Use medical tests judiciously
Use medical tests judiciously: Excerpt from Avoiding Common Pediatric Errors
Author:
Madan Dharmar, MD
What to Do - Gather Appropriate Data,
Interpret the Data, Make a Decision
Using the appropriate test or study for a particular disease helps to optimize
the risk-benefit profile of the diagnostic testing and improve the diagnostic
capabilities of the test.
Medical tests can be classified into two types: screening tests and diagnostic tests. Screening tests are usually applied in the community to identify
a disease early and thus able to appropriately intervene as a measure of secondary prevention; and they are also used as a basis for primary prevention.
Diagnostic tests are applied to an individual in the clinical setting to identify
and provide effective health care to that individual. The results of the tests
could fall into one of the four groups:
• True positive (TP): when the patient's test is positive and is diseased
• False positive (FP): when the patient's test is positive but not diseased
• Truenegative(TN): whenthepatient'stestis negative andis not diseased
• False negative (FN): when the patient's test is negative but is diseased.
An ideal test is one that, when applied to an individual or a population,
can separate patients into two groups (TP and TN). In reality, the majority
of tests are far from ideal and are influenced by many factors, which affects
the way they separate the population into groups. Hence, it is important to
know how the test performs before interpreting the test's results.
The first step in understanding a test is to assess the ability of the test to
correctly diagnose the diseased and nondiseased. Sensitivity and specificity
are test characteristics that help to access the ability of the test to make that
distinction in an individual.
• The sensitivity of a test is the ability of the test to correctly identify the
diseased. It could also be defined as the proportion of TP among the
diseased [TP/(TP+FN)].
• The specificity of a test is the ability of the test to correctly identify the
individual as not diseased. It could also be define as the proportion of TN
among the not diseased [TN/(TN+FP)].
Toillustrate withan example, letusconsiderapopulation of 100people,
of whom 20 are diseased and 80 are not diseased. Let us say that when
each of the individuals was tested using Test A, we found that 18 were
correctly identified as diseased (TP) and that 64 were correctly identified as
not diseased (TN).
Sensitivity of the test = TP/(TP + FN) = 18/20 = 0.90 (90%)
Specificity of the test = TN/(TN + FP) = 64/80 = 0.80 (80%)
It is important to understand that tests may be chosen based on the
situation it is being applied to the individual. It is essential to understand the
impact of false positive and false negative before choosing a test. A positive
result in a screening test may need to be followed with more tests to confirm
the disease, which could be burdensome to the health system; and it could
cause anxiety and worry in the person who had been told that the test was
positive when he or she was, in fact, not diseased. A person who has been
diagnosed as positive based on false-positive results from a screening test
may never be able to remove that label, even when it was later found to be
negative on subsequent evaluation. Similarly, consider the impact of a false-
negative test result in a diseased individual, especially with a serious disease,
which, when diagnosed in a timely manner, could affect management of that
disease (e.g., early stages of cancer).
Sensitivity and specificity of a test provides the probability of the test
resultsinthepresenceorabsenceofdisease,butaphysicianismoreinterested
in knowing the probability of disease in the presence of positive or negative
results from the test. That is, if the test is positive for an individual, then
what is the probability that the individual is truly positive? And similarly,
if a test is negative for an individual, then what is the probability that the
individual is truly negative?
PredictiveValueandPrevalence. Positivepredictivevalue(PPV)isdefined
astheproportionofTPamongthetest-positiveindividuals[TP/(TP+FP)].
Negative predictive value (NPV) is the proportion of TN among the test
negative individuals [TN/(TN+FN)]. Using the sample example above,
PPV = TP/(TP + FP) = 18/24 = 0.53
NPV = TN/(TN + FN) = 64/66 = 0.97
Prevalence of a disease is defined as the number of cases of a disease
presentinapopulationatthatspecifictime dividedbythenumberofpersons
inthepopulationatthattime.Itistheprobabilityofdiseaseinthepopulation.
The predictive value of a test is dependant on the prevalence of the disease
in that population. To illustrate this let us go back to the example above. The
prevalence of disease in the population is 0.20 (20/100) and we know that
the predictive values are 0.53 and 0.97 (positive and negative, respectively).
Usingthesametest,(sensitivity= 90%andspecificity= 80%)letuscalculate
the predictive values for population with prevalence of 0.4 (40/100).
Disease Status
PPV = TP/(TP + FP) = 54/62 = 0.87
NPV = TN/(TN + FN) = 32/38 = 0.84
The PPV increased from 0.53 to 0.87 when the same test was applied
to a population with greater prevalence of the disease. When the prevalence
of a disease is very low, then the PPV of the test will not be even close to 1,
even when the test is highly sensitive and specific. When the test is used to
screen apopulation with low prevalence, it is inevitable that many people will
have false positive results. This shows that test results must be interpreted
only after taking into consideration the prevalence of the disease in that
population.
Most of the tests during development are evaluated under ideal conditions (experimental condition): sensitivity being determined by testing on
diseased individuals, and the specificitydeterminedbytesting on individuals
who do not have the disease. Therefore, test specifications may vary when
the test is used in real-world conditions and applied to a different population with different prevalence of disease. Prevalence affects the positive
and negative predictive value of the test. Prevalence can be considered the
pretestprobabilitythat theindividualcouldhave the disease, andthepositive
and negative predictive values (PPV and NPV) of the test are the revised
estimate for the individuals who are positive or negative on the test, and
are post-test probabilities or posterior probabilities. The positive (PPV) and
negative (NPV) predictive value can be calculated for different population
prevalence for a test by,
PPV =
sensitivity × prevalence /
(((sensitivity × prevalence) + ((1 - specificity) × (1 - prevalence)))
NPV =
specificity × (1 - prevalence) / (((1 - sensitivity) × prevalence) + (specificity × (1 - prevalence)))
The usefulness of the test can be assessed by the difference between the pre-
and posttest probabilities.
In a clinical setting, the physician, following his or her initial evaluation,
arrives at a probability (pretest probability or prevalence) of that individual
being diseased. Based on the initial assessment, the physician could either
decide the individual is not diseased, or test the individual further for the
disease, or treat the individual for the disease. This approach was described
by Pauker and Kassirer in their article about clinical decision making. The
pretest probability for any individual can fall into the category of: (a) no
disease, (b) may have the disease but need to test, and (c) has the disease.
The medical test is used to enable the physician to arrive at a diagnosis. If
a test is used in an individual with a low pretest probability, there is greater
chance for a false positive result, that could in turn result in a misdiagnosis
for the individual. It is essential to avoid testing an individual when there is
a greater likelihood that one or more of the results could be a false positive.
The judicious use of medical testing based on the sensitivity, specificity, and
predictive values of the test in important.
In conclusion, when choosing or interpreting a test result in an individual we need to consider:
• The pretest probability of the disease in the individual.
• Sensitivity and specificity are specific characteristics of a test, which could
help to assess quality of the test.
• Predictive values of a test could help us to assess how good the test is at
identifying diseased patients correctly.
• Impact of false-positive and false-negative results of a test.
Suggested Readings
Altman DG, Bland JM. Diagnostic tests 2: predictive values. BMJ. 1994;309(6947):102.
Gordis L. Epidemiology. 3rd ed. Philadelphia: Saunders; 2004.
Pauker SG, Kassirer JP. The threshold approach to clinical decision making. N Engl J Med.
1980;302(20):1109–1117.
Pictures
Book Source Details
- Book Title: Avoiding Common Pediatric Errors
- Author(s): Anthony D Slonim MD, DrPH; Lisa Marcucci MD
- Year of Publication: 2008
- Copyright Details: Avoiding Common Pediatric Errors, Copyright © 2008 Lippincott Williams & Wilkins.
More About Drug Allergies
More Medical Textbooks Online about Drug Allergies
Review other book chapters online related to Drug Allergies:
Medical Books Excerpts
- Anaphylaxis
- "Professional Guide to Diseases (Eighth Edition)" (2005)
- [ read ]
Copyright notice for book excerpts: Copyright © 2008 Lippincott Williams & Wilkins. All rights reserved.
|
|
More About This Book:
Title: Avoiding Common Pediatric Errors
Authors: Anthony D Slonim MD, DrPH; Lisa Marcucci MD
Publisher: Lippincott Williams & Wilkins
Copyright: 2008
ISBN: 0-7817-7489-6
|
|
» Next page: Surveys relating to Drug Allergies
Rate This Website
What do you think about the features of this website?
Take our user survey and have your say:
Website User Survey
Medical Tools & Articles:
Next articles:
Tools & Services:
Medical Articles:
Forums & Message Boards
- Ask or answer a question at the Boards:
