From Islands to Inference: Building Caribbean Genomic and Clinical Data Networks to Reduce AI Bias and Enable Precision Public Health

Genomics and clinical artificial intelligence (AI) offer major gains in screening, diagnosis, and treatment optimization. Yet the benefits are uneven: models trained on ancestry-skewed datasets often underperform in underrepresented and admixed populations, risking the amplification of inequity. The Caribbean—characterized by high genetic diversity, small and interconnected societies, and complex cross-border care—faces a dual challenge: reducing ancestry and measurement bias while creating the infrastructure for trustworthy data exchange. Here we argue that the region’s highest-leverage move is not a single sequencing project or a single algorithm, but a connected system: interoperable clinical records, harmonized phenotype capture, secure identity resolution, and enforceable governance for secondary use. We use CariGenetics and the Caribbean Genome Programme as an illustrative example of emerging regional genomic capacity and stewardship, and we situate such efforts within broader regional roadmaps for health data exchange. We propose a “networked equity” agenda: multi-ancestry model development and validation using Caribbean cohorts; minimum viable regional interoperability; and transparent benefit-sharing structures that convert data into measurable improvements at the point of care.

1. The problem is not interest in genomics and AI—it is transferability

The equity gap in genomic AI is no longer a theoretical concern. The dominant failure mode is transfer: tools developed in one population and health system are deployed in another with different ancestry structure, different disease profiles, and different measurement pathways. When discovery datasets underrepresent Caribbean-relevant ancestries, model calibration degrades; when clinical phenotypes are inconsistently recorded, ground truth becomes unstable; when governance is unclear, trust collapses before value is realized.

Put simply: a model can be statistically strong yet clinically wrong in the wrong context—especially in small systems where error margins are thin.

2. Why the Caribbean is a distinct testbed for “bias-resistant” precision health

Caribbean settings combine characteristics that intensify both risk and opportunity:

• Population structure and admixture that may differ meaningfully across islands and diaspora networks.
  
  
• Small labor markets and variable specialist access, which shapes who gets diagnosed and how (measurement bias).
  
  
• Overseas referrals and cross-border care, which fragment longitudinal records.
  
  
• High social proximity, making privacy concerns more salient and trust more fragile.

These features mean that “more data” alone is insufficient. What matters is connected data with consistent meaning and credible stewardship.

3. An illustrative signal: CariGenetics as “capacity in motion,” not the endpoint

CariGenetics provides a useful example of what is beginning to emerge in the region: locally anchored genomics work explicitly oriented to Caribbean ancestry and regional relevance. The Caribbean Genome Programme, launched with Oxford Nanopore Technologies, is positioned as a large-scale sequencing initiative for the region (initially sequencing 1,000 men across 10 islands, with stated plans to scale). CariGenetics also frames its research mission around Caribbean genetic diversity and underrepresentation.

The scientific significance here is not that one initiative will “solve bias.” Rather, it signals something more important for the next decade: the region is beginning to generate its own reference data, local partnerships, and governance instincts around sensitive biological data—including attention to data protection for genetic material.

That said, sequencing alone does not create equitable clinical AI. Without linkage to health system data—diagnoses, labs, medications, outcomes—genomic discovery remains partially disconnected from care.

4. The real unlock: connecting systems and data across the region

A meaningful Caribbean precision health future requires a connected ecosystem that can do three things at once:

4.1 Harmonize clinical meaning (phenotypes)

AI bias is often described as a data diversity problem. In practice, it is equally a measurement pathway problem. If one island confirms cancer diagnoses through different diagnostic pathways than another—because imaging access differs or overseas referral is common—the recorded phenotype can encode system constraints rather than biology.

Connecting data networks makes it possible to standardize and audit phenotypes—a prerequisite for any regionally valid model.

Connecting data networks makes it possible to standardize and audit phenotypes—a prerequisite for any regionally valid model.

4.2 Enable secure data exchange through a regional interoperability roadmap

Encouragingly, the region is already moving toward health data exchange planning. The Inter-American Development Bank has reported Caribbean countries agreeing on a roadmap to begin exchanging health data by 2028. CARPHA’s reporting on the ONE Caribbean Connect dialogue similarly emphasizes a regional commitment and priority actions around governance, legal/regulatory enablers, and implementation.

This is the missing connective tissue between genomics projects and clinical impact: the ability to follow patient pathways, treatments, and outcomes across settings in a governed manner.

4.3 Build governance that can sustain trust at small-island scale

The Caribbean has already generated region-specific governance scholarship. The SHARE framework (Safeguarding Health And Research data sharing by promoting Equity) explicitly addresses equitable data sharing and culturally responsive stewardship for Caribbean communities and researchers.

At the operational level, PAHO’s Information Systems for Health work frames digital transformation around governance, interoperability, and quality data for decision-making.

Finally, global governance tools can be adapted regionally; the Global Alliance for Genomics and Health (GA4GH) provides a widely used framework for responsible sharing of genomic and health-related data.

5. What “the promise of the future” looks like if the Caribbean connects systems and data

If Caribbean governments, providers, and research groups can converge on minimum viable interoperability + enforceable governance, several high-impact outcomes become plausible:

5.1 Caribbean-calibrated risk prediction and screening

Rather than importing polygenic risk scores or AI triage models “as is,” the region can re-estimate and recalibrate models using Caribbean data linkages (genomics + labs + diagnoses + outcomes). This reduces systematic misclassification and improves real-world clinical utility.

5.2 Faster, safer learning cycles in small populations

Small systems are often framed as data-poor. But connected small systems can be learning-fast, because feedback loops are shorter and pathway variation is easier to detect. Interoperability makes quality improvement and model monitoring feasible at the pace needed for safe deployment.

5.3 Equity as an engineering outcome, not a slogan

With governance structures like SHARE operationalized, equity can be built into: consent, secondary use permissions, benefit-sharing, and community accountability.

5.4 Regional bargaining power and reduced vendor lock-in

When interoperability standards and data export requirements are mandated, the region reduces dependence on siloed platforms and improves negotiating position. PAHO’s emphasis on aligned governance and standards supports this direction.

Connected small systems can be learning-fast, because feedback loops are shorter and pathway variation is easier to detect.

6. A practical “networked equity” agenda for the next 3–5 years

A Science-level Caribbean strategy is credible only if it is implementable. The near-term agenda can be framed as five commitments:

1. Minimum viable interoperability
   Start with identity, meds, labs, diagnoses, and discharge/referral summaries—then expand. Tie procurement to standards and exportability. (Aligns with regional roadmap momentum.)
   
   
2. Caribbean phenotype harmonization
   Define “gold standard” phenotype rules for priority conditions (diabetes, hypertension, CKD, cancers) accounting for overseas diagnostics and variable access.
   
   
3. Local validation as a safety requirement
   No genomic risk score or diagnostic AI should be scaled without local benchmarking and calibration, reported transparently.
   
   
4. Governance operationalization
   Translate frameworks (SHARE/GA4GH) into data access committees, tiered permissions, audit trails, and plain-language public transparency.
   
   
5. Benefit-sharing with measurable outputs
   Define deliverables that communities can feel: improved screening pathways, reduced diagnostic delays, clinician decision support that works in local context, and workforce development that stays in the region.

7. Conclusion

CariGenetics’ regional work illustrates that Caribbean genomics capacity is no longer hypothetical. But the long-term scientific and equity payoff will depend less on any single initiative and more on whether the region can connect the dots: interoperable clinical systems, harmonized phenotypes, secure identity, and enforceable governance that sustains trust.

The promise is not merely “more Caribbean genomes” or “more AI.” The promise is a Caribbean learning health system—where diversity becomes a methodological strength, not a source of algorithmic error; and where data is transformed into better care across islands, not extracted and exported without return.