Skalerbar Klinisk Overvågning af Store Sprogmodeller Via Usikkerhedstriangulering (SCOUT)
Prospektiv Evaluering af en Model-Agnostisk Meta-Verificeringsramme (SCOUT) til Skalerbar Klinisk Overvågning af Store Sprogmodels Output i Diagnostisering af Koronar Hjertesygdom: Et Multi-Læser, Randomiseret, Crossover-Studie
This prospective, multi-reader, randomized crossover trial evaluates SCOUT (Scalable Clinical Oversight via Uncertainty Triangulation), a model-agnostic meta-verification framework that selectively defers unreliable large language model (LLM) predictions to clinicians by triangulating three orthogonal uncertainty signals: model heterogeneity, stochastic inconsistency, and reasoning critique. The trial assesses whether SCOUT-assisted review can reduce physician review time compared with standard manual review of AI-generated diagnoses while maintaining non-inferior diagnostic accuracy in coronary heart disease (CHD) subtyping.
Studieoversigt
Status
Status
Betingelser
Betingelser
Intervention / Behandling
Intervention / Behandling
Detaljeret beskrivelse
Background: Large language models are increasingly deployed in clinical workflows, yet requiring clinician review of every AI output negates the efficiency gains that motivate their adoption. SCOUT addresses this efficiency-safety paradox through algorithmic meta-verification.
The SCOUT framework triangulates three orthogonal external signals to determine case-level uncertainty: (1) Model Heterogeneity - whether a structurally different auxiliary LLM agrees with the primary model; (2) Stochastic Inconsistency - whether repeated sampling from the same model yields divergent outputs; (3) Reasoning Critique - whether an external checker model identifies logical flaws in the chain-of-thought reasoning.
In this crossover trial, 7 clinicians of varying seniority (2 junior residents, 3 senior residents, 2 attending physicians) each review all 110 cases under both standard manual review and SCOUT-assisted review workflows. The study evaluates workflow efficiency (primary endpoint) and diagnostic accuracy (secondary endpoint).
Undersøgelsestype
Undersøgelsestype
Tilmelding (Faktiske)
Tilmelding
Fase
Fase
- Ikke anvendelig
Kontakter og lokationer
Studiesteder
-
-
Beijing Municipality
-
Beijing, Beijing Municipality, Kina, 100037
- Fuwai Hospital
-
-
Deltagelseskriterier
Berettigelseskriterier
Berettigelseskriterier
Aldre berettiget til at studere
Tager imod sunde frivillige
Køn, der er berettiget til at studere
Beskrivelse
Inclusion Criteria:
- Board-certified or in-training cardiologists at Fuwai Hospital
- Spanning three experience strata: junior residents, senior residents, attending physicians
Exclusion Criteria:
- Clinicians involved in the development or optimization of the SCOUT framework
- Clinicians involved in the gold-standard adjudication process
Studieplan
Hvordan er undersøgelsen tilrettelagt?
Design detaljer
- Primært formål: Diagnostisk
- Tildeling: Randomiseret
- Interventionel model: Crossover opgave
- Maskning: Ingen (Åben etiket)
Antal våben
Våben og indgreb
Deltagergruppe / ArmDeltagergruppe / Arm |
Intervention / BehandlingIntervention / Behandling |
|---|---|
|
Aktiv komparator: Control (Standard Manual Review)
Physicians manually review all cases in the control set (n=54) with access to AI predictions and reasoning. No selective deferral. |
Physicians perform a full manual review of 54 cases using raw medical records with access to the AI model's predictions and reasoning, but without SCOUT uncertainty stratification or selective deferral. |
|
Eksperimentel: Experimental (SCOUT-Assisted Review)
Physicians process the intervention set (n=56) through the SCOUT framework. Low-uncertainty cases are auto-accepted; high-uncertainty cases undergo physician review with full audit trail. |
SCOUT-Assisted Review (Intervention Arm): Physicians review 56 cases processed through the SCOUT framework. For cases classified as low-uncertainty (D(x)=0), the AI prediction is auto-accepted without physician review. For high-uncertainty cases (D(x)=1), the physician reviews the case with access to the main model's chain-of-thought reasoning and the meta-verification audit results. The main model is DeepSeek-V3.1 with chain-of-thought prompting. |
Hvad måler undersøgelsen?
Primære resultatmål
Primære resultatmål
Resultatmål |
Foranstaltningsbeskrivelse |
Tidsramme |
|---|---|---|
|
Mean physician review time per case (minutes)
Tidsramme: Through study completion, an average of 2 hours.
|
Mean time spent by each clinician reviewing and rendering a diagnostic decision per case under each arm. Measured in minutes. |
Through study completion, an average of 2 hours.
|
Sekundære resultatmål
Sekundære resultatmål
Resultatmål |
Foranstaltningsbeskrivelse |
Tidsramme |
|---|---|---|
|
Diagnostic accuracy (%)
Tidsramme: Through study completion, an average of 2 hours.
|
Proportion of correct CHD subtype classifications (STEMI, NSTEMI, unstable angina, chronic coronary syndromes) under each arm. |
Through study completion, an average of 2 hours.
|
|
Computational Return on Investment (ROI)
Tidsramme: Through study completion, an average of 2 hours.
|
Ratio of physician time savings (valued at standardized minute-wages from Sanming healthcare reform benchmarks) to computational cost of SCOUT inference, stratified by clinician seniority level. |
Through study completion, an average of 2 hours.
|
Samarbejdspartnere og efterforskere
Sponsor
Sponsor
Datoer for undersøgelser
Studer store datoer
Studiestart (Faktiske)
Studiestart
Primær færdiggørelse (Faktiske)
Primær færdiggørelse
Studieafslutning (Faktiske)
Studieafslutning
Datoer for studieregistrering
Først indsendt
Først indsendt
Først indsendt, der opfyldte QC-kriterier
Først indsendt, der opfyldte QC-kriterier
Først opslået (Faktiske)
Først opslået
Opdateringer af undersøgelsesjournaler
Sidste opdatering sendt (Faktiske)
Sidste opdatering sendt
Sidste opdatering indsendt, der opfyldte kvalitetskontrolkriterier
Sidste opdatering indsendt, der opfyldte kvalitetskontrolkriterier
Sidst verificeret
Sidst verificeret
Mere information
Begreber relateret til denne undersøgelse
Nøgleord
Yderligere relevante MeSH-vilkår
Andre undersøgelses-id-numre
Andre undersøgelses-id-numre
- 2025-2702-1
Plan for individuelle deltagerdata (IPD)
Planlægger du at dele individuelle deltagerdata (IPD)?
IPD-planbeskrivelse
De-identified individual participant data underlying the results reported in this study will be made available.
IPD-delingstidsramme
Beginning 1 months after publication of the primary results and available for up to 60 months.
IPD-delingsadgangskriterier
Data are available from the corresponding author upon reasonable request. Requestors will need to provide a methodologically sound research proposal and sign a data use agreement.
IPD-deling Understøttende informationstype
- STUDY_PROTOCOL
- SAP
- ICF
- ANALYTIC_CODE
- CSR
Lægemiddel- og udstyrsoplysninger, undersøgelsesdokumenter
Studerer et amerikansk FDA-reguleret lægemiddelprodukt
Studerer et amerikansk FDA-reguleret enhedsprodukt
Disse oplysninger blev hentet direkte fra webstedet clinicaltrials.gov uden ændringer. Hvis du har nogen anmodninger om at ændre, fjerne eller opdatere dine undersøgelsesoplysninger, bedes du kontakte register@clinicaltrials.gov. Så snart en ændring er implementeret på clinicaltrials.gov, vil denne også blive opdateret automatisk på vores hjemmeside .