Kinnary Patel

Screening Exercise

A screening test is used to separate from a large group of apparently wellpersons those who have a high
probability of having the disease under study, so that they may be given a diagnostic work-up and, if diseased, brought
to treatment. A screening test is not intended to be diagnostic; it is designed to rapidly and inexpensively sort out those
who probably have a disease from those who probably do not.

Validity

Validity is the ability of a screening test to give a true measurement of the item being tested, i.e. which individuals have
a disease and which do not. Validity is measured by the sensitivity and specificity of the screening instrument.

Sensitivity is the ability of the test to identify correctly those persons tested who truly have the disease under a study,
and is expressed as a percentage:

Sensitivity = Persons with the disease detected by the screening test X 100
Total number of persons tested with the disease

Specificity is the ability of the test to identify correctly those who do not have the disease. It is also expressed as a
percentage:

Specificity =
Persons without the disease who are negative to the screening test X 100
Total number of person tested without the disease

Measurement of sensitivity and specificity are obtained by comparing the results obtained by the screening test with
those derived from some definitive diagnostic procedure. The extent to which the screening results agree with those
derived from the more definitive tests will determine the sensitivity and specificity, and therefore the validity, of the
screening test.

Before a screening test can be put into operation it must be evaluated. The evaluation of sensitivity and specificity
requires that a true diagnosis for the disease under study be established or ruled out for every person tested by the
screening procedure, regardless of whether the person was found to be positive or negative by the screening test.
TABLE 1

Definitive Diagnosis
Screening Test
Results Diseased Not Diseased Total
 Positive ab a+b
Negative cd c+d

Total a + cb + da + b + c+ d

Note: a =diseased persons found positive on the screening test (true positives)
b =persons without disease, falsely found positive on the screening test (false positives)
c = diseased persons, falsely found negative on the screening test (false negatives)
d =persons without disease found negative on the screening test (true negatives)

Sensitivity (%) = Percentage of persons with the disease = a x 100

who were detected by the test a+c

Specificity (%) = Percentage of nondiseased persons who = d x 100

were negative to the test b+d

False Negative (%) = Percentage of persons with the = c x 100

disease who were not detected by a+c
the test

False Positive (%) = Percentage of persons without the = b x 100

disease who were incorrectly b+d
labelled by the test as having
disease
Positive
Predictive Value (%) = Percentage of persons with positive = a x 100
test who have disease a+b

Negative
Predictive Value (%) = Percentage of persons with negative = d x 100
test who do not have disease c+d

Overlapping theory, screening level, sensitivity, and specificity

As is seen in Figure 6 above, screening test results demonstrate overlapping distributions of an attribute for disease
and nondiseased groups. For some values the distributions overlap and it is not possible to assign individuals with
these values to either the normal or the diseased group on the basis of screening alone. False positives and false
negatives occur in the area of overlap. The degree of sensitivity and specificity may be varied by placing the
screening level at any point in the area of overlap.

Ideally a screening test should establish either the presence or absence of a disease in every individual
screened. Both sensitivity and specificity would then equal 100%. Unfortunately, this is generally impossible to
achieve in practice. The problem is further complicated by the fact that sensitivity is inversely related to specificity,
i.e., if there are few false positives there are likely to be a considerable number of false negatives.

Questions

1.If the screening level for 40-59 year-old women is set at 135 mm Hg, what happens to the sensitivity and specificity
of the test?
If a low cutoff level is chosen then the sensitivity increases. This means the individuals who
have the diseases will be correctly identified. On the other hand, the specificity will decrease,
which means the identification of those who do not have the disease may be identified as
positive by the test.

2.If the screening level is then set at 145 mm Hg, how do the sensitivity and specificity change?
If a higher cutoff level is chosen then the specificity will increase, but the sensitivity will
decrease. Meaning we may miss the individuals with the disease due to the sensitivity being
very low.
3.What general principle does this example illustrate?
This example illustrates that there is a trade-ff between sensitivity and specificity. As stated
in Gordis (p. 94, 2014), if we increase the sensitivity by lowering the cutoff level, we
decrease the specificity; if we increase the specificity by raising the cutoff levels, we
decrease the sensitivity. There is no such thing as free lunch.

The decision on the screening level used is ultimately a subjective judgment about the number of false positives and
false negatives which are tolerable to the population and to the provider of the screening service. This judgment
should, however, be based on the severity of the disease, the cost of the test, the time taken to administer it, the
advantages and probability of success of early treatment, and the cost of further testing (both financial and
psychological for true AND false positives).

Prevalence, sensitivity, and specificity

One important problem in screening concerns the relationship between sensitivity, specificity, and prevalence of
disease. When the prevalence of a disease is low (in the order of 1.0 or 2.0 percent), as it is for many chronic
diseases, most of the population will be free of the disease, and the positive results, even for a highly sensitive and
highly specific test will include a large proportion of false positives. A small decrease in the specificity of a test will
greatly increase the number of false positives, and unless this is offset by a large gain in sensitivity, the proportion of
positives that are false will increase or, at least, remain high. The effect on test results of various screening levels
should be carefully analyzed before a screening project is undertaken, especially when prevalence is low.

With a given sensitivity and specificity, a small increase in prevalence (1.0 or 2.0 percent) results in a negligible
reduction in the number of false positives for the same number of tests, but the yield of new cases increases in
proportion to the increase in prevalence. As a result, the proportion of false positives among persons screening
positive is reduced.

By directing her screening efforts toward high prevalence groups which have been defined by epidemiological
studies, a public health administrator can, in effect, raise the prevalence in the group being screened. For example,
she can limit a diabetes screening program to persons over 40 and make special efforts to screen obese persons and
persons with a family history of diabetes. Screening becomes a two-stage process in which selection of a high
prevalence group is the first stage and application of the screening test is the second.

Questions

4.Why is the two-stage process a useful method?

A two-stage process is useful because it is less expensive and those who screen positive are
recalled for further testing with a greater sensitivity and specificity. This method of testing also
reduces the problem of false positives.
5.The prevalence of unknown diabetes in a population to be screened is approximately 1.5 percent and it is assumed
that 10,000 persons will be screened. The screening will be done using the blood sugar level at non- specified
hours after eating. A measurement of 180 mg percent or higher is considered positive. The sensitivity and
specificity associated with this screening are 22.9 and 99.8 percent respectively.
a.Set up a four-fold table with appropriate numbers in each cell of the table.

+ - Total
+ 34 20 54
- 116 9,830 9,946

Total 150 9,850 10,000

b.Calculate the percentage of false positives: 20/9,850 x 100%= 0.2%

c.Calculate the percentage of false negatives: 116/150 x 100%= 77.3%
6.Let us assume a screening level of 130 mg percent. This level has a sensitivity of 44.3% (and a specificity of
99.0%). Calculate the following values:
a. The % of false positives: 98/9,850 x 100%= 0.99%
b. The % of false negatives: 84/150 x 100%= 56%

7.If you were the director for the diabetes screening program which screening level would you prefer to use? Explain
your choice.
In this circumstance, I would prefer to use the lower cutoff (130 mg percent) level. The
results yielded a lower false negative using the lower cutoff level. The proper identification
of diabetes mellitus is crucial since the disease is associated with lower quality of life and
increase risk for other co-morbidities such as cardiovascular and kidney disease. By lowering
the cutoff level we increase the sensitivity and in this example we also saw fewer false
positives when compared to the higher cutoff level.
Combination of tests

One method for enhancing the sensitivity or specificity of screening is to combine several tests. The combination of
tests may be used in "parallel" or in "series." When used in parallel, screenee is considered positive if he tests
positive to any one of the tests. In series, the screenee must be positive to each successive test to be considered
positive; if he is negative to any of the tests he is considered to have screened negative.

A combination of tests in series enhances the specificity of the testing but reduces the sensitivity, while the
combination of tests in parallel reduces the specificity but enhances the sensitivity. When used in parallel, some
diseased persons are detected by one test but not the other, resulting in detection of a higher proportion of diseased
persons by the combination of tests.

Consider the hypothetical data presented in Table 2, assuming the test population to have been screened for
diabetes using a blood sugar test and a urine sugar test.
Table 2: Test Results

Diabetic Not Diabetic

Positive to urine test, negative to blood 7 3
Positive to blood test, negative to urine 23 11
Positive to both tests 45 7
Negative to both tests124 7,620

TOTAL199 7,641

SOURCE: Thorner and Remein, 1961

8.Calculate the sensitivity and specificity for:

 a. only blood sugar test
(Sensitivity) 23+45/199 x 100%= 34.17
(Specificity) 3+ 7,620/7,641 x 100%= 99.8%
b. only urine sugar test
(Sensitivity) 7 +45/199 x 100%= 26.1%
(Specificity) 11+ 7,620/7,641 x 100% = 99.9%
c. in parallel, both urine sugar and blood sugar done
(Sensitivity) 7+23+45 / 199 x 100%= 37.7%
(Specificity) 7,620/7,641 x 100= 99.7%
d. in series, urine test done and then blood test
(Sensitivity) 45/199 x 100% = 22.6%
(Specificity) 3+11+ 7,620/7,641x 100%= 99.9%

9. How do the sensitivity or specificity of each combination of the tests change when compared with single test?
The sensitivity increases in the parallel test in comparison to the single tests. The specificity is
about the same in the parallel test in comparison to the single test. On the other hand, the series test
decreases the sensitivity, whereas, the single test sensitivity increase. However, the specificity is
about the same when comparing the series and single test.

10.Again, assuming you are the harried director of the diabetes screening program, which test or combination of tests
would you prefer to use?
I would use the series test because a succession of tests must all be positive for the individual to
be considered diabetic. I think this is important because when diagnosing a disease such as
diabetes there are clear clinical manifestation and diagnostic results that determine if the individual
has the diseases or not. Sure, this method would take a longer time because multiples tests are
needed but I think this results in greater patient outcomes and greater quality of life for patients.

11.State your opinion on the ethics involved in screening where your test yields a high percentage of false positives:
a.for breast cancer
This goes against the ethical principle of beneficence. We would be doing harm to the
patient because if we incorrectly diagnose an individual with breast cancer because they would
require invasive treatment that is costly. This would take a toll on their psychological health
because coping with cancer is extremely difficult.
b.for adult-onset diabetes
Typically the treatment involves lifestyle modifications such as diet and exercise. Any
further examinations or interventions would not be invasive to the patient and less likely to do
harm.
One factor which must be considered before starting screening programs are the justifiability,
acceptability, and efficiency of the proposed test. First, is the test justified, scientifically and financially, the resulting
benefit to the community? Each proposal for screening is a scientific hypothesis that earlier diagnosis makes the
present therapies more efficient.

Example: Porphyria variegata is a rare inherited abnormality of the liver with an estimated prevalence of about 4 per
million. If subjects with this abnormality take barbiturates or sulphonamides they run the risk of death by
paralysis. Latent cases can be discovered by a biochemical test using feces and costing about two dollars. If such
patients are warned not to take barbiturates and sulphonamides the incidence and the mortality of acute attacks can
be drastically reduced. Let us assume that an attack would occur in only one of four latent cases, and that of those
experiencing an attack only one in four is fatal.

12.What is the cost of saving the life of a patient with porphyria variegata by population screening?
1 in 16 is fatal. We have to screen 4 millions people to save one life. $2/ tests x 4 millions = $ 8 million
to save one life.

13.Is it worth it?

Unfortunately, this screening is not worth it. Through a cost-benefit analysis perspective this would
not make sense financially.

14. How efficient is the proposed screening test as a method of measurement?

This test is not efficient at all. We would require so much manpower and money to save a few lives. In the
grand scheme of funds distributed in healthcare this would not be feasible as it is not efficient.

A test used in screening must be relatively simple and must be acceptable to the subjects. A cervical smear is more
acceptable than sigmoidoscopy for example. The test should give a true measurement of the attribute under
investigation (accuracy) and provide consistent results in repeated trials (precision, sometimes called repeatability or
reliability).