Picture of Michael GarfinkleMichael Garfinkle is a Nephrology Fellow at the University of Calgary

 

The practice of evidence based diagnosis is what we are taught we should all do, but, in practice, this is quite difficult. EBM diagnosis relies on likelihood ratios, but it’s impossible to remember the diagnostic accuracies of the thousands of tests that exist.

 This is why I made a simple tool to help with the scientific diagnosis of diseases spanning most specialties. It is a database of over 700 likelihood ratios of tests (history, radiology, physical exam etc.). The likelihood ratios are completely free for all to access and are found on my website and on an app I developed for iPhone / iPad called DxLogic. Android users are not left out as the website is mobile friendly!

The following examples will demonstrate how you to can use this resource to approach some common clinical situations in a probabilistic and rational manner. The linked images found below are from the iOS application.

 Example #1:

You are seeing a 40 year old gentleman with new onset confusion. He has a fever and a high white cell count and you wonder if he has meningitis. You remember being told that neck stiffness is a good sign for this disease; however, this patient does not have neck stiffness on exam. You wonder if this significantly decreases his probability of having meningitis or if you should get an LP. By looking this up on the app you find that the absence of neck stiffness does not significantly decrease the probability of meningitis. As your clinical suspicion remains high, you get an LP.

Example #2:

You are reviewing a 70 year old woman for her annual check-up. You notice a systolic murmur on exam and you wonder if she might have significant underlying aortic stenosis. Using the resource we find that her probability of having moderate or severe aortic stenosis, given that she has a systolic murmur and before any tests, is 15%.

 Browsing through the app you notice that the absence of radiation to the right clavicle is the best sign to rule-out significant aortic stenosis. Plugging this into the app we find that her probability of having moderate or severe aortic stenosis on echocardiography is now 2%. You discuss this probability of disease with the patient and decide together whether an echocardiogram is warranted.

Likelihood ratios could play an important role in decision-making in every day medical practice. Give my app/website a try and feel free to contact me with any comments or suggestions.

Editors’ note: We don’t, as a rule, let people ‘advertise their goods’ in the form of a CMAJ guest blog. In this instance we were impressed by Michael’s clever and useful tool and thought that more people should get to know about it, so we invited him to blog for us.