Analysis of Multiple Imputation Techniques in Healthcare Data

Clinical data are rapidly being mined to generate new medical knowledge with the aim of providing increased diagnostic accuracy, more customized treatment regimens, better clinical results, and more effective exploitation of health-related resources. The effect of missing information on the interpretation of statistics differs based on a number of variables, including the amount of missingness, the mechanism for handling missing data, and the technique used to deal with missing values. Therefore, missing data is one of the primary barriers to optimal knowledge elicitation from clinical data. In recent years, pattern-mixture model (PMM) has grown significantly with multiple types of missingness processes and it enables the analysis of incomplete longitudinal data. The root mean square error rate of PMM is around 0.092 which is very low compared to conventional methods like multi-chained equations (MICE) and K-nearest neighbor (KNN).