Analyzing the effect of household income, parent death and spouse death on survival of retired people: a left truncated marginal cox model with SIMEX correction for measurement error

We aimed to investigate the associations and to estimate the effects of the parents death, spouses death, and household income on survival time of retired people. We used left truncated right-censored marginal Cox regression model on HRS dataset in our study. We also implemented the simulation and extrapolation (SIMEX) method in SAS program, and used this method to correct the effect of measurement error of household income. We then applied the bootstrap approach to obtain the standard error and the 95% confidence interval for the SIMEX estimates. We found that the mortality risk is significantly associated with mothers death, fathers death, spouses death, and household income. Parents death and spouses death both have negative impacts on mortality risk. The mortality risk would be less, if an individual has higher household income. The measurement error does not change the effects of these variables significantly.