AI in Caribbean Public Health: Use Cases, Risks and a Regional Action Plan
AI tools can improve disease surveillance, NCD management, dengue outbreak prediction, maternal health triage, mental health outreach, and climate-health response across the Caribbean. The region's most urgent AI opportunity is applying machine learning to the NCD burden, specifically hypertension, diabetes, and cardiovascular disease account for approximately 75–80% of deaths across Caribbean territories. Most high-impact use cases require minimal new infrastructure: structured electronic health records and a mobile-capable interface are enough to start. The principal risk is applying AI tools trained on non-Caribbean data to Caribbean populations without local validation.
The Health System Problem AI Was Built to Address
Caribbean health systems are chronically under-resourced relative to their disease burden. A World Bank and PAHO report published in September 2025 put it plainly: the failure to build resilient primary health care in Latin America and the Caribbean could lead to significant preventable deaths and long-term economic damage that the region cannot afford. During COVID-19, essential health services fell by up to 50% in some countries. In several Caribbean territories, those gaps persisted for two years or longer after the acute phase of the pandemic passed.
The structural constraints are not new. Small populations produce limited health economies of scale. Specialist shortages mean patients travel long distances or go without. Laboratory capacity in the Eastern Caribbean operates in resource-limited environments with personnel shortages that cannot be rapidly corrected through training alone. Hurricane seasons, floods, and Saharan dust events create recurring surges in respiratory, infectious, and trauma cases that health systems must absorb without proportional surge capacity.
AI does not solve any of these problems directly. What it does is extend the reach of the human capacity that does exist, compress the time between data and decision, identify risks before they become crises, and automate documentation work that currently consumes clinical hours that would otherwise go to patients. In a health system with no slack, those gains are not marginal. They are structural.
AI in Caribbean public health is not a technology story. It is a capacity story. The question is not whether the technology is impressive. The question is whether it can extend what a community health worker in rural Guyana can see, what a clinician in an understaffed Haitian clinic can decide, and what a Surinamese Ministry of Health can anticipate three weeks before a dengue surge arrives.
What Caribbean Populations Actually Die From
The Caribbean health burden is concentrated in four areas: non-communicable diseases, vector-borne infectious diseases, maternal and child health outcomes, and mental health. Climate change is not yet a separate disease category in regional statistics, but it operates as an amplifier across all four. Each area has distinct AI application opportunities and distinct data constraints.
Non-Communicable Diseases: The Quiet Emergency
NCDs are the dominant health challenge in virtually every Caribbean territory. A 2025 analysis found Caribbean women face particularly elevated rates of obesity and diabetes relative to global averages, compounded by socioeconomic disparities including food insecurity and unemployment. Hypertension rates above 30% of adults are common across Jamaica, Barbados, Trinidad and Tobago, Guyana, and the OECS. Diabetes prevalence in some territories exceeds 15% of the adult population. These are not conditions that require exotic interventions. They require consistent, well-timed primary care and patient behaviour change at scale. Both of those requirements are exactly what AI tools can address when they are integrated into community health worker workflows and electronic health records.
Vector-Borne Disease: A Regional Intelligence Gap
In 2025, the Caribbean experienced a 7% increase in dengue cases while the wider Americas saw a 66% decline compared to 2024. All four dengue virus serotypes circulated simultaneously across the region, raising the ongoing risk of severe cases and complex outbreaks. This divergence from the broader trend matters. When the rest of the hemisphere improves and the Caribbean does not, the explanation is structural: limited real-time surveillance capacity, delayed reporting, and insufficient inter-island data sharing. AI-powered early-warning systems address each of these gaps directly, and the evidence for their effectiveness now exists at scale. AI models have demonstrated outbreak prediction accuracy of up to 89.25% in dengue surveillance systems in the Caribbean region.
Maternal and Child Health: The Preventable Losses
Maternal mortality remains unacceptably high in Haiti, Guyana, and parts of Suriname. In smaller OECS territories, low patient volumes make maintaining specialist obstetric capacity economically difficult, meaning high-risk pregnancies are often managed at primary care level by practitioners who lack specialist backup. AI-powered antenatal risk assessment tools, identifying pre-eclampsia risk, gestational diabetes flags, and growth restriction signals from routine vital signs and blood work, are technically deployable today on tablet or smartphone interfaces that already exist in most primary care settings.
Mental Health: The Underfunded Crisis
Caribbean health budgets allocate disproportionately small shares to mental health. Post-disaster trauma is a recurring burden in the hurricane belt. As recently as 2025, PAHO deployed mental health and psychosocial services support to Jamaica following Hurricane Melissa. Between hurricane events, depression, anxiety disorders, and substance dependence are poorly screened and undertreated across the region. AI-powered screening tools, deployed through WhatsApp-based chatbots or community health worker tablets, can reach populations who would never attend a formal mental health clinic and route those with moderate-to-severe presentations to the limited specialist capacity that does exist.
Every Caribbean Territory: Key Health Challenges and AI Entry Points
The Caribbean is not a single health system. Haiti's maternal mortality rate is among the highest in the Western Hemisphere. Barbados manages a sophisticated chronic disease registry. Cuba has one of the highest doctor-to-patient ratios in the world but faces pharmaceutical supply constraints. Suriname's interior health access challenge is geographic. The Cayman Islands has per-capita health expenditure that exceeds many European countries. A single AI deployment strategy cannot serve all of these contexts, but each has specific, actionable AI entry points.
The country-level patterns reveal two distinct AI readiness tiers. Territories with existing electronic health record infrastructure, specifically Barbados, Trinidad and Tobago, Jamaica, and the Overseas Territories, can apply AI to structured data that already exists. Territories with fragmented or paper-based records, including much of Haiti, parts of Guyana's interior, and smaller Eastern Caribbean states, need basic health information infrastructure before AI analysis produces reliable outputs. The AI entry point in the second tier is not clinical AI: it is data collection AI, including mobile-based community health worker reporting, image-based diagnostic support, and satellite observation.
Hypertension affects an estimated 30%+ of adults. Diabetes burden is rising sharply. Cervical cancer mortality is among the highest globally for a middle-income country. Hurricane risk is annual.
→ AI priority: NCD risk scoring in primary care; AI-assisted cervical cytology to reduce screening backlogDengue is endemic and cyclical. Obesity rates are among the highest in the Caribbean. An ageing population creates growing demand for chronic disease management and geriatric care.
→ AI priority: Dengue early warning; AI pathology triage; geriatric care decision supportMaternal mortality ratio is among the highest in the hemisphere. Cholera re-emerged following 2021. Gang violence has disrupted health service access in major urban areas. Vaccine cold chain gaps persist.
→ AI priority: Community health agent triage support; offline-capable diagnostic tools; supply chain forecastingMost sophisticated chronic disease registry in the English-speaking Caribbean. Ageing population creates rising NCD and mental health demand. Laboratory quality leader for Eastern Caribbean.
→ AI priority: Predictive modelling on existing registry; AI dermatology screening; regional surveillance hubInterior hinterland communities lack consistent health access. Malaria remains a risk in forest regions. Maternal mortality in interior populations is significantly above coastal rates. Sickle cell disease burden is high.
→ AI priority: Satellite health monitoring; malaria prediction; remote antenatal triage via mobileHIV prevalence is among the highest in the non-Spanish Caribbean. Vector-borne disease is a year-round challenge. Alcohol-related harm prompted a National Alcohol Forum in 2025. Rural access gaps persist.
→ AI priority: HIV treatment adherence support; substance use disorder screening tools; vector forecastingInterior Amazonian communities face severe health access challenges. Malaria is endemic in forest regions. Indigenous population health data is limited. Maternal mortality rates are elevated.
→ AI priority: Satellite-based health monitoring; malaria prediction models; multilingual community health toolsWorld-class physician density but pharmaceutical supply shortages affect treatment continuity. Dengue is endemic. An ageing population requires expanding geriatric care infrastructure.
→ AI priority: Drug demand forecasting; dengue severity triage; AI radiology to extend specialist reachLarge dengue burden. Health equity disparities between urban and rural populations are pronounced. Maternal mortality gaps exist in low-income areas. NCD burden is growing rapidly.
→ AI priority: Dengue outbreak forecasting; maternal risk AI at primary care level; NCD screeningNCDs are the leading cause of death. Zoonotic disease preparedness was prioritised at a 2025 PAHO-FAO workshop. Limited specialist capacity creates referral dependency on Barbados and Trinidad.
→ AI priority: Shared OECS surveillance AI; NCD risk scoring at community level; telemedicine triageSmall populations create health system scale challenges. Hurricane and climate disaster exposure is annual. Zoonotic disease risks are growing with climate change.
→ AI priority: Regional disaster health response AI; zoonotic early warning; post-hurricane mental health toolsVery small populations make standalone AI investment economically difficult. Tourism-driven economies create seasonal health surveillance needs. NCDs are the primary mortality driver.
→ AI priority: Shared regional platform investment; tourism health surveillance; NCD management AIActive ML dengue surveillance model via SEDSS with 89% outbreak prediction accuracy. Mental health burden elevated since Hurricane Maria. Post-COVID chronic disease backlog. Health brain drain to mainland US.
→ AI priority: SEDSS model expansion; mental health AI; post-disaster recovery health monitoringHigh per-capita health expenditure. Tourism creates imported disease risk and seasonal population spikes. Mental health and substance abuse demands are growing.
→ AI priority: Tourism-linked disease surveillance; mental health chatbot deployment; rapid outbreak alert systemsDutch health system integration provides data infrastructure advantages. NCDs and mental health are the primary burden. Dengue and Zika remain active vectors.
→ AI priority: NCD management AI; mental health support tools; dengue monitoring with Dutch health data linksArchipelagic geography creates health access challenges across 700 islands. Hurricane Dorian (2019) health impacts are a model for AI disaster response planning. NCDs are the primary burden.
→ AI priority: Hurricane health response coordination AI; NCD management across islands; pharmacy supply AIEight Specific AI Use Cases the Caribbean Can Deploy Now
The following use cases are not theoretical. Each has a documented analogue in a comparable health system: small island developing states in the Pacific, low-resource settings in sub-Saharan Africa, or active deployments in the Caribbean itself. The deployment requirements are described honestly, including where data infrastructure prerequisites exist and where they do not.
Machine learning models trained on dengue surveillance data, climate variables, and entomological sampling can predict outbreak onset weeks before clinical case counts rise. Puerto Rico's SEDSS-based model achieves 89% accuracy. The same approach is deployable in Jamaica, Trinidad, Barbados, and the Dominican Republic using existing CARPHA surveillance data.
Inputs required: Weekly case counts by municipality, rainfall data, temperature data, and mosquito larval index readings. Alall of these are collected today. The gap is digitization and integration.
Tools: Python scikit-learn · WHO EIOS · CARPHA data feeds · CIMH climate dataA machine learning model that takes age, BMI, blood pressure, blood glucose, and family history from a routine primary care visit and produces a five-year cardiovascular or diabetic complication risk score. This exists in NHS England, the US Veterans Health Administration, and Rwanda's community health system. It requires structured electronic health records.
In Jamaica, Barbados, and Trinidad, EHR systems at the primary care level are expanding. Even a simplified rule-based risk score, deployable as a tablet app for community health workers, would identify high-risk patients who currently go undetected until emergency presentation.
Tools: ODK Collect · KoboToolbox · Microsoft Azure Health Bot · Google Health AIAI models that identify pre-eclampsia risk, gestational diabetes, and foetal growth restriction from routine antenatal care data: blood pressure readings, fundal height, urine dipstick results, and haemoglobin levels, flagging high-risk pregnancies for specialist referral. Deployed in low-resource settings in Uganda and Rwanda with measurable reductions in maternal mortality.
In Caribbean contexts, the target deployment is community health workers and rural primary care nurses in Guyana, Haiti, Suriname, and the interior of larger islands, where specialist obstetric backup is hours away.
Tools: Safe Delivery App · Medic Mobile · WHO SMART guidelines digital toolsWhatsApp-based or SMS-based chatbots that administer validated screening tools: the PHQ-9 for depression, the GAD-7 for anxiety, and route moderate-to-severe presentations to appropriate care levels. Deployed in the Philippines, Kenya, and several Pacific island states. Reaches populations who would never attend a formal mental health clinic.
In the Caribbean, post-hurricane deployment is the highest-value use case. PAHO deployed psychosocial support following Hurricane Melissa in Jamaica. An AI screening and routing layer on top of that human response would extend reach significantly at low marginal cost.
Tools: Woebot Health API · Mindline integration · Claude/ChatGPT on WhatsApp Business · Twilio SMSAI tools that read medical images: chest X-rays for TB and pneumonia, retinal photos for diabetic eye disease, skin photographs for dermatology, at a level that matches specialist accuracy. Google Health's retinal imaging AI has been validated in diabetic screening programs in India and Thailand. Cervical cytology AI reading is reducing cancer screening backlogs in lower-resource settings.
In Jamaica, cervical cancer mortality is disproportionately high relative to income level. The screening backlog at cytology laboratories is a documented bottleneck. An AI reading assistant for Pap smear images would extend the reach of existing laboratory personnel without requiring new specialist hires.
Tools: Google Health AI · Qure.ai · Sight Diagnostics · Microsoft InnerEyeAI demand forecasting for essential medicines, vaccines, and consumables reduces stockouts at facility level and prevents expiry-related waste in small health systems with limited cold chain capacity. The same approach used by the UN Supply Chain Financing Initiative in sub-Saharan Africa applies directly to the Caribbean, where pharmaceutical supply disruption is a recurring problem in Cuba, Haiti, and smaller island states following hurricanes or shipping disruptions.
Tools: Llamasoft / Coupa supply AI · UNICEF COVES · WHO EMP planning tools · custom ML forecastingSeasonal climate forecasts from CIMH (the Caribbean Institute for Meteorology and Hydrology) can be integrated with health surveillance data to produce disease risk forecasts three months in advance. CIMH already publishes Health Implications Bulletins for the Caribbean in partnership with CARPHA and PAHO. Adding an AI layer to this existing workflow would convert probabilistic climate forecasts into specific health system preparation signals: when to pre-position IV fluids, when to increase mosquito vector surveillance, when to brief hospital ERs on respiratory case surge expectations.
Tools: CIMH climate data API · OYA AI (Maestro AI Labs) · PAHO DHIS2 · custom Python integrationAI scribing tools that listen to a clinical encounter and generate a structured clinical note reduce the documentation burden on physicians, nurses, and community health workers by 30–60% in settings where they have been deployed in the US and Australia. In the Caribbean, where physician-to-population ratios are already strained, recovering that documentation time and redirecting it to patient contact is a direct health system capacity gain.
This use case requires the lowest infrastructure investment of any in this list. A smartphone and an API call are sufficient. The constraint is clinical validation, privacy regulation compliance, and the absence of Caribbean-specific medical vocabulary in most AI voice models.
Tools: Microsoft DAX Copilot · Nuance Dragon Medical · Abridge · Claude API with custom promptingThe Caribbean Health Data Gap: Why Generic AI Tools Will Fail Without Local Validation
Most global AI health models are trained on data from North America, Europe, the UK, and parts of Asia. Caribbean populations have distinct disease prevalence ratios, genetic profiles, dietary patterns, and social determinants of health that are not captured in global training datasets. An AI diagnostic model trained on European or American data may produce systematically biased outputs when applied to Caribbean patients.
Specific examples: sickle cell disease prevalence in Afro-Caribbean populations differs significantly from North American Black populations; dengue serotype patterns in the Eastern Caribbean do not mirror patterns used to train most global dengue AI models; the relationship between BMI and cardiovascular risk in Caribbean populations with different adiposity distribution patterns requires local calibration.
Every AI health tool deployed in the Caribbean should include a local validation step using Caribbean patient data before it is used to inform clinical decisions. This is not optional. It is the difference between a useful tool and a harmful one.
The data gap is not just a technical problem. It reflects a structural exclusion: Caribbean populations have been health data subjects for colonial-era health systems and global research programmes, but rarely owners or beneficiaries of the data intelligence that health data generates. The AI models trained on Caribbean health data should be trained by Caribbean institutions and should serve Caribbean health systems first. This is not protectionism. It is the difference between a system optimised for your population and one that applies someone else's parameters to your patients.
CARPHA's expansion of laboratory quality systems across the Eastern Caribbean, advancing efforts to strengthen laboratory quality management and reinforce the region's capacity for disease surveillance, emergency response, and overall health security, is the infrastructure layer on which Caribbean health AI must be built. Without quality data at the laboratory and surveillance level, AI models produce garbage at scale with high confidence.
AI Tools Available to Caribbean Health Practitioners Right Now
Most AI health tools currently accessible to Caribbean practitioners require only an internet connection and an institutional or professional email address. The barriers are awareness, training, and the absence of formal guidance from ministries of health rather than technology cost. The table below covers tools that are deployable today at zero or minimal cost and the specific health applications they support.
The single highest-value AI action a Caribbean health practitioner can take today requires no infrastructure: use Claude or ChatGPT to draft patient communication materials in plain language. Patient education leaflets, post-discharge instructions, chronic disease management plans, and vaccine information sheets that currently take an hour to write can be drafted in five minutes and reviewed in another five. The practitioner edits for clinical accuracy and local context. The total time reduction per document is significant enough to notice within a week of consistent use.
Five Risks the Caribbean Cannot Ignore
The fifth risk in the matrix above deserves more space than a card allows. The term Agent Colonialism describes AI systems deployed into communities without their input, their data, or their consent in the design process. In Caribbean health, this risk is already present: global health AI companies are entering the region with tools trained elsewhere, validated on non-Caribbean populations, and designed to extract operational intelligence from Caribbean health data without building the region's own analytical capacity.
A Jamaican health ministry that purchases an AI epidemiological surveillance system from a US technology company is not building Jamaican health intelligence. It is purchasing access to a system optimised for someone else's context, generating insights that flow to a platform it does not control, and paying for the privilege of having its own health data used to train models it has no governance over. The alternative is not refusing AI. It is insisting on data sovereignty, local validation, and the kind of technical partnership that builds Caribbean capability rather than Caribbean dependency.
What Needs to Happen at Three Levels
Government and Ministry of Health Level
Pass data governance legislation before deploying AI. The absence of clear data protection law in most Caribbean territories means patient data shared with AI platforms has no legal protection framework. Jamaica's Data Protection Act (2020) is the model. Every territory that has not passed equivalent legislation should treat that as the first prerequisite for any clinical AI deployment.
Invest in structured health data collection as the first AI step. AI models are only as good as the data they consume. Mobile data collection tools for community health workers: KoboToolbox, ODK Collect, or equivalent, deployed with basic training and mobile connectivity produce structured data that is the raw material for every use case described in this article.
Mandate CARPHA oversight of any AI tool used in national health systems. CARPHA's expanding laboratory and surveillance infrastructure is the correct regional governance body for AI health tool validation. No clinical AI tool should be deployed at national level without CARPHA technical review and local Caribbean validation data.
Health Practitioners and Institutions
Start with documentation and patient communication, not clinical diagnosis. The lowest-risk, highest-immediate-value AI use cases are in clinical writing: discharge summaries, patient education materials, referral letters, and health promotion content. These do not involve AI making clinical decisions. They involve AI drafting, human reviewing, and clinician finalising. Every health institution can start here without any new infrastructure.
Validate every AI screening or diagnostic tool on your patient population before using it. If you cannot run a retrospective validation study on local patient records, require the vendor to provide Caribbean-population validation data. If they cannot, the tool is not ready for your context.
Report AI errors when they occur. The Caribbean will not build its own AI health evidence base if practitioners do not document cases where AI tools produced wrong outputs for Caribbean patients. A systematic error reporting mechanism, even an informal shared registry within CARPHA's network, would accelerate Caribbean-specific AI learning significantly.
Regional Bodies: CARPHA, PAHO, CARICOM
Commission a Caribbean Health AI Readiness Assessment. No Caribbean-wide assessment of AI health readiness currently exists. CARPHA, in partnership with PAHO and regional university faculties of medicine, should publish a territory-by-territory assessment of data infrastructure, regulatory environment, and AI deployment readiness within the 2026–2027 planning cycle.
Fund a shared Caribbean health AI platform. Small island developing states cannot individually justify the investment in dedicated health AI infrastructure. A shared regional platform, the equivalent of what CIMH provides for climate data, and would allow every territory to access AI-powered surveillance, NCD risk modelling, and maternal health tools at a fraction of the cost of individual procurement.
Establish Caribbean health data sovereignty standards. Any AI company that accesses Caribbean health data for training or analysis purposes should be required to provide training data contributions back to Caribbean health institutions, publish validation results on Caribbean populations, and ensure regional governance oversight of how Caribbean health intelligence is used.
Frequently Asked Questions
-
How is AI being used in Caribbean public health right now?
The most active AI health deployments in the Caribbean are in disease surveillance and dengue prediction. Puerto Rico's Sentinel Enhanced Dengue Surveillance System has piloted machine learning models that predict dengue outbreak onset with 89% accuracy. CARPHA operates regional disease surveillance networks and is advancing laboratory quality systems across the Eastern Caribbean. Most other Caribbean territories have not yet deployed dedicated AI health systems, though WHO's EIOS platform and PAHO's DHIS2 provide AI-enhanced surveillance intelligence accessible to regional health ministries.
-
What is the biggest public health challenge AI could address in the Caribbean?
Non-communicable diseases account for 75–80% of deaths across Caribbean territories. Hypertension, diabetes, and cardiovascular disease are at epidemic levels. AI applications for NCD risk stratification and early warning in primary care offer the highest immediate impact because the disease burden is already enormous and the technology requires only structured health records and a mobile interface, both of which are increasingly available across the region.
-
Why is Caribbean health data a problem for AI systems?
Caribbean populations are critically underrepresented in global AI health training datasets, accounting for under 1% of global health AI training data. Caribbean populations have distinct genetic profiles, dietary patterns, disease prevalence ratios, and socioeconomic determinants of health. An AI model trained on North American or European data may produce systematically biased outputs when applied to Caribbean patients, particularly for conditions like sickle cell disease, dengue, or the specific NCD risk profiles of Caribbean populations.
-
What is CARPHA and how is it relevant to AI in Caribbean public health?
CARPHA, the Caribbean Public Health Agency headquartered in Port of Spain, Trinidad, serves as the primary regional public health body for CARICOM member states. It operates disease surveillance networks, laboratory quality systems, and epidemiological monitoring. Any AI health tool deployed at national level in the Caribbean should include CARPHA technical review as part of its validation process. CARPHA is also the appropriate body to lead a regional Caribbean health AI readiness assessment.
-
Which AI tools are available to Caribbean health practitioners today?
WHO's EIOS provides AI-powered epidemic intelligence at no cost. Google Health AI tools include validated retinal imaging and chest X-ray analysis. KoboToolbox and ODK Collect are free mobile data collection platforms with AI logic capabilities used by PAHO in field operations. Claude and ChatGPT are available on free tiers for clinical documentation drafting and patient communication materials. Microsoft's Azure Health Bot is deployable for patient triage. DHIS2 with AI extensions is the open-source health information system supported by PAHO for Caribbean deployment.
-
What does "Agent Colonialism" mean in the context of Caribbean health AI?
Agent Colonialism describes AI systems deployed into communities without their input, their data representation, or consent in the design process. In Caribbean health specifically, it refers to global AI health companies entering the region with tools trained on non-Caribbean populations, extracting operational intelligence from Caribbean health data, and building the company's models rather than the region's analytical capacity. The mitigation is insisting on data sovereignty agreements, local validation requirements, and technical partnerships that build Caribbean health AI capability rather than Caribbean dependency on external AI platforms.
Adrian Dunkley
Founder, StarApple AI · Co-founder and CEO, Maestro AI Labs
Adrian Dunkley is the founder of StarApple AI, the Caribbean's first AI company, and Co-founder and CEO of Maestro AI Labs. He is a member of Jamaica's National AI Task Force and advises CARICOM governments on national AI strategy. He is a Forbes Technology Council member, Caribbean AI Innovator of the Year, and an EY Entrepreneur of the Year award recipient. He trained as a physicist at UWI, specializing in climate physics.
Published by Adrian Dunkley for Caribbean AI. Statistics sourced from PAHO, CARPHA, CIMH, and peer-reviewed public health literature. Last verified May 2026.
