Updated: 11/28/2020

# Evaluating Diagnostic Tests

0%
Topic
Review Topic
0
0
N/A
N/A
Questions
11
0
0
Topic
 Sensitivity, Specificity, PPV, NPV These 4 measures describe how well diagnostic tests capture the true presence or absence of disease  Sensitivity (SN) % with disease who test positive = a/(a+c) = TP/(TP+FN) Specificity (SP) % without disease who test negative = d/(b+d) = TN/(FP+TN) Positive predictive value (PPV)  % positive test results that are true positives = a/(a+b) = TP/(TP+FP)  Negative predictive value (NPV)  % negative test results that are true negatives = d/(c+d) = TN/(FN+TN) Cut-off point may be adjusted to optimize sensitivity and specificity, which are inversely related (cut-off point with decreased sensitivity is associated with increased specificity and vice-versa)    will also affect NPV and PPV  i.e., decrease in sensitivity associated with decrease in NPV in the same population (due to higher false negative rates) Sensitivity and specificity are intrinsic to the diagnostic test  do not change with prevalence PPV and NPV do change with prevalence  Receiver operating characteristic (ROC) curves are a graphical depiction of a test's performance Y axis: sensitivity X axis: 1-specificity The higher the curve, the better the test This is quantified by the AUC (area under the curve); an AUC of 0.5 states that the test performs no better than chance (bad test!), whereas an AUC of 0.9 suggests a better-performing test
 Odds Ratio, Relative Risk, Attributable Risk These measures describe the relationship between a risk factor and a disease Odds Ratio (OR) odds of having disease in expose group / odds of having disease in unexposed group = ad/bc Relative Risk (RR) probability of getting disease in exposed group / probability of getting disease in unexposed group = [a/(a+b)] / [c/(c+d)] If RR = 1, there is no association between exposure and outcome  Dose-reponse increased level of exposure shows an increased relative risk of developing/odds ratio of having a disease can be used in OR or RR to support causality  Attributable Risk (AR) risk in exposed group - risk in unexposed group = a/(a+b) - c/(c+d)