Page 98 - Journal of Special Operations Medicine - Spring 2017

An Ongoing Series

Basic Biostatistics and Clinical Medicine

Joshua Banting; Tony Meriano, MD

OBJECTIVES

In this column of Clinical Corner, we are going to talk Consider that you want to develop a test, a nonmedical
about a higher-level concept and cover some basic bio- test. You are trying to build a metal detector to screen
statistics related to clinical medicine. It is not as exciting people for weapons. It sends “rays” to screen a person
as penetrating trauma, but it is essential for the elite- for metal, which might indicate a weapon. We want to
level clinician to understand because these concepts are detect all the weapons that people have but we do not
critical to taking your medical decision-making to the want the machine to falsely identify zippers on clothes
next level. or coins in pockets. If we set the machine’s rays too low,
then we might not detect the metal present in a weapon.
It is common for inexperienced medical personnel to If we set it too high, it will beep constantly and create
think of medical tests to be binary, being either posi- huge waiting lines because it is detecting items that are
tive or negative. In actuality, it is not so simple. Even not weapons.
a simple qualitative test like a colormetric urine preg-
nancy test is more complicated than this. It is true that When a person walks through the scanner, four possibili-
it can be positive or negative, but it must have a cutoff. ties can occur. The machine can beep and we can find a
This cutoff is the amount of detectable human chorionic weapon; this is called a TP. The machine could also beep
gonadotropin (hCG) in the urine that is able to change but the person does not have a weapon; this is called
the color of the strip. The manufacturer sets this cutoff an FP. The machine could clear someone who does not
to provide the highest yield, and this yield is defined by have a weapon, which is a TN. However, if someone
the terms “sensitivity” and “specificity.” gets through with a weapon without the machine detect-
ing the weapon, that would be called an FN. These four
possibilities can be represented in a grid (Figure 1). Ulti-
Sensitivity and Specificity
mately, to calculate SN and SP, you need a secondary test
to confirm the results of the first test. In our theoretical
Sensitivity (Sn) measures how often the test will example, this would be a pat-down search. This type of
be positive in a person with the disease. More for- confirmatory test is often called the gold standard.
mally, it is a measure of the number of true posi-
tives (TPs) of the total number of diseased persons This is SN and SP: super high SN will detect all metal
including those whom the test identifies as falsely but every person would be pulled out. Super high SP will
negative. not accidentally identify spare change as a weapon but
TPs might miss a true firearm. All ideal tests, even medical
Sn = ones, have a combination of high SN and high SP.
(TPs + false negatives [FNs])
Tests with a high degree of SN can be used (with your
Specificity (Sp) measures how often a test will be clinical judgment) to rule out a medical condition. For
negative in a person without disease. It measures example, a serum hCG has a high SN. If it is negative,
the number of TNs correctly identified by a test it is most likely that a woman is not pregnant. This is
of the total of disease-free persons including those commonly written as sensitivity rules out, or “Sn-out.”
whom the test identifies as falsely positive.
TNPs Tests with a high degree of AP can be used (with your
Sp = clinical judgment) to rule in a medical condition. For
TNs/(TNs + false positives [FPs]) instance, a positive Western blot test for HIV infection
is highly specific. This is sometimes written as “Sp-in.”

76