Weighted Estimating Equations For Longitudinal Studies With Missing Data And Binary Outcomes

Objective:Longitudinal studies of medicine effect evaluation experiment is a common method,because of its long research time, repeatedly research, continuous testing, the subjects may not be in time to participate in the survey, resulting in missing data. Based on the likelihood of missing data methods for generalized estimating equation linear, nonlinear mixed models, etc.Based on semi-parametric estimation methods for generalized estimating equations and so on, this methods for missing completely at random mechanism missing data’s dealing provide a more accurate and effective research approach.Methods:According to longitudinal study missing data problem, in the base of generalized estimating equation introducing weight analysis, and further clarify the weighted estimating equation analysis principles, through a small sample method, gradual method simulation comparative study,which is characterized as missing at random provides a methodological reference.In preoperative, intraoperative and postoperative cardiac rehabilitation implement comprehensive secondary prevention interventions, while collecting patient Symptom Checklist 90(SCL-90), Self-rating Anxiety Scale(SAS) and Self-rating Depression Scale(SDS) measurement results were evaluated to Comprehensive intervention when patients is in CCU. Because patients detection index in CCU existing missing data,in this study, based on comparative analysis of the Weighted Estimating Equations(WEE) and Last Observation Carried Forward(LOCF) this two methods,analysis the similarities and differences. Through simulation studies,comparative analysis of the effects of sample size and the proportion of missing data for model parameter estimation results.Results:1. Two missing data padding method related to sample size and the proportion of missing dataAfter the sample size between 50 and 300, the ratio of missing data is between 5%and 40%,the results show that when the sample size is constant, with the proportion of missing data increased,the two methods parameter estimates of standard error are growing,the deviation between parameter estimation error and simulated true value is also growing.In the case of absence of a certain proportion, the larger the sample size, the two standard error of the estimate method parameters are getting smaller and smaller, when the sample is greater than 150, the parameter estimation results are stabilized. Covariates and the time deviation analysis shows, LOCF method is more WEE great law, and the estimation results WEE method to intercept parameters of large deviations. When the sample size of 150 or more, different percentage of missing parameter estimation results differ less, closer to the true value, the larger the sample size, the smaller the bias parameter estimates, parameter estimates are more closely simulate true value stable. In short, LOCF and WEE Both approaches need to ensure that there is a certain amount of samples, analysis is more robust.When the percentage of missing at 25% or less, the sample size is greater than 150, the weighted estimating equations for longitudinal monitoring of binary random missing data analysis, we can get a more robust parameter estimation results. WEE method considers not only the conditions of the model, the marginal structural model and associated data, but also by the number of weighted thought the missing unit exploded right to a non-deletion unit, by increasing the weight of the sample observations, to reduce the estimates of missing data offset produced better treatment than the LOCF method.2. WEE treatment of acute coronary syndrome dichotomous missing data more robust interpretation of resultsDuring cardiac rehabilitation after acute coronary syndrome check CCU integrated intervention for secondary prevention weighted estimating equation analysis example, the results indicate that smoking, applying a comprehensive cardiac rehabilitation secondary prevention interventions are early changes in the patient’s mental state emergency influential crown, also That smoking is the impact of early rehabilitation of patients with acute crown risk factors and secondary prevention of acute crown integrated intervention in favor of the patient’s early psychological rehabilitation. Effective use of MI and WEE method to fill the missing data processing, new ideas and new methods can be used asindividualized comprehensive cardiac rehabilitation intervention standardized management and evaluation.3.Secondary prevention intervention model in favor of a comprehensive cardiac rehabilitation patients with acute crownComprehensive cardiac rehabilitation secondary prevention intervention evaluation,patients with acute crown repeatedly test results may be affected by multiple covariate(such as individual BMI, smoking, previous history of coronary artery stenosis, treatment methods, whether applied intervention) effects the control group with conventional treatment, the experimental group implemented in the conventional treatment based on a comprehensive secondary prevention intervention. After intervention by psychological rehabilitation in a variety of analysis methods comparison study results showed that the two groups of patients can be anxious crown with varying degrees of anxiety, depression and other psychological problems after the arrival of CCU. By applying a targeted individual intervention trials in patients seen in the intervention group than the control group SCL-90, SAS scale scores when compared CCU stay were significantly improved.After multiple imputation for missing data processing, before and after treatment score of repeated measurement data generalized linear mixed model analysis, two groups of SCL-90, SAS scale scores were P value higher than 0.05, still can not believe that early intervention SCL-90, SAS scale scores have improved, statistical analysis and interpretation of actual intervention were quite different.Conclusions:For the psychological rehabilitation after the intervention make a comprehensive evaluation, this study within one week SCL-90, SAS and SDS repeat the score of the test results, according to the diagnostic evaluation criteria for each scale(normal or not) to organize, and the lack of any mode multiple imputation of missing data and other treatment carried out, using the three integrated intervention in patients with acute crown psychological state judge result the weighted estimating equations analysis. The results showed that smoking, treatment, if applied to cardiac rehabilitation integrated intervention for secondary prevention are the main factors in patients with early acute psychological rehabilitation Crowns, smoking is not conducive to the process of cardiac rehabilitation patients with acute coronary syndrome psychological rehabilitation, cardiac rehabilitation is applied Comprehensive secondary prevention interventions in favor of psychologicalrehabilitation crown urgent patients, but whether drug therapy or PCI surgery,comprehensive intervention for patients with acute mental rehabilitation crown still can not believe that there is a difference.In short, the paper through comparison study in longitudinal binary missing data processing method using multiple imputation, further confirmed the sample size and the proportional are two key issues of the missing data analysis. Moderate sample size, the lack proportion at 25% or less, weighted estimating equations can consider the relationship between the variables and the lack of time, make full use of the information provided by the missing data, effective decomposition of weight to make the model analysis more robust.Weighted estimating equations based on an integrated analysis of the intervention, further validates the multiple imputation model and analysis method longitudinal dichotomous data Absence, missing data for the longitudinal study provides a new analysis methods and analysis of new ideas for secondary prevention in patients with acute cardiac rehabilitation crown syndrome.