Introduction
A diagnostic test is designed to help determine whether a person has a particular disease or condition. But a test is not simply “good” or “bad.” Its performance depends on how accurately it identifies people with the target condition and distinguishes them from people without it. Two of the most important measures used to evaluate this performance are sensitivity and specificity. Read More
Understanding these measures is essential when developing, evaluating, or interpreting diagnostic technologies. They are commonly used in diagnostic accuracy studies alongside other measures such as predictive values, likelihood ratios, and receiver operating characteristic (ROC) curves. Read more
Sensitivity and Disease Detection
Sensitivity describes how well a diagnostic test identifies people who actually have the target condition.
Among all individuals who are truly positive according to a reference standard, sensitivity tells us what proportion receive a positive result from the test.
Sensitivity = True Positives / (True Positives + False Negatives)
A highly sensitive test produces relatively few false-negative results. This makes sensitivity particularly important when the objective is to minimize the chance of missing a condition.
Sensitivity in Practice
Imagine that a diagnostic test is evaluated in 100 people who truly have a disease.
If the test correctly identifies 95 of them as positive but gives a negative result for 5 people:
Sensitivity = 95 / 100 = 95%
The test therefore has a sensitivity of 95% in this example.
The five people who were not detected represent false-negative results.
The video below explains more the concept .
Specificity and Identification of True Negatives
Specificity describes how well a diagnostic test correctly identifies people who do not have the target condition.
Among individuals who are truly negative according to the reference standard, specificity represents the proportion who receive a negative test result.
Specificity = True Negatives / (True Negatives + False Positives)
A highly specific test produces relatively few false-positive results.
Specificity in Practice
Imagine that 100 people are known not to have a disease.
If a diagnostic test correctly identifies 97 of them as negative but incorrectly reports 3 as positive:
Specificity = 97 / 100 = 97%
The three incorrect positive results are false positives.
Core Measures of Diagnostic Accuracy
The easiest way to distinguish sensitivity and specificity is to consider the population each measure focuses on.
| Measure | Main focus | Incorrect result |
|---|---|---|
| Sensitivity | Correct identification of people with the condition | False negatives |
| Specificity | Correct identification of people without the condition | False positives |
A useful way to remember the difference is:
Sensitivity → detection of the condition
Specificity → exclusion of the condition
These measures are calculated by comparing the diagnostic test, often called the index test, with a reference standard that is used to establish the target condition.
False Positives and False Negatives
No diagnostic test is necessarily perfect.
A false positive occurs when a test result is positive even though the target condition is absent.
A false negative occurs when a test result is negative even though the target condition is present.
These errors can have different implications depending on the diagnostic context. Missing a disease because of a false-negative result can be particularly important when early detection is critical. Conversely, excessive false-positive results can lead to unnecessary follow-up testing or other interventions.
This is why diagnostic test evaluation requires more than simply assessing whether a test produces positive or negative results.
The Balance Between Sensitivity and Specificity
A diagnostic test may perform very well in one characteristic while performing less well in another.
For example, a test may be designed to detect as many true cases as possible. Increasing the ability to detect positive cases can sometimes result in more people without the condition also receiving positive results.
This means that diagnostic test evaluation often involves considering the balance between sensitivity and specificity, rather than focusing on a single number. ROC curves are one method used to examine this relationship across different test thresholds.
Diagnostic Thresholds and Test Classification
Many diagnostic technologies do not simply detect a target as present or absent. Instead, they produce a measurement.
A threshold, or cut-off value, can then be selected to classify results as positive or negative.
Changing this threshold can change the sensitivity and specificity of the test. A different threshold may detect more true cases but also produce more false positives, while another threshold may reduce false positives but miss more true cases.
This concept is particularly relevant for quantitative diagnostic technologies, where the measured signal may be continuous rather than simply positive or negative.
ROC Curves and Diagnostic Performance
A receiver operating characteristic (ROC) curve provides a way to visualize how sensitivity and specificity change across different decision thresholds.
The curve plots sensitivity against the false-positive rate, which is calculated as:
False-positive rate = 1 − specificity
By examining the ROC curve, researchers can assess how well a test separates individuals with the target condition from those without it across different thresholds.
The area under the ROC curve (AUC) is another commonly used measure of diagnostic test performance.
Open-access scientific paper: Florkowski, 2008
Predictive Values and Diagnostic Interpretation
Although sensitivity and specificity are fundamental measures, they do not answer every question about how a test will perform in a real population.
Two other measures are particularly useful: positive predictive value (PPV) and negative predictive value (NPV).
Positive predictive value describes the probability that a person actually has the condition when their test result is positive.
Negative predictive value describes the probability that a person does not have the condition when their test result is negative.
Unlike sensitivity and specificity, predictive values can be strongly influenced by the prevalence of the condition in the population being tested.
This distinction is important because a test's performance in a research study may not translate directly to the same predictive values in another population with a different prevalence of the condition.
Likelihood Ratios and Diagnostic Probability
Likelihood ratios (LRs) provide another way to interpret diagnostic test results.
The positive likelihood ratio is:
LR+ = Sensitivity / (1 − Specificity)
The negative likelihood ratio is:
LR− = (1 − Sensitivity) / Specificity
Likelihood ratios can be used together with the probability of disease before testing to estimate the probability after receiving the test result.
This makes likelihood ratios useful for connecting laboratory or diagnostic test performance with the broader diagnostic decision-making process.
Diagnostic Accuracy Study Design
Diagnostic accuracy studies generally compare an index test with a reference standard in a defined population.
A well-designed study needs to describe how participants were selected, how the tests were performed and interpreted, and how the results were analyzed. These methodological factors matter because weaknesses in study design can produce biased estimates of sensitivity and specificity.
The STARD 2015 framework was developed to improve the transparency and completeness of reporting diagnostic accuracy studies. It encourages researchers to provide sufficient information for readers to assess the reliability and applicability of reported diagnostic performance.
The Role of Population and Testing Context
Sensitivity and specificity should not be treated as completely fixed properties of a diagnostic technology.
Diagnostic accuracy can vary according to the population studied, the clinical setting, the reference standard, disease spectrum, testing procedures, and other methodological factors.
Consequently, a test that performs well in one study population may not necessarily produce exactly the same performance in another setting.
This is why diagnostic technologies need to be evaluated in populations and contexts that are relevant to their intended use.
From Biological Detection to Diagnostic Application
Understanding sensitivity and specificity helps researchers move from simply developing a detection method to evaluating whether that method can provide meaningful diagnostic information.
A diagnostic workflow can therefore be viewed as a sequence:
Biological target → Detection technology → Test result → Performance evaluation → Diagnostic interpretation
The target might be a nucleic acid, protein, antibody, pathogen, or biomarker. The detection technology may involve molecular diagnostics, immunodiagnostics, biosensors, or other analytical approaches. The resulting test then needs to be evaluated using appropriate measures of diagnostic performance.
Key Concepts to Remember
- Sensitivity measures how effectively a test identifies people who have the target condition.
- Specificity measures how effectively a test identifies people who do not have the target condition.
- False negatives are particularly relevant when considering sensitivity.
- False positives are particularly relevant when considering specificity.
- Changing a diagnostic threshold can affect both sensitivity and specificity.
- ROC curves help visualize the relationship between sensitivity and specificity across thresholds.
- Predictive values depend on the population and prevalence of the condition.
- Likelihood ratios help connect test results with pre-test and post-test probability.
- Diagnostic performance should always be interpreted in the context of how and where the test was evaluated.
Further Scientific Reading
Sensitivity, Specificity, ROC Curves and Likelihood Ratios
Florkowski, 2008 — an open-access overview of key concepts in diagnostic test interpretation.
https://pmc.ncbi.nlm.nih.gov/articles/PMC2556590/
Assessing Diagnostic and Screening Tests: Part 1 — Concepts
An open-access resource covering sensitivity, specificity, predictive values, and related concepts.
https://pmc.ncbi.nlm.nih.gov/articles/PMC1071430/
STARD 2015 Guidelines for Diagnostic Accuracy Studies
Guidance for transparent reporting of diagnostic accuracy research.
https://bmjopen.bmj.com/content/6/11/e012799
STARD 2015 — EQUATOR Network
Reporting guidance and resources for diagnostic accuracy studies.
https://www.equator-network.org/reporting-guidelines/stard/