Page 49 - PATIENT REGISTRY DATA FOR RESEARCH: A Basic Practical Guide

During statistical analysis

e. For a descriptive study (using census data) with no missing values, it is still

acceptable to provide just a descriptive statistical analysis of the registry data

alone, which means that no inferential statistical analysis will be conducted.

f. However, if there are missing values in the data set, then it will become necessary

to conduct both descriptive and inferential statistical analysis of the data set since

it will now be necessary to infer these findings about the target population from

the sample data.

g. It is important to observe and take proactive steps to ensure that a list of

underlying assumptions which must hold true in order for all the statistical

computations to be valid, especially with regard to parametric test and regression

modelling. This is because it will be a futile attempt to conduct statistical analysis

by performing all statistical computations and yet violating these underlying

statistical assumptions (or failing to ensure that these assumptions actually hold

true), as this will yield invalid conclusions.

h. Conducting inferential statistical analysis on patient registry data (such as the use

of disease modelling methods as research tools) can be a very complicated task

because these patient registries can have many dozens of variables which are

regarded as the independent variables for a measurable outcome. It will not be

possible to label all these variables as risk factors in a single analysis by using a

multivariate model because such an analysis will not be efficient due to issues

which have arisen from not being able to (i) fulfil all the assumptions in regression

modelling and also to (ii) meet the minimum sample size requirements. Therefore,

it is strongly recommended for the researcher to carefully select a list of

independent variables that qualify to be regarded as risk factors to be tested in the