Abstract:

Objective To develop a knowledge graph-based voice-interactive intelligent alarm management system for blood purification and evaluate its application in a pediatric intensive care unit（PICU），with the aim of enhancing the efficiency of healthcare professionals in recognizing and managing blood purification alarms and ensuring patient safety. Methods By integrating multi-source information and establishing a knowledge framework through structured expert meetings，a knowledge ontology was constructed. Knowledge extraction and fusion were completed via manual review，forming the entity framework of the blood purification alarm management knowledge graph. Based on this knowledge graph，a voice-interactive intelligent alarm management system for blood purification was developed，encompassing 3 functional modules：prevention，response，and optimization. Using convenience sampling，pediatric patients undergoing blood purification therapy in the PICU of a tertiary A hospital in Jilin Province from May 2024 to June 2025 were enrolled as study subjects. Patients admitted after system implementation（January-June 2025） constituted an experimental group and received system-assisted management；those admitted before system implementation（from May to November 2024） constituted a control group and received routine blood purification nursing protocols. Outcomes compared between groups included the incidence of unplanned circuit downtime，filter lifespan，alarm frequency，and alarm response time. System-level metrics included response accuracy and system usability. Additionally，the degree of clinical alarm fatigue among nurses was assessed before and after system implementation. Results Compared with the control group，the experimental group demonstrated significantly lower incidence of unplanned circuit downtime，alarm frequency，and alarm response time，along with significantly prolonged filter lifespan（all P<0.05）. The system achieved a response accuracy rate of 90.15% and a System Usability Scale score of（80.30 ± 4.32）points. Nurses’ Clinical Alarm Fatigue Scale scores decreased significantly from 20.00（14.00，25.00） to 13.00（10.00，19.00） points following system implementation（P<0.001）. Conclusion The knowledge graph-based voice-interactive intelligent alarm management system provides effective clinical decision support for blood purification alarm management in the PICU，facilitating rapid alarm response and standardized clinical practice. This translates to a reduced alarm burden，shortened alarm response time，and overall improvement in the quality of blood purification alarm management.

Key words: Pediatric Intensive Care Unit, Continuous Blood Purification, Alarm Management, Voice Interaction System, Nursing Care

摘要：

目的 基于知识图谱，构建语音交互式血液净化智能警报护理管理系统，并在儿童重症监护病房进行应用，旨在提升医护人员对血液净化警报的识别与处理效率，保障患儿安全。方法 通过整合多源信息，经专家会议确定知识框架，进而构建知识本体，采用人工审核完成知识抽取与融合，形成血液净化警报管理知识图谱的实体框架；最终构建了涵盖预防、应对与优化3个模块的语音交互式智能警报护理管理系统。采用便利抽样法，选取2024年5月—2025年6月在吉林省某三级甲等医院儿童重症监护病房接受血液净化的患儿作为研究对象。将2025年1—6月系统应用后的患儿作为试验组，接受语音交互智能警报护理管理系统辅助血液净化护理流程；2024年5—11月系统应用前的患儿作为对照组，接受常规血液净化护理流程。干预后比较两组非计划性下机发生率、滤过器使用时间、报警次数及报警处理时间、系统应答正确率、系统可用性，以及系统应用前后护士的临床警报疲劳程度。结果 试验组非计划性下机发生率、报警次数及报警处理时间均低于对照组，滤过器使用时间长于对照组，差异具有统计学意义（均P<0.05）；系统应答正确率为90.15%、系统可用性评价量表得分为（80.30±4.32）分；与系统应用前相比，护士临床警报疲劳量表得分由20.00（14.00，25.00）分降至13.00（10.00，19.00）分，差异具有统计学意义（P<0.001）。结论 应用该系统为儿童重症监护病房血液净化警报管理提供了有效的决策支持，有助于实现报警的快速响应与操作规范化，从而减少报警次数，缩短处理时间，提升血液净化警报管理的质量。

关键词: 儿童重症监护病房, 连续性血液净化, 警报管理, 语音交互系统, 护理