CLINICAL IMPACT

From triage to discharge.
Every step governed. Every decision supported.

Follow a patient through a BulletTrain-powered hospital where data flows through the OpenHIE (Open Health Information Exchange) mediator — an orchestration layer that coordinates 25 AI agents, clinical decision support, and operational systems for standards-compliant interoperability across every step of care.

Schedule an Architecture Walkthrough Follow the Patient Journey Watch the Journey Demo
THE PATIENT JOURNEY

Eight steps. Zero gaps.

Scroll through each step to see how data flows through the BulletTrain Health Information Exchange — an OpenHIE mediator that orchestrates clinical, operational, and AI systems to eliminate the failures that cost lives.

01 TRIAGE 02 BED 03 DIAGNOSIS 04 IMAGING & LABS 05 TREATMENT 06 MONITORING 07 DISCHARGE 08 FOLLOW-UP
STEP 01

Arrival & Triage

Patient arrives at the emergency department. The Triage Agent (Nexus) assesses using NEWS2 scoring. APEX Strategy runs mortality and deterioration models simultaneously. Patient classified as Category 2 (urgent) within 4 minutes.

In some health systems, clinical diagnosis takes 3–12 hours after triage due to workforce shortages and missing equipment. With AI-native triage, projected time is 35 minutes. (PMC, national community-health information system data)

Ensemble diagnostics — 4 strategies vote on every assessment, eliminating single-clinician bottlenecks.

ENSEMBLE DIAGNOSTICS
STEP 02

Bed Allocation

The Bed Manager Agent checks real-time availability across all wards. Patient matched to the appropriate bed based on clinical need, acuity, and proximity to required services.

In one major teaching hospital, 60 patients were sharing a 36-bed unit — treated in plastic chairs, on bare floors. A patient died after 3 hours being turned away from multiple hospitals with no beds. Real-time bed visibility prevents this. (Ghana GMA 2026)

BulletTrain orchestrator coordinates across departments. Digital bed management replaces phone calls and guesswork.

REAL-TIME ORCHESTRATION
STEP 03

Diagnosis

The Diagnostic Reasoning Agent generates a differential diagnosis using 4 ensemble strategies: Bayesian inference, heuristic patterns, LLM reasoning, and codified clinical guidelines (HEART, Wells, CURB-65, qSOFA). Cognitive biases countered: anchoring, availability, and confirmation bias.

Studies found major diagnostic errors in 40.3% of maternal deaths in one LMIC setting. Ensemble diagnostics reduces single-clinician bias by requiring multiple independent strategies to converge. (PMC, Mozambique study)

Confidence scores and evidence citations accompany every diagnosis. No black-box decisions.

CONFIDENCE SCORES + EVIDENCE CITATIONS
STEP 04

Imaging & Labs

The Imaging Agent orders appropriate investigations. The Investigation Planner optimises test sequence by cost and risk-benefit. Results feed directly into the Patient360 longitudinal view.

Duplicate tests are routinely ordered because previous results cannot be found across siloed systems. Patient360 eliminates this by providing a unified, consent-gated clinical record. (Frontiers, SSA interoperability)

Patient360 provides a consent-gated, longitudinal view of all investigations, eliminating redundancy.

PATIENT360 CONSENT-GATED VIEW
STEP 05

Treatment

The Treatment Recommendation Agent proposes interventions. The Prescribing Agent checks allergies, drug interactions (serotonin syndrome, QT prolongation), renal dosing, and contraindications. The Clinical Pathways Engine personalises guidelines — skips penicillin for allergic patients, adjusts for renal function.

Systematic reviews show medication errors reduced by 54% on average with digital prescribing safety checks. (JMIR 2025)

Drug interaction checking with Hunter criteria plus 7 personalisation rules ensure every prescription is safe for this patient.

DRUG INTERACTION CHECKING · 7 PERSONALISATION RULES
STEP 06

Ongoing Monitoring

APEX Strategies run 6 predictive models continuously: prognosis, risk, progression, complication, mortality, and deterioration. Threshold alerts fire to the clinician when the patient deteriorates. SignalBox orchestrator monitors all agent interactions.

Healthcare workers are overwhelmed — projected shortage of 5.3 million by 2030. Grounded Clarity gives them exactly the right clinical context at the right time, reducing cognitive load. (WHO AFRO workforce data)

Real-time predictive monitoring means deterioration is detected before it becomes critical.

REAL-TIME PREDICTIVE MONITORING
STEP 07

Discharge Planning

The Discharge Agent coordinates when the patient is clinically ready. The Pharmacy Agent handles medication reconciliation. The Follow-up Scheduler books a GP appointment. The Care Coordinator arranges home care if needed.

Studies show 24.7–63% of hospital patients in Africa experience prolonged stays — significantly higher than other regions. Automated discharge planning reduces this. (PMC 2024)

Multi-agent coordination through Nexus A2A ensures no discharge step is missed or delayed.

NEXUS A2A MULTI-AGENT COORDINATION
STEP 08

Follow-up & Continuity

The Primary Care Agent receives the handover. The Continuity Agent tracks ongoing care plans. The CCM Agent manages chronic conditions. The Public Health Surveillance Agent monitors for notifiable conditions.

In one major health system, 92.9% of patients bypass referrals and report directly to tertiary hospitals. Tracked follow-up with delegation chains closes this gap. (Frontiers, Nigeria 2025)

Complete delegation chain — every step auditable from discharge through community follow-up.

COMPLETE AUDITABLE DELEGATION CHAIN
THE AGENT NETWORK

25 agents. One patient journey.

Every agent is registered in the Nexus Agent Registry, governed by clinical policies, and coordinated through the BulletTrain orchestrator. Here is how they map to each step.

STEPS 01 – 02

Assessment & Allocation

Triage AgentAPEX StrategyBed ManagerOrchestrator
STEPS 03 – 04

Diagnosis & Investigation

Diagnostic ReasoningImaging AgentInvestigation PlannerPatient360
STEPS 05 – 06

Treatment & Monitoring

Treatment RecPrescribingClinical PathwaysAPEX (6 models)SignalBoxAmbient Scribe
STEPS 07 – 08

Discharge & Continuity

DischargePharmacyFollow-up SchedulerCare CoordinatorPrimary CareContinuityCCMSurveillance
MEASURABLE OUTCOMES

What the evidence shows.

3–12 hours
35 min

Triage Time

AI-native triage with NEWS2 scoring and ensemble diagnostics reduces time from arrival to clinical assessment.

Projected · national CHIS data, peer-reviewed
Baseline error rate
−54%

Medication Errors

Average reduction in medication errors with digital prescribing safety checks, drug interaction screening, and renal dosing adjustments.

JMIR 2025 · Systematic review
Single-clinician decisions
4-strategy voting

Diagnostic Errors

Ensemble voting across Bayesian, heuristic, LLM, and guideline-based strategies eliminates single-clinician diagnostic bias.

APEX ensemble architecture
Phone calls & guesswork
Real-time visibility

Bed Utilisation

Digital bed management with real-time availability across wards reduces blocked beds and prevents overcrowding.

Ghana GMA 8-point strategy · 2026
24.7–63% prolonged stays
Automated planning

Discharge Delay

Multi-agent discharge coordination reduces prolonged hospital stays through automated medication reconciliation, follow-up scheduling, and care coordination.

PMC 2024 · Africa hospital stay data
Manual documentation burden
−22.4% time

Documentation

Ambient clinical scribe reduces documentation time, returning clinician focus to patient care during a projected shortage of 5.3 million health workers.

Ambient scribe evidence base
GROUNDED CLARITY

Supporting the clinician, not replacing them.

Clinical decision-making fails when cognitive barriers go unaddressed. BulletTrain's Grounded Clarity framework identifies 6 barriers and provides structured support for each.

Anchoring Bias

Over-relying on the first piece of information. Ensemble voting forces consideration of multiple diagnostic hypotheses before convergence.

Availability Bias

Judging likelihood by how easily examples come to mind. Bayesian inference grounds probability in population-level evidence, not recall.

Confirmation Bias

Seeking information that confirms an existing hypothesis. Differential diagnosis agents actively generate competing explanations.

Information Overload

Too much data, too little context. APEX Strategy surfaces only clinically relevant signals through threshold-based alerting.

Fragmented Records

Critical information scattered across systems. Patient360 provides a unified, consent-gated longitudinal view of the entire clinical record.

Alert Fatigue

Desensitisation from too many notifications. SignalBox orchestrates alerts with clinical priority, delivering context-appropriate notifications.

Schedule an Architecture Walkthrough View Platform Architecture
CROSS-BORDER CARE

When the patient crosses borders.

Patient data does not stop at national boundaries, but sovereignty must. GHARRA (Governed Health Agent Resource and Routing Architecture) and sovereign BulletTrain instances handle cross-border care with full data governance.

Country A

Sovereign BulletTrain instance. Patient data remains within national jurisdiction. Local agents, local governance.

GHARRA

Country B

Sovereign BulletTrain instance. Receives only consented, governed clinical summaries. Full audit trail preserved.

See the full platform in action.

From triage to discharge and across borders — see how BulletTrain transforms clinical workflows with governed connectivity and AI-native decision support.

Schedule an Architecture Walkthrough Cross-Border Demo Agent Topology Demo