Page 39 - PATIENT REGISTRY DATA FOR RESEARCH: A Basic Practical Guide

scenarios, the researcher will be losing some vital information. Unfortunately, there is no

universal method for handling such missing data (Rubin, 1987; Greenland & Finkle, 1995).

Therefore, it is advisable to take proactive measures to avoid or at least minimize the chance

of having such MNAR data in patient registries.

These missing data can also occur at any stage of the collection of registry data.

Firstly, they can often arise because the required information has not been collected, which

could possibly happen due to a variety of reasons such as the loss of a case record form (i.e.

resulting in a MCAR) or the unavailability of resources (i.e. resulting in a MNAR). Secondly,

it is also possible that the information has not been recorded on the case report form (CRF),

even though it is easily available. Thirdly, it is also possible that although the data are already

available in the CRF, the procedure of data entry has inadvertently omitted the recording of

such data within the registry database. To understand how to deal with such missing data, the

researcher should determine at which stage the data have gone missing in order to decide the

right course of action to take. Since most researchers are keen to proceed with data analysis

as soon as possible, they may adopt any one of the three possible ways to handle these

missing data, namely (i) by using a validation technique, (ii) by using an imputation

technique and the last and also the easiest way will be to (iii) simply use a specific code to

designate the missing data as 'missing'.

The validation technique for handling missing data refers to a method whereby the

missing data are being substituted by using relevant supplementary information obtained by

the other variables in the registry data set. Using the Malaysian identification card number as

an example, it is possible for a researcher to realistically determine the date of birth (by

referring to the first five digits), the gender (i.e. an odd last digit indicates male and an even

last digit indicates female) and the province/state in which the patient was born (by referring