Diagnostic studies: Does my patient have the disease or not?
Diagnostic tests are used to identify or rule-out the presence of a disease (or a particular condition) in a patient.
The result of a diagnostic test may be dichotomous (e.g. pregnancy, fracture, HIV) or non-dichotomous (e.g. troponin value, exercise tolerance).
- A diagnostic test result may be made dichotomous by setting a cut-off point for abnormal, e.g. defining PSA > 8 ng/ml as elevated (positive).
- Varying the cut-off that defines a positive versus negative result changes the perceived test performance; this is illustrated by a receiver-operator characteristic (ROC) curve.
The 2 x 2 table for a diagnostic test
|
|
Disease (+) |
Disease (-) |
|---|---|---|
|
Test (+) |
True Positive (TP) |
False Positive (FP) |
|
Test (-) |
False Negative (FN) |
True Negative (TN) |
Characteristics of the diagnostic test (not affected by prevalence of the disease in the population)
|
Sensitivity |
TP/(TP+FN) |
Tests with high sensitivity (Sn) help rule out (Out) disease if negative (i.e., few false negatives) - “SnOut” |
|
Specificity |
TN/(TN+FP) |
Tests with high specificity (Sp) help rule in (In) disease if positive (i.e. few false positives) - “SpIn” |
Characteristics of diagnostic test in a population (dependent on the prevalence of the disease in the population)
|
Positive predictive value (PPV) |
TP / (TP+FP) |
proportion of positive tests that represent true disease. OR If my patient has a positive test result, how likely is she to have the disease? |
|
Negative predictive value (NPV) |
TN / (TN+FN) |
proportion of negative tests that represent true absence of disease OR If my patient has a negative test result, how likely is he to still have the disease? |
The PPV and NPV are what we need to know in order to interpret the clinical implications of a positive or negative test – and these are completely dependent on our estimate of the likelihood that the patient has the disease, prior to performing the test (Pretest probability). Likelihood Ratios (LR) provide the most efficient way to calculate PPV and NPV. See next section!