ChiCTR2500114698 版本V1.1 版本创建时间2025/12/16 16:44:00 中国临床试验注册中心

审核状态：

Project audit state：

通过审核

Successful

注册号：

Registration number：

 ChiCTR2500114698 

最近更新日期：

Date of Last Refreshed on：

 2025-12-16 16:36:08 

注册时间：

Date of Registration：

 2025-12-16 00:00:00 

注册号状态：

预注册

Registration Status：

Prospective registration

注册题目：

评估大语言模型生成心血管疾病患者护理诊断的准确性和可靠性的横断面研究

Public title：

A Cross-Sectional Study to Evaluate the Accuracy and Reliability of Large Language Models in Generating Nursing Diagnoses for Patients with Cardiovascular Diseases

注册题目简写：

English Acronym：

研究课题的正式科学名称：

评估大语言模型生成心血管疾病患者护理诊断的准确性和可靠性的横断面研究

Scientific title：

A Cross-Sectional Study to Evaluate the Accuracy and Reliability of Large Language Models in Generating Nursing Diagnoses for Patients with Cardiovascular Diseases

研究课题代号(代码)：

Study subject ID：

在二级注册机构或其它机构的注册号：

The registration number of the Partner Registry or other register：

申请注册联系人：

陈媛 

研究负责人：

陈媛 

Applicant：

Yuan Chen 

Study leader：

Chen Yuan 

申请注册联系人电话：

Applicant telephone：

 +86 592 299 3237

研究负责人电话：

Study leader's
telephone：

+86 592 299 3237

申请注册联系人传真 ：

Applicant Fax：

研究负责人传真：

Study leader's fax：

申请注册联系人电子邮件：

Applicant E-mail：

 28837445@qq.com

研究负责人电子邮件：

Study leader's E-mail：

 28837445@qq.com

申请单位网址(自愿提供)：

Applicant website(voluntary supply)：

研究负责人网址(自愿提供)：

Study leader's website(voluntary supply)：

申请注册联系人通讯地址：

福建省厦门市湖里区金山路 2999号

研究负责人通讯地址：

福建省厦门市湖里区金山路 2999号

Applicant address：

2999 Jinshan Road, Huli District, Xiamen City, Fujian Province

Study leader's address：

2999 Jinshan Road, Huli District, Xiamen City, Fujian Province

申请注册联系人邮政编码：

Applicant postcode：

研究负责人邮政编码：

Study leader's postcode：

申请人所在单位：

厦门大学附属心血管病医院

Applicant's institution：

Xiamen Cardiovascular Hospital of Xiamen University

研究负责人所在单位：

厦门大学附属心血管病医院

Affiliation of the Leader：

Xiamen Cardiovascular Hospital of Xiamen University

是否获伦理委员会批准：

是

Approved by ethic committee：

Yes

伦理委员会批件文号：

Approved No. of ethic committee：

 （2025）医伦科第（33）号

伦理委员会批件附件：

Approved file of Ethical Committee：

 查看附件View

批准本研究的伦理委员会名称：

厦门大学附属心血管病医院医学伦理委员会

Name of the ethic committee：

Ethics Committee of Xiamen Cardiovascular Hospital of Xiamen University

伦理委员会批准日期：

Date of approved by ethic committee：

 2025-06-15 00:00:00

伦理委员会联系人：

严妍

Contact Name of the ethic committee：

Yan Yan

伦理委员会联系地址：

福建省厦门市湖里区金山路 2999号

Contact Address of the ethic committee：

2999 Jinshan Road, Huli District, Xiamen City, Fujian Province

伦理委员会联系人电话：

Contact phone of the ethic committee：

 +86 592 2292562

伦理委员会联系人邮箱：

Contact email of the ethic committee：

 1039734296@qq.com

研究实施负责（组长）单位：

厦门大学附属心血管病医院

Primary sponsor：

Xiamen Cardiovascular Hospital of Xiamen University

研究实施负责（组长）单位地址：

福建省厦门市湖里区金山路 2999号

Primary sponsor's address：

2999 Jinshan Road, Huli District, Xiamen City, Fujian Province

试验主办单位(项目批准或申办者)：

Secondary sponsor：

国家：

中国

省(直辖市)：

福建省

市(区县)：

Country：

China

Province：

Fujian

City：

单位(医院)：

厦门大学附属心血管病医院

具体地址：

福建省厦门市湖里区金山路 2999号

Institution
hospital：

Xiamen Cardiovascular Hospital of Xiamen University

Address：

2999 Jinshan Road, Huli District, Xiamen City, Fujian Province

经费或物资来源：

自选课题（自筹）

Source(s) of funding：

Self-raised

研究疾病：

心血管疾病  

Target disease：

Cardiovascular diseases

研究疾病代码：

Target disease code：

研究类型：

观察性研究

Study type：

Observational study

研究所处阶段：

 其它 

Study phase：

N/A

研究设计：

横断面 

Study design：

Cross-sectional 

研究目的：

主要目标：评估大型语言模型生成的护理诊断相对于人工参考标准的准确性，确定其在心血管专科护理诊断中的正确识别率。 次要目标：评估大型语言模型生成护理诊断的可靠性，包括模型对同一病例重复生成诊断的一致性，以及不同评估者对模型输出结果判断的一致性。大型语言模型生成护理诊断的幻觉和可能对患者造成的风险。  

Objectives of Study：

Primary objective: To assess the accuracy of nursing diagnoses generated by large language models compared to human reference standards, and to determine their correct identification rate in cardiovascular specialty nursing diagnoses. Secondary objectives: To evaluate the reliability of nursing diagnoses generated by large language models, including the consistency of the model in generating diagnoses for the same case and the agreement among different evaluators regarding the model's output. To investigate hallucinations generated by large language models in nursing diagnoses and the potential risks they may pose to patients.

药物成份或治疗方案详述：

 

Description for medicine or protocol of treatment in detail：

 

纳入标准：

①为注册护士，研究对象须为注册护士，且在本研究启动前，已在需常规书写护理诊断的心血管疾病病房等临床病区工作满 1 年及以上；②持有有效护士执业证书；③自愿参与本研究，并签署知情同意书。

Inclusion criteria

1. The research subjects must be registered nurses, and have worked for at least one year in clinical wards such as cardiovascular disease wards where nursing diagnoses need to be routinely written before the start of this study; 2. Hold a valid nurse practice certificate; 3. Voluntarily participate in this study and sign the informed consent form.

排除标准：

①进修护士或轮转护士（非正式编制、非长期工作于指定病区者）；②目前处于产假、病假或长期请假状态。

Exclusion criteria：

1. Nurses on further study or rotation (non - formally established and not working in the designated ward for a long - term); 2. Currently on maternity leave, sick leave or long - term leave.

研究实施时间：

Study execute time：

从 From 2025-06-01 00:00:00至 To 2025-11-30 00:00:00  

征募观察对象时间：

Recruiting time：

从 From 2025-12-16 00:00:00 至 To 2025-09-30 00:00:00

干预措施：

Interventions：

组别：

护士

样本量：

 33

Group：

Nurse

Sample size：

干预措施：

无

干预措施代码：

Intervention：

None

Intervention code：

研究实施地点：

Countries of recruitment and research settings：

国家：

 中国

省(直辖市)：

 福建省 

市(区县)：

  

Country：

China

Province：

Fujian

City：

单位(医院)：

 厦门大学附属心血管病医院 

单位级别：

 三级甲等 

Institution
hospital：

Xiamen Cardiovascular Hospital of Xiamen University

Level of the institution：

Tertiary A

测量指标：

Outcomes：

指标中文名：

大型语言模型生成的护理诊断的准确性

指标类型：

主要指标

Outcome：

Accuracy of nursing diagnoses generated by large language models

Type：

Primary indicator

测量时间点：

测量方法：

Measure time point of outcome：

Measure method：

指标中文名：

大型语言模型生成的护理诊断的可靠性

指标类型：

主要指标

Outcome：

Reliability of Nursing Diagnoses Generated by Large Language Models

Type：

Primary indicator

测量时间点：

测量方法：

Measure time point of outcome：

Measure method：

采集人体标本：

Collecting sample(s)
from participants：

标本中文名：

无

组织：

Sample Name：

NA

Tissue：

人体标本去向

 其它  

说明

Fate of sample：

0thers  

Note：

征募研究对象情况：

Recruiting status：

尚未开始

Not yet recruiting

年龄范围：

Participant age：
最小 Min age 岁 years
最大 Max age 岁 years

性别：

男女均可

Gender：

Both

随机方法（请说明由何人用什么方法产生随机序列）：

无

Randomization Procedure (please state who generates the random number sequence and by what method)：

None

是否公开试验完成后的统计结果:

Calculated Results after the Study Completed public access:

不公开/Private

盲法：

无

Blinding：

None

是否共享原始数据：

IPD sharing

否No

共享原始数据的方式（说明：请填入公开原始数据日期和方式，如采用网络平台，需填该网络平台名称和网址）：

无

The way of sharing IPD”(include metadata and protocol, If use web-based public database, please provide the url)：

N/A

数据采集和管理（说明：数据采集和管理由两部分组成，一为病例记录表(Case Record Form, CRF)，二为电子采集和管理系统(Electronic Data Capture, EDC)，如ResMan即为一种基于互联网的EDC：

数据收集 护士书写每份病例的护理诊断：8份病历按顺序提供给护士，护士在问卷网页上使用电脑作答，随机展示病历顺序防止因疲劳作答速度减慢，分别记录8份病历作答的时间及字数。收集护士的基本信息，年龄，性别，学历，职称，工作年限，是否是心血管专科护士、有无接受护理诊断书写的相关培训。 随后，研究人员根据标准化的提问模板，将患者资料输入大型语言模型以生成护理诊断建议。模型采用OpenAI GPT-4o和DeepSeek-R1模型。我们采用了零样本方法作为引导提示，即使用“假如你是一名心血管疾病专科护士，请根据文件中的病历写出该患者的NANDA护理诊断，且按照首优中优次优顺序排序前五个，定义诊断的因素，以PES格式书写，输出为中文。”的提示，不提供任何示例。 对于每个患者的病历，模型独立运行三次（间隔至少1天，重新发起会话）以生成三份护理诊断列表，用于评估模型输出的一致性。模型生成的所有诊断建议将原文保留并标记时间戳。 所有答案都收集在一个 Excel 电子表格中，为了确保评分者无法区分不同的 LLM-Chatbot，我们将所有生成的响应格式化为纯文本。 数据管理 数据采集：经过培训的研究护士负责病例资料的收集与录入，由研究护士从医院电子病历系统调取相关信息。所有资料填写在统一设计的病例报告表（CRF）上，包括患者基本信息表、临床资料汇总表、专家护理诊断表和模型输出记录表等。为了确保数据一致性，研究开始前对参与收集的人员进行统一培训，明确各字段定义和填写规范。 数据核查：数据录入后，将采取双人核对和随机抽查相结合的方法保证准确性。每份病例资料由另一名研究成员独立复核，对照原始病历或CRF检查录入的准确性。如发现遗漏或错误及时更正。 数据存储与安全：所有研究数据将在数据采集后及时汇总进入电子数据库。采用加密的Excel数据库或专业数据管理软件（如EpiData）进行录入和管理，并定期备份至安全服务器。每例患者的数据以唯一识别码编码，不包含姓名、住院号等能识别个人身份的信息，以保护患者隐私。数据访问权限仅限研究主要成员，使用账号和密码控制。纸质原始资料（如打印的病历摘要、专家诊断记录等）统一由各中心保存于加锁文件柜中，研究结束后由项目负责人集中封存。依据国家和医院科研档案管理要求，研究数据和资料将在项目完成后至少保存5年。期间如有需要再次核对，将在伦理审批下进行。项目完成分析发表后，个人敏感信息将销毁，仅保留匿名化的数据用于学术用途。

Data collection and Management (A standard data collection and management system include a CRF and an electronic data capture：

Data Collection Nurses write the nursing diagnoses for each medical record: 8 medical records are provided to the nurses in sequence. The nurses answer on the questionnaire web page using a computer. The order of the medical records is presented randomly to prevent a slowdown in answering speed due to fatigue. Record the answering time and word count for each of the 8 medical records respectively. Collect the basic information of the nurses, including age, gender, education level, professional title, years of work experience, whether they are cardiovascular specialist nurses, and whether they have received relevant training in writing nursing diagnoses. Subsequently, the researchers input the patient data into large language models according to a standardized question template to generate nursing diagnosis suggestions. The models used are OpenAI GPT - 4o and DeepSeek - R1. We adopted a zero - shot method as a guiding prompt, that is, using the prompt "If you are a cardiovascular disease specialist nurse, please write the NANDA nursing diagnoses for the patient according to the medical record in the document, and rank the top five in the order of highest priority, medium priority, and low priority. Define the factors of the diagnosis and write it in PES format, with the output in Chinese." without providing any examples. For each patient's medical record, the model runs independently three times (with an interval of at least 1 day, starting a new session each time) to generate three nursing diagnosis lists, which are used to evaluate the consistency of the model output. All diagnosis suggestions generated by the model will be retained in the original form and marked with a timestamp. All answers are collected in an Excel spreadsheet. To ensure that the raters cannot distinguish different LLM - Chatbots, we format all the generated responses into plain text. Data Management Data Collection: Trained research nurses are responsible for the collection and entry of medical record data. The research nurses retrieve relevant information from the hospital's electronic medical record system. All data is filled in a uniformly designed Case Report Form (CRF), including the patient's basic information form, clinical data summary form, expert nursing diagnosis form, and model output record form, etc. To ensure data consistency, the personnel involved in the collection are uniformly trained before the start of the research, and the definitions and filling specifications of each field are clarified. Data Verification: After data entry, a combination of double - checking and random sampling will be adopted to ensure accuracy. Each piece of medical record data is independently reviewed by another research member, who checks the accuracy of the entry against the original medical record or CRF. Any omissions or errors are corrected in a timely manner. Data Storage and Security: All research data will be promptly summarized and entered into an electronic database after data collection. An encrypted Excel database or professional data management software (such as EpiData) is used for entry and management, and regular backups are made to a secure server. The data of each patient is encoded with a unique identification code and does not contain information that can identify personal identity, such as name and hospital number, to protect patient privacy. Data access rights are limited to the main research members, controlled by accounts and passwords. Paper - based original materials (such as printed medical record summaries, expert diagnosis records, etc.) are uniformly stored in locked file cabinets by each center. After the end of the research, they are sealed centrally by the project leader. In accordance with national and hospital scientific research archive management requirements, research data and materials will be retained for at least 5 years after the completion of the project. If re - checking is required during this period, it will be carried out under ethical approval. After the completion of the project analysis and publication, personal sensitive information will be destroyed, and only anonymized data will be retained for academic purposes.

数据与安全监察委员会：

Data and Safety Monitoring Committee：

有/Yes

注册人：

Name of Registration：

  2025-12-16 16:35:52