In the post JCI Standard MCI.19.4 – Patient Clinical Record, Medical Records Review, Medical Records Review Protocol, I had proposed the simple random sample technique to select the representative sample.

I introduced the use of a random table to select the sample.

An intuitive approach might be to uniquely identify all the units in a (finite) population, by writing a 3 digit number starting from 001 to let’s say 400(finite population) on small pieces of paper, put all the pieces of paper in a hat(use a Texan hat if you like), mix well and draw out enough numbers for a required sample size, for example 100 numbers starting from 001 to 100 to give a sample size of 100. Do not replace the picked pieces of paper back into the hat and ignore all the numbers greater than 100, mixing the hat after each selection of number.

This is the principle used in the selection of winning tickets in a raffle or lottery, and it is the model underlying  the simple random sample.

A simple random sample is a sample chosen in such a way that, at each draw, every number in the hat has the same chance of being chosen. Everybody in the population has the same chance of getting into the sample.

Such samples are representative of the population in so far as no particular block of the population is more likely to be represented than any other. The general term ‘random sample’ refers to the situation when every member of the population has a known (non-zero, but not necessarily the same) probability of selection. Random is thus a term that describes how the sample is chosen, rather than the sample itself.

You could of course choose other sampling techniques.

You could have picked the stratified random sample. The population is divided into groups, or strata, on the basis of certain characteristics, for example age or sex. A simple random sample is then selected from each stratum and the results for each stratum are combined to give the results for the total sample. The object of this type of sample design is to ensure that each stratum in the population is represented in the sample in certain fixed proportions, which are determined in advance. For example, I could have divided the admissions or inpatients into different groups representing the practitioners providing care and the types of care provided. A simple random sample is then selected from each stratum and the results for each stratum are combined to give the results for the total sample.

Then are other ways of sampling you also choose to use in the protocol, like multistage and cluster random sampling, and quota sampling.

Find a good statistics book or books and do some good reading before deciding on the sampling technique to use in the protocol.

Before I leave this post, I leave you with a sample page of a 5 digit random table from a statistic book. Click this link to view a random table from the textbook Basic Concepts in Statistics and Epidemiology, Appendix F, Random Numbers, page 198

References:
Leslie E. D., and Geoffrey J. B., Interpretation and uses of medical statistics, 5th ed, Blackwell Science, UK

Theodore H.M., Basic Concepts in Statistics and Epidemiology, 2007, Radcliffe Publishing, UK