Deciding when a test is useful: how to interpret the jargon
The usefulness of any laboratory test is determined by the clinical context. For example, a study of diagnostic tests
ordered by 87 General Practitioners for over 1200 patients found that when a test was ordered purely for patient reassurance,
approximately 66% of results outside the reference range were interpreted as normal, however, when a test was ordered
to confirm a suspected diagnosis, only 28% of results outside the reference range were interpreted as normal. 11
To determine the likelihood that a patient has a specific condition, based on a test result, the clinician must first
- How likely is it that the patient has this condition? This is termed the pre-test probability, and is based on the
clinical characteristics of the patient, the local prevalence of the diseases being considered, and the clinician’s
- How accurate is this diagnostic test? This is determined by the sensitivity and specificity of the test.
Pre-test probability is defined as the probability that the condition being tested for is the cause
of the symptoms, before a diagnostic test result is known. The pre-test probability helps clinicians to decide whether
it is worthwhile requesting a diagnostic test. This probability may be altered during the consultation as symptoms and
signs are weighted as being “somewhat more suggestive” or “somewhat less suggestive” of the suspected medical condition.
The sensitivity of a test is defined as the proportion of people with the disease who have a “positive”
result (above or below the diagnostic threshold used), i.e. the ability of the test to correctly identify patients with
the condition. Because the number of false-negatives decreases as the sensitivity of the test increases, a highly sensitive
test is useful for “ruling out” a disease if the patient tests negative. Highly sensitive tests, with deliberate use
of an appropriate diagnostic threshold for follow-up, are used when the consequences of missing a particular disease
are potentially very serious, such as for an acute myocardial infarction.
The specificity of a test is defined as the proportion of people without the disease who have a “negative”
result, i.e. the ability of the test to correctly identify patients without the condition. Because the number of false-positives
decreases as the specificity of the test increases, a test with a high specificity is useful in “ruling in” a disease
if a person tests positive. As with sensitivity, the specificity of a test will vary somewhat depending on the diagnostic
Unfortunately, almost no test is perfect with complete (100%) sensitivity and specificity. The choice of what threshold
is used depends on the parameters of the test and what the purpose is when using it. Deliberately setting the threshold
for optimum sensitivity can result in increased numbers of false positives (above or below the threshold) as well, resulting
in reduced specificity. Conversely, in other circumstances optimising specificity may be more relevant, at the cost of
Performing several tests serially increases the overall specificity for detecting a particular disease, with each test
being sequentially more specific than the previous one.
Positive predictive value
The positive predictive value is defined as the probability that a patient with a positive test result really does
have the condition for which the test was requested. Unlike sensitivity and specificity which are independent of the
population being tested, the positive predictive value of a test changes depending on the prevalence of the disease in
the population being tested.
For example, a theoretical ELISA test for HIV may have a sensitivity and specificity of 99.9%. Among 1000 intravenous
drug users with an HIV prevalence of 10%, the test will correctly detect approximately 100 (99.9) people with the disease,
but incorrectly label one person (0.9) without the disease as being HIV-positive. This is a positive predictive value
of 99%. 13 However, in a population of blood-donors (already screened for HIV) the prevalence of HIV would
be much lower, closer to 0.1%. 13 For every 1000 blood-donors screened for HIV the test would correctly detect
one person (0.9) with HIV, but incorrectly label one person (0.9) as being falsely-positive for HIV. In this second population
the positive predictive value of the test falls to 50%. 13
The negative predictive value is defined as the probability that a patient with a negative test result
really is free of the condition for which the test was conducted.
The probability of an abnormal result increases when the number of tests increases
The risk of a healthy individual having a result outside the reference interval increases as the number of tests selected
increases. This is because the normal reference interval for most biochemical tests is defined as being two standard
deviations from the mean of a healthy population. 5 Therefore, an average of 5% of all test results from healthy
patients will fall outside the normal range and be recorded as abnormal (Table 1). 5
False-positive results are more likely when people with a low probability of a condition undergo testing. Although
false positive results can cause significant anxiety to the patient, false-negative results can often have more serious
health consequences. Test results should always be interpreted in the context of other information gained from the clinical
history and physical examination. Results which are borderline need to be interpreted with caution as the inter-test
variability could mean the result is either normal or abnormal, so may need to be repeated after a period of time. If
there is doubt, consultation with a pathologist about the test results can be helpful.
Table 1: Probability of a healthy person returning an abnormal biochemical test result, adapted from
Deyo (2002) 5
|Number of tests
|Probability of at least one abnormal test (%)*
*Assuming each test outcome is independent
An example of pre-test probability, sensitivity and specificity
A D-dimer test can be used in conjunction with the Wells Rule or Primary Care Rule to determine the probability of
a patient having a deep vein thrombosis (DVT). The sensitivity of the D-dimer test is 88% and the specificity is 72%. 12 Because
of the low specificity, D-dimer is most useful as a “rule-out” test for DVT, i.e. a patient with a low or normal D-dimer
level, whose symptoms and signs suggests a low pre-test probability of DVT, is unlikely to have a DVT. A patient with
a high pre-test probability of DVT should be referred for ultrasound irrespective of the results of the D-dimer test.
For further information see: “The
role of thrombophilia testing in general practice” , Best Tests (Mar, 2011).