Clinical Reasoning--Diagnosis (Stevens) Flashcards
Sensitivity
Percent of patients with disease who are correctly identified
Sn_N_Out: want a SeNsitive test to rule Out a disease if Negative result
(Few false negatives, many false positives)
a/(a+c) = Disease positive/total disease
High sensitivity means if they have the disease they’ll be positive
Ex: D-dimer test
Specificity
Percent of patients without disease who are correctly identified
Sp_P_In: want a SPecific test with Positive result to rule disease In
(Few false positives, many false negatives)
d/(b+d) = No disease and negative/total no disease
High specificity means if they don’t have the disease they’ll be negative
Confidence Interval
Plausible range of results from the study
Use diagnosis calculator online to calculate
For 95% confidence interval that INCLUDES 0, means that if the two treatments are truly the same, a difference as large (or larger) than found in this study occurs by chance in greater than 5% of studies like this
Positive likelihood ratio
Sensitivity/(1 - Specificity)
Likelihood of + test in diseased/Likelihood of + test in undiseased
(don’t need to memorize)
Negative likelihood ratio
(1 - Sensitivity)/Specificity
Likelihood of - test in diseased/Likelihood of - test in undiseased
(don’t need to memorize)
Positive predictive value
Percent with positive test who in fact have disease
a/(a+b) = Disease positive/total positive
Negative predictive value
Percent with negative test who don’t have disease
d/(c+d) = No disease negative/total negative
Verification bias
When you decide to do the reference test ONLY if your study test is positive
You’ll miss a lot of false negatives (that are actually positive, but you’d never know)
When do you use a sensitive versus specific test?
Sensitive test first for screening, to get rid of the people you know don’t have the disease
Specific test second to confirm people who definitely do have the disease
Prevalence
People in the population who have the disease
(a+c)/(a+b+c+d)