基于患者画像的知识库问答智能体在急性胰腺炎患者中的应用与30天自我管理相关因素分析

Application of Knowledge Base Question-Answering Agent Based on Patient Portrait in Patients with Acute Pancreatitis and Analysis of Related Factors of 30-Day Self-Management

摘要: 目的：探讨基于患者画像的知识库问答智能体在急性胰腺炎患者中的应用情况，并分析与患者出院后30天自我管理水平相关的因素。方法：采用单组前瞻性观察性研究设计，纳入218例急性胰腺炎患者。患者出院前收集结构化信息以构建患者画像；出院后30天内，患者通过智能体获取个体化问答服务。收集患者一般资料、电子健康素养、智能体使用行为及自我管理评分等数据。采用配对t检验比较基线与30天自我管理评分差异，采用多元线性回归分析与30天自我管理评分相关的因素。结果：患者30天自我管理评分为(74.37 ± 9.08)分，高于基线的(65.16 ± 6.68)分，差异有统计学意义(t = 33.62, P < 0.001)。多元线性回归分析显示，年龄(B = −0.066, P = 0.003)、30天活跃天数(B = 0.158, P = 0.005)、回答理解度评分(B = 0.798, P < 0.001)及基线自我管理评分(B = 0.991, P < 0.001)与30天自我管理评分独立相关。结论：急性胰腺炎患者使用基于患者画像的知识库问答智能体后，30天自我管理评分较基线升高。年龄、活跃使用天数、回答理解度及基线自我管理评分与30天自我管理评分相关。临床应用中应关注高龄患者的使用适配性，并持续优化智能体回答的可理解性。

Abstract: Objective: To explore the application of knowledge base question-answering agent based on patient portrait in patients with acute pancreatitis, and analyze the factors related to the self-management level of patients 30 days after discharge. Methods: 218 patients with acute pancreatitis were enrolled in a single-group prospective observational study design. Collect structured information before the patient leaves the hospital to construct a portrait of the patient; within 30 days after discharge, patients get personalized question and answer service through agents. Data such as patients’ general information, electronic health literacy, agent use behavior and self-management score were collected. Paired t-test was used to compare the difference between baseline and 30-day self-management score, and multiple linear regression was used to analyze the factors related to 30-day self-management score. Results: The 30-day self-management score of patients was (74.37± 9.08), which was higher than the baseline score of (65.16 ± 6.68), and the difference was statistically significant (t = 33.62, P < 0.001). Multiple linear regression analysis showed that age (B = −0.066, P = 0.003), 30-day active days (B = 0.158, P = 0.005), answer comprehension score (B = 0.798, P < 0.001) and baseline self-management score (B = 0.991, P < 0.001) Conclusion: The 30-day self-management score of patients with acute pancreatitis is higher than the baseline after using the knowledge base question-answering agent based on patient portrait. Age, days of active use, understanding of answers and baseline self-management score are related to 30-day self-management score. In clinical application, we should pay attention to the adaptability of elderly patients and continuously optimize the intelligibility of agent answers.