The impact of artificial intelligence on the prescribing, selection, resistance, and stewardship of antimicrobials: a scoping review

Abstract

Background

Antimicrobial selection, prescribing, and resistance are global health issues resulting from the overuse and misuse of antimicrobials in the healthcare and agricultural sectors. It raises healthcare costs, prolongs diseases, and escalates mortality.

Objective

The current study objective was to specifically explore how Artificial Intelligence and Machine Learning affect the selection of antimicrobials, address antimicrobial resistance, and strengthen antimicrobial stewardship programs through a structured scoping review. The aim was to clarify what direct impacts AI/ML have in these areas and how they contribute to improvements and challenges in practice.

Method

A literature search was conducted in PubMed, Cochrane Library, Ovid Embase, Scopus, and CINAHL. A detailed search approach was developed to guarantee that all relevant studies were included. The entire electronic search strategy included terms such as “Artificial intelligence-AI,” “digital health,” “selection/prescribing of antimicrobials, “antimicrobial stewardship-AMS, “antimicrobial resistance-AMR, “Machine Learning-ML”, and “telemedicine,”.

Results

A critical appraisal of sources of evidence from the included studies was conducted using the Newcastle-Ottawa Quality Assessment Form. For this review, 70 sources related to artificial intelligence’s impact on antimicrobial selection/prescribing, resistance, and stewardship were initially screened. Of these, 33 were assessed for eligibility, resulting in 16 studies included in the review. Seventeen were excluded for lack of direct information relevant to AI’s effect on antimicrobial prescribing, resistance, and stewardship. This scoping review summarizes how artificial intelligence improves the accuracy of therapy selection, helps reduce inappropriate prescriptions by predicting necessity, and aids clinical decision-making (CDSS). It also details specific barriers, such as integration challenges, and facilitators like improved workflow, to incorporating artificial intelligence technologies in real-world clinical settings.

Conclusion

The reviewed studies showed that Artificial Intelligence and Machine Learning improve selection, prescribing, antimicrobial resistance, and antimicrobial stewardship. The use of artificial intelligence and Machine Learning models in selection, prescribing, antimicrobial resistance, and antimicrobial stewardship has a profound impact on clinical outcomes. The utilization of Artificial Intelligence and Machine Learning enhances prescription accuracy in AMS programs. The use of Machine Learning optimizes antimicrobial selection and predicts resistance. Future research should examine the implementation of Artificial Intelligence, Machine Learning, and AI-CDSS over a more extended period to understand its long-term effects on professional practices and organizational structures.

Supplementary information

The online version contains supplementary material available at 10.1186/s12879-025-12320-4.

Background

Antimicrobial resistance (AMR)

Antimicrobial resistance (AMR) occurs when microorganisms adapt to resist the effects of antimicrobial agents, rendering them ineffective. These agents, initially designed to kill or inhibit microbes, have, over time, resulted in adaptations in microbes due to indiscriminate selection/prescribing. The first signs of resistance appeared shortly after penicillin was introduced in 1941, with emerging penicillin-resistant strains such as Staphylococcus aureus [1]. The overuse and misuse of antimicrobials for agriculture and healthcare are among the factors contributing to resistance. This has led to the global health crisis of AMR, elevating mortality rates, extending sickness durations, and driving up healthcare expenditures [2, 3]. AMR raises mortality rates and healthcare costs; therefore, it has a substantial impact on human health. Infectious disease treatment becomes more complex, resulting in more extended hospital stays and significant medical expenses [4]. Novel computational modeling and predictive analytics with advanced Artificial Intelligence (AI), Machine Learning (ML), and clinical expertise play a pivotal role in combating AMR and improving clinical outcomes [5]. Recognizing the growing problem of AMR is becoming one of the significant challenges in 21st-century clinical care and necessitates an understanding of the mechanisms of resistance, reasons, and effects [6]. AMR presents profound social, economic, and health issues internationally. According to a study, the impact of AMR varies by region, and effective intervention techniques are desperately needed. It highlights the necessity of thorough environmental management and monitoring to halt the development of AMR [7]. AI and decision-support technologies have all helped to improve antimicrobial selection, and interdisciplinary collaboration is critical for addressing the complex issues of AMR. However, the evidence for their specific impact could be more precise.

Antimicrobial stewardship (AMS)

Antimicrobial stewardship (AMS) refers to coordinated interventions aimed at improving and monitoring the appropriate use of antimicrobials to enhance patient outcomes, reduce microbial resistance, and decrease the spread of infections caused by multidrug-resistant organisms. Effective stewardship programs are vital in combating antimicrobial resistance AMR¸ which is a growing global health crisis. This can complicate treatment options and increase healthcare costs. Key components dictate the foundational knowledge regarding AMS practices, and the importance of addressing AMR is essential to understand. AMS keys include prescribing practices, which encourage appropriate prescribing by ensuring that the choice of antibiotics is based on evidence-based guidelines and susceptibility data [8]. Education and training components within the AMS provide ongoing education for healthcare professionals about the principles of appropriate antimicrobial use and the implications of AMR [9]. Regular monitoring and feedback assess prescribing patterns and provide feedback to clinicians to promote adherence to stewardship protocols [10]. Policy development is essential to establish institutional policies that promote the judicious use of antimicrobials, incorporating evidence-based recommendations [11]. Further, AMS underscores collaborative approaches, which entail interdisciplinary teams in stewardship efforts, including infectious disease consultants, pharmacists, microbiologists, and infection prevention specialists, to optimize patient care [12].

Knowledge gap

Despite the promising potential of AI in antimicrobial management, there remains a significant knowledge gap regarding its practical applications and effectiveness in real clinical settings. Specifically, there is insufficient understanding of how AI tools influence prescribing patterns, impact patient outcomes, and contribute to the spread of antimicrobial resistance. Moreover, research on the integration of AI into existing antimicrobial stewardship programs is limited. This gap highlights the need for comprehensive studies that evaluate both the pros, cons, and challenges associated with implementing AI in antimicrobial stewardship practices, as well as the ethical implications and barriers to adoption in different healthcare settings. Addressing these uncertainties is crucial for maximizing the benefits of AI while ensuring safety and efficacy in antimicrobial use.

Rationale

The increasing prevalence of antimicrobial resistance poses a significant threat to public health globally. As healthcare systems seek effective solutions, AI has emerged as a promising tool for enhancing the prescribing, selection, and prediction of antimicrobial resistance. Via leveraging AI, we can potentially improve treatment outcomes, reduce inappropriate prescriptions, and enhance antimicrobial stewardship initiatives. This scoping review aims to explore the various ways in which AI impacts antimicrobial prescribing, selection processes, resistance patterns, and overall stewardship efforts, thereby identifying gaps in current knowledge and guiding future research in this critical area. Therefore, a scoping review is required to map the wide range of research, identify major themes, and investigate knowledge gaps to drive future research and policy development.

Objective

The current study’s primary objective was to explore the impact of AI and ML on the selection/prescribing of antimicrobials, AMR, and AMS via a structured scoping review reporting. The following research questions are designed to report the study objectives: -

1. How does AI influence antimicrobial prescribing practices in clinical settings?

2. What role does AI play in the selection of antimicrobials, and how does it influence treatment outcomes for patients?

3. How can AI contribute to understanding and mitigating AMR patterns?

4. What are the effects of AI-driven interventions on AMS?

Methods

Unpublished protocol developed before conducting the current Scoping review. Inclusion criteria: The eligibility criteria included heterogeneous studies published in English in the last five years (2018–2024) and addressing digital health interventions in selecting antimicrobials, AMS, and avoiding AMR. Exclusion criteria: Studies without full text or non-peer-reviewed sources were excluded to ensure high-quality evidence. We conducted a comprehensive literature search in PubMed, Cochrane Library, Ovid Embase, Scopus databases, and CINAHL. A detailed search approach was developed to guarantee that all relevant studies were included. The entire electronic search strategy included terms like “digital health,” “selection/prescribing of antimicrobials, “antimicrobial stewardship programs-AMS, “antimicrobial resistance-AMR, “telemedicine, “Artificial Intelligence-AI,” and Machine Learning (ML). Filters applied in the last five years. The search strategy is made available upon request for database reproducibility. The selection process involved two stages: title and abstract screening for relevance by two independent reviewers and full-text review for eligibility. Discrepancies were resolved through discussion or by a third reviewer. Data charted using a calibrated extraction form tested by the research team. Data extracted independently by two reviewers and compared for consistency. Any discrepancies were resolved through discussion. Variables extracted included author, year of publication, country, and type of digital health intervention, trial design, participants, outcome measures, and critical findings. We have ensured that the chosen databases provide adequate coverage of the research question. The data synthesis involved summarizing the findings based on the types of digital health interventions used and their reported impact on selection/prescribing of antimicrobial, AMR, and AMS. The main findings of the current review were reported based on the identification of three themes that imposed beneficial effects on antimicrobial selection/prescribing, AMR, and AMS.

Results

For the current review, 70 sources related to the impact of AI on the selection/prescribing of antimicrobials and AMS were initially screened. Of these, 33 were deemed relevant and assessed for eligibility. Ultimately, 16 studies were included in the review. At the same time, 17 were excluded due to a lack of relevance or inadequate data specific to the impact of AI on the selection of antimicrobials, AMR, and AMS. The characteristics of the included sources were charted based on crucial information such as the authors and publication year, the study design, the digital health interventions, and the outcome measures. The Newcastle-Ottawa Quality Assessment Form, adapted for practice settings, is used to evaluate the clarity of digital health interventions and the reliability of patient care outcomes. Twelve out of 16 studies scored highly in participant selection and intervention clarity, while four had less clear outcome measures [Fig. 1].

Fig. 1.

PRISMA flow diagram

Results of individual sources of evidence

The transformative potential of AI in healthcare, particularly in precision medicine and AMS, is highlighted in the current review. The ability of AI to analyze vast amounts of data allows for more appropriate treatment decisions, which improves patient outcomes and reduces AMR. AI-clinical decision support systems (AI-CDSS) improve antimicrobial prescribing by reducing improper use of antimicrobials, shortening treatment duration, and minimizing side effects. Integrating AI allows personalized treatment plans based on individual patient data and microbial resistance patterns, optimizing antimicrobial therapy [6].

Prescribing and selection of antimicrobial

In one study, the potential of AI to revolutionize current practices and lessen the threat posed by multidrug-resistant (MDR) pathogens was highlighted, as it examines how AI can aid in detecting, diagnosing, and treating AMR. Using AI to detect and manage AMR is a revolutionary way to tackle this worldwide health issue. AI has the potential to enhance treatment plans, reduce healthcare expenses, produce precise diagnoses, and enhance patient outcomes through its capacity to analyze vast amounts of data. On the other hand, data accuracy, infrastructure limitations, and ethical issues require addressing. To realize AI’s potential in addressing AMR, more research should concentrate on improving predictive models, promoting data sharing, developing ethical frameworks, expanding capacity, and allocating resources for research and development. Future research should focus on improving predictive models, data sharing, developing ethical frameworks, expanding capacity, strengthening real-time surveillance, and investing in research and development to take full advantage of AI’s potential in tackling AMR [7]. The article aims to improve AMS by utilizing artificial neural networks (ANN) to select antimicrobials for recurrent UTIs in women. Prior usage of fluoroquinolone and cephalosporin, as well as the existence of antimicrobial-resistant Escherichia coli, influenced the ANN’s predictions significantly. The study found that previous antimicrobial use predicted therapeutic failure; this reflected the importance of considering treatment history when managing UTI. ANNs have an advantage over standard models and may help with antibiotic stewardship by improving UTI decision-making and predicting beneficial therapies. Nonetheless, more research is needed to validate these findings in more extensive clinical settings. As reported by Cai T and colleagues [13], this approach may improve patient outcomes by incorporating resistance patterns and prior therapies into prescribing recommendations [13–15], [Fig. 2, Table 1].

Fig. 2.

Illustration of the concepts of the impact of AI on antimicrobial prescribing, selection, resistance, and stewardship. Understanding the significant role AI plays in enhancing antimicrobial stewardship and addressing resistance challenges

Table 1.

PICOs criteria summary for studies on the impact of artificial intelligence on the selection of antimicrobials, antimicrobial resistance, and antimicrobial stewardship

Author, Year	Study Design/Author country	Population	Intervention	Comparison	Outcomes
George Edison 2024 [6]	Review Article (France)	In healthcare settings, patients with AMR	AI-driven decision support systems in antibiotic Stewardship	Standard antibiotic management practices	Provide new prospects to improve antimicrobial stewardship, increase precision medicine, and speed up the development of novel antibiotics and antimicrobial agents through predictive modelling, clinical decision support, genomic analysis, and drug discovery.
Rony Ibne Masud et al. 2024 [7]	Review Article (Bangladesh)	General population, AMR patients	AI in the detection, diagnosis, and treatment of AMR	Current AMR management practices	Significant advancements in the rapid and accurate identification of AMR pathogens, Informed decisions regarding patient treatment and containment measures by healthcare professionals. Enhanced diagnostics, reduced healthcare costs, improved patient outcomes, challenges in data accuracy, ethical concerns
Tommaso Cai et al. 2023 [13]	Observational Study (Italy)	Women with recurrent urinary tract infections (UTIs)	Artificial Neural Networks (ANN) predicting effective antibiotic treatment	Traditional statistical linear models of treatment	It provided significant benefits regarding adherence to antibiotic stewardship principles in managing recurrent UTIs in adult women. Improved prediction of treatment effectiveness, the importance of considering past antibiotic use, and the need for further research in broader clinical contexts
Alexandra Harry. 2024 [16]	Review Article (USA)	General population	AI in antibiotic management (diagnosis, treatment, infection surveillance)	Current antibiotic management practices	Promising for improving antimicrobial stewardship initiatives’ efficacy, equality, and efficiency. Enhanced diagnostic precision, personalized treatment plans, proactive monitoring, challenges in technical complexity and financial sustainability
Amy Chang et al. 2024 [17]	Review Article (USA)	Frontline healthcare providers	AI for real-time personalized antibiotic recommendation	Standard antibiotic prescribing practices	Improved accuracy in empiric antibiotic prescribing or better than the clinician’s choice while choosing narrower-spectrum antibiotics based on resistance profile. Has the potential to improve both safety and stewardship
Bethany A. Van Dort, et al. 2022 [18]	Systematic Review (Australia)	Patients in hospitals	Digital interventions in antimicrobial stewardship (AMS)	Traditional AMS practices	Reduced antimicrobial use, improved appropriateness, unclear impact on clinical outcomes (mortality, length of stay), variability in outcomes, limited evidence on sociotechnical factors
Ji Lv, et al. 2020 [19]	Review Article (China)	General population, AMR patients	AI algorithms in AMR management (predictive models, AMP identification, antibiotic combinations)	Traditional AMR management practices	Enhanced prediction accuracy, optimized antibiotic use, identification of new practices, faster diagnostic methods, improved clinical decision support
Doaa Amin et al. 2024 [20]	Systematic Review (Ireland)	General population	Supervised machine learning (ML) models for antibiotic prescribing	Traditional antibiotic prescribing practices	ML improves antibiotic prescribing in different clinical settings. Promising results in prescription accuracy, challenges in clinician involvement, concerns about usability, bias reproduction and overfitting
Rafaela Pinto-de-Sá et al. 2024. [21]	Systematic Review (Portugal)	Inpatients and outpatients	AI algorithms for antimicrobial stewardship (predicting resistance, prescribing, patient outcomes)	Clinical and antimicrobial stewardship teams’ decisions	Effective in identifying inappropriate prescriptions, choosing therapies, and estimating outcomes but requires external validation
Zhilian Huang. 2024 [22]	Cross-sectional Survey (Singapore)	Doctors in Singapore	Ethical perspectives on AI-based clinical decision support systems (CDSS) for antibiotic prescribing	Traditional decision-making practices	There is a lack of awareness and trust in AI concerns about ethics (privacy, justice), and there is a need for a robust ethical and legal framework.
Andy Wallman et al. 2024 [23]	Observational Study (Sweden)	Patients in internet-based primary health care (PHC) vs. physical-PHC	Comparison of antibiotic prescribing by internet-PHC and traditional physical-PHC providers	Physical-Primary Health Care providers (PHC)	Lower antibiotic prescribing in internet-PHC, aligned with guidelines, raised concerns about care quality in digital healthcare.
Kelvin Smith 2024 [24]	Review Article (Indonesia)	General population	AI in antibiotic stewardship for improving resistance management	Traditional methods (AMS)	Enhanced diagnostic accuracy, reduced resistance spread, ethical and privacy concerns, and need for proactive AI governance.
Ismail Rabiu et al. 2024 [25]	Review Article (Taiwan)	Nigerian healthcare settings	AI in Antibiotic Prescribing in Nigeria	Traditional prescribing practices	Promising advancements in antibiotic stewardship, decision support, combating antibiotic resistance, and achieving efficiencies in new antimicrobial development, diagnostics, therapeutics, and cost reduction in both economic and health personnel aspects challenges need for post-implementation strategies and community involvement
José M. Pérez de la Lastra et al. 2024 [26]	Review Article (Spain)	General population	AI/ML in antimicrobial resistance (AMR) and drug discovery	Current AMR management practices	Faster diagnoses, better stewardship, reducing the need for manual testing and minimizing human error. AI/ML facilitates targeted treatments based on individual patient and pathogen data; challenges in data quality and ethical concerns require multidisciplinary collaboration.
Pinar Tokgöz et al. 2024 [27]	Observational Study (Germany)	Healthcare professionals	Implementation of AI-based clinical decision support systems (CDSS) in hospitals for antibiotic prescribing	Traditional decision-making	Trust in AI-based CDSS, but concerns about knowledge gaps and implementation uncertainty, the importance of education and user engagement for successful adoption
Andrea Kwa Lay Hoon et al. 2021 [28]	Observational Study (Singapore)	Healthcare providers in Singapore	Automation and AI to reduce antibiotic overuse in Singapore	Current antimicrobial stewardship (AMS)practices	Shift to holistic pre-emptive strategy, reduced inappropriate prescriptions, enhanced awareness of patient-centered approaches in antibiotic use.

Another study highlights the potential of AI to manage antimicrobials, combat resistance, and improve patient outcomes. It draws attention to the role of AI in several domains, such as infection surveillance, personalized treatment, diagnosis, and possible future uses. AI-driven approaches allow improved diagnostic precision, personalized treatment strategies, and proactive monitoring of AMR patterns. However, addressing issues such as clinical acceptability, data quality, technical complexity, financial viability, and regulatory concerns is essential. AI’s potential to improve patient outcomes, fight AMR, and transform medical procedures is enormous in antimicrobial management. The use of AI in antimicrobial management has the potential to significantly alter medical practices, combat AMR, and enhance patient outcomes. Regardless of technical challenges, data quality, clinical acceptability, regulatory problems, and financial sustainability, strategic investments and stakeholder collaboration allow for AI’s transformative potential. AI-driven antimicrobial management offers a new era of sustainable healthcare, enhancing worldwide efforts to fight AMR and improving patient lives [16]. One more study explores how AI can improve AMS. It emphasizes the difficulties that frontline physicians face in selecting proper antimicrobials. AI can improve AMS by providing personalized, real-time recommendations, enhancing the accuracy of empiric antimicrobial prescribing while reducing AMR [17].

Antimicrobial resistance (AMR)

A study assessed the effectiveness of digital interventions in promoting AMS in hospitals. Antimicrobial use and efficiency, clinical results, microbiological outcomes, and economic indicators were standard outcome measures. In general, digital interventions decreased antimicrobial use while improving its appropriateness. Digital interventions reduce antimicrobial use while improving appropriateness. However, no apparent conclusion was drawn about how different types of digital interventions achieve these optimized outcomes [18]. A study explores AI’s impact on managing AMR through predictive modelling, optimizing antimicrobial use, and identifying effective antimicrobial peptides. AI methods like naïve Bayes, decision trees, random forests, support vector machines, and ANN were evaluated for their speed, interpretability, and strengths in AMR applications. AI facilitates the integration of advanced methods such as flow cytometry and infrared spectroscopy, which improves both traditional and whole-genome sequencing-based antimicrobial susceptibility testing. Furthermore, it improves clinical decision-making by integrating biomarkers with AI-CDSS. In addition, it enables responsible use of antimicrobials, accurate forecasting, and the development of innovative treatments. The review suggests that future research should focus on comprehensive databases, biomarker integration, and AI optimization for complicated, small datasets to utilize AI to tackle AMR [19]. A systematic analysis evaluates AI’s impact on improving antimicrobial prescribing practices to reduce AMR. Out of 3692 records, only five research studies met the inclusion criteria. These studies used machine learning (ML) models to reduce inappropriate antimicrobial use, including logistic regression, random forests, gradient-boosting decision trees, support vector machines, and K-nearest neighbors. Although the models demonstrated potential, the need for clinician involvement in their development raised concerns about usability and reliability. Only one study showed a decrease in the overall number of prescriptions, and the majority anticipated incorrect prescriptions using electronic medical records. The study identified vital ML development steps in healthcare, including identifying problems, data collection, model generation, evaluation, and implementation. Challenges highlighted include overfitting, bias, and insufficient clinician engagement, with a recommendation for clinician involvement to enhance real-world effectiveness and model performance [20], [Fig. 2, Table 1].

Antimicrobial stewardship (AMS)

A systematic review compares the predicted accuracy of ML algorithms to traditional clinical choices in AMS for both inpatient and outpatient settings. AI algorithms have proven helpful in various settings, including detecting incorrect prescriptions, choosing suitable antimicrobials, inspecting prophylactic procedures, predicting the risk of resistance, and evaluating treatment results. Random forest models were very effective at predicting antimicrobial susceptibility. The results demonstrate how AI can assist AMS in optimizing the use of antimicrobials and enhancing patient outcomes. The authors recommend more research and validation to ensure that AI models are applicable and practical across different healthcare conditions [21].

One study examines doctors’ perceptions of using AI-CDSS for antimicrobial prescription in Singapore, highlighting ethical issues. Many medical professionals need to learn more about the moral issues surrounding AI in healthcare, as this affects their confidence and motivation to use these tools. When challenged with ethical dilemmas, most people prefer traditional methods to AI-based recommendations. The report suggests that clear regulations on AI ethics are necessary to overcome these issues and ensure ethical AI implementation in healthcare [22].

Research compares patient characteristics and prescribing patterns across internet-based and traditional primary healthcare practitioners. During the COVID-19 pandemic, more patients used internet-PHC, and infection diagnoses decreased, indicating a change in health-seeking behavior. Patients with illness diagnoses were less likely to receive antimicrobials from internet-PHC providers than physical-PHC providers, even after controlling for demographic factors. Internet-PHC providers prescribed antimicrobials more carefully, which aligned with prescription standards. The report showed digital providers may encourage appropriate antimicrobial usage. However, this pattern raises concerns about the quality and dimensions of online healthcare [23].

Studies explore the potential benefits of integrating AI into AMS in preventing resistance and improving patient outcomes. AI is a crucial tool in the fight against the spread of resistant microorganisms because of its capacity to increase diagnostic accuracy, direct treatment, and anticipate resistance patterns. AI application challenges include data privacy, bias, transparency, and ethical dilemmas. The future of AI-powered stewardship is promising, with real-time data, digital platforms, and readily available AI tools all contributing to precision medicine. Collaboration between researchers and healthcare practitioners is required to build open-source, cost-effective AI systems that maximize antimicrobial effectiveness while minimizing resistance. Healthcare systems that use AI properly can help to find long-term solutions to AMR, resulting in better healthcare outcomes [24].

According to a study on AI integration in Nigerian healthcare, it can improve public health and antimicrobial management. AI can improve decision support systems, fight AMR, and optimize diagnostic and antimicrobial development, potentially resulting in better patient outcomes and lower healthcare costs. To monitor the impact of AI on patient outcomes, AMR rates, and prescribing practices, it is suggested to set up a post-implementation monitoring unit. Standardized data collection and antimicrobial ethics education are necessary to evaluate AI’s efficacy. This article points out the necessity for strategic planning, community involvement, and financial support [25].

One study on AI and ML in AMR demonstrates how these technologies are improving drug discovery and infectious disease control. AI and ML allow faster, accurate diagnoses critical for improving treatment outcomes by analyzing massive datasets to find AMR patterns, predict collection outcomes, and discover new antimicrobials. This encourages targeted treatments and efficient medication use, which improves AMS. High-quality datasets, including clinical, phenotypic, and genomic data, are necessary to increase prediction accuracy. AI and ML applications are becoming more advanced as data quality improves [26].

Another study on using AI-CDSS in hospitals focuses on antimicrobial prescription and identifies critical elements for successful implementation. It emphasizes the value of both the organizational environment and user perceptions. A supportive hospital environment, along with satisfying customer demands, is critical for the successful implementation of these technologies. Concerns about implementation and outcomes remain despite the widespread trust in AI-CDSS. Hospital administrators should highlight the benefits of AI-CDSS and include clinicians in its development. Future studies should examine the long-term effects of AI-CDSS on professional practices and organizational structures [27].

Singapore’s healthcare system uses automation and artificial intelligence to decrease the misuse of antimicrobials. In 2018, 12,000 audits found that 20–30% of broad-spectrum antimicrobial prescriptions were incorrect.AMS necessitates a broader, preventative approach. This approach entails evaluating clinical syndromes and patient risk factors before prescribing. Authors revealed that current practices are reactive, with antimicrobials administered before a proper assessment; for instance, pneumonia, which accounts for 20% of infections, is a critical area for improvement in stewardship, as not all pneumonia cases require antimicrobials, underscoring the importance of patient-centered methods [28]. The synthesis of results from these studies highlights the significant impact of digital health tools on selecting antimicrobials, AMS, and avoiding AMR. Digital health technologies are improving the selection/prescribing of antimicrobials, AMS, and avoiding AMR [Fig. 2, Table 1].

Results in supplementary material (Appendices)

In order to enrich the presented results, we have provided more information on the pros and cons of the impact of artificial intelligence on the prescribing, selection, resistance, and stewardship of antimicrobials [Appendix 1]. Further, we have exemplified the satisfaction of healthcare professionals, tools, and brands for the impact of artificial intelligence on the prescribing, selection, resistance, and stewardship of antimicrobials [Appendices 2, 3].

The clinical, humanistic, and economic outcome measures for the impact of AI on the prescribing, selection, resistance, and AMS were presented in the appendices [Appendices 4, 5, and 6]. The challenges for the impact of AI on the prescribing, selection, resistance, and AMS in clinical practice were depicted in Appendix 7.

Discussions

The main findings of the current review included the identification of three themes that imposed beneficial effects on antimicrobial selection/prescribing, AMR, and AMS. The main findings from the current scoping review indicated that AI and ML improve selection, prescribing, AMR, and AMS. The utilization of AI and ML models in selection, prescribing, AMR, and AMS has a profound impact on clinical outcomes. The utilization of AI and ML enhances prescription accuracy in AMS programs. Finally, the utility of ML optimizes antimicrobial selection and predicts resistance.

Prescribing and selection of antimicrobials

In the first theme of selection/prescribing of antimicrobials, Tommaso Cai’s observational study compared women with recurrent UTIs using an ANN to predict effective antibiotic treatment. The review provided significant benefits regarding adherence to antibiotic stewardship principles in managing recurrent UTIs in adult women and improved prediction of treatment effectiveness [13]. The challenges call for considering past antibiotic use, and the need for further research in broader clinical contexts [13–15]. Another review article by Amy Chang et al. 2024 compared frontline healthcare providers’ standard antibiotic prescribing practices versus AI for real-time personalized antibiotic recommendations. The result indicated improved accuracy in empiric antibiotic prescribing, or better than the clinician’s choice, while choosing narrower-spectrum antibiotics based on resistance profile. The intervention has the potential to improve both safety and stewardship programs [17]. Doaa Amin et al. 2024 conducted a systematic review of studies with supervised ML models for antibiotic prescribing versus traditional antimicrobial prescribing practices. The intervention with ML improves antimicrobial prescribing in different clinical settings with promising results in prescription accuracy. However, challenges included clinician involvement, concerns about usability, bias reproduction, and overfitting [20]. Ismail Rabiu et al. 2024, review Nigerian healthcare settings, AI in antibiotic prescribing in Nigeria versus traditional prescribing practices. Authors reported promising advancements in antibiotic stewardship, decision support, combating antibiotic resistance, and achieving efficiencies in new antimicrobial development, diagnostics, therapeutics, and cost reduction in both economic and health personnel aspects. The challenges needed for post-implementation strategies and community involvement [25]. Pinar Tokgöz et al. (2024), observational study on healthcare professionals for the implementation of AI-CDSS in hospitals for antibiotic prescribing versus traditional decision-making. Trust in AI-CDSS, but concerns knowledge gaps and implementation uncertainty. The author raises the importance of education and user engagement for successful adoption [27].

In Summary AI has the potential to transform antimicrobial prescribing by enhancing precision medicine approaches, improving adherence to evidence-based guidelines, and ultimately reducing the impact of AMR [29–32]. However, challenges such as data privacy, integration into existing systems, and the need for clinician training addressed to realize its full potential [33–36]. Our scoping review provides a structured approach to explore the impact of AI on antimicrobial prescribing.

Antimicrobial resistance (AMR)

In the second theme of AMR, Rony Ibne Masud 2024 reviewed the article in the general population, comparing AMR patients to AI in the detection, diagnosis, and treatment of AMR. The review concluded significant advancements in the rapid and accurate identification of AMR pathogens, informed decisions regarding patient treatment and containment measures by healthcare professionals, enhanced diagnostics, reduced healthcare costs, and improved patient outcomes. The challenges identified in the review included data accuracy and ethical concerns [7]. Ji Lv et al. (2020) performed a review Article in patients with AMR, where they compared AI algorithms in AMR management (predictive models, AMP identification, antibiotic combinations) versus traditional AMR management practices. The review concluded that the intervention enhanced prediction accuracy, optimized antimicrobial use, faster diagnostic methods, and improved clinical decision support [18]. Zhilian Huang, 2024, conducted a cross-sectional survey for Doctors in Singapore, has reported a lack of awareness and trust in AI (Ethics perspectives on AI-based clinical decision support systems-CDSS for antibiotic prescribing), concerns about ethics (privacy, justice), and there is a need for a robust ethical and legal framework [22]. Andy Wallman et al. (2024) observational study on patients in internet-based primary health care (PHC) versus physical-PHC. The authors compared antibiotic prescribing by internet-PHC and traditional physical-PHC providers. PHC providers showed lower antibiotic prescribing in internet-PHC, aligned with guidelines, and raised concerns about care quality in digital healthcare [19–23]. José M. Pérez de la Lastra et al. 2024, review AI/ML in AMR and drug discovery. They compared current AMR management practices versus AL/ML practices, where the latter presented faster diagnoses, better stewardship, reduced the need for manual testing, and minimized human error. AI/ML facilitates targeted treatments based on individual patient and pathogen data; challenges in data quality and ethical concerns require multidisciplinary collaboration [26]. AI and ML models can predict resistance patterns by analyzing vast datasets, thus aiding in the anticipation of outbreaks and guiding empirical treatment choices [37]. Real-time surveillance of AMR trends using AI provides health agencies with timely data to respond to emerging threats effectively [38]. Machine learning algorithms improve the speed and accuracy of diagnosing infections and determining resistance profiles, facilitating personalized antibiotic therapy [39]. AI can help optimize therapeutic regimens by analyzing patient data and clinical outcomes, ultimately reducing the incidence of resistance development [40]. AI and ML have significant potential to transform the landscape of antimicrobial resistance management. By harnessing these technologies, stakeholders can improve surveillance, diagnostics, and treatment strategies, ultimately contributing to combating AMR effectively. This outline summarizes key concepts around the integration of AI and ML in addressing AMR.

Antimicrobial stewardship (AMS)

In the third theme of AMS, George Edison 2024 review compared patients with AMR versus AI-driven decision support systems in AMS. The review emphasized new prospects to improve AMS, increase precision medicine, and speed up the development of novel antibiotics and antimicrobial agents through predictive modelling, clinical decision support, genomic analysis, and drug discovery [6]. Alexandra Harry’s 2024 review compared AI in antibiotic management (diagnosis, treatment, infection surveillance) versus current antibiotic management practices. The findings indicated prospects for improving AMS initiatives’ efficacy, equality, and efficiency. Further, AI resulted in enhanced diagnostic precision, personalized treatment plans, and proactive monitoring. The challenges rectified as technical complexity and financial sustainability [16]. Bethany et al. 2022 conducted a systematic review on hospitalized patients with Digital interventions in AMS versus traditional AMS practices. The findings emphasize reduced antimicrobial use and improved appropriateness; however, the impact on clinical outcomes (mortality, length of stay) was unclear [17]. Rafaela Pinto-de-Sá et al. (2024) performed a systematic review in inpatients and outpatients via AI algorithms for AMS (predicting resistance, prescribing, and patient outcomes). The findings reported clinical and AMS teams’ decisions were effective in identifying inappropriate prescriptions, choosing therapies, and estimating outcomes [21]. Kelvin Smith (2024) conducted a review article in general population for use of AI in antibiotic stewardship for improving resistance management versus traditional methods for AMS. The author reported enhanced diagnostic accuracy and reduced resistance spread. However, challenges included ethical and privacy concerns, and the need for proactive AI governance [23]. Andrea Kwa Lay Hoon et al., observational study on healthcare providers in Singapore for automation and AI to reduce antibiotic overuse in Singapore versus current AMS practices. The study reported that shift to holistic pre-emptive strategy, reduced inappropriate prescriptions, and enhanced awareness of patient-centered approaches in antibiotic use [28]. AMR poses a significant threat to global health, necessitating effective AMS programs to optimize antibiotic use and combat resistance. The integration of AI and ML into AMS programs represents a promising approach to enhancing decision-making and improving patient outcomes. AI and ML can analyze vast datasets from electronic health records, clinical decision support systems, and laboratory results to identify patterns and predict potential outbreaks of resistant infections [41]. By utilizing algorithms, healthcare providers can receive real-time recommendations on appropriate antibiotic use, tailored to individual patient characteristics and local resistance patterns [42]. Several studies have demonstrated the efficacy of these technologies. For instance, a study by Shrestha et al. (2021) [43] found that an ML model significantly reduced unnecessary antibiotic prescriptions in a hospital setting, leading to a notable decrease in AMR rates. Similarly, Yang et al. (2022) [44] highlighted the role of AI in optimizing empirical treatment choices, showing improved adherence to AMS guidelines. Despite the benefits, challenges remain in the implementation of AI and ML within AMS.

Summary of evidence

The findings of the current scoping review strongly suggest that using digital tools to prescribe/select antimicrobials, combat AMR, and implement digital AMS is beneficial and necessary, especially in settings where patient safety and timely decision-making are critical. Furthermore, digital literacy and competencies are becoming increasingly important for healthcare providers to successfully navigate and use new technologies in their daily practice.

The challenges of AI and ML in antimicrobial prescribing, AMR, and AMS

The integration of AI and ML into antimicrobial prescribing, AMR, and AMS presents several challenges that need careful consideration. These technologies have the potential to enhance the prescribing process and optimize antimicrobial selection; however, they also pose significant hurdles related to resistance patterns, clinical data interpretation, and implementation in healthcare systems. One of the foremost challenges in leveraging AI and ML is the quality and availability of data. These technologies rely on large datasets for training algorithms, yet clinical data regarding antimicrobial prescriptions can be incomplete or inconsistent. This can lead to biases in the AI models, which may affect their predictive power and, ultimately, clinical decision-making [45].

AMR is a complex and multifaceted issue. AI systems must account for various genetic, environmental, and patient-specific factors that contribute to AMR. Failures to incorporate these complexities could lead to ineffective or inappropriate treatment recommendations, exacerbating the resistance problem [46]. The deployment of AI in clinical settings raises ethical and regulatory challenges. The use of AI algorithms must ensure patient safety, privacy, and informed consent. Moreover, determining accountability for AI-driven decisions, especially in cases of adverse outcomes, remains a legal gray area that must be addressed [47]. Implementing and integrating AI and ML tools into existing clinical workflows can be difficult. Health professionals may be hesitant to adopt new technologies due to a lack of familiarity or trust in machine-generated recommendations. Ensuring that AI tools are user-friendly and easily integrated into electronic health records is crucial for successful adoption [48]. Healthcare practitioners must be educated about the limitations of AI and ML algorithms. Misinterpretation of algorithm recommendations or over-reliance on technology without clinical judgement can lead to suboptimal patient outcomes. Continuous training and professional development will be essential to bridge the gap between technology and clinical practice [49]. The rapid evolution of pathogens poses a challenge for AI and ML systems to remain effective over time. Models must be updated frequently to encompass new strains and resistance patterns, requiring an agile and responsive approach to data collection and algorithmic adjustments [50]. While AI and ML hold significant promise for improving antimicrobial prescribing and stewardship, addressing these challenges is essential for their successful implementation. Through a careful approach that considers data integrity, ethical implications, and clinician engagement, the healthcare field can better leverage these technologies to combat AMR.

Despite the potential benefits, challenges exist, including data privacy concerns, the need for high-quality data, and the integration of AI tools into existing healthcare systems [51, 52]. Future research should focus on developing standardized protocols for AI applications in AMR, enhancing collaboration between technologists and healthcare professionals, and addressing ethical considerations related to AI in medicine [52].

Issues such as data privacy, the need for high-quality datasets, and the potential for algorithmic bias must be addressed [53]. Furthermore, training healthcare professionals for the effective use of these technologies is essential for their success in clinical practice. In summary, while AI and ML hold great potential in strengthening AMS programs, further research is required to fully understand their impact and address the associated challenges. Issues associated with AI and ML, such as ethics, scientific bias, interdisciplinary collaboration, and regulatory considerations in the development of AI and ML solutions for healthcare, deserve further emphasis [54–58].

The strengths, weaknesses, and limitations of the current scoping review

The strengths of the current scoping review

The scoping review provides a broad understanding of how AI is influencing various aspects of antimicrobial management, including prescribing practices and resistance patterns. By reviewing a wide range of literature, it highlights areas where further research is needed, guiding future studies on the topic. The review incorporates versatile perspectives, offering a well-rounded analysis of AI’s impact on antimicrobials. Our findings can inform clinical practice by illustrating effective AI applications in antimicrobial stewardship and improving patient outcomes. The review suggests implications for health professionals and policy makers that promote the integration of AI technologies in antimicrobial prescribing.

The weaknesses of the current scoping review

Our current review might include studies of varying quality, which could affect the reliability of the findings and conclusions drawn. The rapid advancement of AI technologies may lead to a review becoming outdated quickly as new findings emerge. The review may not fully address the challenges and barriers to implementing AI-driven solutions in antimicrobial stewardship in real-world clinical settings. By considering these strengths and weaknesses, readers can gain a clearer understanding of the implications of the scoping review on the impact of AI in the context of antimicrobials.

The limitations of the current scoping review

The limitations for the current scoping review included that the appraisal of methodological quality was limited, an overlap between the studies in the three thematic reporting, for instance, antibiotic prescribing and selection/prescribing sometimes fall within the AMS programs. However, when conducting a scoping review to explore the impact of AI and ML on antimicrobial prescribing, selection, resistance, and stewardship programs, it is essential to consider several limitations. Further limitations, such as the effectiveness of AI and ML algorithms, heavily depend on the quality and comprehensiveness of the data used for training. In many cases, the available datasets may be limited in scope or may not accurately reflect real-world clinical settings, potentially leading to biased results. Hence, we report the limitations related to literature identification and retrieval.

The use of AI in healthcare raises significant ethical issues, such as privacy concerns, informed consent, and potential biases in algorithmic decision-making. These ethical challenges can impact the implementation of AI technologies in AMS programs. The integration of AI and ML into clinical practice often requires collaboration between data scientists, clinicians, and healthcare administrators. Lack of interoperability and communication between these groups can hinder the successful implementation of AI for antimicrobial prescribing.

The rapidly evolving landscape of AI and ML technologies poses challenges for regulation and standardization in healthcare. Uncertainty in regulatory processes can delay the adoption of beneficial AI applications in antimicrobial stewardship. There is currently no consensus on best practices for implementing AI and ML in antimicrobial prescribing. This lack of standardization can lead to inconsistencies in results across different studies and hinder generalizability. Through acknowledging the above-mentioned limitations, researchers can better frame their scoping review and identify areas for further study and improvement in the application of AI and ML in antimicrobial prescribing, resistance, and stewardship.

The clinical implications of the current review

• The reviewed studies showed that AI and ML improve selection, prescribing, AMR, and AMS.

• The use of AI and ML models in selection, prescribing, AMR, and AMS has a profound impact on clinical outcomes.

• The utilization of AI and ML enhances prescription accuracy in AMS programs.

• The use of ML optimizes antimicrobial selection and predicts AMR.

Conclusion

The evidence from the current Scoping review shows that the utilization of digital technology tools (AI/ML) improves antimicrobial selection, halts AMR, and enables digital AMS while improving healthcare providers’ ability to provide adequate care in increasingly complex healthcare settings.

The reviewed studies showed that AI and ML improve selection, prescribing, AMR, and AMS. The use of AI and ML models in selection, prescribing, antimicrobial resistance, and antimicrobial stewardship has a profound impact on clinical outcomes. The utilization of AI and ML enhances prescription accuracy in AMS programs. The use of ML optimizes antimicrobial selection and predicts resistance. Future research should examine the implementation of AI, ML, and AI-CDSS over a more extended period to understand their long-term effects on professional practices and organizational structures.

Electronic supplementary material

Below is the link to the electronic supplementary material.

Abbreviations

ANN

Artificial Neural Networks (ANN)

AMP

Antimicrobial Peptides

AI-CDSS

Artificial Intelligence-based clinical decision support systems

AI

Artificial Intelligence

AMR

Antimicrobial Resistant

AMS

Antimicrobial Stewardship

ML

Machine Learning

MDR

Multidrug-resistant

PHC

Primary health care

UTIs

Urinary Tract Infections

Author contributions

All authors contributed to conceptualization, study design, data curation, formal analysis, methodology, project administration, software, supervision, validation, writing – original draft, writing – review and editing. All authors approve of the final version to be published and agree to be accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved.

Funding

No funding provided for the current scoping review.

Data availability

All data supporting the findings of this study are available within the paper and its Supplementary Information.

Declarations

Ethics approval and consent to participate

Review exempted from ethics approval. All author consented to participate in the review.

Consent for publication

All author consented for publication of the review.

Competing interests

The authors declare no competing interests.