Abstract:

Objective To construct a risk prediction model for post-traumatic stress disorder（PTSD） in ICU transferred patients and to conduct internal and external validation. Methods Convenient sampling method was used to select 514 ICU patients from 2 tertiary hospitals in Jiangxi Province from October 2022 to May 2023 as the survey subjects. Univariate factor and multivariate logistic regression analysis were employed to determine risk factors and construct predictive models. The discrimination and calibration of the prediction model were evaluated through the area under the receiver operating characteristic（ROC） curve and the Hosmer-Leme-show test，and the clinical usefulness of the prediction model were evaluated through the Decision Curve Analysis. Results The predictive factors for PTSD in ICU transferred patients were gender，history of mental illness，sedatives，mechanical ventilation，ICU hospitalization time，and delusional memory. The constructed model was as follows：Logit P=Gender × 1.366+History of Psychological Disorders × 1.221+sedatives × 1.018+mechanical ventilation × 1.378+ICU hospitalization time × 0.546+delusional memory × 0.929-6.793. The area under the ROC curve was 0.898（95% CI 0.865~0.932），with sensitivity of 82.9%，specificity of 83.2%，maximum Youden index of 0.661，critical value of 0.393，and critical score of 60 points. The Hosmer-Leme-show test results display χ² value=11.312，P=0.185. The internal and external validation showed that the areas under the ROC curve were 0.855 and 0.867，respectively. The calibration curve showed that the observed curve fitted well with the predicted curve，and the model calibration was good. The DCA shows that the model has good clinical effectiveness. Conclusion The prediction model constructed in this study has good predictive performance and can provide theoretical basis for clinical medical staff to identify high-risk populations of PTSD in the early stage.

Key words: Critical Patients, Stress Disorder,Post-Traumatic, Prediction Model, Nursing Care

摘要：

目的 构建ICU转出患者应激障碍（post-traumatic stress disorder，PTSD）风险预测模型，并进行内外部验证。 方法 采用便利抽样法，于2022年10月至2023年5月选取江西省2所三级甲等医院514例ICU患者为调查对象。通过单因素和Logistic回归分析确定危险因素并构建预测模型。通过受试者工作特征（receiver operating characteristic，ROC）曲线下面积和Hosmer-Leme-show检验评价预测模型的区分度和校准度，临床决策曲线评价预测模型的临床有效性。 结果 ICU转出患者PTSD的影响因素为性别、心理疾病史、镇静剂、机械通气史、ICU住院时间、妄想记忆。预测模型为Logit P=性别 × 1.366+心理疾病史 × 1.221+镇静剂 × 1.018+机械通气史 × 1.378+ICU住院时间 × 0.546+妄想记忆 × 0.929-6.793。ROC曲线下面积为0.898（95%CI 0.865~0.932），灵敏度为82.9%，特异度为83.2%，最大约登指数为0.661，临界值为0.393，临界分数为60分。Hosmer-Leme-show检验显示，χ²=11.312，P=0.185。内外部验证显示，ROC曲线下面积分别为0.855、0.867。校准曲线显示观察曲线与预测曲线拟合良好，模型校准度较好。临床决策曲线显示该模型有良好的临床有效性。 结论 该研究构建的预测模型具有较好的预测效能，能够为临床医护人员早期识别ICU转出患者发生PTSD的高危人群和进行及时干预提供理论依据。

关键词: 危重症患者, 应激障碍,创伤后, 预测模型, 护理