Page 22 - GHES-2-3

Global Health Economics and
Sustainability
AI in antibiotic prescribing in Nigeria

Table 1. AI application strategies against antimicrobial resistance
AI applications for AMR Concepts Advantages Drawbacks
Appropriate antibiotic Appropriate therapy selection, dose, and • Automatic support for decisions and • Biasness in operation
prescription correct administration route review of antimicrobial prescriptions • Little labor
• Automatic feedback input and relevant • Need for health funds
improvement
• Directed operation
Prediction of antibiotic ML techniques to predict early AMR • Genomic exploitation to predict the • Lack of genotypes and genome data
resistance or the probability of a microbial agent phenotype in NCBI or other databases
becoming resistant • Ability to support clinician’s decision • Challenge of large data integration
Prediction of infection Machine/deep learning tools for • Efficiency in distinguishing between • Challenge in collecting accurate data
severity infectious pathology recognition and infectious and noninfectious diseases • Insufficient relevant laboratory
appropriate management • Decision support provision information
• Mortality reduction
Source: Adapted from Ali et al. (2022).
Abbreviations: AI: Artificial intelligence; AMR: Antimicrobial resistance; ML: Machine learning.

LMICs, especially in the context of the escalating threat combining CDSS with machine learning, as demonstrated
of antibiotic resistance (Valderrama-Rios et al., 2023). by researchers from the Université de Sherbrooke. While
Despite the proven efficacy of antibiotic stewardship promising, integrating machine-learned rules into existing
programs in various regions globally, their success rate knowledge bases and automating rule maintenance pose
has been notably lower in LMICs, possibly attributed to significant challenges (Ali et al., 2022; GAO, 2020).
challenges such as insufficient human resources, lack of The use of AI has significantly been used to explore
local expertise and knowledge, inadequate funding, and antimicrobial stewardship. This is based on a system of
limited institutional support. Setting aside health problems information collation about the clinical records of patients.
as a priority will mandate the government to reduce the cost This is then employed in CDSS. This will assist clinicians in
of governance as well as focus on the implementation of monitoring all stewardship parameters such as the guidelines
novel technologies including AI in the health-care setting and protocols for the development and implementation of
as this will reduce the overdependence on and demand evidence-based guidelines and protocols for antimicrobial
of skilled health personnel. The implementation of these prescribing, including recommendations for appropriate
technologies, with proper maintenance, may save costs in drug selection, dosing, duration of therapy, and route of
the long run. Besides, AI introduces unique opportunities administration; regular review of antimicrobial prescribing
to enhance and augment the current antibiotic stewardship practices; implementation of infection prevention and
program model (Okeowo et al., 2020; Chang & Chen, control measures to reduce the spread of antimicrobial-
2022). Integrating AI into antimicrobial stewardship efforts resistant pathogens; and antimicrobial surveillance to
in LMICs could potentially overcome existing barriers and monitor AMR patterns and antimicrobial use data to identify
foster more effective strategies to combat the rising menace trends, patterns of resistance, and emerging threats, and to
of antibiotic resistance. inform antimicrobial stewardship interventions. This will
AI for Antimicrobial Stewardship focuses on the use assist health authorities in promoting sensible applications
of computerized systems in reviewing prescriptions, taking into consideration all aforementioned parameters (Ali
particularly in identifying patients requiring a review, a et al., 2022). Figure 1 shows a schematic flow of the possible
task often time-consuming for human reviewers. CDSS use of AI and the dataflow integration for clinical decisions.
are commonly employed, but their reliance on expert and
rule-based knowledge bases poses challenges in adapting 6.1. Pros and cons of using AI in antimicrobial
to changing guidelines (Valderrama-Rios et al., 2023). stewardship
Resource constraints may lead to selective evaluation While the use of AI has tremendous potential to
of antibiotics. To address this, scientists attempted to revolutionize the health-care service providers to patients,
develop models using health-care data to identify patients it also has some disadvantages (Marra et al., 2023).
for prescription review. Although these models had fair As such, its implementation in the health-care sector
discriminatory power, the concept of automating the should be accompanied by careful consideration of its
evaluation of all patients with prescribed antimicrobials risks, implications and possible ways to overcome such
holds enormous potential. Another approach is challenges. The use of AI in health-care settings offers

Volume 2 Issue 3 (2024) 4 https://doi.org/10.36922/ghes.2602