The Insight Corner Hub: Application, similarities, difference and challenge in positivity rate, negativity rate, specificity and sensitivity with interpretation

Positivity rate, negativity rate, specificity and sensitivity are statistical measures of the performance of a binary classification test, which reports the presence or absence of a condition[2]. These measures are widely used in various fields, such as medicine, epidemiology, information retrieval, and machine learning. In this article, we will explain the definitions, similarities, differences and challenges of these measures, as well as how to interpret them.

Definitions

- Sensitivity (true positive rate) is the probability of a positive test result, conditioned on the individual truly being positive[2]. For example, if a test for COVID-19 has a sensitivity of 90%, it means that 90% of people who have COVID-19 will test positive, and 10% will test negative (false negatives).

- Specificity (true negative rate) is the probability of a negative test result, conditioned on the individual truly being negative[2]. For example, if a test for COVID-19 has a specificity of 95%, it means that 95% of people who do not have COVID-19 will test negative, and 5% will test positive (false positives).

- Positivity rate (positive predictive value) is the probability of having the condition, conditioned on the test result being positive[1]. For example, if a test for COVID-19 has a positivity rate of 80%, it means that 80% of people who test positive actually have COVID-19, and 20% do not have COVID-19 (false positives).

- Negativity rate (negative predictive value) is the probability of not having the condition, conditioned on the test result being negative[1]. For example, if a test for COVID-19 has a negativity rate of 99%, it means that 99% of people who test negative do not have COVID-19, and 1% have COVID-19 (false negatives).

Similarities

The similarity between these measures is that they all describe the accuracy of a test in different ways. They are all expressed as percentages or decimals between 0 and 1. They are all dependent on the prevalence of the condition in the population[1]. Prevalence is the number of cases in a defined population at a single point in time[2].

Differences

The difference between these measures is that they focus on different aspects of the test. Sensitivity and specificity are characteristics of the test itself, and are independent of prevalence[2]. They allow us to rule conditions in or out based on the test results. A high sensitivity means that a negative result can rule out the condition with high confidence. A high specificity means that a positive result can rule in the condition with high confidence.

Positivity rate and negativity rate are best thought of as the clinical relevance of the test[1]. They depend on prevalence and tell us how likely it is that we have or do not have the condition based on the test results. A high positivity rate means that a positive result is very likely to indicate the presence of the condition. A high negativity rate means that a negative result is very likely to indicate the absence of the condition.

Challenges

The challenges in using these measures are that they are often confused or misinterpreted by users or decision-makers. Some common misconceptions are[2,3,4]:

- Sensitivity and specificity are fixed properties of a test. In reality, they may vary depending on how the test is performed or interpreted, or how the condition is defined.

- Positivity rate and negativity rate are fixed properties of a test. In reality, they vary depending on prevalence and may change over time or across populations.

- A high sensitivity or specificity means that a test is accurate or reliable. In reality, accuracy depends on both sensitivity and specificity, as well as prevalence.

- A high positivity rate or negativity rate means that a test is accurate or reliable. In reality, accuracy depends on both positivity rate and negativity rate, as well as prevalence.

Another challenge is that these measures may not be available or reported for some tests or conditions. This may limit their usefulness or applicability in practice.

Interpretation

To interpret these measures correctly, we need to consider their definitions, similarities, differences and challenges. We also need to use them in conjunction with other information sources, such as clinical judgment, expert opinion, guidelines and recommendations. We should not rely on them alone to make decisions or draw conclusions.

Some examples of how to interpret these measures are:

- If a test for COVID-19 has a high sensitivity but low specificity, it means that it can detect most cases of COVID-19 but also produce many false positives. This may be useful for screening large populations to identify potential cases for further testing or isolation.

- If a test for COVID-19 has a low sensitivity but high specificity, it means that it can confirm most cases of COVID-19 but also miss many cases. This may be useful for confirming diagnoses or clearing cases for discharge or travel.

- If a test for COVID-19 has a high positivity rate but low negativity rate, it means that it is very likely that a positive result indicates COVID-19 but not very likely that a negative result excludes COVID-19. This may be the case when the prevalence of COVID-19 is high in the population.

- If a test for COVID-19 has a low positivity rate but high negativity rate, it means that it is not very likely that a positive result indicates COVID-19 but very likely that a negative result excludes COVID-19. This may be the case when the prevalence of COVID-19 is low in the population.

Applications

These measures are useful for evaluating the performance of different tests or instruments for diagnosing, screening, or monitoring a condition or disease. They can help us compare the strengths and limitations of different tests or instruments, and choose the most appropriate one for a given purpose or setting. They can also help us interpret the results of a test or instrument, and estimate the probability of having or not having a condition or disease based on the test results. These measures can inform clinical decision making, health policy, and research design.[5,6,7,8]

References

1. Diagnostic Testing Accuracy: Sensitivity, Specificity, Predictive .... https://www.ncbi.nlm.nih.gov/books/NBK557491/.
2. Accuracy and precision - Wikipedia. https://en.wikipedia.org/wiki/Accuracy_and_precision.
3. What are diagnostic test accuracy reviews? | Cochrane. https://www.cochrane.org/news/what-are-diagnostic-test-accuracy-reviews.
4. Validation accuracy vs Testing accuracy - Cross Validated. https://stats.stackexchange.com/questions/401696/validation-accuracy-vs-testing-accuracy.
5. Sensitivity and specificity - Wikipedia. https://en.wikipedia.org/wiki/Sensitivity_and_specificity.
6. Sensitivity, Specificity, PPV and NPV - Geeky Medics. https://geekymedics.com/sensitivity-specificity-ppv-and-npv/.
7. Sensitivity and specificity - HandWiki. https://handwiki.org/wiki/Sensitivity_and_specificity.
8. MedCalc's Diagnostic test evaluation calculator. https://www.medcalc.org/calc/diagnostic_test.php.