The purpose of a new microbiologic test during the SARS-CoV2 pandemic is to differentiate between a patient with infection, and a patient without infection. The test is considered a screening test if performed on asymptomatic patients or patients with symptoms unrelated to infection, and a diagnostic test if performed on patients with infection-specific symptoms and signs. The epidemiologic principles underlying the ability of a test to detect an infection are reviewed assuming 1000 patients are tested in our healthcare facilities in each of the following different hypothetical scenarios and the test result is reported as positive or negative.  

Scenario #1. The available test is absolute gold, with a Sensitivity and Specificity of 100%. All future tests will be compared to this.

 Infection PresentInfection AbsentTotal
Test Positivea (True Positive)b= 0 (False Positive)a + b
Test Negativec= 0 (False Negative)d (True Negative)c + d
Totala + cb + d1000 (=a+b+c+d)

Prevalence of infection = ((a+c)/(a+b+c+d))*100.

Positive Predictive Value = a/(a+b) = 100%; Negative Predictive Value = d/(c+d) = 100%.

Practical interpretation: “The test result is positive; my patient has the infection.” “The test result is negative; my patient does not have the infection.”

Note that when the test is “gold,” the prevalence of infection does not have an impact on either positive or negative predictive value.

Scenario #2a. The available test is “awesome,” with a Sensitivity as well as Specificity of 90%. The “best educated guess” Prevalence in the patient being tested is about 60%. “My patient more likely than not has SARS-CoV2 infection.”

 Infection PresentInfection AbsentTotal
Test Positive54040580
Test Negative60360420
Total6004001000

Positive Predictive Value = 540/580 = 93.1%; Negative Predictive Value = 360/420 = 85.7%

Practical interpretation: “The test result is positive; the chance of my patient having infection is 93.1%.” “The test result is negative; the chance of my patient having infection is 14.3%.”

Scenario #2b. The “awesome” test is used to detect infection in a patient with “best educated guess” Prevalence of 10%. “It is not uncommon for a patient like this to have SARS-CoV2 infection.”

 Infection PresentInfection AbsentTotal
Test Positive9090180
Test Negative10810820
Total1009001000

Positive Predictive Value = 90/180 = 50%; Negative Predictive Value = 810/820 = 98.8%

Practical interpretation: “The test result is positive; the chance of my patient having infection is 50%.” “The test result is negative; the chance of my patient having infection is 1.2%.”

Scenario #2c. The “awesome” test is used to detect infection in a patient with “best educated guess” Prevalence of 1%. “It is rare for a patient like this to have SARS-CoV2 infection.” (e.g., asymptomatic patient with no known exposure to SARS-CoV2)

 Infection PresentInfection AbsentTotal
Test Positive999108
Test Negative1891892
Total109901000

Positive Predictive Value = 9/108 = 8.3%; Negative Predictive Value = 891/892 = 99.9%

Practical interpretation: “The test result is positive; the chance of my patient having infection is 8.3%.” “The test result is negative; the chance of my patient having infection is 0.1%.” In this scenario, 11 out of 12 positive tests are false positive. It means that for every patient with true infection, eleven additional patients are likely to receive prevention measures (e.g., personal protective equipment, isolation) or medication for treatment (if given) unnecessarily.

Scenario 3. A new “rapid” test is available that gives results in 30 minutes as opposed to the “awesome” test which gives results in 24 hours. The test has a sensitivity of 60% and a specificity of 99%. If this rapid test is followed by the awesome test with the ultimate goal of identifying all patients with infection and the prevalence of infection is 1%, the test will identify 60% of the expected infections in 30 minutes (while generating 16.5 times as many false positives), and the remaining ~40% of the infections in 24 hours (when all the patients with negative result are tested again using the “awesome” test). You can calculate how these numbers play out using this open source calculator. https://www.openepi.com/DiagnosticTest/DiagnosticTest.htm

Summary: The usefulness of a test depends on the purpose it is used for. For the purpose of treating an infection, it may be more important to know that the patient truly has infection (high positive predictive value). For the purpose of preventing transmission of infection, it may be more important to know that the patient truly does not have an infection (high negative predictive value) at the time of admission (rapid results important unless the admission or clinic visit is elective). Positive predictive value of a test increases with increasing prevalence of disease. Increasing the likelihood of positive test result by testing a high risk population will help the cost factor (e.g., use of personal protective gear for elective surgery).

Additional Reading:

1. Epidemiology. Gordis, Leon. 1996. W.B. Saunders Company, Philadelphia, PA.

2. Modern Epidemiology. Rothman, Kenneth J., Greenland, Sander. Second Edition. 1998. Lippincott Williams and Wilkins, Philadelphia, PA.