Case in the Spoltight 12
FORMAL REPORTED FINDINGS
There are multiple pulmonary nodules scattered throughout both lungs as well as a general coarsening of the lung markings. The cardiomediastinal contours and subdiaphragmatic regions are normal. There are no suspicious bone lesions.
Multifocal pulmonary nodules. I note a previous history of colorectal cancer, that these are likely metastatic lesions.
Comment from the DCT
I think one of the concepts that is really important here is pre test probability and how it influences accuracy of diagnosis. Imagine that you don’t know about the bowel cancer, or you don’t let the radiologist know that for their reporting consideration
The history of bowel cancer changes the pre test probability of finding metastases
If your interpretation skills had a sensitivity and specificity of 85 and 90%
Pre test probability 50% (a guesstimate for high risk patient)
- Positive predictive value = 90%
- Negative predictive value = 14%
These results mean positive or negative diagnosis supported, a further test might help accuracy
- Improving test/reporting performance to sensitivity and specificity to 95% would improve PPV to 95%, and NPV to 0.5%
- This could be using expert reporters, or better technology e.g CT
Pre test probability 2%
- Positive predictive value = 14%
- Negative predictive value = 0.3%
This result means that starting from a low probability, a negative test is reassuring, a positive test really just poses a question
Improving test/reporting performance to sensitivity and specificity to 95% would improve PPV to 28%, and NPV to 0.1%
This result shows that when starting from a low probability, that improving test accuracy helps, but does not provide the final answer
SO, including relevant information in pre test information, improves accuracy.