Construction and Validation of a Coronary Artery Disease-Specific Large Language Model and Its Evaluation Framework (CorAI) (CorAI)
A Prospective, Multicenter, Observational Cohort Study for the Construction of a Multimodal Coronary Artery Disease Dataset and the Development, Evaluation, and External Validation of a Disease-Specific Large Language Model (CorAI)
Coronary artery disease (CAD) affects an estimated 11.39 million people in China, and mortality has continued to rise since 2012. Quality-of-care data indicate persistent gaps in diagnostic and therapeutic guideline adherence, incomplete implementation of secondary prevention, and suboptimal risk-factor control. Large language models (LLMs) can rapidly retrieve and integrate literature, medical records, and multimodal clinical data, and may support clinical decision-making. However, most existing medical LLMs are general-purpose, are trained predominantly on published literature and web text rather than authentic clinical records, and have not been validated with real patient data. No disease-specific LLM for CAD currently exists, and no evaluation framework tailored to CAD has been established; existing benchmarks are general-purpose, lack specialty-specific risk grading, are not rigorously anchored to current guidelines, and emphasize literal accuracy over clinically critical reasoning steps and prescription concordance.
This prospective, multicenter, observational study will enroll approximately 640 patients with confirmed or suspected CAD at Fuwai Hospital and collaborating centers. No study-specific intervention, examination, biospecimen collection, or additional follow-up visit is performed; all clinical decisions are made independently by the treating physician according to routine standards of care. For consenting participants, routinely generated clinical documentation from the index outpatient visit or hospitalization - including medical history, laboratory results, imaging and coronary angiography/coronary CT angiography reports, physician notes, online consultation records, and follow-up records - will be de-identified and used to construct a high-quality multimodal CAD dataset. This dataset will support the development of a CAD-specific large language model (CorAI), the construction of an evaluation framework spanning eight predefined clinical scenarios and multiple assessment dimensions, and external validation of the model at participating centers.
The primary objectives are to characterize the completeness and quality of the constructed prospective cohort dataset, and to quantify the guideline concordance of CorAI-generated clinical recommendations as adjudicated by a blinded panel of cardiologists.
調査の概要
研究の種類
入学 (推定)
連絡先と場所
研究連絡先
- 名前：Prof. Gao
- 電話番号：+861088322413
- メール：sophie_gao@sina.com
研究場所
-
-
Anhui
-
Hefei、Anhui、中国、230001
- The First Affiliated Hospital of USTC
-
コンタクト：
- Hongwu Chen, MD
- 電話番号：+86-551-62283114
- メール：1025707914@qq.com
-
-
Beijing Municipality
-
Beijing、Beijing Municipality、中国、100037
- Fuwai Hospital, Chinese Academy of Medical Sciences/National Center for Cardiovascular Diseases
-
コンタクト：
- Prof. Gao, MD, PhD
- 電話番号：+861088322413
- メール：sophie_gao@sina.com
-
コンタクト：
- Prof. Gao, MD, PhD
- メール：sophie_gao@sina.com
-
-
Guangdong
-
Shenzhen、Guangdong、中国、518057
- Fuwai Hospital Chinese Academy of Medical Sciences, Shenzhen
-
コンタクト：
- Fenghuan Hu, MD
- 電話番号：0755-25509566
- メール：hufenghuanvip@163.com
-
-
参加基準
適格基準
就学可能な年齢
- 大人
- 高齢者
健康ボランティアの受け入れ
サンプリング方法
調査対象母集団
説明
Inclusion Criteria:
- Age ≥ 18 years;
- Patients with confirmed or suspected coronary heart disease;
- Complete case information (including medical history, test results, imaging, follow-up records, etc.);
- Consent for the use of case data for model evaluation and research analysis.
Exclusion Criteria:
- Cases with severe data gaps that prevent the formation of a complete clinical scenario;
- Cases where the primary diagnosis is not coronary heart disease;
- Cases with serious data errors or where data cannot be anonymized.
研究計画
研究はどのように設計されていますか？
デザインの詳細
コホートと介入
グループ/コホート |
介入・治療 |
|---|---|
|
CorAI development and evaluation cohort
Adults aged 18 years or older with confirmed or suspected coronary artery disease presenting for routine outpatient or inpatient care at a participating center.
Routinely generated clinical documentation is de-identified and used for construction of a multimodal CAD dataset and for development and evaluation of a disease-specific large language model.
No study-specific intervention, testing, or follow-up is performed.
|
介入なし: 観察コホート
|
この研究は何を測定していますか？
主要な結果の測定
結果測定 |
メジャーの説明 |
時間枠 |
|---|---|---|
|
Completeness of the constructed multimodal CAD dataset
時間枠：From enrollment of the first participant through completion of dataset construction, up to 12 months
|
Proportion of enrolled participants whose de-identified record satisfies all pre-specified data-completeness criteria, defined as the presence of medical history, comorbidity and medication documentation, laboratory results, electrocardiography, coronary CT angiography and/or invasive coronary angiography findings, and at least one structured follow-up or discharge record sufficient to reconstruct a complete clinical scenario.
Reported as a percentage with 95% confidence interval.
|
From enrollment of the first participant through completion of dataset construction, up to 12 months
|
二次結果の測定
結果測定 |
メジャーの説明 |
時間枠 |
|---|---|---|
|
Guideline concordance of CorAI-generated clinical recommendations across eight predefined coronary artery disease scenarios
時間枠：At completion of model development and evaluation, up to 18 months
|
Proportion of CorAI responses judged fully concordant with the coronary artery disease clinical practice guidelines in force at the time of adjudication, assessed on the held-out evaluation set across eight predefined clinical scenarios: (1) initial evaluation of chest pain; (2) coronary artery assessment; (3) antithrombotic therapy; (4) revascularization decision-making; (5) lipid-lowering therapy; (6) post-procedural management; (7) post-procedural complications; and (8) patient education and physician-patient communication.
Each item is scored as concordant, partially concordant, or discordant by cardiologists blinded to response source using a pre-specified rubric, with disagreements resolved by an independent third adjudicator.
The primary metric is the pooled percentage of items scored fully concordant across all eight scenarios, reported with a 95% confidence interval.
|
At completion of model development and evaluation, up to 18 months
|
協力者と研究者
研究記録日
主要日程の研究
研究開始 (推定)
一次修了 (推定)
研究の完了 (推定)
試験登録日
最初に提出
QC基準を満たした最初の提出物
最初の投稿 (実際)
学習記録の更新
投稿された最後の更新 (実際)
QC基準を満たした最後の更新が送信されました
最終確認日
詳しくは
本研究に関する用語
キーワード
その他の研究ID番号
- 2025-GSP-GG-18-3
- 2025-GSP-GG-18 (その他の助成金/資金番号：the National High Level Hospital Clinical Research Funding (China))
個々の参加者データ (IPD) の計画
個々の参加者データ (IPD) を共有する予定はありますか?
IPD プランの説明
医薬品およびデバイス情報、研究文書
米国FDA規制医薬品の研究
米国FDA規制機器製品の研究
この情報は、Web サイト clinicaltrials.gov から変更なしで直接取得したものです。研究の詳細を変更、削除、または更新するリクエストがある場合は、register@clinicaltrials.gov。 までご連絡ください。 clinicaltrials.gov に変更が加えられるとすぐに、ウェブサイトでも自動的に更新されます。