Diagnostic Accuracy: Statistical Tools to Enhance Clinical Decision Making Flashcards
reliability
produces precise, accurate & reproducible information
diagnostic accuracy
ability to discriminate between patients with or without a specific disorder
pretest probability
likelihood that a patient exhibits a specific disorder before the physical exam starts
uses prevalence rates utilized or clinical demographics, medical history, results of previous tests, clinician expertise, expressed as % or qualitative
posttest probability
likelihood that a patient has a specific disorder after the clinical exam test has been performed
what is nominal data?
groupings/place holder
what is ordinal data?
quantities that have a natural order, but can’t state with any certainty whether intervals between values are equal
for nominal & ordinal data what is kappa coefficient (k)?
measure of agreement (%) between two or more individuals
-kappa = measure of agreement after chance has been removed
-weighted Kappa used if more than 2 categories exist
-always < 1.0
( <0.50 = poor reliability 0.50-0.75= moderate reliability. >0.75= good reliability )
what is interval data?
order with equal intervals but without an absolute zero (Ex temp)
what is ratio data?
order with equal intervals and an absolute zero point (Ex ROM)
what does it mean if a test has sensitivity?
true positive rate
a sensitive test when negative test will certainly rule a patient out
what does it mean if a test has specificity?
true negative rate
specific test when a positive result definitely rules a patient in
What is used to find interval and ratio data correlation coefficient?
Pearson (r)
- analyzes association between 2 measurements
-hypothesis: there is or is not an association between the variables
-analyzes magnitude & direction of associate ranging -1 to 1
- negative values indicate an inverse relationship
- positive values indicate a direct relationship
- zero demonstrates no relationship
what is used for interval and ratio data?
ICC (intraclass correlation)
investigates the amount of variance between 2 or more repeated measures
based on the number of raters and specific measurement obtained
what is spectrum bias?
may be influenced by prevalence or total instances of a diagnosis in the population at a given time, sex ratios, or severity of the disease
what is a positive predictive value?
estimates the likelihood that a patient with a (+) test result actually has the disease
-a higher PPV indicates that a (+) result is strong indicator that the patient has the diagnosis
what is a negative predictive value?
estimates the likelihood that a patient with a (-) test result actually does not have the diagnosis
- a high NPV indicates that a (-) result is a strong indicator that the patient does not have the disease
what is likelihood ratio?
- a test result is valuable only if it alters the pretest probability that a patient has a disorder
-sensitivity & specificity are combined to indicate this shift give a certain test result
-clinical decision making is guided by this statistic
Easy to remember LR’s
- 2 & 1/2 (small shift)
- 5 & 1/5 (moderate shift)
- 10 & 1/10 (large shift)
how do you judge the quality of a test?
QUADAS - quality assessment of diagnostic accuracy studies
what does a positive likelihood ration (LR) indicate?
shift in probability that favors the existence of the disorder >1 LR = increase the odds of a disorder given a positive test