13 sites
Production clinical AI QA evidence base in the Sayvant preprint.
Research surface
Research that starts in the ED and has to survive the handoff.
This page is the public map of my current research work: clinical AI evaluation at Sayvant, a Stanford chest pain decision-support practicum, pediatric video laryngoscopy economics, and translational airway research around IntuBlade.
The common thread is practical validity. The work needs evidence, labels, workflow fit, and economic realism before it deserves trust.
1,089 iterations
Optimization loop tracked against physician complaints and regression checks.
2 preprints
2026 medRxiv outputs spanning documentation QA and airway procedural telemetry.
30 runs
Public IntuBlade manikin evidence-package audit for video-linked airway telemetry.
400+ agencies
IntuBlade field adoption informing translational airway research.
Active Research Programs
These are the active lanes described in the CV data, with completed outputs separated from work still in preparation.
Chest pain practicum
Training clinical decision support from ambient conversation
Stanford MCiM capstone / AIMI poster / AMIA 2026 podium abstract accepted
This work asks whether emergency chest pain decision support can be trained from the clinical conversation and physician-authored decision traces instead of only from downstream chart text.
Chest pain is a useful stress test for clinical AI because the danger is not a pretty note. The danger is false certainty around ACS risk, missing history, disposition, and what the physician actually knew at the bedside.
Public trail
- AIMI / Health AI Week poster presented at Stanford on Jun 3, 2026.
- AMIA 2026 podium abstract #15227 accepted.
- Public control repo for ED chest pain analytics and validation planning.
Current work
- Separate patient signal from documentation bias.
- Design validation around calibration, expected utility, and physician workflow.
- Keep the output contract modest: decision support, not physician replacement.
Sayvant evaluation work
Production documentation QA, ScribeBench, and SQS
medRxiv preprint, public benchmark artifact, and continuing methods work
The Sayvant research line is about evaluating clinical documentation AI in the setting where it can actually cause harm: signed charts, physician complaints, billing expectations, hallucination risk, and post-launch drift.
Fluent clinical text is cheap. Defensible documentation requires explicit scoring, source fidelity, dangerous-fabrication review, regression checks, and a loop that can turn physician complaints into product changes.
Public trail
- First-author medRxiv preprint posted May 29, 2026 with Justin Wiley and Mark Heslin.
- Production QA across 13 hospital sites, 42 tracked physician complaints, and 1,089 optimization iterations.
- ScribeBench and Sayvant SQS include physician-written checks for source fidelity, narrative quality, leaks, and dangerous fabrication.
Current work
- Turn SQS and emergency medicine benchmark artifacts into publishable methods material.
- Extend evaluation beyond fluency into chart grounding, clinical defensibility, and repair loops.
- Keep privacy, PHI routing, and HIPAA/BAA-aware deployment constraints visible in the evaluation story.
IntuBlade translational research
Human-in-the-loop airway procedural guidance
medRxiv preprint posted Jun 29, 2026 and translational protocol development
IntuBlade started as an access problem: make video laryngoscopy affordable enough to be present where airways actually happen. The active translational work extends that platform into video-linked telemetry, training, review, and procedural guidance.
Airway AI should not pretend to be autonomous. The credible path is connected video, event telemetry, computer vision observations, generated reports, and trainer-confirmed safety fields that keep the human accountable.
Public trail
- Computer-vision procedural telemetry preprint posted to medRxiv on Jun 29, 2026.
- Public 30-run manikin evidence package with structured JSON, QC status, package checks, and app/model metadata.
- FDA-regulated Class I single-use USB-C video laryngoscope platform.
- Related airway IP includes U.S. Patent No. 12,653,393, issued Jun 16, 2026.
Current work
- Keep the current claim bounded to package completeness, traceability, and assigned-condition consistency.
- Prepare the next validation step around independent review, report quality, and training workflow.
- Use generated reports, JSON exports, and trainer-confirmed safety fields without implying autonomous airway guidance.
Pediatric video laryngoscopy CEA
Cost-effectiveness and budget impact for pediatric video laryngoscopy
Active manuscript and budget-impact appendix buildout
This manuscript lane treats device adoption as a clinical and economic question: when does pediatric video laryngoscopy create enough value to justify coverage, training, stocking, and replacement costs?
Pediatric airway equipment decisions are not made in a vacuum. A credible analysis has to connect device cost, first-pass success assumptions, adverse-event avoidance, training burden, and the buyer economics that determine whether video is actually available.
Public trail
- CV-listed active manuscript lane for pediatric video laryngoscopy cost-effectiveness.
- Budget-impact appendix path in progress.
- Field-data strategy tied to IntuBlade deployment and airway training context.
Current work
- Model the cost and outcome assumptions transparently enough to survive sensitivity analysis.
- Separate hospital and EMS adoption economics where the purchasing constraints differ.
- Keep the claim focused on access and implementation rather than unsupported superiority.
Public Outputs
Preprints, benchmark artifacts, and presentations with enough public detail to inspect.
Computer-Vision Procedural Telemetry for Airway Guidance: A Public 30-Run Manikin Evidence-Package Audit
Posted Jun 29, 2026
medRxiv preprint
First-author preprint with Serge Klement and Ben Fedeles. The work audits a public post-reconciliation evidence package from 30 simulated airway runs using an IntuBlade device and iPhone workflow. The claim is intentionally bounded: video-linked structured JSON, QC status, app/model metadata, package hashes, and assigned-condition consistency for formative airway telemetry, not autonomous guidance, model accuracy, training effectiveness, clinical efficacy, or patient outcomes. DOI: 10.64898/2026.06.26.26356677.
Closed-Loop Quality Assurance for Production Clinical AI Documentation
Posted May 29, 2026
medRxiv preprint
First-author preprint with Justin Wiley and Mark Heslin. The work turns physician complaints from production AI documentation across 13 hospital sites into reproducible QA artifacts, driving 1,089 optimization iterations with zero regressions on resolved complaints. Evidence base: 42 tracked complaints, 282 binary checks, 166 deterministic repair rules, and 6/6 fabrication-class failures eliminated in ablation. DOI: 10.64898/2026.05.27.26353977v1.
Training Clinical Decision Support from Ambient Conversation
AIMI poster Jun 3, 2026; AMIA abstract #15227 accepted
Stanford AIMI / AMIA 2026 / manuscript in preparation
Stanford MCiM capstone work with a public control repo, AIMI poster, and AMIA 2026 podium abstract #15227. The project pairs ambient patient signal with physician-authored decision traces so decision-support models are not trained only on downstream chart text. Manuscript and validation plan in preparation with Ashley Griffin and Mark Musen.
Health-System Implementation of AI in Acute Care Delivery
Presented May 5, 2026
HITLAB Spring Summit, Sayvant session
Sayvant session on acute-care AI rollouts, workflow ownership, clinician review, rollout friction, feedback loops, and product lessons from health-system deployment. Final title: Planning for the Future: Lessons Learned Partnering with Health Systems to Implement AI in Acute Care Delivery.
ScribeBench and Sayvant SQS Framework
Released Jun 2026
Public benchmark and framework artifacts
Public clinical documentation benchmark and scoring framework for narrative quality, source fidelity, leak detection, dangerous fabrication, and rubric-based review of ambient scribe output.
Manuscript Pipeline
This list mirrors the CV pipeline. It is intentionally conservative: private or unfinished work stays labeled as such until there is a preprint, abstract, manuscript, or public artifact.
- Production clinical AI documentation QA: medRxiv preprint posted May 29, 2026, with peer-review package and reproducibility materials in preparation.
- ED chest pain decision support: Stanford AIMI poster presented Jun 3, 2026 and AMIA 2026 abstract #15227 accepted, with manuscript and validation plan in preparation.
- ScribeBench / SQS / EM benchmark: public benchmark and scoring framework for clinical documentation quality, source fidelity, leak detection, dangerous fabrication, and rubric-based review.
- Diagnosis-to-ICD-10 code selection: private research/software with Geanderson Santos using California HCAI code-frequency data, synthetic EM/HM examples, frequency-aware retrieval, constrained LLM selection, and error analysis.
- Computer-vision airway procedural telemetry: medRxiv preprint posted Jun 29, 2026 with a public 30-run manikin evidence package, package checker, QC records, and repository-level reproducibility materials.
- Pediatric video laryngoscopy cost-effectiveness analysis: manuscript and budget-impact appendix in progress.
Research Interests
The method lane is narrow on purpose: high-acuity clinical AI, documentation reliability, and airway systems where the cost of being wrong is real.
- Clinical AI evaluation and implementation science in emergency and hospital medicine
- Ambient documentation QA, dangerous-fabrication detection, and physician oversight
- Clinical conversation signals for risk stratification, decision support, and validation
- Physician-authored rubrics, labels, and adjudication workflows for model evaluation
- Clinical coding, structured labels, retrieval-plus-selector methods, and error analysis
- Video-linked procedural telemetry and human-in-the-loop guidance for airway and other high-risk procedures
Recent Presentations
Talks and presentations that connect the research to clinical operations, implementation, and device work.
- Emergency Chest Pain Clinical AI Work. AMIA 2026 podium abstract accepted, abstract #15227, Fall 2026.
- IntuBlade pitch presentation. Stanford Emergence Annual Impact Summit, Stanford, CA, Jun 5, 2026.
- Training Clinical Decision Support from Ambient Conversation: Separating Patient Signal from Documentation Bias in Emergency Chest Pain. Stanford AIMI Symposium / Health AI Week, poster presentation, Stanford, CA, Jun 3, 2026.
- Planning for the Future: Lessons Learned Partnering with Health Systems to Implement AI in Acute Care Delivery. HITLAB Spring Summit, Sayvant session, May 5, 2026.
- The HIT Will Come. Guest lecture, Eastern Kentucky University Emergency Medical Care Program, Richmond, KY, Apr 2026.
Publication Record
The formal CV language is below, followed by selected older clinical publications and book-length work.
CV-listed publications
- Napier A., Klement S., Fedeles B. Computer-Vision Procedural Telemetry for Airway Guidance: A Public 30-Run Manikin Evidence-Package Audit. medRxiv preprint. Posted Jun 29, 2026. DOI: 10.64898/2026.06.26.26356677.
- Napier A., Wiley J., Heslin M. Closed-Loop Quality Assurance for Production Clinical AI Documentation. medRxiv preprint. Posted May 29, 2026. DOI: 10.64898/2026.05.27.26353977v1.
- Napier A., Zitek T. Decreased time to intubation by experienced users with a new lens-clearing video laryngoscope in a simulated setting. American Journal of Emergency Medicine. 2021. PMID: 33632548.
- Azimi-Ghomi O., Napier A. "Getting Under Your Skin": First Reported Case of Transient Reactive Phlebitis Involving Vancomycin Infusion Therapy. Cureus. 2021. PMID: 34754636. DOI: 10.7759/cureus.18471.
- Napier A. Digital Nerve Block. StatPearls. 2018. PMID: 30252367.
- Napier A. Supraorbital Nerve Block. StatPearls. 2019. PMID: 30725622.
- Napier A. The Expected D-Dimer Values During Each Trimester of Pregnancy. Annals of Emergency Medicine abstract. 2019. DOI: 10.1016/j.annemergmed.2019.08.049.
- Napier A. A Disappearing Act: The curious case of Lemierre's Syndrome. EMPulse. 2020.
- Napier A. Emergency Medicine Ultrasound Pocketbook Guide. Independently published. 2020.
Older clinical work
Decreased time to intubation by experienced users with a new lens-clearing video laryngoscope in a simulated setting
American Journal of Emergency Medicine · 2021
Publication with Tony Zitek. PMID: 33632548.
"Getting Under Your Skin": Transient Reactive Phlebitis With Vancomycin
Cureus · 2021
Case report on transient reactive phlebitis involving vancomycin infusion therapy. PMID: 34754636. DOI: 10.7759/cureus.18471.
Emergency Medicine Ultrasound Pocketbook Guide
Independently published · 2020
Compact emergency medicine ultrasound reference.
Digital Nerve Block and Supraorbital Nerve Block
StatPearls · 2018-2019
Procedural reference chapters. PMIDs: 30252367 and 30725622.
The Expected D-Dimer Values During Each Trimester of Pregnancy
Annals of Emergency Medicine abstract · 2019
ACEP abstract. DOI: 10.1016/j.annemergmed.2019.08.049.
A Disappearing Act: The curious case of Lemierre's Syndrome
EMpulse · 2020
Clinical case article for Florida College of Emergency Physicians.
For the formal version
The CV carries the official record for publications, patents, presentations, education, awards, and clinical roles. This page is the readable public map of the research work in motion.