Evaluating diagnostic tests Flashcards
How is prevalence calculated from a 2x2 table?
a+c/N
a = true positive, c = false negative
N = total population
How is sensitivity calculated from a 2x2 table?
True positive rate
a/(a+c)
1- proportion of false negatives
P(T+|D+) (test positive given disease positive)
How is specificity calculated from a 2x2 table?
True negative rate d/(b+d) d = true negative b = false positive 1-proportion of false positives P(T-|D-)
How is the positive predictive value (PPV) calculated from a 2x2 table?
a/a+b
Probability that someone has the disease when the test is positive
How is negative predictive value (NPV)calculated from a 2x2 table?
d/c+d
Proportion of individuals without the disease when the test is negative
What is the effect of prevalence of disease on PPV and NPV?
PPV - increased with prevalence
NPV - decreased with prevalence
What is more important in a screening test?
High sensitivity to rule out those without disease
What is more important in a diagnostic test?
Needs high specificity to include those with the disease
What is the likelihood ratio for a positive test?
Probability of a positive result with a disease/probability of a positive result without disease
How is LR+ calculated and what does this mean?
Sensitivity/1-specificity
LR+>1 means that the test is more likely to give a positive result if patient has the disease than if they did not
How is LR- calculated and what does this mean?
1-Sensitivity/Specificity
How are pre-test odds calculated?
prevalence/(1-prevalence)
How are post-test odds calculated and what does this mean?
pre-test odds x LR+
Quantifies what a positive test reults provides about whether an individual truly has the disease
How can a cut-off be decided for diagnosis based on continuous measurement?
Using an ROC curve (receiver operating characteristic)
How is an ROC curve plotted?
X axis = 1-specificty (false positive probability)
Y axis = sensitivity |(true positive probability)
For each diagnostic cut-off