Abstract:

Objective To construct a knowledge graph based intelligent Q&A system for ventilator alarm management in intensive care units，to improve healthcare professionals’ ability to recognize and handle ventilator alarms，and to enhance patient safety and treatment outcomes. Methods We integrated knowledge from various sources such as thesis，guidelines，standards，clinical practice cases，device user manuals，etc.，and developed the knowledge framework through expert interviews and expert meetings. We used a combination of automation and manual review to construct entities，extract and integrate knowledge，and input data into the Neo4j graph database as “entity-relationship-entity” ternary data for storage. And for the semantic understanding and parsing of natural language questions，the knowledge graph Cypher query statement was created，so that the knowledge graph has intelligent question and answer functions. The system usability scale was used to test the usability of the Q&A system. Results The constructed knowledge graph covered six fields，including ventilation mode，parameter settings，alarm types，alarm causes，treatment measures，and equipment maintenance，and contains 206 knowledge entities，15 kinds of inter-entity relationships，and 256 linkages of knowledge points，and a web-side intelligent Q&A system has been developed based on the knowledge graph. The mean score of the System Usability Scale was （76.44 ± 13.17）. Conclusion The knowledge graph based intelligent Q&A system for ventilator alarm management constructed in this study has certain scientific validity and usability. Through the visualization of knowledge system functions，intelligent question and answer，and personalized recommendation，it provides an auxiliary decision support tool for healthcare professionals to effectively identify and accurately deactivate the alarms，and improves the safety of clinical alarms.

Key words: Intensive Care Units, Ventilators, Mechanical, Alarm Management, Knowledge Graph, Intelligent Question Answering System