Jakov Kožić
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Jakov Kožić has a total of 5 co-authors. Here are the top co-authors by number of shared publications:
- Marta Krpan
School of Medicine, University of Zagreb, Zagreb 10000, Croatia
1 shared publication - Martin Bobek
School of Medicine University of Zagreb Zagreb Croatia
1 shared publication - Mila Kostić
School of Medicine, University of Zagreb, Zagreb 10000, Croatia
1 shared publication - Paula Bulić
School of Medicine, University of Zagreb, Zagreb 10000, Croatia
1 shared publication - Robert Likic
Division for Clinical Pharmacology and Therapeutics University Hospital Centre Zagreb Zagreb Croatia
1 shared publication
Latest Publications
- Open Access
Safety Monitoring of High-Risk Antibiotics Using Artificial Intelligence: A Narrative Review with Focus on Real-World Evidence
High-risk antibiotics remain indispensable in contemporary infectious diseases practice, yet they account for a disproportionate share of preventable toxicity, therapeutic drug monitoring complexity, and antimicrobial stewardship workload. Vancomycin, aminoglycosides, colistin, linezolid, daptomycin, selected beta-lactams, and amphotericin B are particularly challenging because clinically relevant exposure-toxicity relationships coexist with marked inter-patient variability and fragmented post-marketing safety surveillance. Artificial intelligence and real-world evidence are increasingly proposed as complementary approaches to address these limitations, although the evidence base remains heterogeneous and predominantly retrospective. This narrative review synthesises literature from PubMed/MEDLINE, Scopus, and Web of Science published between 2019 and April 2026, supplemented by citation chaining and regulatory pharmacovigilance resources, with studies prioritised by implementation maturity, external validation status, and stewardship relevance. Current evidence indicates that artificial intelligence may improve safety monitoring when embedded within clinically rich data environments: machine learning models show promising discrimination for nephrotoxicity and haematological toxicity in vancomycin, colistin, and linezolid therapy; natural language processing may enhance adverse drug event extraction from clinical text; and Bayesian, model-informed tools already demonstrate clinical utility in vancomycin and aminoglycoside dosing. However, prospective implementation data remain sparse, external validation is uncommon, and evidence that these tools improve real-world antibiotic safety outcomes, as opposed to predictive discrimination alone, remains limited. Artificial intelligence-enabled antibiotic safety monitoring is therefore transitioning from methodological promise towards conditional clinical utility rather than proven benefit. Near-term value is most likely to arise from integration with therapeutic drug monitoring, antimicrobial stewardship, and pharmacology-led clinical review rather than autonomous decision-making, with clinical pharmacologists and stewardship teams leading local implementation, validation, and governance of these tools.
Open AccessSafety Monitoring of High-Risk Antibiotics Using Artificial Intelligence: A Narrative Review with Focus on Real-World Evidence
High-risk antibiotics remain indispensable in contemporary infectious diseases practice, yet they account for a disproportionate share of preventable toxicity, therapeutic drug monitoring complexity, and antimicrobial stewardship workload. Vancomycin, aminoglycosides, colistin, linezolid, daptomycin, selected beta-lactams, and amphotericin B are particularly challenging because clinically relevant exposure-toxicity relationships coexist with marked inter-patient variability and fragmented post-marketing safety surveillance. Artificial intelligence and real-world evidence are increasingly proposed as complementary approaches to address these limitations, although the evidence base remains heterogeneous and predominantly retrospective. This narrative review synthesises literature from PubMed/MEDLINE, Scopus, and Web of Science published between 2019 and April 2026, supplemented by citation chaining and regulatory pharmacovigilance resources, with studies prioritised by implementation maturity, external validation status, and stewardship relevance. Current evidence indicates that artificial intelligence may improve safety monitoring when embedded within clinically rich data environments: machine learning models show promising discrimination for nephrotoxicity and haematological toxicity in vancomycin, colistin, and linezolid therapy; natural language processing may enhance adverse drug event extraction from clinical text; and Bayesian, model-informed tools already demonstrate clinical utility in vancomycin and aminoglycoside dosing. However, prospective implementation data remain sparse, external validation is uncommon, and evidence that these tools improve real-world antibiotic safety outcomes, as opposed to predictive discrimination alone, remains limited. Artificial intelligence-enabled antibiotic safety monitoring is therefore transitioning from methodological promise towards conditional clinical utility rather than proven benefit. Near-term value is most likely to arise from integration with therapeutic drug monitoring, antimicrobial stewardship, and pharmacology-led clinical review rather than autonomous decision-making, with clinical pharmacologists and stewardship teams leading local implementation, validation, and governance of these tools.