OBJECTIVES: Healthcare researchers often encounter missing values in many datasets. Ultimately, a patient record with missing fields can still carry valuable information. This extra information becomes more important to keep in oncology and other rare disease studies where sample size is typically limited. The purpose of this study is to demonstrate that researchers can benefit from using Multiple Imputation (MI) approach to tackle missing value problems. METHODS: The model data from IMS claims (Dx) and retail prescription (Rx) contained year 2011 patient level CKD stage indications, longitudinal drug therapies, days of supplies, titration rates, Demographic characteristics, payment type, and physician specialties, etc. We built multivariate logistic models to identify Chronic Kidney Disease (CDK) patient stages using prescription data in order to further evaluate the prevalence, economic burden and market opportunities. Under the general assumption of missing at random (MAR), we used MI with regression method to impute the missing monotone and categorical values before the modeling process. RESULTS: The pooled results from 5 MI imputed datasets were reported. Compared with the results from deterministic missing imputation approach, the MI showed larger standard error and wider 95% confidence interval. The wider CI reflected the additional data uncertainty from the missing values. CDK stage 4 (11.2%) had smallest proportion and it had lowest hit rate in the prediction model. MI approach showed more CDK stage 4 identifications than those from deterministic complete case analysis. CONCLUSIONS: This study demonstrated that MI is capable of reflecting the underline uncertainty associated with the data by introducing random errors into the imputation process. MI can generate unbiased results and good standard error estimation when using appropriately. On the other hand, with the advent of modern computational technology, the MI becomes computationally simple and easy to use.