Page 48 - PATIENT REGISTRY DATA FOR RESEARCH: A Basic Practical Guide

4.1 A Proposed Guide for Statistical Analysis of Patient Registry Data

Depending on the research question and the study objectives, the statistical analysis of

the registry data can range from a simple analysis such as the calculation of percentage

frequency and other descriptive statistical analyses to more advanced statistical analysis such

as multivariate regression and disease modelling. The following is a compilation of all the

recommended steps for planning a robust way to conduct statistical analysis of patient

registry data.

Bef e statistical analysis

a. Data must be free from invalid duplicates, inconsistency and outliers before

beginning to conduct any of the proposed statistical analyses. Only outliers which

occur due to incorrect data entry must be removed from the data set [i.e. this

means that the true observations (albeit they resemble outliers) should be retained

for subsequent statistical analysis.]

b. Researchers should be fully equipped with prior knowledge on the design of these

registries (including their clinical parameters, definitions of all the variables and

their relevant terminologies), the design of its case report form and all the data

collection procedures.

c. The proposed statistical analysis should itself align with the research question, the

study objectives and the layout of the data in the dummy tables.

d. Researchers must have the full understanding of how the estimation of the

required sample size of registry data has been specifically tailored for the

subsequent statistical analysis of registry data, as this data set shall become the

analysis set for the study.