Now that you’ve selected which approach you will use; the following steps will help you develop and conduct your assessment.
Selecting an appropriate sample population is a crucial step in collecting quality data and, depending on your sampling strategy, can allow you to draw conclusions about your entire community, not just the individuals who participate in the assessment.
A sample population is a portion of the total population you are interested in for your assessment. A sampling strategy is the process you use to determine your sample.
First, determine who would be able to provide the necessary information to address your objectives. Those who can answer your questions to address the overall objectives are your target population.
Based on the objectives that you identified earlier, complete the following exercise to identify your target population.
Be specific in terms of the characteristics you would like included in your target population (also known as inclusion criteria). This will help you determine who should and should not be invited to participate in your assessment. This may include age, occupation, exposure to disaster, living in certain region, have certain health conditions, etc. In determining about your target population, it may also help to think about who should not be considered (also known as exclusion criteria).
Exercise:
I would like to ask questions of the following group(s):
- ______________________
- ______________________
- ______________________
Example(s):
I would like to ask questions of the following group(s):
- Women living in my city/town who are 5-9 months pregnant
- Parents/Caretakers with infants under the age of 1 living in a shelter after a disaster in my city/town
The In-depth approach suggests a modified cluster sampling strategy with referral. Cluster sampling strategies have been used extensively in public health research and in reproductive health research. The CDC Community Assessment for Public Health Emergency Response (CASPER), Reproductive Health Assessment (RHA) Toolkit for Conflict-Affected Women, and the RHAD Toolkit 1.0 have recommended cluster sampling strategies for similar assessments.
The sampling strategy suggested in the In-depth approach is based on the following resources if you would like to read more to inform your sampling strategy. Because the size of regions and density of population across the US varies, exact numbers for sampling are not provided in this toolkit. Guidance from resources are provided but it is strongly encouraged to calculate the appropriate sample based on your target region and needs.
- “Modified Cluster-Sampling Method for Post-Disaster Rapid Assessment of Needs” by J. Malilay, et al. (Bulletin of the World Health Organization,1996, 74 (4) 399-405)
- “Cluster Sampling with Referral to Improve the Efficiency of Estimating Unmet Needs among Pregnant and Postpartum Women after Disasters” by J.Horney et al. (Women’s Health Issues, 2012, 22(3) 253-257)
- CDC Reproductive Health Assessment Toolkit (RHA) for Conflict-Affected Women
- Examples of random selection for cluster sampling from RHAD for Conflict-Affected Women
Overview of cluster sampling steps:
- Define your geographic region (e.g., specific counties hit by a hurricane, a state under stay-at-home orders due to a pandemic)
- Define clusters within the geographic boundaries. Determine groupings of populations such as blocks, districts, neighborhoods, villages, and camps.
- Determine the total number of households within each cluster. The RHA toolkit suggested for each of the clusters, you must know the total number of households. If you have a listing of households, total up the number of households per cluster. If you do not have a listing of all the households, then the total number can be estimated from the total population and the average size of each household. Divide the total population by the average size of households to estimate the number of households, as follows: Total population ÷ average size of households = total number of households.
- Randomly select the clusters to be sampled. The RHAD Toolkit 1.0 recommended selecting a sample of 30 groups (census blocks/groups) and the RHA toolkit recommended that at least 25 clusters with at least 25 households in each cluster. Check out this resource from Michigan State University Center for Rural Health on how to create a random sample using Microsoft Excel. If you are interested in identifying a statistically proportional or significant sample, please contact a statistician at your agency or a public health researcher at a university closest to you.
Randomly select the households within each cluster – This will depend on how large your region is, the number of clusters, and the number of households within each cluster. For example, the RHAD Toolkit 1.0 suggested randomly selecting 7 households within the 30 census blocks. As data is collected, ensure each household includes a member of your target population for your assessment. Each team will proceed to one of the randomly selected housing units in their first cluster to try and conduct an interview. If an eligible resident is willing to participate at the first home, the team then moves to another randomly selected housing unit in their cluster. If this first attempt does not result in an interview, the team then goes to the next nearest house and continues in this fashion until the first interview in that cluster is complete. Once an interview is collected near the first randomly selected housing unit, the interview team moves to another point in their cluster. Because this is a modified two-stage cluster sampling method with referral, interview teams will also be asking each participant if they know of any pregnant or recently postpartum women (or other target population for the assessment) in the area. If the team secures a referral, they then proceed directly to the referral address. If the referral is unavailable or unwilling to participate, the interview team then continues to the next nearest house if the referral’s home is within the cluster boundaries. If the referral home is not within the team’s cluster, they return to a randomly selected housing unit in their cluster and attempt an interview. This method increases the proportion of pregnant and postpartum women in the sample population.
Helpful tip 1: While you may use households to calculate an appropriate sample size, keep in mind that people make have moved during the disaster. Therefore, you may need to collect data in other methods beyond going to households in person (phone interviews, interviews at community organizations/shelters, sending online surveys through local groups’ and health department social media) and assess if you have collected a sample that represents your population.
Helpful tip 2: If you have questions on how to calculate an appropriate sample for a more rigorous assessment, consult your agency’s data analyst or statistician.
