Sc re eni ng a nd Dia gnos tic Te st

Definition of Screening:
Th e pre sump tiv e ide ntificatio n of unrecogniz ed disease or def ect by th e app lica tion of te sts, exa mina tions or other proce dure s which ca n be ap plie d rapi dly.

Screening program

Goa l: secon da ry p reven tion

det ec t di sea se befo re clini cal poi nt for cure or impr ov ed ou tcome ge t peop le with di se ase int o appr op ri ate trea tm ent

1. Design of screening test
  

gol d stan dard Se lect ion of objecti ves Samp le siz e
cas es:
2 zα Se(1 − Se) n1 = d2

co nt rols

2 zα Sp (1 − Sp ) n2 = d2

Asse ssing rel iab ility

th e

validity

an d

2. Assessment of Screening Test
  

valid ity reliabil ity predictive valu e

True value and test results
(True value: gold standard)
True results in the populations No disease b (false positive) d (true negative) b+d

Screening test results Positive Negative Total

Disease a (true positive) c (false negative) a+c

Total a+b c+d a+b+c+d

(1) Validity of screening test
The abi lity of a tes t to disting uish bet ween who has a disease an d wh o does not have th e di sease.

Two m ain comp on en ts: sen sitiv ity an d sp ecif icity

Sensitivity (Se): Pro portion of positiv e test wh o have th e disease. Se = a / (a+c )

Specificity (Sp) : Pro portion of neg ative test wh o do n ot h ave th e diseas e. Sp = d / (b+ d)

False positive rate (FP): FP = b / (b +d) = (1 - S p)

False negative rate (FN):
FN = c / (a +c) = (1 Se )

+ LR =   = 1 − Sp  b /(b + d ) 

Youden’s index (YI): YI = (Se + Sp) Likelihood ratio (LR):  a /(a + c ) Se

- 1

 c /(a + c )  1 − Se − LR =   = Sp  d /(b + d ) 

Agreement rate:
a+d Agreement rate = ×100% a+b+c+d

Diagnostic index (DI):
a d DI = Se + Sp = + a+c b+d

Diagnostic availability (DA):
DA = 2( Se 2 + Sp 2 ) − 1

ROC c urve a nd i ts applic ati ons

ROC

cur ve (Recei ve

Oper at or

ca n show the rel ati ons hi p bet we en sens it ivi ty and spec ifi city i nt ui ti vel y.
Char act er isti c cu rve)

Ap pli cation s:
 

Se lect ion of cut off value (cu t poi nt ) Diag no st ic val ue co mpar is on amo ng se ver al di agno st ic tes ts

(2) reliability ( Repeatability )

The test rep ea table.

is

reli abl e

or

Intra-subject

(var iati on wi thin i ndivi du al sub jec ts)
Va riance , C. V. etc . sta nda rd dev iati on ,

variation

Inter-observer variation (var iati on
between resu lts)

th ose rea di ng th e test

agr ee men t, Kap pa val ue

Dr. A Positive Dr. B Positive a agreement Negative c Agreement

Negative b d agreement

= (a+d) / (a+b+c+d)

Ka ppa va lue: (sel f readi ng )
Kappa = 2(ad-bc)/(a+b)(b+d)(a+c)(c+d)

(3) predictive value (PV)

Positive PV (+ PV): The likel ih ood th at a pati ent wi th a pos itive test resu lt really has th e diseas e. + PV = a / (a+b) Negative PV (- PV) : The likel ih ood th at a pati ent wi th a nega ti ve test result really does not ha ve th e di sea se. - PV = d / (c +d)

PV and disease prevalence

P : disease prevalence

3.Two-stage screening

To increase sen siti vity sp ec ifi city : Two-stage screening

or

Series Test: mo re sp ec ifi ci ty ( eg. di abe tes, HIV inf ect ion ) Parallel Test: mor e sensi tivi ty ( eg . blood do no r sc ree ni ng )

Multiple test multiple test in parallel Test 1 + + + + Test 2 + + + Not necessary Result + + + + -

in series