A Joint Model Of Longitudinal And Survival Data Based On Stochastic EM Algorithm And Its Application In HIV Clinical Data Analysis

In the evaluation of clinical efficacy of AIDS,the main concern of researchers is to determine the risk factors associated with AIDS and the survival time of patients.Longitudinal observation data such as CD4 cell count are often used as potential markers for human immunodeficiency virus(HIV)tests.Meanwhile,in clinical studies,we are also interested in survival time(such as the time of death of patients,etc.),but the correlation between between the longitudinal data and the survival data is not considered,which leads to a large deviation in data analysis results.Therefore,researchers propose that the major problem of large error can be better solved by joint longitudinal data and survival data modeling.In the traditional joint model,when longitudinal data is measured,there are measurement errors due to some external reasons,which may also lead to data missing or deletion.The simple linear mixed effect model may lead to a large deviation in the likelihood estimation of parameters.Therefore,we proposed to use the nonlinear mixed effects model instead of the linear mixed effects model as the longitudinal sub-model and the Cox proportional risk model as the survival sub-model for joint modeling.Considering the complexity of the nonlinear mixed effect model,a new parameter estimation method,namely the stochastic approximate expectation maximization algorithm(SAEM),was used to estimate the parameters of the joint model.The fitting effect based on actual data proved the advantages of the model and parameter estimation method,which provided alternative methods for the subsequent research.