We explored when a clinician should order a diagnostic test in this post about evidence-based medicine. At its core, the principles follow this reasoning:
A) all diagnosis in medicine are probabilities, examples:
- a college student with chest pain after failing a test - 1% of coronary artery disease
- a 57yo obese, male with history of high blood pressure with chest pain - maybe 50% further depending on the type of pain, duration, etc.
B) a test should significantly increase or decrease the probability, otherwise it shouldn't be ordered
C) clinicians set a threshold probability for treating diseases, based on
- the harm of the disease (e.g spinal meningitis kills a lot more than a small skin infection)
- the benefit of the treatment (we'll explore this later)
- the harm of the treatment (every procedure or medication may result in significant adverse reactions or side effects)
D) There are caveats to even ordering a test
- If the pre-test probability already crosses the threshold, don't order the test...just treat
- If, given a certain pre-test probability, the post-test probability can't cross the threshold with a particular test even if came back positive, then don't order the test (i.e. even with a positive test, the clinician wouldn't change the treatment, therefore don't order the test).
And this all makes sense theoretically. And evidence-based medicine continues to develop clinically useful likelihood ratios and help clinicians define pre-test probabilities.
What I don't get however is "C" - how to set a threshold level for deciding when to treat patients. Can anyone help? Maybe using pulmonary embolus as an example...
Comments