Page 53 - PATIENT REGISTRY DATA FOR RESEARCH: A Basic Practical Guide

After having presented the final results, the findings obtained from these results will

have to be reviewed again by the investigators to ensure that they are scientifically sound and

valid, logical, relevant, and be reflective of the total population. Although the statistical

analysis is run by using a programming code, it is still possible for mistakes to occur. These

mistakes can arise at any stage during the analysis; for example, from the data itself or during

the process of analysis (such as the use of the wrong programming code).

To promote best practices for data analysis, the researchers are encouraged to

carefully review and validate all the results obtained from statistical analysis. If possible, it is

strongly recommended to validate the results obtained from all the tables or at least from a

few tables chosen randomly to ensure that all the analyses are correct. In addition, it is also

advisable to conduct an external validation check by a colleague(s), a senior statistician or the

supervisor. Any necessary amendments and/or rectifications should be made immediately

whenever errors have been found. This is an important validation step to ensure that an

effective presentation of accurate and significant research findings can be delivered to key

stakeholders and/or research clients later on.

4.3 Development of the Statistical Analysis Plan

Statistical Analysis Plan (SAP) is a document that records all the important steps for

the overall framework of statistical analysis. This plan provides a documentation which

describes the procedures for (i) handling duplicates, (ii) matching and combining data sets,

(iii) setting the conditions and requirements for the statistical analysis, (iv) handling missing

values and (v) developing the framework for the statistical analysis (which includes a detailed

description of the all steps to be taken for the statistical analysis).

Hence, the SAP is usually prepared after the output obtained from the analysis has

been finalised. This plan will then provide a documentary evidence of how the analysis has

been prepared, conducted and presented. The programming codes that are used to produce the