This Sample Size section of an outbreak surveillance plan will take considerable time to develop because it will involve calculating the number of premises and number of animals to sample. Optionally, this section may also provide specific details about the actual premises and animals that will be sampled and the methods used to select those premises / animals.
A fundamental objective in controlling or eradicating any disease is identifying all infected animals in a target population as quickly as possible. However, because a list (sampling frame) of each animal (sampling unit) in this population will likely be difficult to construct, animals are typically aggregated into herds/flocks as the initial sampling unit. Thus, your sampling plan should first focus on selecting the appropriate number of premises (sampling unit) from which to then sample animals for the disease in question. Within these selected premises, a sample of animals should then be selected from among all animals (sampling frame) in the herd/flock to determine their infection status.
The proportion of infected animals per premises is not of major interest when developing the sampling plan because for the purposes of disease control or eradication, a herd/flock is considered to be infected when at least one of its animals is infected.
Thus, the Sample Size section should include, at minimum:
The minimum number of samples taken in each premises (herd/flock) should reflect the degree of confidence you have set (e.g. 95%) that, based on the limitations (its Se) of the diagnostic test being used to identify truly infected animals, you will find one or more of these animals at the time of testing if infection is present in the herd/flock at or above the level (design prevalence) set by you.
Determine how many premises to sample in each disease control zone. Read the guidelines on this topic (below).
Determine the number of animals to sample on each premises. Read the guidelines on this topic (below).
Write the Sample Size section of the surveillance plan - Open your draft surveillance plan and modify the sample sizes content. Click here to preview the template content in your browser.
The number of premises that should be sampled to detect existing and new cases of disease at the herd or flock level during a disease outbreak depends on the disease control area / zone where the premises resides and their classification within this area / zone with respect to probable disease status.
||Number of Premises to Sample
|Sample ALL if resources are available
||Sample ALL if resources are available
*As a starting point, the premises sample size should reflect the desire to be 95% confident of detecting a 1 or 2% inter-premises disease threshold from among all premises in the disease control area / zone.
If resources (e.g. money, personnel) are limited, an option to consider is one of first sampling surveillance area / zone premises most distant from the control area and work inwards toward the center of the infected zone. See Sampling Priorities section for more information.
Tip: Use the output from the calculators as a guide for the sampling plan. Unique disease situations, veterinary knowledge, epidemiologic expertise, or other factors such as resource limitations (e.g. monetary, personnel) may indicate a need for sample sizes that are more or less conservative than the estimates provided from these calculators.
Four pieces of information are necessary to use the Animal Sample Size Calculator to provide a meaningful minimum number of animals to test on each premises to detect the disease in question, if it is present:
The first two pieces of information should not be difficult to ascertain. However, the threshold amount of disease (disease prevalence) in each herd/flock that you want to be able to detect may be more problematic to establish. If you are unsure about what this design prevalence should be, it is recommended to follow the USDA APHIS VS. National Animal Health Emergency Management Foreign Animal Disease Preparedness and Response Plan (FAD PReP) guidelines as a starting point. According to FADPReP, the within herd/flock disease threshold should be set at 5%. Keep in mind, however, that the uniqueness of the disease situation, veterinary knowledge, epidemiologic expertise, or other factors may indicate a need to vary from these recommendations.
Confidence level describes the extent to which a number or other assumption is likely to be true. In this case, it is the probability of finding at least one positive animal from those tested from each premises chosen to be sampled. Confidence level is also a proxy for what you would like your herd sensitivity (HSe) of detection to be for the disease in question.
The HSe is the probability that an infected herd/flock will yield a positive result, i.e. one or more samples are positive to your screening test (taking into account its specific diagnostic Se), given that the herd/flock is infected at a design prevalence equal to or greater than the one set by you. The aforementioned FADPreP document suggests setting confidence level at 95%. At this setting, it means that, on average, 5 out of every 100 herds with a given minimum prevalence of disease (your design prevalence) would not be detected as being infected when tested.
If there are more animals on a premises than are required to be sampled, animals should be considered for sampling in the following priority categories:
Tip: If external factors (e.g. monetary, personnel) dictate the need to reduce your sample size estimate for each herd/flock, a matrix (Probability of Failure to Detect Diseased Animals) is provided to evaluate what the effect various downward adjustments in sample size have on the herd sensitivity (HSe) of detection you have set for the disease in question.
Once the veterinary epidemiologist has determined the sample size, the next step is to actually choose the specific premises or animals to sample to meet the predetermined sample size. There are two broad types of sampling approaches for choosing these premises or animals: probability sampling and non-probability sampling.
Probability sampling uses statistical theory in such a manner to insure that every member in the population of interest has a known (and nonzero) probability of being included in the sample. In contrast, non-probability sampling is not based on statistical theory. Examples are convenience sampling, purposive (or judgment) sampling, and haphazard sampling. Use of these latter sampling methods is discouraged. Each introduces bias (sampling error) and hampers the ability of a veterinary epidemiologist to make valid inferences about the prevalence of disease in the general population.
This section focuses on probability-based sample selection techniques. References for this information are provided below.
Simple random sampling is generally considered to produce a sample representative of the population of interest because every member in this population has an equal chance of being included in the sample. The name implies that this method of probabilistic sampling is simple, but in reality it may be logistically impractical to implement due to geographic dispersion of the study sample. Nonetheless, simple random sampling still serves as the basis of all good probabilistic sampling techniques.
The outbreak surveillance toolbox provides a random sampling calculator. In order to use this calculator properly, the total number of premises (or animals) in the population of interest, i.e. the sampling frame, must be known as well as the total number of samples that must be taken from this population.
As the name implies, interval sampling involves systematically sampling premises (or animals) at set intervals from among the population to be sampled. The sampling interval is determined by dividing the population to be sampled by the predetermined sample size. Employing a random starting point, premises or animals are then systematically drawn at this interval from among the population members until the sample size is met.
The outbreak surveillance toolbox provides an interval sampling calculator. In order to use this calculator properly, the total number of premises (or animals) in the population of interest (sampling frame) must be known as well as the total number of samples that must be taken from this population.