基于预测模型的大型水面舰艇航海疲劳风险筛查评分量表的构建与验证  

Construction and Verification of Navigation Fatigue Risk Screening Scale for Large Surface Warship Based on Predictive Model

作　　者：李栋 周涛 李珍 董东方 赵琦 董文珠 LI Dong;ZHOU Tao;LI Zhen;DONG Dongfang;ZHAO Qi;DONG Wenzhu(Department of Gastroenterology,NO.971 Hospital of the Navy,Qingdao Shandong 266071,China)

机构地区：[1]海军971医院消化内科,山东青岛266071 [2]海军971医院驻91197部队医务中心

出　　处：《联勤军事医学》2024年第5期412-418,432,共8页Military Medicine of Joint Logistics

基　　金：海军后勤部自主项目(CHJ23J026)。

摘　　要：目的 本研究拟构建基于预测模型的航海疲劳风险筛查评分量表,为航海疲劳人群早期识别及提前干预提供依据。方法 前瞻性选取参加航海任务的某部官兵进行问卷调查,通过单因素分析、多因素分析筛选出航海疲劳独立危险因素,并将其代入Logistic回归模型、神经网络模型、分类树模型3种模型中,对模型的受试者工作特征(receiver operating characteristic, ROC)曲线及曲线下面积(area under the curve, AUC)进行比较以选出最优模型,在此基础上形成航海疲劳风险筛查评分量表。结果 舰员在航行第30天时疲劳程度高于航行前1天(P<0.001)。单因素分析显示,舰艇官兵运动频次、靠港期间吸烟情况、碳水化合物摄取情况、发酵饮食摄入情况、压力情况、每天工作总时长、环境不利影响因素数量、每天最大连续工作时长以及匹茨堡睡眠质量指数量表(Pittsburgh sleep quality index, PSQI)得分9个因素与疲劳评定量表14(fatigue scale 14,FS 14)评分相关(P均<0.05)。Logistic回归模型、神经网络模型、分类树模型3种模型预测舰艇官兵航海疲劳的AUC分别为0.953、0.915、0.811,均具有较好的诊断能力,最终选定Logistic回归模型为最佳预测模型,其在训练集中的灵敏度、特异度分别为78.95%、84.62%。本研究将Logistic回归模型以列线固的形式可视化,最终形成航海疲劳风险筛查评分量表。结论 本研究依据Logistic回归模型构建的航海疲劳风险筛查评分量表,可客观有效预测航海疲劳,并为航海疲劳早期风险筛查和针对性干预提供依据。Objective To construct rating scale for navigation fatigue risk screening based on predictive model,so as to provide a basis for early identification and intervention of navigation fatigue population.Methods The officers and soldiers of a certain department participating in navigation tasks were prospectively selected for questionnaire survey,in-dependent risk factors of navigation fatigue were screened out through univariate analysis and multivariate analysis,which then substituted into 3 models,Logistic regression model,neural network model and classification tree model,the receiv-er operating characteristic(ROC)curve and area under the curve(AUC)of the models were compared to select the opti-mal model,and on this basis,the navigation fatigue risk screening scale was formed.Results The degree of fatigue of the crew on the 30th day of the voyage was higher than that on the 1st day before the voyage(P<0.001).Univariate analysis showed that 9 factors including exercise frequency,smoking during port stay,carbohydrate intake,fermented di-et intake,stress,total daily working hours,number of adverse environmental factors,maximum continuous daily work-ing hours and Pittsburgh sleep quality index(PSQI)score of officers and soldiers on the naval vessel were related to fa-tigue scale 14(FS-14)score(all P<0.05).The AUC of Logistic regression model,neural network model and classifica-tion tree model were 0.953,0.915 and 0.811 respectively,showing good diagnostic ability.The Logistic regression mod-el was selected as the best predictive model,and the sensitivity and specificity were 78.95%and 84.62%in the training set.In this study,the Logistic regression model was visualized in the form of a nomogram,and finally navigation fatigue risk screening score scale was formed.Conclusion In this study,the risk screening scale for navigation fatigue constructed based on Logistic regression model can predict navigation fatigue objectively and effectively,which pro-vide a basis for early risk screening and targeted interven-tion for na

关 键 词：官兵 航海疲劳 危险因素 预测模型 列线图 

分 类 号：R82[医药卫生—临床医学]