Multiple Imputation For Missing Data That Including A Ratio And Evaluation And Application The Effect Of Intervention That Implement The Secondary Prevention Of Cardiac Rehabilitation

Objective:Evaluation of cardiac rehabilitation is often needed for the follow-up monitoring,but comprehensive intervention of patients need a long time, Different stages take different interventions, and the time of follow-up and intervals of patients are different. There are a lot of hazardous factors affecting patients. Because of less cooperative or other reasons like limited mobility, Such that in the acquisition process of the longitudinal monitoring data inevitably result in missing data in the hazardous factors about Cardiovascular disease.Overweight and obesity is very important in disease-causing process. These kinds of indicators were used to evaluate the effect of intervention. The missing of denominator or numerator will happen at the process of collection the information about patients. It is missing ratio,we also called incomplete ratio. Because of a covariate that is a ratio contains a certain functional relationship between denominator and numerator.When we imputate a covariate that is a ratio,it maybe a new problem in contemporary time.Methods:This thesis mainly studies in patients with BMI, BNP and c Tn I.Using multiple imputation methods for missing data will form a complete data set and analysis it,Explore methods and strategies to deal with incomplete covariates that are ratios.We imputate missing data which in cardiac rehabilitation comprehensive intervention of research.we select the key of indicator that is BMI.If the numerator anddenominator does not miss at the same time.In imputating the model directly observable weight and height seems reasonable.We use a passive imputation.making use of the numerator and denominator of logarithmic conversion, and then calculate the body mass index(BMI).In addititon, we always use active imputation to imputate the missing indicator that is BMI. They have the same power. Stimulation studies can prove the different features between model M1-M6,and figure out the factors of models.Analyzing the instance which was collected from a hospital in Shanxi Province. We can use stata software to compile program for solving the sequence researching.Results:The thesis select BMI to reflect the management weight. BMI is a ratio, which was built up two variables, as a covariate was chosen in the analysis of regression. The result of stimulation studies:1、When the coefficient of variation of denominator in covariates that are ratio equal to 10%, using model(M1~M6) to imputate covariates that are ratios and effects are basically the same,but in the absence of mechanisms for missing completely at random(MCAR), the can produce negative bias.2、When the coefficient of variation of denominator in covariates that are ratio equal to 20%, useing model(M1~M6) imputate it.in the absence of mechanisms for missing completely at random(MCAR),negative bias is-0.610,Other imputation model M1~ M4,M6 deviations were-0.01,-0.013,-0.011,-0.01 and-0.058, the absolute value Significantly less than-0.610.Imputation models are strongly depended on the denominator of variation of coefficient. when CV become larger, the model M5 has the poorly effect.3、With the CV of denominator in increasing, The effect of imputation using Model M6 is better than that by using Model M6,Not only we make use of logarithmic transformation to make denominator and numerator to fit normal distribution, but also change the algorithm of ratio.this model can fully mine the information of the data which was researched by us, and improve the robust of the method of imputating.We analyze the actual case, CV(height2)=10.2%,the results of candidate models are consistency. the differences of means are(0.21 ~ 1.31)kg/m2,because of the model M6 can fully mine the information of data, we opt the result of model M6. The complete date which was imputated of BMI is(0.21~1.31)kg/m2,we use the multiple variable linear regression to hand in the compete data.the outcome show the relationship of between some indicators,they have active correlation,the standard coefficient of c Tn I equal to0.273, and have the biggest influence.Conclusions:When we encounter the problems of incomplete covariates that are ratios in medical researching, using the method of simulation studies to explore the incomplete covariates that are ratios which was utilized mathematical conversion, the different models judging by the coefficient of variation of denominator, and analyzing the case,and finished the imputation of BMI, we will obtain compete data, utilizing the Corresponding model to analyze it.examples are analyzed to achieve a ratio of incomplete filling of multiple covariates. It can provide the way of dealing with incomplete covariate that is a ratio, it can as the reference to solve the problems in medicinal studies.