Digital Health and Artificial Intelligence in Resource-Constrained Health Systems: What the Evidence Says

Background: Digital health tools and artificial intelligence (AI) are increasingly promoted as ways to extend the reach of under-resourced health systems, particularly in low- and middle-income countries (LMICs). Global digital health strategy and AI governance frameworks have matured substantially over the past five years, yet uptake and rigorous evaluation continue to lag behind enthusiasm.

Policy and implications: The evidence base is real but uneven. Global guidelines already distinguish which digital interventions are supported by sufficient evidence and which are not, and clinical trials in LMICs demonstrate effectiveness for select conditions such as hypertension management and childhood vaccination uptake. However, AI evaluations remain concentrated in high-income settings, models frequently lose accuracy when transferred to new contexts, and connectivity, workforce capacity, and data affordability continue to constrain deployment for billions of people, particularly in low-income countries.

Recommendations: This brief recommends that governments, donors, and implementers: adopt existing global guidance as a common reference framework for national digital health plans; fund only evidence-matched interventions accompanied by a registered evaluation plan; require local validation before deploying clinical AI; invest in connectivity, interoperability, and workforce digital skills as foundational infrastructure; institutionalize AI governance and accountability mechanisms; and prioritize open, reusable digital platforms over one-off pilots.

Conclusions: Digital health and AI can meaningfully extend health system capacity in resource-constrained settings, but only when treated as governed public health investments rather than novel technology. Countries do not need new frameworks; they need to adopt, fund, and enforce the ones that already exist, while investing in the underlying digital foundations that determine whether any tool can reach the populations that need it.