The diagnostic study sample size calculator gives the number of people with and without the disease and the total to recruit to estimate a test’s sensitivity and specificity with a chosen precision.
How the calculation works
- Enter the expected sensitivity and specificity.
- Enter the precision, the confidence level and the prevalence.
- The calculator gives the cases, the non-cases, the totals and the measure that drives the size.
Example
A new test is expected to have a sensitivity of 90 % and a specificity of 85 %. You want each within ±7 points at 95 % confidence, in a sample where 20 % have the disease. The cases needed: 1.96² × 0.9 × 0.1 / 0.07² = 71. The non-cases needed: 1.96² × 0.85 × 0.15 / 0.07² = 100. Dividing by the shares gives 355 and 125, so you recruit 355 people.
Why it matters
The calculator helps researchers who plan studies of diagnostic tests, medical students, and laboratories that validate a new assay.
Good to know
The expected sensitivity and specificity come from pilot data or the literature. If you are unsure, using 50 % gives the largest, most cautious size. The prevalence should be that of the people who will really be tested, since a test often performs differently in a hospital clinic and in the general population. The formula uses a normal approximation, which is poor when the expected value is near 0 % or 100 % or the numbers are small, and an exact method is then better. Remember that precision here is about the width of the interval, not about a hypothesis test. Report the design using the STARD guidelines.
Related tools
The sensitivity and specificity calculator evaluates a test, the Bayes theorem calculator links prevalence and predictive values, the proportion confidence interval calculator brackets a proportion and the margin of error calculator sizes surveys.