Screening and Early Detection

Epidemiological Basis for Disease

Control – Fall 2001

Joel L. Weissfeld, M.D. M.P.H.

 Objectives

1. Theoretical justification for disease

control through early detection of
disease.
2. Concepts and quantitative methods used
to evaluate a screening test.
 Definitions
1. Screening program -- comprehensive disease
control activity based on the identification and
treatment of persons with either unrecognized
disease or unrecognized risk factors for
disease.
2. Screening test -- specific technology (survey
questionnaire, physical observation or
measurement, laboratory test, radiological
procedure, etc.) used to help identify persons
with unrecognized disease or unrecognized
risk factors for disease.
 Definitions

3. Primary prevention -- disease control

approach based on the elimination or
reduction of risk factors for disease.
Primary prevention aims to prevent the
occurrence of disease. Primary
prevention may use screening tests to
identify persons with risk factors.
 Definitions

4. Secondary prevention -- disease control

approach based on the active identification
and treatment of persons with unrecognized
disease. Secondary prevention aims to
prevent the occurrence of adverse outcomes
from disease (such as fatal outcomes), without
necessarily reducing the occurrence of
disease. Secondary prevention must screen to
identify persons with unrecognized disease.
 Generalities
1. Screening often implies a public health
related activity involving asymptomatic or
healthy subjects coming from the general
population.
2. Case-finding refers to special clinical
efforts to recognize disease among
persons who consult a health
professional.
 Two essential attributes of
successful secondary prevention
1. Screening must accomplish early detection of
disease. The screening program must detect
disease in asymptomatic persons or in sympto-
matic persons not recognized to have disease.
Relative to background conditions, screening
must identify affected persons at an earlier
time point in the natural history of disease.
2. The act of early detection increases the
effectiveness of treatments for the disease.
Evaluation framework used by US
Preventive Services Task Force

1. Assess burden of suffering.

2. Assess properties of the screening test.
3. Determine the efficacy, effectiveness,
and cost-effectiveness of early detection.
 Important properties of a
screening test
1. Screening is simple, inexpensive, and
easily diffused through the population.
2. The act of screening is safe and
acceptable.
3. The screening test is reliable.
4. The screening test is accurate.
 Reliability
1. Kappa -- an appropriate reliability
measure (or measure of agreement) for
a screening test which gives a categor-
ical result.
2. Intraclass correlation coefficient -- an
appropriate reliability measure for a
screening test, which gives a quantitative
result.
 Cohen’s Kappa
2nd obs
pos neg
1st obs pos a b P1
neg c d Q1
P2 Q2 1.00

Po = a + d = proportion observed agreement

Pe = P1∙P2 + Q1∙Q2 = proportion expected agreement
kappa = (Po - Pe)/(1 - Pe)
Subject X1 X2 0.5*d^2 X1 X2
1 213 189 288,0 Mean 205 202
2 185 194 40,5 SD 56 54
3 193 155 722,0 CV 0,27 0,27
4 162 191 420,5
5 256 277 220,5
6 174 178 8,0 SD(Err) 15,7
7 165 203 722,0 CV(Err) 0,077
8 198 185 84,5
9 228 213 112,5
10 164 134 450,0 Tot var 3001,0
11 217 229 72,0 ICC 0,918
12 195 195 0,0
13 240 217 264,5
14 191 153 722,0 Calculating the intraclass
15 250 203 1104,5 correlation coefficient (ICC)
16 220 214 18,0
17 118 146 392,0
18 165 178 84,5
19 314 303 60,5
20 189 202 84,5 Click for larger picture
21 299 293 18,0
22 186 206 200,0
23 88 83 12,5
24 174 179 12,5
25 330 324 18,0
Sum 5114 5044 6132,0
 Origin of the term
1. Screening implies a simple process, which classifies
persons into one of two groups.
2. Members of the group with a positive (or abnormal)
screening test result have disease with a probability
high enough to justify diagnostic testing (for
confirmation of disease presence) and/or medical
therapies (for prevention of adverse outcomes from
disease).
3. Members of the group with a negative (or normal)
screening test result have disease with a probability
too low to justify diagnostic testing (for confirmation
of disease presence) and/or medical therapies (for
prevention of adverse outcomes from disease).
 Comparison of screening test
results against gold standard
1. A screening test may produce only a
categorical (positive or negative) result.
Important concepts include disease
prevalence, test sensitivity, test
specificity, positive predictive value,
negative predictive value, proportion test
positive, proportion test negative, and
disease yield.
 Classification of screening test
results against gold standard

Screening test results

Positive Negative
Gold standard Positive TP FN
Negative FP TN

TP true positive
TN true negative
FP false positive
FN false negative
 Classification of screening test
results against gold standard
Screening test results
Positive Negative Total
Gold standard Positive a b a+b
Negative c d c+d
Total a+c b+d 1.00

a+b Disease prevalence

a+c Proportion test positive
a Yield
a/(a+b) Sensitivity
d/(c+d) Specificity
a/(a+c) Positive predictive value (PPV)
d/(b+d) Negative predictive value (NPV)
 Comparison of screening test
results against gold standard

2. Sensitivity and specificity are said to be

properties of the screening test. That is,
sensitivity and specificity of a screening
test may be relatively independent of the
population being screened.
 Comparison of screening test
results against gold standard
3. For rare conditions, specificity has a major impact on positive
prediction (Bayes Theorem). Despite a direct effect on yield,
sensitivity has quantitatively less impact on positive
prediction.
 Deriving Bayes theorem
Step 1: Specify the prevalence
(P) of disease

Screening test results

Positive Negative Total
Gold standard Positive P
Negative 1-P
Total 1.00
 Deriving Bayes theorem
Step 2: Use sensitivity (Se) to
distribute test results among the
diseased

Screening test results

Positive Negative Total
Gold standard Positive SeP (1-Se)P P
Negative 1-P
Total 1.00
 Deriving Bayes theorem
Step 3: Use specificity (Sp) to
distribute test results among
the non-diseased

Screening test results

Positive Negative Total
Gold standard Positive SeP (1-Se)P P
Negative (1-Sp)(1-P) Sp(1-P) 1-P
Total 1.00
 Deriving Bayes theorem
Step 4: Determine the proportion
testing positive and the proportion
testing negative

Screening test results

Positive Negative Total
Gold standard Positive SeP (1-Se)P P
Negative (1-Sp)(1-P) Sp(1-P) 1-P
Total SeP + (1-Sp)(1-P) (1-Se)P + Sp(1-P) 1.00
 Deriving Bayes theorem
Step 5: Equate PPV and
NPV with appro-priate
expressions from Step 5

PPV = SeP/[SeP + (1-Sp)(1-P)]

NPV = Sp(1-P)/[(1-Se)P + Sp(1-P)]

 Comparison of screening test
results against gold standard
4. A screening test may produce quantitative values.
However, for decision-making purposes, a
screening program must convert quantitative test
results into two (or more) discrete categories. The
screening program has a choice of different
cutpoint(s), which may be used to categorize test
results. In general, the choice of cutpoint(s) will
affect values for sensitivity and specificity. Analysis
of receiver operating characteristic (ROC) curves
provides a graphical method for studying the effects
of different cutpoints on the performance (sensitivity
and specificity) of a screening test. ROC curve
analysis also provides a method for comparing
competing screening tests.
 Net effects from screening
1. True negatives are exposed to the costs,
inconvenience, and hazards of screening.
True negatives may be reassured by
knowledge of a negative screening test result.
2. False positives are exposed to the costs,
inconvenience, and hazards of screening and
follow-up diagnostic evaluations. The falsely
positive screening test result may cause
psychological and emotional distress.
 Net effects from screening
3. False negatives are exposed to the costs,
inconvenience, and hazards of screening.
False negatives may be falsely reassured by
knowledge of a negative screening test result.
False negatives represent lost opportunities to
prevent adverse outcomes from disease.
4. True positives are exposed to the costs,
inconvenience, and hazards of screening,
follow-up diagnostic evaluations, and
therapeutic interventions. Only true positives
have an opportunity to benefit from medical
therapy.
 Disease pre- Test accuracy Complication? Treatment
Population valence (P) (Se & Sp) Screen Diag outcome?

True positive No No Good

Se No Yes Good
Yes No Good
Yes Yes Good
No No Poor
Diseased No Yes Poor
P Yes No Poor
Yes Yes Poor

False negative No
Screen (1 - Se) Yes

False positive No No
(1 - Sp) No Yes
Yes No
Healthy Yes Yes
(1 - P)

True negative No
Sp Yes
 Generalities

1. Specificity is often the major determinant

of the costs and feasibility of a screening
program.
2. Sensitivity establishes the maximum
extent (upper bound) to which a
screening program is capable of
producing health benefit.