DEFINITION OF STUDY POPULATION AND NUMBERS ESTIMATION

Your study population can be patients, people, microbes, institutions, students, any number of things, depending on the research you need to do.  It is very important to:

1. Describe the place where you will do the word
2. Describe the population itself—who are they, where are they, what are they like?
3. Show how you can access the population
4. How big is the total population
5. How will you sample the population
6. What is your case definition:  inclusion/exclusion criteria

The following examples will help you.  Not all proposals have all these elements, but many will have at least one if not more..

Here is an example of how you might describe a population.  In this example we are studying rat infestations in a Colonia in Matamoros.

The clinic operated by Dr. Rodriguez in Derechos Humanos includes a small dispensary, and is located in the center the colonia.  The settlement is about three years old, and houses between two hundred and two hundred and fifty families.  Though outside official Matamoros city limits, it is entirely surrounded by areas which are incorporated, including the city airport.  The nearest running water is from a spigot about two blocks from the dispensary.   The nearest Matamoros city dumpsite is a few hundred yards from the furthest point south of the encampment.

Here is an example of how you might design a survey.  This example is a study to identify genes associated with diabetes in Brownsville.

Following updating block maps from the Brownsville metropolitan area , we will randomly select blocks according to dwelling unit density (i.e., number of units per block as opposed to numbers of individuals per dwelling unit). On selected blocks, the field team will enumerate all individuals to determine age, gender and diabetes status based on history. Utilizing a permutation algorithm, we will select one individual at random aged 15 to 54 years. These individuals will be invited for a physical examination at the Regional Campus.

Inclusion/exclusion criteria are critical and these are also described as the case definition.  Don’t forget you need to define both cases and controls.  Here is an example

Patient definition: Patients who have visited the physician or clinic and whose charts have been coded with the following ICD-9 diagnoses:

            155.0         Liver cancer

    571.0-9        Cirrhosis and chronic liver disease

Controls          Two patients age and sex matched for each control from the same data source (same clinic/hospital) with neither of the above coding

Here is an example of a numbers estimation for a chart review of end-stage liver disease in Brownsville.  This kind of estimation is essential for most if not all proposals.  If you don’t have exact numbers you have to make your best estimate based on what you know and state that these are assumptions.  To make the calculations the easiest way is to use the StatCalc function of EpiInfo—all you have to do is put in the numbers and it gives you the answers.

Assuming a rate of HCV infection in the population of 2.5% and the lower limits of the HCV rate in chronic liver disease at 10%, with power of 80% and 95 confidence levels, taking 2 controls for each case, we will need to sample 131 cases and 262 controls.  We therefore plan to sample a minimum of 200 cases of chronic liver disease, and 30 cases of hepatocellular carcinoma to ensure adequate numbers.  We will start with charts of patients seen within the six months prior to the study.  If this does not yield sufficient patients we will continue backwards six months at a time until we have the required numbers.  Since stratified analyses will examine end points likely to require larger numbers (for instance separate male and female analyses), it will be desirable to exceed these sampling numbers if time permits.