Just published online today by Meyers and Smith.
This is our best and most important work ever. Another nail in the STEMI/NonSTEMI coffin. Accuracy of OMI ECG findings versus STEMI criteria for diagnosis of acute OMI. Full text:
https://www.sciencedirect.com/science/article/pii/S2352906721000555
Hi Dr Smith
ReplyDeleteI have been an avid reader of your blog for several years now and am very glad to see your work has been published.
After reading your paper I would like to add a comment which hopefully is helpful.
The sensitivity of STEMI criteria in predicting OMI is so low that it is very likely that this will cause a much higher false positive rate than suggested by its high specificity. This is because the ability of a test to rule out a disease is determined by both its sensitivity and specificity (the same is true to rule in a disease). Therefore, in this case I think that positive and negative likelihood ratios would provide a more accurate method to compare expert interpretation and STEMI criteria accuracy (apologies if you understand this already!).
I have therefore calculated the positive and negative likelihood ratios (based on the sensitivity and specificity reported in your manuscript) of the STEMI criteria versus expert interpretation. For STEMI vs interpreter 1 this is 6.83 vs 9.56 and 0.63 vs 0.15; and for STEMI versus interpreter 2 this is 4 vs 10 and 0.7 vs 0.22. This suggests that (in addition to the vast improvement in identifying STEMI(−) OMI) expert ECG interpretation is less likely to produce a false positive than STEMI criteria (I am unable to calculate a p value comparison without the full 2x2 tables but I imagine this would be significant if it were calculated). Therefore the similar specificities of STEMI criteria and expert ECG interpretation is probably misleading and expert interpretation most likely has (much) better rule in and rule out performance.
(Likelihood ratios should be interpreted in the following way - the larger the positive LR the larger the post test probability of having the disease of interest with a positive test result; and smaller the negative LR the lower the post test probability with a negative test result (https://doi.org/10.1016/S0140-6736(05)66422-7)).
Ravi
(Junior doctor working in the UK)
Very good point. We should have calculated + and - LR. of course I know this, but we did not do it. Thanks! Will share with Pendell.
Delete