ACCURACY OF TEST

BY:
NAZIHAH MOHD YUNUS
NOORUL HUDA MOKHTAR
NUR ASYILLA
WAN HASMUNI
 CONTENT
• INTRODUCTION
• SENSITIVITY
• SPECIFICITY
• PREDICTIVE VALUE
• ROC CURVE
• CLINICAL SIGNIFICANCE
 INTRODUCTION
What is accuracy of test?
PART OF THE QUALITY CONTROL IN THE
DIAGNOSTIC LABORATORY
- measures or procedures taken by the
laboratory to recognize and minimize errors so as to
produce test results that are precise and accurate
enough to be reliable and useful .

SENSITIVITY,SPECIFICITY,
PREDICTIVE VALUE
 SENSITIVITY
• Analytical sensitivity:
- measures low concentration of analytes

• diagnostic sensitivity ( true positive rate )

- ability of a test to detect disease and is expressed as the
proportion of person with disease who have the positive
result

TP X 100 = SENSITIVITY (%)

TP + FN
 SPECIFICITY
• Analytical specificity
- not subject to interference by other substance
- related to accuracy

• Diagnostic specificity
- ability of the test to detect absence of disease and is
expressed as the proportion of persons without disease who
have the negative test
TN X 100 = SPECIFICITY (%)
FP + TN
Calculating sensitivity and specificity
from a 2x2 table
Truly have disease
+ -
Screening test
+ a ( TP ) c ( FP )

- b ( FN ) d (TN )

a+b c +d
a Among those with true
Sensitivity  disease, how many test
a b positive?

d Among those without the

Specificity  disease, how many test
cd negative?
 PREDICTIVE VALUE THEORY
 Predictive value for positive test
-The probability that the patient with a positive test result has
the disease.
TP X 100 %
TP + FP

 Predictive value for negative test

- The probability that the patient with a negative test result does
not have the disease.
TN X 100%
FN + TN
 Depends on the characteristics of the test (sensitivity, specificity ) and the
prevalence PV+ = [ prevalence . Sensitivity] x 100%
[prevalence.sensitivity]+ [(1 – prevalence)(1-speciificity)]
Calculating PPV and NPV from a 2x2
table
Truly have disease
+ -
Sceening Test
+ a ( TP ) c ( FP )
a+c
- b (FN ) d ( TN )
b+d

Among those who test

a positive, how many truly have
PPV 
ac the disease?

d Among those who test

NPV  negative, how many truly do
bd not have the disease?
 Relationship between terms
CONDITION
(AS DETERMINED BY “GOLD STANDARD”)
Positive Negative

Positive → Positive Predictive

TEST True Positive False Positive Value (PPV)

OUTCOME
→ Negative Predictive
Negative False Negative True Negative Value (NPV)

↓ ↓
Sensitivity Specificity
 Hypothetical Example
Breast cancer (on biopsy)
+ -
Mammography
+ 9 109

- 1 881

10
990
Sensitivity=9/10=0.90 1 false negatives out of 10
cases ( 1 – sensitivity)

Specificity= 881/990 =0.89 109 false positives out of 990

(1 – specificity )
 Hypothetical Example
Breast cancer (on biopsy)
+ -
Screening test
+ 9 109 118
- 1 881 882

PPV=9/118=7.6%
NPV=881/882=99.9%
Prevalence of disease = 10/1000 =1%
 What if disease was twice as
prevalent in the population?
Breast cancer (on biopsy)
+ -
Screening test
+ 18 108

- 2 872

20 980
sensitivity=18/20=.90
specificity=872/980=.89
Sensitivity and specificity are characteristics of the test, so they don’t
change!
 What if disease was more
prevalent?
Breast cancer (on biopsy)
+ -
Screening test
+ 18 108 126
- 2 872 874

PPV=18/126=14.3%
NPV=872/874=99.8%

Prevalence of disease = 20/1000 =2%

A worked examples:
The fecal occult blood ( FOC ) screen test was used in 203 people to look for
bowel cancer.

Patients With Bowel Cancer

(As Confirmed On Endoscopy)
Positive Negative
Positive TP=2 FP=18 → +Ve Predictive
Value
FOB test = TP/(TP+FP)
= 2/(2+18)
= 2/20
=10%
Negative FN=1 TN=182 → -Ve Predictive Value
= TN/(FN+TN)
↓ ↓ = 182/ (1+182)
Sensitivity Specificity =182/183
= TP/(TP+FN) =TN/(FP+TN) =99.5%
=2/(2+1) =182/(18+182)
=2/3 =182/200
=66.67% =91%
 Efficiency
• Percentage of the time that the test give the
correct answer compared to the total number
of test.
TP + TN X 100%
TP + TN + FP + FN
EFFECT OF ALTERING CUTOFF
 Effect of altering the test cutoff

• When a cutoff is altered there is an inverse relationship

between sensitivity and specificity.
• Increased specificity will result in decreased sensitivity,
that is more false negatives.
• Increased sensitivity will result in decreased specificity,
that is more false positives.
• Then, “What cutoff should be used?”
• Receiver operator characteristic (ROC) curves will help.
Receiver operator characteristic (ROC)
curves
 The ROC curve is a graphic representation of the varying
sensitivities and specificities that are possible by varying
cutoff of a test.
 It is a plot of the sensitivity versus 1-specitivity ( FP rate).
 The ability of a test, using a specific analytes concentration,
to discriminate disease from non disease.
 A test with perfect discrimination ( no overlap in the 2
distributions) has a ROC plot that passes through the upper
left corner (100% sensitive, 100% specificity).
 Therefore, the closer the ROC plot is to the upper left corner,
the higher the overall accuracy of the test. ( Zweig & campbell, 1993)
-represent the cuttoff limit value or decision limit
• The diagnostic performance of a test, or the accuracy of a test
to discriminate diseased cases from normal cases is evaluated
using ROC curve analysis. (Metz,1978 ; Zweig & Campbell, 1993).

• ROC curve can also be used to compare the diagnostic

performance of two or more laboratory or diagnostic tests.
( Griner et al,1981)
 Receiver operator characteristic (ROC)
curve
Sensitivity •The point closest to the upper left
corner represents the cutoff with
maximum sensitivity and specificity
and is generally selected.
•Method A is superior to Method B
Method A
since it always has higher sensitivity
Method B and specificity than Method B.
•Method C cannot be used to
Method C discriminate the disease and non-
disease states at all.

1-specificity
 CLINICAL SIGNIFICANCE
Determination of an optimal cutoff value or medical decision limit
for clinical laboratory test –
- help clinicians make choices regarding diagnosis, follow-up and need for
adjunct diagnostic testing
Eg : low medical decision limit used for CK-MB, the diagnostic sensitivity of the test
may approach 100% for Dx of MI (few or no FN results),diagnostic specificity may
decrease to a range of 50-60% ( a large no. of FP results).

The perfect test is both sensitivity and specificity of 100% and a diagnostic efficiency of
100%.

Evidence –based clinical biochemistry

- ideally, test should be chosen on the basis of evidence of their
utility, and their results used on the basis of outcomes measures
 CONCLUSION
• Specificity, sensitivity and predictive values of the test are
tools to ensure if the data are accurate and precise and thus a
reliable result can be produced that the clinician can use in
making medical decision.
• Positive predictive value increases with increasing prevalence
of disease.
• However, in practice, the clinician will combined clinical
information and often results of several investigation to make
the diagnosis.
 References
 Lawrence A.Kaplan, Amadeo J. Pesce & Steven C.Kazmierczak: Clinical
Chemistry,theory,analysis,correlation:Reference Intervals and Clinical
Decision limits, 4th edition, Mosby,

 William J. Marshall & Stephen K Bangert: Clinical Chemistry:

Biochemical investigations in clinical medicine, 6th edition, Mosby,
2008.

 Carl A. Curtis & Edward R. Ashwood : Tietz Fundamentals of Clinical

Chemistry : Evaluation of Methods,5th edition,Saunders

 From free Wikipedia: sensitivity, specificity and predictive value;

MedCalc, ROC curve analysis; (date: 10 July 2010)
 Sensitivity 1-Specificity
= True positive rate = false positive rate
= TP/(TP+FN) = FP/(TN+FP)

Disease presence Disease absence

Test positive True positive (TP) False positive(FP)
Test negative False negative (FN) True negative (TN)

1-Sensitivity Specificity
= False negative rate = True negative rate
= FN/(TP+FN) = TN/(TN+FP)
 Predictive value
• Given the sensitivity and specificity of a test, prevalence of
a disease, construct the “truth” table and derive the
predictive value formula.

Disease presence Disease absence

N x (Prevalence) x (Sensitivity) N x (1-Prevalence) x (1-Specificity)

Test positive
Test negative
N x (Prevalence) x (1-Sensitivity) N x (1-Prevalence) x (Specificity)

N x (Prevalence) N x (1-Prevalence) N