Page 99 - Journal of Special Operations Medicine - Spring 2017

Positive and Negative Predictive Values inexpensive and the results come back quickly. It has a
high SN, so if the result is negative, you can effectively
Positive predictive value (PPV) is the term used rule out that your team member contracted HIV. If the
to refer to the chance the test correctly identified tests come back all negative, that is great; however, the
someone who truly has the disease. ELISA test does not have a very high SP, which results
in a percentage of FPs. If you are going to tell someone
TNs that he or he has a serious illness, like HIV infection,
NPV = you want to be very certain of the diagnosis. Therefore,
(TNs + FNs)
a secondary confirmatory test is used. This test is the
Negative predictive value (NPV) refers to the Western blot, which has a very high degree of SP and
chance that a negative test truly means the person can rule in a disease (Sp-in).
does not have the disease.
In your case, you screen 10 members on your team and
TPs one has a positive ELISA test result. You have ruled out
PPV =
(TPs + FPs) infection in nine members. You perform a Western blot
test on the one positive member, and it returns negative.
You explain the results to him and verify the tests and
successfully confirm he has not been infected.
Every test and even every physical examination tech-
nique has an SN and an SP. Knowing these concepts lets
you interpret the test to determine if it can confirm (rule SN and SP are important concepts but are a little ab-
in) or refute (rule out) a diagnosis. In some cases, a test stract for day-to-day clinical medical decision-making.
or an examination will not be helpful and you could What you really want to know when evaluating a pa-
consider an alternate diagnostic algorithm. tient is, if the test is positive, what percentage of people
will have the disease. Or, if the test is negative, what
percentage of people will truly not have the disease. This
Let’s look at a medical example.
concept can be express by using predictive values. This
is graphically calculated in Figure 1.
Case Presentation: Blood Exposure
The problem with PPV and NPV is that these numbers
Your team responds to a mass casualty incident in can vary depending on the prevalence of the disease,
which many people are injured. The team pitches in to whereas SN and SP do not vary with the prevalence of
provide medical care and triage the casualties. After the disease.
situation is stabilized, you realize that many of your
team members were exposed to blood from these vic- Prevalence is defined as the percentage of those with the
tims. You as the medical clinician on this team have disease divided the total population.
a responsibility to make sure all your team members Prevalence = all positive cases (TP + FN)/total population
are cared for. One of your concerns is HIV exposure.
You provide postexposure prophylaxis (a topic for an- If your head is swimming after reading this, you are not
other time) but need to screen your team members over alone. Let’s look at a real-world example.
the next few months to ensure that they have not con-
tracted the infection. You are deployed to a Third World location during an
outbreak. You have a test to detect the illness. The test
There are a few different tests to help with this screen- has a 90% SN and 94% SP.
ing. One strategy is to use an initial screen test with a
high SN to rule out infection (Sn-out). This is an ELISA During the outbreak, you test 1000 people. The results
(enzyme-linked immunosorbent assay) test. It is fast and of their test follow.

Figure 1 Test Results of Theoretical Metal Detector (Weapon Confirmed by Search)
Had a Weapon Did Not Have a Weapon
(Positive)
(Negative)
(Positive)
Detector Metal (Negative) TP TN NPV = TN/(FN + TN)
PPV
FP
PPV = TP/(TP + FP)
NPV
FN
Sp
Sn
Sn = TP/(TP + FN)

Basic Biostatistics and Clinical Medicine Sp = TN/(FP + TN) 77