Diagnostic tests Flashcards
Screening vs. Diagnostic Tests
Screening
- preventative
- measures “healthy”
- sub-clinical signs
- early detection
- early treatment & increased prognosis of disease
- focus: population
- during incubation & latent period
- important to have it pick up signs earlier
diagnostic
- clinical signs showing
- identify disease faster
- measures “sick”
- guide treatment & prognosis of disease
- take more reactive approach
- focus: individual
- from clinical disease evident to outcome (after incubation period)
valid & reliable
narrow curve w/ true value
sensitivity & specificity
inversely proportional
2x2 tables
True disease status: +
Test result: +
true positive (want these!)
- truly sick
True disease status: +
Test result: -
False negative
- not truly healthy
- timing: diurnal variation in what test is measuring
- factors suppressing body’s reaction to pathogen: little production of antibodies, early in subclinical stage
- lab/test error (too early)
True disease status: -
Test result: +
False positive
- not truly sick
- similar disease agent
- previous exposure
- lab/test error
True disease status: -
Test result: -
true negative
- truly healthy
proportion exposed
P = (a+b)/n
proportion diseased
P = (a+c)/n
proportion diseased & exposed
P = a / n
what do we look for in epidemiology
disease risk in exposed group & diseased risk in unexposed group
True prevalence
- actual level of disease present in population
TP = (a+c)/n
apparent prevalence
- what the prevalence “appears to be” if you use this particular test
- positive test results
AP = (a+b)/n
sensitivity
- proportion of diseased that test positive
- never have perfect sensitivity
Sn = a / (a+c)
specificity
- the proportion of nondiseased that test negative
Sp = d / (b+d)
sensitivity => false negatives
- animals that test negative but actually have disease
1 - Sn = % of false negatives
- “Snout” - if test is highly SeNsitive, you can rule out a Negative test result, you can be confident in ruling the disease OUT
- highly sensitive test = few false negatives
- highly infectious diseases that cause serious illness/death
specificity => false positives
- animals that test positive but aren’t diseased
1 - Sp = % of false positive
- “Spin” - if test is highly SPecific, and you get a Positive test result, you can be confident in ruling disease IN
- disease w/ costly treatments, treatments cause suffering, etc.
What if apparent prevalence if higher than true prevalence?
- poor specificity, lots of false positive so non-diseased animals are falsely diagnosed
If true prevalence is higher than apparent
good sensitivity, more false negatives so diseased animals are missed by test
Predictive Value
probability test results being currect depend on Sn & Sp, prevalence
predictive value of positive test (PPV)
probability of TEST POSITIVES are actually DISEASED
PPV = a / (a+b)
- usually aim for 90%
1 - PPV = % false positives (disease free)
predictive value of negative test (NPV)
probability of TEST NEGATIVES are truly negatives
NPV = d / (c+d)
- aim for 90% or higher
1 - NPV = % false negatives (actually be diseased)
Prevalence affects predictive values
prevalence increase, PPV increase, NPV decrease
prevalence decrease, PPV decreased, NPV increase