Assessment of Screening
Test
Epidem iology in Medical
Pr act ice
By DJP Bar ker and G Rose
Two basic components
Repeat abilit y / Reliabilit y
Whet her single m easur em ent can guide act ions
Poor r epeat abilit y Æ r epeat ed m easur em ent s
Validit y
Scr eening t est usually cheap, sim ple, and quick
Ther efor e quest ionable validit y
Validity: two types of errors
False posit ive
Needless alar m
Wast ed inv est igat ion
F + is high w hen pr evalence is low
Br east cancer : Se 80% , Sp 96% , F+ : 75%
False negat ive
Neglect t he needy
Diagnostic criteria to assess screening
test
Repeat abilit y / r eliabilit y
Validit y
Pr edict ive values
Posit ive pr edict ive value
Negat ive pr edict ive value
Sensit ivit y / specificit y
Refer ence t est
+ -
Scr eening Test
+ Tr ue + ( a)
False+ ( b)
PV+ a/ a+ b _ False –
( c)
Tr ue – ( d)
PV- d/ c+ d Se
a/ a+ c
Sp d/ C+ d
ROC
Pr evalence %
Predictive value
A scr eening t est w it h 95% se and sp.
Relat ionship bet w een disease
pr evalence and t he pr edict ive values
Sensitivity and specificity
Se
Pr opor t ion of t r ue posit ive cor r ect ly ident ified
Hi Se is necessar y w hen false – is ser ious.
Sp
Pr opor t ion of t r ue negat ive cor r ect ly ident ified
Hi specificit y is necessar y w hen false + is undesir able.
Se Æ r educe F- Æ incr ease F+
Sp Æ r educe F+ Æ incr ease F-
Balance: a m at t er of j udgem ent
Evaluation of a screening tes
Cheap
Safe
Cost of t he t est ( consider ing t he t hr eat of t he disease)
Tim e lost
Discom for t
Anxiet y
Com pet e r esour ces
Yield
The num ber of t est s done t o ident ify one case
Low pr evalence Æ poor yield Æ scr een hi r isk gr oup
Balance above fact or s