基于随机森林算法的肿瘤科护士工作倦怠影响因素分析:一项多中心研究

目的 采用随机森林算法探讨肿瘤科护士工作倦怠的影响因素。 方法 于2018年3—8月对广西全区25所三级综合医院和专科医院1253名肿瘤科护士采用一般资料问卷、工作倦怠量表、心理资本量表进行横断面调查，分析其工作倦怠现状及其影响因素，使用随机森林分析对可能影响肿瘤科护士工作倦怠的自变量进行重要性排序，将袋外估算误差率最小的变量集纳入logistic回归模型，分析自变量的作用方向和相对危险度。 结果 逐步随机森林分析显示，变量数为8时的袋外估算误差率最低，重要性排名前8的变量依次为收入状况满意度、护患关系满意度、从事肿瘤科护理工作年限、心理资本、年龄、参加院级以上培训次数、群众认可与尊重程度、夜班频次；Logistic回归分析显示，收入状况满意度为一般（OR=1.228，95%CI:1.019~1.497）和不满意或很不满意（OR=1.276，95%CI:1.005~1.621）、护患关系满意度为一般（OR=1.158，95%CI:1.036~1.295）和不满意或很不满意组（OR=1.196，95%CI:1.020~1.402）、从事肿瘤科护理工作5~10年（OR=1.365，95%CI:1.001~1.860）和>10年（OR=1.702，95%CI:1.222~2.371）、心理资本（OR=0.904，95%CI:0.824~0.991）、年龄25~30岁（OR=1.347， 95%CI:1.010~1.797）和>30岁（OR=1.397，95%CI:1.007~1.937）、参加院级以上培训次数1~2次/年（OR=1.107，95%CI:1.008~1.217）、群众认可与尊重程度为一般（OR=1.105，95%CI:1.008~1.212）和不尊重（OR=1.260，95%CI:1.010~1.572）、夜班频次<1次/周（OR=1.115，95%CI:1.025~1.213）和≥1次/周（OR=1.397，95%CI:1.146~1.702）是肿瘤科护士工作倦怠的独立危险因素（P<0.05）。 结论 影响肿瘤科护士工作倦怠的前8位因素依次为收入状况满意度、护患关系满意度、从事肿瘤科护理工作年限、心理资本、年龄、参加院级以上培训次数、群众认可与尊重程度、夜班频次，建议医院和护理管理者应制定相应策略，减少肿瘤科护士工作倦怠。

Influencing factors of job burnout of oncology nurses based on Random Forest Algorithm: a multicenter study

To explore the influencing factors of job burnout of nurses in oncology department by using random forest algorithm. Methods From March 2018 to August, 1 253 oncology nurses in 25 level 3 general hospitals and specialized subject hospitals in Guangxi were introduced to this cross-sectional study, and were surveyed with general information questionnaire, job burnout scale, and psychological capital scale. And the present situation of job burnout and its influencing factors were analyzed by using the random forest that was likely to affect the oncology nurse’s job burnout, which was the independent variable importance. The variable set with the smallest error rate was incorporated into the logistic regression model, and the direction of action and relative risk of independent variables were analyzed. Results Stepwise random forest analysis showed, the lowest out-of-bag estimation error rate was found when the number was 8, and the top eight variables in order of importance were income satisfaction, satisfaction with nurse-patient relationship, number of years in oncology nursing, psychological capital, age, number of training sessions at the faculty level or above, degree of recognition and respect of people, night shift frequency. Logistic regression analysis showed that the degree of satisfaction with income status was average(OR=1.228, 95%CI: 1.019-1.497), dissatisfied or very dissatisfied(OR=1.276, 95%CI: 1.005-1.621), and the degree of satisfaction with nurse-patient relationship was average(OR=1.158, 95%CI: 1.036-1.295), dissatisfied or very dissatisfied(OR=1.196, 95%CI: 1.020-1.402), and engaged in oncology nursing work for 5-10 years(OR=1.365, 95%CI: 1.001-1.860) and >10 years(OR=1.702, 95%CI: 1.222-2.371), psychological capital(OR=0.904, 95%CI: 0.824-0.991), age 25-30 years old(OR=1.347, 95%CI: 1.010-1.797) and >30 years old(OR=1.397, 95%CI: 1.007-1.937), participated in college level or above training for 1-2 times/years(OR=1.107, 95%CI: 1.008-1.217), general recognition and respect from the public(OR=1.105, 95%CI: 1.008-1.212) and disrespect(OR=1.260, 95%CI: 1.010-1.572), night shift frequency <1 time/week(OR=1.115, 95%CI: 1.025-1.213) and ≥1 time/week(OR=1.397, 95%CI: 1.146-1.702) were independent risk factors for job burnout of oncology nurses. Conclusion The top 8 factors affecting oncology nurses’ job burnout are income satisfaction, nurse-patient relationship satisfaction in turn, years of nursing work in oncology department, psychological capital, age, times of attending training above hospital level, degree of public recognition and respect, frequency of night shift. It is suggested that hospitals and nursing managers should formulate corresponding strategies to reduce the job burnout of nurses in oncology department.