonisation, baseline classifiers (logistic regression, random forests) were compared against sophisticated ML models, including gradient boosting, deep neural networks, and Long Short-Term Memory (LSTM) networks. Performance was measured via ROC-AUC, accuracy, and F1-score, with SHapley Additive exPlanations (SHAP) used to identify key resistance predictors.

Results

Advanced ML models outperformed traditional approaches, achieving ROC-AUC values of 0.86–0.91. Accuracy for high-burden pathogens like E. coli and S. aureus ranged from 82% to 89%. Temporal models improved the early detection of rising resistance trends by 20–30%. Analysis identified antibiotic class, historical resistance rates, pathogen species, and geographic location as the strongest predictors, with geographic variations in prevalence reaching 35%.

Conclusions

This research demonstrates that advanced AI models can significantly enhance AMR surveillance and prediction in Sub-Saharan Africa. Implementing ML tools into public health systems will facilitate timely interventions, optimize antimicrobial use, and support evidence-based policy. These findings provide a scalable procedure for strengthening regional surveillance and infection prevention.

Key messages

• Advanced machine learning models significantly outperform traditional methods in predicting resistance patterns in Sub-Saharan Africa.

• Integrating AI into community health surveillance can optimize antibiotic use and guide evidence-based public health policy.