The Marginal Structure Model Based On Martingale Residual Method Was Used To Evaluate The Effect Of Lifestyle On Rehospitalization Of Patients With Chronic Heart Failure

Objective:The purpose of this study was to evaluate the effects of tea drinking,hypnotics,and physical exercise on readmission in patients with chronic heart failure.The effect of the marginal structure model based on martingale residual imputation method in the evaluation of the effectiveness of the treatment factors in the verification center is constructed and verified.Methods:All the data were used to calculate the martingale residuals,and the martingale residuals were incorporated into the validation data set to build a model about the unmeasured confounding factors,and the unmeasured factors were filled in the main database.The superiority of the martingale residual-based imputation method was verified by simulation marginal structural models were used to assess the potential risk of rehospitalization in patients with chronic heart failure by using martingale residual-based imputation,some of which were measured repeatedly in only a small sub-cohort.Logistic probability estimation was carried out in the probability estimation stage of the edge structure model.In the inverse probability weighting stage,stable weighting and truncated weighting were used to calculate the weight,and the virtual population was constructed.In the utility estimation stage,the marginal structure model is compared with the traditional statistical analysis method to verify whether it can effectively deal with time-dependent confounding.Results:1.The simulation results show that the martingale residual-based method can systematically reduce the confounding bias compared with the traditional method,the propensity scoring calibration method and the log(t)imputed method,and the overall accuracy is the highest.The deviation of estimation by the calibration method of propensity score depends on whether the hypothesis is violated and the degree of the violation.When the hypothesis is violated,the estimation deviation will be serious.In contrast,the martingale residual-based imputation method is unbiased even if the assumption is seriously violated.In all cases,the SD of the martingale residual-based method is comparable to that of the log(t)imputed SD;In most cases,the SD of the martingale residual-based method is comparable to that of the prodigality scoring calibration method.However,when the validation sample is small,the validation sample is external and random deletion,the PSC has a larger standard deviation than the martingale residual-based imputed method.Compared with the traditional Cox model,the martingale residual-based method has a good advantage in both bias and SD.2.Using traditional Cox regression analysis,only excluding the influence of baseline confounding factors,physical exercise [HR=0.749 95%CI(0.560,1.003)]and tea drinking [HR=1.149 95%CI(0.855,1.546)] had no significant effect on rehospitalization of patients with chronic heart failure,while hypnotics increased the risk of rehospitalization [HR=1.652 95%CI(1.195,2.284)].Based on the marginal structure model after the martingale residual-based imputation,after excluding the influence of baseline confounding factors and time-dependent confounding factors,the results show that drinking tea can increase the risk of rehospitalization in patients with heart failure[HR=1.257 95%CI(1.036,1.525)].Physical exercise can reduce the risk of rehospitalization in patients with heart failure [HR=0.784 95%CI(0.625,0.985)],while hypnotics have no significant effect on patients’ rehospitalization [HR=1.142 95%CI(0.922,1.413)].Conclusion:1.In most cases,the martingale residual-filling method performed well in the verification set and could effectively reduce the bias of model analysis after the filling.Marginal structure model can effectively solve the time dependent confounding factors and the bias in the evaluation of the effectiveness of treatment factors.2.Physical exercise can significantly reduce the rehospitalization of patients with chronic heart failure,and drinking tea will increase the risk of rehospitalization of patients,and provide a new direction for guiding the daily life style of patients with chronic heart failure after discharge.No hypnotic drugs have been found to improve patients with chronic heart failure after they are discharged from the hospital.