A tweetorial.
One of my favorite points to making when teaching diagnostic reasoning it to stress that positive findings are immensely more important than negative ones. I believe this so strongly that it is point 4 on my diagnostic reasoning yellow card.
One of my favorite points to making when teaching diagnostic reasoning it to stress that positive findings are immensely more important than negative ones. I believe this so strongly that it is point 4 on my diagnostic reasoning yellow card.
As an example, hearing an S3 is very powerful when it comes to diagnosing heart failure (LR+, 11) but the lack of tachycardia in a patient suspected of having a PE is meaningless when it comes to excluding PE (LR-,1.0).
In an effort to support (and better quantify) this point we did some "research". We pulled together pretty much all the test characteristics we could find in the published literature.
Our richest sources of data were JAMA's Rational Clinical Examination series, Symptom to Diagnosis (of course), and Steven McGee’s transcendently good book, Evidence-Based Physical Diagnosis
We identified positive LRs for 250 clinical findings and negative LRs for 241. These findings were associated with 67 distinct diagnoses. We considered a diagnostic test to be useful if it had a positive LR ≥5 or negative LR ≤0.2.
Aside: I know the arguments for and against likelihood ratios. Let's not get distracted. Please see @f2harrell great posts on the topic for why we should be suspicious. If you really want to doubt LRs, just carefully read some diagnostic test studies.
The results? First, generally speaking, individual clinical findings are not powerful diagnostic tests. The median LR+ was 3.4 and the median LR- was 0.59.
Second, supporting my teaching point, positive findings are more informative than negative ones. 28% of positive findings had LRs > 5 while only 12% of negative findings had LRs < 0.2. The mean positive and negative likelihood ratios were 6.4 and 0.57, respectively.
This is my favorite finding: considering the most impactful diagnostic findings, there were 31(12.4%) positive findings with LRs > 10 while only 13 (5.4%) negative findings with LRs < 0.1. Here are a few powerful findings:
Take home message? Most signs and symptoms are not diagnostically powerful and thus do little to alter disease probability. Few signs or symptoms are pathognomonic of a diagnosis and even fewer can definitively exclude a diagnosis.
Ah hell, let’s put that on a slide.
Ah hell, let’s put that on a slide.
Relish this tweetorial as you’re unlikely to see it in a journal. I think the associated article may have been rejected by about a dozen journals and I wore out. Thanks to Keven Kennedy, 4th year student here at Pritzker and Scott Stern.
