Student or educator? Sign up with your school email and get 6 months of Pro free.
CoChat

The DOI matches its Crossref record; 22 of the 22 references it cites were found in a registry or an index; 26 of the 38 sources its claims cite were checked for support.

An auditable and source-verified framework for clinical AI decision support: integrating retrieval-augmented generation with data provenance

Open access
Alu, Fidelis Fidelis, Oluwadare, Sunkanmi · 2026 · Frontiers in Artificial Intelligence

Abstract

Conceptual framework for auditable, source-verified AI-CDSS integrating curated evidence retrieval, provenance metadata, and tamper-evident audit logging.

Overview

CoChat Summary

## TL;DR

This paper is a conceptual framework for clinical AI decision support that combines retrieval-augmented generation with explicit provenance and tamper-evident audit logging. It does not present a prototype, experiments, or empirical results; instead, it proposes a trust-oriented architecture and a staged evaluation roadmap.

## Key findings

  • The authors propose three core trust requirements for an AI-CDSS: every recommendation must be backed by identifiable sources, every stage of processing must be recorded, and outputs must be delivered in a user-friendly manner.
  • The framework has three main components: a data/knowledge layer, a retrieval and reasoning engine, and a user interface with audit logging.
  • In the paper’s sepsis walkthrough, the system retrieves the Surviving Sepsis Campaign 2021 guideline and returns the recommendation: 30 mL/kg IV crystalloid fluid immediately for suspected sepsis-induced hypoperfusion, labeled “Strong recommendation, Low quality of evidence.”
  • The authors’ planned evaluation includes PubMedQA, MedQA, and clinician studies with about 50–60 clinicians; they state this sample size should provide around 80% power to detect meaningful differences in trust and cognitive workload at α = 0.05.

## Methods

  • Design: Conceptual, future-oriented system design developed by synthesizing prior literature on evidence-based medicine, provenance, trustworthy AI, and clinical decision support; no prototype or empirical experiments were conducted.
  • Data/sample: No study sample was collected for the paper itself. The framework is described using curated medical sources such as clinical guidelines, PubMed abstracts, textbooks, and terminology systems including SNOMED-CT and UMLS; a clinical simulation uses a suspected sepsis case.
  • Main analysis: Architectural proposal and walkthrough of a retrieval-augmented reasoning pipeline with provenance metadata, semantic mapping, audit logging, and planned evaluation phases (simulation testing, clinician usability study, and decision-quality proxies).

## Limitations

  • The paper explicitly states that it is purely conceptual and that the system has not been implemented, tested, or empirically validated; its practical effectiveness is unknown.
  • The authors note a dependency on a comprehensive, up-to-date medical knowledge base, which they say would be a major challenge to build and maintain.
  • They also identify citation misattribution in RAG systems, latency/usability trade-offs, bias in underlying evidence, and privacy/security concerns as feasibility issues; these are limitations stated in the paper.

Model-written summary. Check it against the paper before relying on it.

Integrity signals

Registry

Crossref · 10.3389/frai.2026.1737532

Cites

22 of 22 references verified0 citation errors · 0 not checkable

Statistics

Nothing to recompute.

Figures

Figure 1 Illustrates the workflow: (A) Clinician enters patient data and question; (B) system retrieves relevant evidence from its curated database; (C) reasoning engine synthesizes an answer, marking each point with reference IDs; (D) final answer and references are presented to the user, while a background module logs all steps.
Figure 1 Illustrates the workflow: (A) Clinician enters patient data and question; (B) system retrieves relevant evidence from its curated database; (C) reasoning engine synthesizes an answer, marking each point with reference IDs; (D) final answer and references are presented to the user, while a background module logs all steps.CC BY · PubMed Central
Figure 2 Shows a block diagram with modules and data flow: user interface, knowledge base, retrieval engine, AI model, citation generator, and audit logger. The diagram shows the major components and data flow. The clinician interacts via a user interface which sends the query (with relevant patient data) to the system. The nowledge ase (curated medical repository) is queried by the etrieval odule, which returns relevant documents. The AI easoning ngine (LLM) then generates an answer using those documents, and a itation enerator ensures references are attached to the answer. Meanwhile, an udit ogger records each step (query, retrieved sources, generated answer, etc.) into a secure log storage. Finally, the answer with its citations is displayed to the clinician. The numbered sequence (1–4) indicates the workflow: (1) ser’s query in, (2) evidence retrieved, (3) answer produced with citations, (4) answer and references returned to user, with all steps logged in the background.
Figure 2 Shows a block diagram with modules and data flow: user interface, knowledge base, retrieval engine, AI model, citation generator, and audit logger. The diagram shows the major components and data flow. The clinician interacts via a user interface which sends the query (with relevant patient data) to the system. The nowledge ase (curated medical repository) is queried by the etrieval odule, which returns relevant documents. The AI easoning ngine (LLM) then generates an answer using those documents, and a itation enerator ensures references are attached to the answer. Meanwhile, an udit ogger records each step (query, retrieved sources, generated answer, etc.) into a secure log storage. Finally, the answer with its citations is displayed to the clinician. The numbered sequence (1–4) indicates the workflow: (1) ser’s query in, (2) evidence retrieved, (3) answer produced with citations, (4) answer and references returned to user, with all steps logged in the background.CC BY · PubMed Central

Verification

Claims the paper makes

25 Valid · 0 Not valid · 12 Not checked · 35 Own results and interpretations

12 more claims · 1 with their cited sources checked

Create a free account to see all 72 claims and to verify the 5 references still open.

See all claims — free account
  • Probably not a substantial claim. 45
  • It may state two separate claims as one. 24

Create a free account to see these claims.

Citations

Cites

Show all 15 references

Research works printed but not linked: 4

  1. 2024, Regulation (EU) 2024/1689 of the European Parliament and of the Council of 13 June 2024 laying down harmonised rules on artificial intelligence (Artificial Intelligence Act)
  2. 2021, 2020 drug trials snapshots summary report
  3. 2020, Drug Trails Snapshot Summary Report
  4. 2024, Transparency for machine learning-enabled medical devices: Guiding principles

Cited by

See all 22

Metadata

Authors
Alu, F. F., & Oluwadare, S.
Affiliations
University of Cincinnati (US)
Corresponding author
Alu, Fidelis Fidelis
Venue
Frontiers in Artificial Intelligence
Published
2026-02-04
Volume
9
Pages
1737532-1737532
ISSN
2624-8212 (ISSN-L)
OpenAlex
W7127628418
Cited by
22
References
15
Citations by year
2026: 22
FWCI
28.60
Type
journal article
Language
en
Open access
Locations
  • Frontiers in Artificial Intelligence (publishedVersion, cc-by)
  • PubMed (publishedVersion)
  • PubMed Central (publishedVersion, cc-by)
  • DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion, cc-by-sa)
Topics
  • Health Sciences › Medicine › Health Informatics › Artificial Intelligence in Healthcare and Education
  • Physical Sciences › Computer Science › Artificial Intelligence › Explainable Artificial Intelligence (XAI)
  • Health Sciences › Health Professions › Health Information Management › Electronic Health Records Systems
Keywords
clinical decision support, retrieval-augmented generation, data provenance, audit logging, source verification, trustworthy AI, evidence-based medicine, medical knowledge base, clinical guidelines, AI transparency, regulatory compliance, software as a medical device, EU AI Act, clinician trust

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

The claim verdicts this report counts are experimental: their accuracy is not yet measured.

A language model drew the claims from the paper on 2026-10-10, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).

A second model labelled each claim on 2026-10-10 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).

Cited sources are checked with a language model’s help (gpt-6-sol).

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review

How to cite

Cite CoChat for the checks on this page, and the paper for its own findings.

This page
CoChat. (2026). Citation check: An auditable and source-verified framework for clinical AI decision support: integrating retrieval-augmented generation with data provenance. https://app.cochat.ai/papers/10.3389/frai.2026.1737532
The paper
Alu, F. F., & Oluwadare, S. (2026). An auditable and source-verified framework for clinical AI decision support: integrating retrieval-augmented generation with data provenance. Frontiers in Artificial Intelligence. https://doi.org/10.3389/frai.2026.1737532

Data sources

Bibliographic data: Semantic Scholar, OpenAlex, Crossref.

Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).

Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.