Short Answer
The ROC curve is a graph that shows, across all possible threshold values of a diagnostic test, the relationship between sensitivity (the true positive rate) and 1 minus specificity (the false positive rate); the AUC is the area under this curve, and it summarizes the overall discriminative power of the test in a single number ranging from 0.5 (random) to 1.0 (perfect). ROC analysis is used to evaluate how well a continuous or ordinal measurement separates a binary outcome (diseased, healthy) and to select the most appropriate threshold value. For the AUC, a rough interpretation is that above 0.90 is excellent, between 0.80 and 0.90 is good, and between 0.70 and 0.80 is acceptable, but the value must always be reported together with its 95 percent confidence interval. The most common mistake is thinking that the ROC automatically gives you the best threshold; the threshold is chosen according to the clinical context (for example, in screening where a false negative is dangerous, sensitivity is prioritized), and when comparing the AUC of two tests, the correct method, such as the DeLong test, should be used instead of the Youden index.
Serteser Consulting is run by a biomedical engineer (BME MSc) with peer-reviewed publications and PROSPERO-registered systematic reviews; it designs and carries out thesis, article, and clinical research statistics, including diagnostic accuracy analysis with the ROC curve and AUC, using SPSS, R, and Python, in a form that is manuscript-ready and defensible before a jury or reviewer.
You are trying to detect a disease with a biomarker you have newly developed. How will you decide above which threshold value to call a result "positive"? Is sensitivity more important, or specificity? This is exactly what ROC analysis answers.
What Is the ROC Curve?
The ROC (Receiver Operating Characteristic) curve is a graph that shows the sensitivity and 1-specificity values of a diagnostic test for different threshold values.
- Sensitivity: The proportion of diseased individuals correctly identified (True Positive Rate)
- Specificity: The proportion of healthy individuals correctly identified
- 1-Specificity: False Positive Rate (horizontal axis)
The closer the curve is to the top-left corner, the better the test. The diagonal line (y=x) represents random prediction power.
What Is the AUC and How Is It Interpreted?
The AUC (Area Under the Curve), the area under the curve, is a single number that summarizes the overall diagnostic performance of the test.
| AUC Value | Interpretation |
|---|---|
| 0.90 – 1.00 | Excellent |
| 0.80 – 0.90 | Good |
| 0.70 – 0.80 | Acceptable |
| 0.60 – 0.70 | Poor |
| 0.50 – 0.60 | Fail (random prediction) |
The confidence interval of the AUC must always be reported. For example, 0.72 (95% CI: 0.64–0.81).
How Is the Threshold Value Selected?
The ROC curve does not automatically give you the best threshold value; you need to make the selection according to the clinical context.
Youden Index: The point that maximizes Sensitivity + Specificity - 1. It is the most commonly used mathematical criterion.
Clinical priority: If a false negative (missing a diseased person) is more dangerous than a false positive, keep sensitivity high; sacrifice a bit of specificity. This logic applies in cancer screening.
Cost analysis: In some situations, the threshold is determined by taking into account the cost of a false positive (unnecessary biopsy, patient anxiety).
ROC Analysis with SPSS
You can calculate the Youden Index manually from the coordinate table: for each row, select the threshold that has the highest (Sensitivity + Specificity - 1) value.
Comparing Two ROC Curves
If you want to compare the diagnostic performance of two different tests, the DeLong method is used. SPSS does not support this directly; the pROC package in R is ideal for this analysis.
For your study involving ROC analysis, request a free consultation.
Where Do People Get Stuck Most in This Analysis?
- The AUC came out as 0.72, but you cannot write whether this is "good" or "acceptable" and what it means in the clinical context.
- You need to compare two tests, but the DeLong test is not in SPSS, and you cannot switch to R.
- Threshold selection should be made with clinical justification, but you do not know anything beyond the Youden index.