epi.ssdetect {epiR}

Sample size to detect an event

Description

Sample size to detect at least one event (e.g., a disease-positive individual) in a population. The method adjusts sample size estimates on the basis of test sensitivity and can account for series and parallel test interpretation.

Usage

epi.ssdetect(N, prev, se, sp, interpretation = "series", covar = c(0,0), 
   finite.correction = TRUE, nfractional = FALSE, conf.level = 0.95)

Arguments

N

a vector of length one or two defining the size of the population. The first element of the vector defines the number of clusters, the second element defining the mean number of sampling units per cluster.

prev

a vector of length one or two defining the prevalence of disease in the population. The first element of the vector defines the between-cluster prevalence, the second element defines the within-cluster prevalence.

se

a vector of length one or two defining the sensitivity of the test(s) used.

sp

a vector of length one or two defining the specificity of the test(s) used.

interpretation

a character string indicating how test results should be interpreted. Options are series or parallel.

covar

a vector of length two defining the covariance between test results for disease positive and disease negative groups. The first element of the vector is the covariance between test results for disease positive subjects. The second element of the vector is the covariance between test results for disease negative subjects. Use covar = c(0,0) (the default) if these values are not known.

finite.correction

logical, apply finite correction? See details, below.

nfractional

logical, return fractional sample size.

conf.level

scalar, defining the level of confidence in the computed result.

Value

A list containing the following:

performance

The sensitivity and specificity of the testing strategy.

sample.size

The number of clusters, units, and total number of units to be sampled.

Note

Sample size calculations are carried out using the binomial distribution and an approximation of the hypergeometric distribution (MacDiarmid 1988). Because the hypergeometric distribution takes into account the size of the population being sampled finite.correction = TRUE is only applied to the binomial sample size estimates.

Define se1 and se2 as the sensitivity for the first and second test, sp1 and sp2 as the specificity for the first and second test, p111 as the proportion of disease-positive subjects with a positive test result to both tests and p000 as the proportion of disease-negative subjects with a negative test result to both tests. The covariance between test results for the disease-positive group is p111 - se1 * se2. The covariance between test results for the disease-negative group is p000 - sp1 * sp2.

References

Cannon RM (2001). Sense and sensitivity — designing surveys based on an imperfect test. Preventive Veterinary Medicine 49: 141 - 163.

Dohoo I, Martin W, Stryhn H (2009). Veterinary Epidemiologic Research. AVC Inc, Charlottetown, Prince Edward Island, Canada, pp. 54.

MacDiarmid S (1988). Future options for brucellosis surveillance in New Zealand beef herds. New Zealand Veterinary Journal 36, 39 - 42. DOI: 10.1080/00480169.1988.35472.

Examples

## EXAMPLE 1:
## We would like to confirm the absence of disease in a single 1000-cow 
## dairy herd. We expect the prevalence of disease in the herd to be 5%.
## We intend to use a single test with a sensitivity of 0.90 and a 
## specificity of 1.00. How many samples should we take to be 95% certain 
## that, if all tests are negative, the disease is not present?

epi.ssdetect(N = 1000, prev = 0.05, se = 0.90, sp = 1.00, interpretation =
   "series", covar = c(0,0), finite.correction = TRUE, nfractional = FALSE, 
   conf.level = 0.95)

## Using the hypergeometric distribution, we need to sample 65 cows.


## EXAMPLE 2:
## We would like to confirm the absence of disease in a study area. If the 
## disease is present we expect the between-herd prevalence to be 8% and the 
## within-herd prevalence to be 5%. We intend to use two tests: the first has 
## a sensitivity and specificity of 0.90 and 0.80, respectively. The second 
## has a sensitivity and specificity of 0.95 and 0.85, respectively. The two 
## tests will be interpreted in parallel. How many herds and cows within herds 
## should we sample to be 95% certain that the disease is not present in the 
## study area if all tests are negative? There area is comprised of 
## approximately 5000 herds and the average number of cows per herd is 100.

epi.ssdetect(N = c(5000, 100), prev = c(0.08, 0.05), se = c(0.90, 0.95), 
   sp = c(0.80, 0.85), interpretation = "parallel", covar = c(0,0), 
   finite.correction = TRUE, nfractional = FALSE, conf.level = 0.95)

## We need to sample 46 cows from 40 herds (a total of 1840 samples).
## The sensitivity of this testing regime is 99%. The specificity of this 
## testing regime is 68%.


## EXAMPLE 3:
## You want to document the absence of Mycoplasma from a 200-sow pig herd.
## Based on your experience and the literature, a minimum of 20% of sows  
## would have seroconverted if Mycoplasma were present in the herd. How many 
## sows do you need to sample?

epi.ssdetect(N = 200, prev = 0.20, se = 1.00, sp = 1.00, interpretation =
   "series", covar = c(0,0), finite.correction = TRUE, nfractional = FALSE, 
   conf.level = 0.95)

## If you test 15 sows and all test negative you can state that you are 95% 
## confident that the prevalence rate of Mycoplasma in the herd is less than
## 20%.