Page 51 - PATIENT REGISTRY DATA FOR RESEARCH: A Basic Practical Guide
P. 51
attained by the researcher since he/she is usually the subject-matter expert who
has a sound understanding of the subject, and is therefore able to decide which
variables will need to be analysed.
After statistical analysis
l. For descriptive analysis, it is recommended that sample size (n) for each variable
will also be reported, in order for the readers to obtain a rough estimate of the
proportion of missing values.
m. It is recommended to insert a footnote at the bottom of the page for alerting the
readers that those findings obtained from the variables with a large number of
missing values (say, the total number of data reported is less than 50% of the total
population data) will have to be interpreted with caution.
n. It is necessary to validate the results again by another senior statistician or
statistical consultant. At the bare minimum, it will be necessary to run the same
analysis at least twice, in order to maximize the chance for detection of errors.
o. It is important to pre-specify any limitation(s) of the registry-based study
pertaining to its achievement of research objectives that can possibly result from
either the estimation of the variables found in the patient registry database, or
from their sample size requirements, or from their proposed statistical analysis.
These are very useful tips which are highly recommended for a statistician to adopt
when formulating an approach for statistical analysis of registry data. In order to serve as
a quick guide to all statisticians and researchers alike, this e-book/book provided a
summary checklist for the whole process of collecting and analysing registry data
