Related Experiment Videos
The misuse of predictive value--or why you must consider the odds
1Columbia University, New York, NY.
Annali Dell'Istituto Superiore Di Sanita
|January 1, 1991
Summary
Predictive values of medical tests are often misunderstood. Likelihood ratios provide a more accurate interpretation of test results, especially when considering disease probability and confirmatory testing.
Area of Science:
- Laboratory medicine
- Diagnostic test evaluation
Background:
- Predictive values of diagnostic tests are frequently misinterpreted due to their presentation as percentages.
- Low percentages are often perceived as poor test performance, regardless of other relevant metrics.
Purpose of the Study:
- To clarify the misinterpretation of predictive values in diagnostic testing.
- To highlight the superior utility of likelihood ratios over predictive values and ROC curves.
Main Methods:
- Comparison of predictive values with positive likelihood ratios using a clinical example.
- Discussion on the influence of signal strength on test sensitivity and specificity.
- Emphasis on the role of confirmatory and repeat testing.
Main Results:
- A positive predictive value of 20% with a positive likelihood ratio of 50 demonstrates a significant increase in disease odds.
- Likelihood ratios offer a more informative assessment of diagnostic test accuracy than predictive values or ROC curves.
- Varying signal strengths impact sensitivity and specificity, affecting interpretation of positive results.
Conclusions:
- Likelihood ratios are more informative than predictive values for interpreting diagnostic test results.
- Confirmatory testing and consideration of signal strength are crucial for accurate diagnosis.
- A nuanced understanding of statistical metrics is essential for effective laboratory medicine.