Explainable AI for early detection of Antimicrobial Resistance: An Approach to Reduce Clinical Risk

• Mariana Timbó de Oliveira UFRN
• Israel Martins Valcareggi UFRN
• Jean Paes Landim de Lucena UFRN
• Marise Reis de Freitas UFRN
• Ícaro Boszczowski USP
• Patrick Terrematte UFRN

DOI: https://doi.org/10.5753/eniac.2026.31449

Resumo

The appropriate prescription of antimicrobials is critical for the clinical outcome of patients with infectious diseases. However, the advancement of Antimicrobial Resistance (AMR) contributes to therapeutic ineffectiveness and worse prognoses. In this scenario, machine learning models have emerged as a promising approach to support the early assessment of AMR risk and assist clinical decision-making. Based on hospital admission records from the MIMIC-IV database, this work evaluates supervised machine learning models for predicting antimicrobial resistance. The features used in this study include demographic, clinical, microbiological, and derived variables. The target variable represents whether the bacterial isolate was resistant or susceptible to the tested antimicrobial. Random Forest, Logistic Regression, and LightGBM were evaluated. To improve model performance, class imbalance strategies and hyperparameter optimization with Optuna were applied. In addition, explainable AI methods were used to interpret the model outputs through SHapley Additive exPlanations (SHAP). The Random Forest achieved the best result, with an AUC of 0.88, approximately. The results suggest that the proposed approach has potential to support AMR risk assessment within the MIMIC-IV context, while further external and clinical validation remains necessary.

Palavras-chave: Antimicrobial Resistance, Machine Learning, Explainable AI, SHAP, MIMIC-IV, Random Forest

Referências

Chahine, E. B., Watanabe, J. H., Dong, M. T., and Andrade, R. M. (2026). Beyond resistance: Prioritizing safety in antimicrobial stewardship for older adults. The Senior Care Pharmacist, 41(3):88–97.

Geraldes, C., Tavares, L., Gil, S., and Oliveira, M. (2022). Enterococcus virulence and resistant traits associated with its permanence in the hospital environment. Antibiotics, 11(7):857.

He, H. and Garcia, E. A. (2009). Learning from imbalanced data. IEEE Transactions on Knowledge and Data Engineering, 21(9):1263–1284.

Hollenbeck, B. L. and Rice, L. B. (2012). Intrinsic and acquired resistance mechanisms in enterococcus. Virulence, 3(5):421–433.

Johnson, A., Bulgarelli, L., Pollard, T., Gow, B., Moody, B., Horng, S., Celi, L. A., and Mark, R. (2024). MIMIC-IV. PhysioNet. Version 3.1.

Lewin-Epstein, O., Baruch, S., Hadany, L., Stein, G. Y., and Obolski, U. (2021). Predicting antibiotic resistance in hospitalized patients by applying machine learning to electronic medical records. Clinical Infectious Diseases, 72(11):e848–e855.

Ministry of Health (Brazil) (2026). Antimicrobial resistance (AMR). Gov.br Portal.

Rosa, S. d. C. L., Carvalho, N. I., and Duani, H. (2024). Machine learning for microbial resistance prediction. Journal of Health Informatics, 16(Especial).

Santos, N. d. Q. (2004). Bacterial resistance in the context of hospital infection. Texto & Contexto – Enfermagem, 13(spe):64–70.

Silveira, B. P. (2023). Use of machine learning to predict antibiotic susceptibility through hospitalized patient data. Master’s thesis, Federal University of Rio Grande do Sul, Porto Alegre.

World Health Organization (2023). Global antimicrobial resistance and use surveillance system (GLASS) report 2023. Technical report, World Health Organization, Geneva.

World Health Organization (2024). WHO updates list of drug-resistant bacteria most threatening to human health. WHO News Release.