关键词: 非小细胞肺癌, 倾向性评分, 逆概率加权, 双重稳健估计, 治疗质量

Abstract: Objective The aim of this study was to estimate the causal effects of non-small cell lung cancer(NSCLC)treatment process on in-hospital mortality based on the double robust estimation(DR)method,and provide a reference basis for reducing in-hospital mortality of NSCLC. Methods According to the quality evaluation system of NSCLC treatment,the utilization rate of treatment process indicators was calculated,and patients were divided into the high-quality or low-quality groups based on the average score of treatment process quality.In-hospital mortality was used as the outcome indicator,Kaplan-Meier method and Cox regression adjusted for propensity score inverse probability of treatment weighting(IPTW)correction were used to analyze the impact of treatment process quality on in-hospital mortality in NSCLC.DR was combined to estimate the causal effects of the treatment process on in-hospital mortality. Results The median utilization rate of treatment process indicators was 66.88%,and the mean and standard deviation of patients′ treatment process quality scores were 0.270±0.124,including 0.358±0.069 in the high-quality group,and 0.158±0.081 in the low-quality group.After the IPTW weighting,the standardized mean difference(SMD)of patients′baseline characteristics decreased;The difference in survival curves between the two groups of patients before and after ITPW was statistically significant(P＜0.05),and the prognosis of patients in the high-quality group was better than that of patients in the low-quality group(pre-IPTW:HR=0.367,95% CI:0.275-0.491;post-IPTW:HR=0.228,95% CI:0.167-0.312).Compared with the low-quality group,the average causal effect of treatment process on in-hospital mortality was -0.026 in the high-quality group. Conclusion DR can compensate for the shortcomings of logistic or IPTW,avoid the risk of model error,and obtain for the causal effect of treatment process on in-hospital mortality.In medical practice,the utilization rate of treatment process indicators should be increased to improve patient prognosis;The study of causal effects suggests that besides the treatment process,other factors that affect in-hospital mortality cannot be ignored.

Key words: Non-small cell lung cancer, Propensity score, Inverse probability of treatment weighting, Doubly robust estimation, Treatment quality