Bridge Tbi Protocol

Steps

Step 1: Understand the Biomarkers

Both are FDA-cleared (Abbott i-STAT TBI Plasma test) when used together with GCS 13-15 within 12 hours of injury.

Step 2: Map the Decision Protocol

Patient presents with mild TBI (GCS 13-15, within 12 hours)
│
├── Draw blood for GFAP + UCH-L2
│
├── IF GFAP < threshold AND UCH-L2 < threshold
│   └── LOW RISK → Observe, no CT needed
│       (Sensitivity ≥ 97% for intracranial lesions)
│
├── IF either biomarker ≥ threshold
│   └── ELEVATED RISK → CT scan indicated
│
└── Clinical judgment always overrides biomarker result
    (e.g., anticoagulant use, deteriorating GCS)

Step 3: Evidence Chain Assessment

Apply the evidence chain framework to BRIDGE-TBI:

Phase 1 — Technical Validation:

• Multi-site derivation and validation (Bazarian et al., Lancet 2023)
• 28,612 patients across multiple studies
• GFAP+UCH-L2 AUROC for detecting CT-positive TBI

Phase 2 — Clinical Validation:

• Prospective, multi-center studies
• Sensitivity ≥ 97% for traumatic intracranial lesions
• 32% CT reduction (meta-analysis)

Phase 3 — Clinical Utility:

• Decision Curve Analysis showing net benefit at relevant thresholds
• NNT calculations (how many patients tested to prevent one unnecessary CT?)
• Comparison to existing rules (Canadian CT Head Rule, New Orleans Criteria)

Phase 4 — Implementation:

• FDA clearance (2023) for point-of-care testing
• Emergency department workflow integration studies
• Turnaround time: ~15 minutes (point-of-care)

Phase 5 — Monitoring:

• Post-market surveillance ongoing
• Population-specific calibration questions (do thresholds hold in African populations?)

Step 4: Calculate Clinical Impact

Using the meta-analysis data:

• 28,612 patients with mild TBI
• CT reduction: 32% (patients safely diverted from CT)
• Sensitivity: ≥ 97% for intracranial lesions requiring intervention
• Missed lesion rate: < 0.5% (and none requiring neurosurgical intervention in major studies)

Calculate for your setting:

• How many mild TBI patients does your facility see per year?
• At 32% reduction, how many CTs would be avoided?
• What is the cost saving? (CT cost × avoided scans)
• What is the time saving for patients? (CT wait time × avoided scans)

Step 5: LMIC Adaptation Design

For a resource-limited setting (e.g., Liberia, Cameroon), design an adaptation:

Key adaptation questions:

1. Is the biomarker threshold valid in this population? (need local validation study)
2. Can clinical officers use the protocol safely? (task-shifting study needed)
3. What's the referral pathway when CT is indicated but not on-site?
4. What's the cost-effectiveness in this resource context?

Artifacts

1. Protocol Flowchart — Visual decision flowchart for BRIDGE-TBI adapted to your setting
2. Evidence Chain Map — Complete 5-phase evidence assessment for BRIDGE-TBI
3. Impact Calculator — Spreadsheet showing CT reduction, cost savings, and time savings for a specific facility
4. LMIC Adaptation Proposal — 2-page protocol adaptation for a named resource-limited setting

Assessment Criteria

Common Mistakes

• Confusing the CT reduction rate (32% from meta-analysis) with sensitivity (≥ 97%)
• Assuming HIC biomarker thresholds automatically apply in LMIC populations (they may not — genetic, nutritional, and injury pattern differences)
• Ignoring the "clinical override" component — biomarkers assist but don't replace clinical judgment
• Treating BRIDGE-TBI as "solved" — Phase 5 monitoring and population-specific validation are ongoing

Related Skills

• evidence-chain-assessment — Apply the framework to other clinical AI tools
• evaluate-model-calibration — Calibration of biomarker thresholds
• regulatory-landscape-analysis — Regulatory pathway for biomarker tests in Africa
• model-card-generator — Document the BRIDGE-TBI evidence in a standardized card

References

• Bazarian et al. (2023). Serum GFAP and UCH-L1 for prediction of absence of intracranial injuries on head CT (ALERT-TBI): Lancet. https://doi.org/10.1016/S0140-6736(23)01554-5
• FDA Clearance: Abbott i-STAT TBI Plasma test (2023)
• Korley et al. (2022). Performance of GFAP in TBI: https://doi.org/10.1016/S1474-4422(21)00083-5
• EvidenceOS BRIDGE-TBI Module: See awesome-health-ai-evaluation for full model registry entry