Abstract
Background: Methods for handling missing data in clinical research are getting more attention since last few years. Contemplation of missing data in any study is crucial as they may lead to considerable biases. It can be handled by simply excluding any patients with missing values from the analysis; this will result in a diminution in the number of cases available for analysis and hence have the impact on statistical power. Hence every attempt should be made to minimise the amount of missing data. The missing data handling technique such as single imputation methods are attractive, but do not reflect the uncertainty about the predictions of the unknown missing values, and hence estimated variance of the parameter will be biased toward zero. The palliative care treatment is a specialised medical care for people with serious illness and it focuses on providing relief from symptoms, and can be used at any stage of an illness if there are troubling symptoms, such as pain or sickness.
Objective: This manuscript presents different imputation techniques to handle missing observations obtained from a repeatedly measured pain score data on palliative cancer.
Methods: Imputation methods such as Regression, Predictive Mean matching, Propensity Score, EM algorithm and Markov Chain Monte Carlo (MCMC) methods were adopted and compared to find out the appropriate imputation method on pain score data. The appropriate imputation method is decided based on the lowest standard error (SE) calculated during the Regression analysis.
Results: The mean (SD) of observed data was 3.638 (3.175) whereas the imputed mean (SD) values were 3.356 (2.6603), 3.502 (2.6100), 3.406 (2.4334), 3.474 (2.6285) and 3.264 (2.6336) respectively, for the methods with Regression, Predictive Mean matching, Propensity Score, EM algorithm and MCMC methods for pain score values at visit three. The mean (SD) of observed data was 3.528 (3.1112) whereas the imputed mean (SD) were 3.231 (2.8715), 3.253 (2.8691), 3.278 (2.7935), 3.268 (2.8725) and 3.227 (2.8952) respectively, for the methods Regression, Predictive Mean Matching, Propensity Score, EM algorithm and MCMC methods for pain score values at visit two.
Conclusion: Accordingly, to our methodology, the Propensity Score Method has appeared to be the most appropriate imputation method for pain score data. The multiple imputation techniques have few advantages; the imputed values are drawn from a distribution, so they inherently contain some variation by introducing an additional form of error in the parameter estimates across the imputation.