When healthcare becomes infrastructure: AI, digital identity, and the quiet risk of exclusion

Across much of the Global South, digital health is increasingly being framed as the solution to a familiar governance problem: how to deliver healthcare at scale in systems marked by workforce shortages, fragmented records, uneven institutional capacity, and rising public demand.

From interoperable health records and biometric-linked insurance systems to AI-enabled health platforms, governments are investing heavily in digital public infrastructure to make healthcare delivery more efficient, targeted, and continuous.

The logic is compelling. Once individuals are connected to a unified health identity, systems can theoretically coordinate care better, reduce duplication, streamline welfare delivery, and improve public health planning.

AI further strengthens this promise. Predictive systems can identify risk earlier, optimise resource allocation, personalise recommendations, and automate decisions that historically moved slowly through administrative layers.

In policy conversations, this is often described as healthcare modernisation. But beneath the language of efficiency, a quieter transformation is taking place: healthcare is gradually shifting from institutional interaction to infrastructural dependence.

That distinction matters more than it initially appears.

When access depends on recognition

Historically, healthcare systems functioned through institutions and intermediaries. Patients interacted with hospitals, local clinics, frontline workers, pharmacists, and administrative officers. Access was imperfect and often unequal, but decisions still passed visibly through human systems.

Increasingly, however, healthcare access is becoming conditional upon successful participation within digital infrastructures. To receive care, individuals must first be recognised by the system itself.

At first glance, this appears administrative rather than political. Yet the implications are profound.

Consider a patient in a rural district attempting to access subsidised treatment through a digitally linked health insurance programme. Connectivity is unstable. A biometric mismatch occurs. A record seeded years earlier contains an error. Authentication fails. From a systems perspective, this is treated as a technical issue. For the individual, however, it becomes something far more immediate: exclusion from care.

This is the paradox increasingly embedded within digital health identity systems. Systems designed to increase inclusion can also institutionalise exclusion at scale.

Importantly, these exclusions rarely emerge dramatically. They accumulate quietly through interoperability gaps, poor-quality datasets, infrastructural failures, and assumptions embedded within system design. Women, migrant populations, rural communities, older persons, and individuals with lower levels of digital literacy often experience these frictions disproportionately because digital inclusion itself remains unevenly distributed.

The digital divide therefore cannot be understood merely through internet penetration or smartphone access. The deeper issue is infrastructural legibility: who is able to remain consistently visible, authenticated, and actionable within increasingly data-driven systems of governance.

From health records to decision-making systems

This challenge becomes even more consequential once AI is integrated into these systems.

Traditionally, health identity platforms functioned largely as repositories of information. AI transforms them into decision-making infrastructure. Algorithms can now influence eligibility verification, prioritisation pathways, fraud detection, insurance assessments, behavioural nudges, and predictive risk scoring.

In other words, systems originally designed to organise healthcare information are gradually evolving into systems that shape healthcare outcomes themselves.

The significance of this shift is often underestimated because AI in healthcare continues to be discussed primarily through the language of innovation. Public debate tends to focus on privacy, bias, or model accuracy.

These are important concerns, but they capture only part of the transformation underway. A prominent example was found within the Arkansas DHS Medicaid waiver program, where an unstable optimization algorithm inadvertently compromised patient care by drastically curtailing essential care hours based on minor data variances

The deeper issue is that AI increasingly functions as an interpretive layer between citizens and public systems.

Individuals are not merely receiving healthcare through digital systems; they are beginning to understand eligibility, risk, health behaviour, and institutional legitimacy through algorithmic mediation. Over time, authority subtly shifts away from visible institutional actors toward technical infrastructures that operate at a scale and speed difficult to meaningfully contest.

The sociologist James Scott once argued that modern states seek to make societies “legible” in order to govern them. Digital health identity systems represent a new form of legibility where citizens increasingly exist within healthcare systems through data presence and system recognition.

But legibility is never neutral. What systems can recognise, they can govern efficiently. What they fail to recognise risks becoming administratively invisible.

Why the Global South faces a different reality

This is particularly important in the Global South, where healthcare decisions are often made under conditions of uncertainty, economic pressure, fragmented trust, and institutional scarcity. In such settings, AI systems may not simply supplement healthcare infrastructure; they may gradually substitute for institutional absence.

That distinction changes the governance challenge entirely.

In many high-income settings, AI-enabled health tools operate alongside relatively stable healthcare ecosystems. In lower-resource environments, however, individuals may increasingly rely on digital systems because alternatives are limited, overburdened, or geographically inaccessible.

The AI system therefore acquires a form of perceived authority not necessarily because it is always superior, but because institutional alternatives remain weak.

Over time, this creates a subtle but significant asymmetry. A recommendation generated by an AI-enabled health system may be technically accurate while remaining economically impossible, geographically inaccessible, or culturally impractical. Yet the system itself may possess limited capacity to recognise these realities.

What emerges is a form of computational rationality detached from operational context.

Economist Amartya Sen’s capability approach becomes particularly relevant here. Development, Sen argued, should not be understood merely through formal access to systems, but through the actual freedom individuals possess to utilise opportunities meaningfully.

Digital health infrastructures often collapse this distinction. The existence of digital access is frequently treated as evidence of empowerment, even when meaningful capability remains deeply uneven.

Data, dependency, and digital power

This becomes even more complex once health identity systems are integrated into broader data ecosystems involving cloud providers, AI developers, insurance systems, financial architectures, and platform intermediaries. Health data increasingly functions not merely as a public health asset, but as infrastructure with economic, political, and strategic value.

Scholars writing on digital colonialism have warned that many Global South countries risk becoming sites where data is extracted, processed, and valorised elsewhere under conditions of regulatory asymmetry. There is validity to this concern. Yet framing all digital health expansion purely through extraction risks oversimplifying reality.

The challenge is not that governments are irrationally embracing digital systems. In many contexts, healthcare scarcity is real, administrative fragmentation is real, and continuity of care remains structurally weak.

The governance challenge therefore is not whether digital health identity systems should exist. Increasingly, they are becoming foundational infrastructure. The real question is whether governance systems can evolve quickly enough to ensure that infrastructural dependence does not outpace accountability.

Governing infrastruture before it governs us

Because ultimately, the greatest risk may not be technological failure itself. It is the gradual normalisation of systems where healthcare access becomes conditional upon successful participation within digital infrastructures whose assumptions, incentives, and decision-making logics remain only partially visible to the individuals governed through them.

This is why digital health governance cannot remain a purely technical conversation. Questions around exclusion safeguards, procedural transparency, correction mechanisms, and institutional accountability are not peripheral concerns; they are central to whether these systems remain democratically legitimate over time.

Healthcare systems have always reflected broader political choices about inclusion, participation, and public trust. The difference now is that these choices are increasingly being embedded into infrastructure itself.

And infrastructures, once normalised, are far harder to contest than policies alone.