slide deck 6, exam 2

screening

screenings

the use of quick, simple and inexpensive testing procedures to examine apparently healthy people and determine who among them probably has a given disease and who probably does not have the disease

goal of screening

secondary prevention of disease

to stop or slow down the progression of a disease in the earliest stage possible

positive screening tests…

are typically followed by a diagnostic test which can be lots of time, money and distress

negative screening tests…

are not typically followed up

but what if it is inaccurate??

examples of cancer screenings

mammogram, pap smear and fecal occult blood test (all follow diagnostic test of a biopsy)

examples of heart disease and diabetes screenings

lipid profile test, blood pressure test and fasting plasma glucose test (none of these have a diagnostic test)

screening examples for new borns

pulse oximetry and PKU test (diagnostic tests include echocardigrams’s and blood analysis)

example of infectious disease screening

tuberculosis test (diagnostic follows a chest xray)

true positive (TP)

the screening test is positive and the disease is present

true negative (TN)

the screening test is negative and the disease is absent

False positive (FP)

the screening test is positive but the disease is absent

false negative (FN)

the screening test is negative but the disease is present

what is the tradeoff of screening tests

they generally have more false positives and false negatives versus a diagnostic test

dichotomous results

either a clear yes or no answer with no interpretation

there are unimodal and bimodal curves

example is pregnancy tests

continuous variable test results

the test is designed so that a result is a value

large range of values

cut off level or cut point in continuous variable tests

must be chosen that divides the values that yield a positive result from values that yield a negative result

ex. lower than this is good but higher than that is bad

continuous variable unimodal test

the cut off is difficult to determine

the example is what blood pressure reading yields hyper tension??

continuous variable bimodel test results

the cut off level is somewhat easier to determine but the results close to the cutoff misclassify some patients

example includes what size of diameter yields a positive TB skin test???

what does reliability =

repeatability

will you get the same result if the test is repeated?

three main reasons for lack of reliability

intrasubject variation

intraobserver variation

interobserver variation

intrasubject variation (within)

occurs when a test result from a given person varies over time with repeated testing

example is blood pressure testing

intraobserver variation (within)

occurs when a given test evaluator conducts a test differently, or reads, or interprets a given test result differently upon repeat evaluation

example is people reading a slide

interobserver variation (between)

occurs when different test evaluators conduct tests differently or read or interpret a given test result differently

what does validity =

accuracy

can the test distinguish between those who have the disease and those who do not have the disease?

lacks of test validity

poor sensitivity and poor specificity

poor sensitivity

significant proportion of disease cases are missed by the test

the test has false negative results (the test reads negative but the disease is actually there)

poor specificity

significant proportion of the non diseased are incorrectly identified as disease cases by the test

the test has lots of false positives (the test says you are positive but the disease is not actually there)

what is sensitivity

the proportion of the diseased who test positive

measure of how frequently the test accurately identifies people with the disease (finding the cases)

if the test has a low sensitivity…

there are a large number of false negatives ( the test says negative but the disease is present) and many cases are missed by the test

small FN means value is closer to 100%

if a test has high sensitivity…

there are a small number of false negatives (the test says negative but the disease is present) and most cases have been identified by the test

if there is small amount of false negatives in a sensitivity test…

the number will be closer to 100%

specificity

proportion of the non diseased who test negative

measure of how frequently the test accurately identifies people without the diseases (finds the non cases)

if a test has as low specificity…

there is a large number of false positives (the test says you are positive but the disease is not present) and many non cases are incorrectly identified

if a test has a high specificity…

there is a small number of false positives (the test says you are positive but the disease is not present) and most non cases are identified

if the FP is low in specificity…

the number is closer to 100%

what do PPV and NPV tell you

measures how frequently the test results match actual disease status

influences by both spec and sens

influences by prevalence

what is PPV

the proportion of people with positive test results who actually have the disease

a test with low PPV…

have a lot of false positives (the test is positive but the disease is absent)

no meaningful to clinician or patient

a test with a high PPV…

there is a small number of false positives results (the test is positive but the disease is absent)

meaningful to the patient and clinician

what does high number of false negatives (the test is negative but the disease is present) mean for NPV and sensitivity

=poor NPV= poor sensitivity

what does high number of false positives (the test is positive but the disease is absent) mean for PPV and spec

=poor PPV= poor specificity

what is NPV

the proportion of people with negative test results who do not have the disease

tests with low NPV…

have a large amount of false negatives (the test is negative but the disease is present)

may not be meaningful to the patient and clinician

tests with high NPV…

have a small number of false negatives (the test is negative but the disease is present)

these are meaningful to the clinician and patient

low sens=

=low NPV

low spec=

low PPV

mammograms and specificity

the screening of this test has lots of false positives (test says positive but the disease is absent) which means low spec and low PPV (not a lot of FN)

why dont young women get mammograms

because there is a prevalence that influences the predictive values

this is why the screening is targeted for a specific sub group (women over 40)

PPV and prevalence in a specified population

result is the probability that a person from this specific group who had a positive test result actually has the disease

NPV and prevalence in a specified population

result is the probability that a person from this specific group who had a negative test result actually do not have the disease

benefits for true positives

there can be recovery, decrease chance of disability and death if there is an effective and acceptable treatment or cure

costs for true positives

emotional distress when the disease is fatal and incurable

benefits to true negatives (no costs)

peace of mind

costs for false positives (tests says positive but the disease is absent) (no benefits)

emotional distress from thinking you have a disease, being labeled as sick, physical harm from diagnostic test, lots of money

costs for false negatives (the test says negative but the disease is present) (no benefits)

delayed diagnosis and treatment, may have ignored previous symptoms

where on disease timeline does screening need to be done

in early pathogenesis (before the critical point)

lead time with screening tests

provide lead time by detecting the disease prior to the usual time of diagnosis after the clinical horizon

problem with over diagnosis bias (ex is mammograms)

if new screening leads to diagnosis of increased number of non fatal cases of a disease, the total cases will increase yet the number of deaths is unchanged

the stats look improved but they are not

second example of over diagnosis with the table chart in class

without screening the survival was 10% but with screening lots of people were falsely diagnosed which made the survival increase to 78% just because lots of people were not sick and didn’t die

lead time bias problem

if a new screening program detects a disease earlier on the disease timeline than the pre existing tests, but the patient continues to die at the same time of progression, there is a falsely made “increase lead time”