The problem of missing data in randomized control trials a quick and easy guide

Abstract

Evidence-based research in health care has been developed well in recent years. One of the biggest challenges of the researchers is the management of missing data. Missing data is defined as a data value that is not available, and that if it was observed, it would make a difference to the analysis. Missing data may affect the value of the research findings and scientific information provided. It can reduce the statistical power and introduce bias in the estimation of various parameters. The participants who withdrew or were lost from a study may have different characteristics, and can thus diversify the sample, compared with the completely adherent participants. In such case, the study sample may no longer be representative. For example, the range of some of the participant´s characteristics may be changed, such as age, gender, socioeconomic variables or other measurements. There are several causes of missing data; a few may be due to the study design, and others simply due to missing value, especially in the case of the questionnaire, which is commonly used in studies in the medical and nursing sciences. For instance, older people may avoid answering specific questions related to sexual activities or leisure activities. In addition, according to Alm-Roijer and colleagues, some people may have difficulty in understanding the questions, and for that reason do not provide an answer. In addition, the time available may not be enough for the respondent to complete the questionnaire. Each study has a particular design, so there is no universal method for managing missing data. The management of missing data, however, needs to be approached by considering three aspects before choosing the most appropriate way of analysis, namely (a) the pro-portion of missing data, (b) the mechanism of the missing data, and (c) the pattern of the missing data.