Diagnostic test evaluation and comparison

Calculate test Sensitivity and Specificity and ROC curves

Test name

Data format

Unstacked

Stacked

Confidence level

0.9 0.95 0.98 0.99

Paste the columns of data to be summarised in the space below.

Download example data

Submit

This utility calculates test sensitivity and specificity for a test producing a continuous outcome. Suggested cut-points are calculated for a range of target values for sensitivity and specificity. A ROC curve and two-grah ROC curve are generated and Youden's index (J and test efficiency (for selected prevalence values (are also calculated).

Inputs:

• the desired level of confidence in the resulting sensitivity and specificity estimates; and
• two columns of data for analysis. Data required is a series of test results for both infected and uninfected individuals. This data can be pasted in either of two formats:
  • stacked - the first column contains status identifiers as either "Infected" or "Uninfected" and the second column contains the corresponding test result; or
  • unstacked - separate columns contain test results for infected and uninfected individuals. Column order is unimportant but columns must be labelled appropriately as "Infected" or "Uninfected" in a header row;
  Regardless of the format used, the first row must contain column headers. Additional columns of data will be ignored.

Outputs:

• numerical and graphical summaries of testing results for both infected and uninfected groups;
• cut-point values to achieve minimum target values for both sensitivity and specificity along with corresponding estimates and Wilson binomial confidence intervals;
• one and two-graph ROC curves, with estimated AUC for the one-graph curve;
• area under the ROC curve (AUC) and associated DeLong confidence limits and Z test. See DeLong et al. (1988). Comparing the areas under two or more correlated receiver operating characteristic curves: a nonparametric approach. Biometrics 44:837-845;
• graphs of Youden's J Index and test efficiency for a range of prevalence values;
• graphs of mis-classification cost terms for a range of prevalence values and relative costs of false negative/false positive; and
• detailed sensitivity and specificity results in a downloadable spreadsheet file.

For more information, see Greiner, M, Pfeiffer, D and Smith, RD (2000). Principles and practical application of the receiver-operating characteristic analysis for diagmostic tests. Preventive Veterinary Medicine 45:23-41.