likelihood ratio

[jfgavina]

---

Members don't see this ad.

just reviewing my uw notes come across this "likelihood ratio" which is = sensitivity/(1-specificity) in what context do we use this? in the boards will they ask directly to calculate this or we need to know when we should use it?

 

[lemonade90]

I think you take the pretest odds and multiply by the LR you will get the post test odds. I would imagine that it should be used in the context of some sort of test or examination technique used for assisting in diagnosing a disease or determining a course of action. E.g. you think somebody has had a heart attack and assign a pretest probability that that is indeed the case...turn that into pretest odds and multiply by LR ratio for a particular test (e.g. cardiac enzyme assay) and get the post test odds which you can then convert to post test probability (i.e. probability a person has the particular problem/disease). There are of course negative and positive LR.

 

[OveractiveBrain]

just reviewing my uw notes come across this "likelihood ratio" which is = sensitivity/(1-specificity) in what context do we use this? in the boards will they ask directly to calculate this or we need to know when we should use it?

On the boards: you will be given a 2x2 (Test +/- and then Dz +/-) and you will be asked to calculate it.

In life: Liklihood ratio is the only relevant factor for clinical medicine. A large positive likelihood ratio means that if the test, the question, or the exam manuver is positive, your diagnosis is more likely. A very small negative likelihood ratio (closest to 0) means that if it is negative, the disease is less likely.

Most people think "hear story, get test, done." Thats a mediocre physician who doesn't understand clinical reasoning. Likelihood ratios are what actually matter: increasing or decreasing the likelihood that your diagnosis is or is not what you think it is.

 

[kaleerkalut]

On the boards: you will be given a 2x2 (Test +/- and then Dz +/-) and you will be asked to calculate it.

In life: Liklihood ratio is the only relevant factor for clinical medicine. A large positive likelihood ratio means that if the test, the question, or the exam manuver is positive, your diagnosis is more likely. A very small negative likelihood ratio (closest to 0) means that if it is negative, the disease is less likely.

Most people think "hear story, get test, done." Thats a mediocre physician who doesn't understand clinical reasoning. Likelihood ratios are what actually matter: increasing or decreasing the likelihood that your diagnosis is or is not what you think it is.

Isn't sensitivity/(1-specificity) the same thing as Positive Predictive Value (PPV) = True Positive/True Positive + False Positive?

Thanks.