随机生存森林模型在肺癌患者预后分析中的应用

Application of Random Survival Forests Model in Prognosis Analysis of Lung Cancer Patients

目的应用随机生存森林模型探讨肺癌患者预后影响因素的重要性并对预测结果进行评价。方法对山西省某三甲医院342例确诊的肺癌患者进行随访研究,建立随机生存森林模型,并与传统的Cox回归模型进行比较。结果342例肺癌患者中226例患者发生死亡,中位生存时间为28.23月。治疗方式、肿瘤大小、临床分期等变量是影响肺癌患者预后的重要因素,淋巴结转移、分化程度、病理分型、年龄是中度预测因素,并分析了变量之间的交互作用。二者的模型比较结果显示随机生存森林模型预测错误率以及预测误差均低于Cox回归模型。结论随机生存森林模型拟合效果好,可用于右删失生存数据的分析,不但能发现重要的影响因素,还能发现变量之间的交互作用,为肺癌患者预后状况的改善,提升生命质量提供科学依据。

Objective To explore the importance of prognostic factors of lung cancer patients using random survival forests model and predict the prognosis.Methods A follow-up study was conducted on 342 confirmed lung cancer patients in a top three hospital in Shanxi Province.A random survival forests model was established and compared with the traditional Cox model.Results Of the 342 lung cancer patients,226 died,with a median survival time of 28.23 months.Treatment methods,tumor size,clinical stage are important factors affecting the prognosis of patients with lung cancer.Lymph node metastasis,degree of differentiation,pathological classification,and age are moderate predictive factors.The interaction between variables was analyzed.The comparison results show that the consistency error rate and prediction error of the RSF model are lower than the Cox regression model.Conclusion The random survival forests model can be used for the analysis of right censored survival data.It can not only find important influencing factors,but also discover the interaction between variables,and provide a scientific basis for improving the prognosis of lung cancer patients and improving the quality of life.