Abstract A027: Artificial intelligence for detection of treatment-related adverse events using electronic health records in oncology: Focus on Sub-Saharan African (SSA) countries

Abstract Background: The rapid influx of cancer drugs in Sub-Saharan African (SSA) Countries necessitates robust adverse event (AE) identification and reporting. Identification and reporting of AEs is mainly hindered by 1) high patient volumes; as oncologists see 325 new consults annually, nearly double the global 175. 2) Only 22% Electronic Health Records (EHR) access; 80% of data are in unstructured notes hence toxicity monitoring is failing. Low adherence (as low as 21.2% in Tanzania) highlights the need for AI and Natural Language Processing (NLP) to automate AE detection. Methods: This review evaluates current AI methodologies for AE detection in SSA, focusing on the transition from manual monitoring to automated extraction. We analyse evidence from high-income countries (HICs) and appraise the feasibility of implementing these tools in the SSA healthcare infrastructure. The study explores data source integration (clinical notes, labs, radiology) and validation requirements using the Common Terminology Criteria for Adverse Events (CTCAE) as a gold standard. Results: AI applications in SSA have mostly prioritized in infectious diseases like tuberculosis and HIV and cancer is left behind despite the continent facing a projected 102% increase in cancer incidence by 2040. Tanzania possesses only 8.42 clinicians and nurses per 10,000 population, well below the WHO-recommended threshold of 22.8. This scarcity makes manual toxicity monitoring nearly impossible. NLP models demonstrate high accuracy in identifying toxicities in HIC settings, but a significant gap remains for SSA. Conclusions: AI-based AE detection is a promising approach for improving patient safety in resource-limited oncology settings. To translate these tools into SSA countries clinical practice, research must prioritize locally relevant, annotated datasets and prospective validation within real-world workflows. Citation Format: Elinda Kuhoga, Agness Ndunguru, Prosper Mgomi. Artificial intelligence for detection of treatment-related adverse events using electronic health records in oncology: Focus on Sub-Saharan African (SSA) countries [abstract]. In: Proceedings of AACR Drug Discovery and Development (AACR D3) Conference; 2026 Jul 21-24; Boston, MA. Philadelphia (PA): AACR; Clin Cancer Res 2026;32(14_Suppl):Abstract nr A027.