Winner of the Most Innovative Award for HL7’s 2026 AI Challenge
Non-communicable diseases (NCDs) are no longer only a clinical challenge. They are a health-system, economic, and development challenge.
NCDs account for 75% of non-pandemic-related deaths worldwide. Cardiovascular disease, cancer, chronic respiratory disease and diabetes cause 80% of all premature NCD deaths, and 73% of NCD deaths occur in low- and middle-income countries. [1,4]
Botswana shows why the next phase of NCD prevention must combine clinical care with better data use. The latest national STEPS ( STEPwise approach to NCD risk factor surveillance) survey found that hypertension prevalence rose from 31.1% in 2014 to 39.6% in 2024. Nearly half of adults are now overweight, and more than one in five are obese. [2]
These diseases do not occur independently. Hypertension, diabetes, obesity, cardiovascular disease and chronic kidney disease (CKD) share risk factors, biological pathways and, ultimately, the same patient. Yet our digital systems often still manage them in separate clinical and administrative silos.
That fragmentation is one of the problems we set out to address with the BioESC Health Intelligence Platform (B-HIP). B-HIP received the Most Innovative Award in the 2026 HL7 International AI Challenge, selected from 73 submissions across 25 countries. It was the category's first year and the first time an African-led platform has received the award.
B-HIP began with a simple question. What if the same interoperable data used to estimate one patient's future risk could also help a health system understand its future clinical and economic burden?
The answer is an architecture designed around prevention rather than late-stage disease management. B-HIP brings together a patient-facing mobile app, a clinician portal, a FHIR-based longitudinal health record, disease registries, clinical decision support, e-referral pathways, and population-health analytics.
Rather than building separate digital tools for hypertension, diabetes, kidney disease and cardiovascular risk, we aim to create one common health-intelligence layer in which these conditions are understood together.
We call the individual-level view the Patient Twin. It brings together observations such as blood pressure, weight, glucose, kidney function, medications, diagnoses, and laboratory results to build a longitudinal picture of each person’s NCD risk.
Validated clinical models then sit within that common architecture. For kidney disease, B-HIP uses established measures such as the CKD-EPI equation and the Kidney Failure Risk Equation (KFRE), which has been externally validated across international populations. [3] The same architecture can progressively add cardiovascular risk, hypertension control, diabetes risk, and other validated models, without building a new digital system for each disease.
This distinction matters. B-HIP does not rely on a single artificial-intelligence algorithm that claims to predict everything. It is a standards-based environment in which validated clinical equations, machine-learning tools and decision-support pathways coexist, with full transparency about where each prediction comes from.
Prediction has limited value unless it changes what happens next, so B-HIP is designed around intervention. A clinician who sees uncontrolled hypertension, declining kidney function, or rising cardiometabolic risk can trigger the right pathway: treatment optimization, repeat investigations, lifestyle support, referral, or closer monitoring.
The platform can then model how different scenarios might change projected clinical outcomes and, where the evidence allows, their economic consequences. The principle is simple: detect earlier, intervene earlier and make the consequences of those decisions visible.
For chronic disease, this matters enormously. The clinical and financial costs of hypertension, diabetes and kidney disease often appear years after the first opportunity for prevention.
The second B-HIP perspective is the National Twin. FHIR-enabled data, appropriately governed and de-identified, can be aggregated to show patterns of hypertension, diabetes, CKD and wider NCD risk across populations and regions.
Instead of seeing only today's disease burden, planners can begin to ask forward-looking questions:
This turns interoperability from a technical exercise into a planning capability. The information that supports a decision in one consultation can also, with the right governance, help a country understand population need.
HL7 FHIR is the common language beneath both views. Resources such as Patient, Observation, Condition, MedicationRequest, DiagnosticReport, RiskAssessment and Provenance represent clinical information consistently, and every risk estimate B-HIP produces is written back as a RiskAssessment with its Provenance.
SMART on FHIR lets the application launch inside existing clinical systems, so clinicians need no second login. CDS Hooks brings decision support into the workflow at the moment of decision. FHIR Bulk Data supports population-level aggregation for the National Twin, and the International Patient Summary offers a route to cross-border continuity of care. Legacy HL7 Version 2 (V2) laboratory and administrative feeds can be mapped to FHIR, so the platform works with health-system infrastructure as it is today.
FHIR therefore does more than let systems exchange records. It makes it possible to build reusable clinical and analytical services on the same structured data. That matters most where resources are constrained and building a separate digital system for every disease is neither affordable nor sustainable.
Building B-HIP taught us that in lower-resource health systems, the binding constraint is rarely model accuracy. The data and the science often already exist. The hard part is assembling them into a timely decision that a clinician will act on and a funder will pay for.
Three lessons stood out:
We also found gaps where the HL7 community could make a real difference:
This thinking now informs a broader Chronic and Non-Communicable Disease Prevention Program being advanced through Botswana's Economic Transformation Program. The proposed model combines a mobile app, clinician portal, FHIR-based health record, disease registry, e-referrals, prevention pathways, remote monitoring and analytics within a single national digital-health architecture.
The ambition is to shift the center of gravity from late-stage treatment towards earlier detection, prevention and long-term management. Botswana's rising rates of hypertension, overweight and obesity make this especially timely. [2]
The opportunity extends well beyond Botswana. Emerging health systems that build on interoperable standards today can avoid much of the fragmentation that more mature systems are now spending heavily to undo.
Winning the HL7 International AI Challenge is not the end point for B-HIP. It validates an idea that now needs rigorous real-world testing.
The next phase is implementation: validating clinical models in the populations they will serve, testing clinical workflows, defining economic assumptions, strengthening data governance, evaluating usability and measuring whether the platform genuinely improves early identification and intervention. B-HIP is designed as clinical decision support with a clinician always in the loop, and its economic modelling informs planning, never individual care decisions.
We welcome collaboration with clinicians, researchers, HL7 implementers, health informaticians, technology partners and ministries of health who share an interest in standards-based NCD prevention.
The ultimate ambition of B-HIP is not simply to predict disease. It is to connect the decisions made for one person today with a clearer view of what a whole health system will face tomorrow. If interoperability can help us do that, FHIR becomes far more than a data standard. It becomes infrastructure for prevention.