The article, written by Abdulrasheed Ibrahim of the Department of Surgery at Ahmadu Bello University and Ahmadu Bello University Teaching Hospital in Zaria, Nigeria, challenges the question that currently dominates the field. Researchers typically ask whether AI-supported stewardship improves outcomes, a question that demands an average effect measured across many settings. Ibrahim contends that this framing is fundamentally mismatched to sub-Saharan Africa. When an intervention’s effectiveness depends on infrastructure that some hospitals possess and others lack, averaging results across dissimilar settings does not merely blur the picture; it can systematically disadvantage the region and convert an infrastructure gap into a verdict about the technology itself.

The core of the argument is that an AI tool is not a single intervention but a bundle of preconditions that must hold together for the tool to help at all. These preconditions span three dimensions, and each is unevenly met across the region. The first is infrastructural: a predictive tool needs digital records, reliable connectivity, and local microbiology data both to train on and to act upon. A recent systematic review of stewardship programmes in sub-Saharan Africa found limited diagnostic and laboratory infrastructure, weak data systems, and underfunded services to be among the most common barriers to implementation. Where these foundations are absent, a tool has nothing to read, and no amount of algorithmic sophistication can compensate.

The second dimension is epidemiological, and it concerns whether a model trained on one population can transfer to another. This is not a hypothetical worry. A hepatitis B decision-support tool developed on Nigerian data performed well in a Nigerian cohort but far less well in an Australian one, a binational external validation study reported in the journal Viruses in 2023. Pathogen mixes, resistance patterns, and comorbidities differ between settings, so local validity is itself a precondition rather than a given. A model that predicts resistance accurately in Lagos may mislead prescribers in a rural district where the circulating organisms and their susceptibility profiles are entirely different. Local validation, in other words, must be treated as a requirement for every deployment, not as a detail to be assumed.

The third dimension is human, and it may be the most decisive of all. An accurate model changes nothing if clinicians do not trust it, if it is not integrated into the clinical workflow, or if it lacks supporting training and governance. Research by Kapatsa and colleagues, published in Frontiers in Microbiology, identified insufficient staff training, weak governance, and limited leadership as recurring obstacles to antimicrobial stewardship across the region. In many settings, these human conditions, rather than the technology itself, decide whether a tool is used at all. A sophisticated resistance-prediction engine that prescribers routinely ignore is, in practical terms, no intervention whatsoever. Trust, workflow fit, and governance are not soft add-ons; they are load-bearing conditions on which the entire enterprise rests.

These preconditions co-occur unevenly, both between countries and within them. A tertiary referral centre in a capital city may satisfy most of them, while a district hospital in the same country satisfies few. No single average can adequately capture this variation, and this is where the methodological critique sharpens into an equity argument. Net-effects evaluation assumes that an intervention acts more or less independently of its context. For an infrastructure-dependent intervention, that assumption fails, because the conditions are interdependent: an accurate model achieves little without clinician trust, trust goes untested without data to run the model, and data remains unused without a workflow that acts on it. What matters is not the average but which combinations of conditions are necessary or sufficient for the tool to help.

Ibrahim’s proposed remedy is a configurational approach, and Qualitative Comparative Analysis in particular. This method, developed by the sociologist Charles Ragin, treats each deployment as a case, scored on a shared outcome and on a set of conditions. The outcome would ideally be a reduction in inappropriate prescribing or in resistance, the shared goal of stewardship. The conditions would span the three dimensions already described: infrastructural factors such as data-system maturity and laboratory capacity, epidemiological factors such as local model validation, and human factors such as clinician trust and training, workflow integration, and governance. The analysis then asks configurational questions. Is local microbiology data present in every successful deployment, making it necessary? Which combinations of conditions are together sufficient for success?

The method is not speculative. Configurational techniques have already been applied across public health interventions, and two examples cited in the article show their practical value. In an integrated mental healthcare trial in South Africa, fuzzy-set Qualitative Comparative Analysis identified which combinations of conditions produced successful implementation. A cluster-randomised physical activity programme was analysed the same way, separating the pathways that led to success from those that did not. Crucially, configurational analysis allows for equifinality, the principle that more than one combination of conditions can produce the same outcome. A hospital with limited digital records might succeed with a simpler, well-validated AI tool, a functioning laboratory, and strong clinician trust, while another hospital with richer data and connectivity might succeed with a more data-intensive system. Both reach the goal through AI, by different routes. Equifinality reframes the problem entirely: the question is not whether a hospital can replicate one ideal, data-rich deployment, but which combination of conditions would let AI-supported stewardship work there.

For hospitals in the region, that shift has concrete consequences. It replaces a single, data-rich template with guidance about what can be done now, at the level of infrastructure actually available. It turns the discouraging statement that a hospital lacks the infrastructure for artificial intelligence into an actionable question about which combination of conditions would let it begin. It also protects the field from a dangerous misreading: a pooled result that looks modest across heterogeneous settings would otherwise be reported as a verdict on the technology, when it actually reflects the uneven presence of the conditions the technology needs. The configurational framing distinguishes a technology that has failed from a hospital that was never resourced to run it, and that distinction is what turns a global template into local, achievable guidance.

Serious challenges remain, and the article does not minimise them. Few AI-supported stewardship programmes in sub-Saharan Africa have been studied configurationally, so the evidence base is thin, and stewardship outcomes are reported inconsistently, making studies difficult to compare. Each enabling condition must also be scored by clear, stated rules: laboratory capacity might be scored as full where a functioning microbiology service is always available, partial where testing is intermittent, and absent where none exists, while clinician trust might be scored by whether prescribers act on a tool’s advice routinely, occasionally, or not at all. Reporting these rules openly lets others see how each hospital was judged and reveals what a hospital needs, whether more laboratory capacity or greater clinician confidence, before an AI tool can help there. The remedy, Ibrahim argues, is within reach: studies of AI-supported stewardship in the region should record not only whether a tool worked but which conditions were present and which combinations accompanied success, building on implementation research already under way. Such reporting would make configurational analysis possible as the evidence grows. The question for AI-supported stewardship in sub-Saharan Africa, the article concludes, is not whether it works but under what conditions, and answering that question configurationally is at once the sounder science and the fairer one.

Subject of Research: Evaluating AI-supported antimicrobial stewardship in sub-Saharan Africa using configurational analysis

Article Title: Whether, or under what conditions? A configurational approach to AI-supported antimicrobial stewardship in sub-Saharan Africa

Article References: Ibrahim, A. (2026). Whether, or under what conditions? A configurational approach to AI-supported antimicrobial stewardship in sub-Saharan Africa. Frontiers in Antibiotics, 5, Article 1954916. https://doi.org/10.3389/frabi.2026.1954916

Image Credits: AI Generated

DOI: 10.3389/frabi.2026.1954916

Keywords: antimicrobial resistance, antimicrobial stewardship, artificial intelligence, machine learning, Qualitative Comparative Analysis, sub-Saharan Africa, health systems, diagnostic infrastructure, implementation science, equifinality, clinical workflow, health equity