Calling Dr. House: Help With Interpreting Diagnostic Tests

One important step in the assessment of a patient is determining what diagnostic test should be used in that clinical scenario. The decision is usually based on the physician's experience and knowledge of the case. During an ARHP session Sunday at the ACR 2016 meeting, Dr. Amy Nowacki and Dr. Brian Mandell discussed principles in test interpretation and provided practical examples.

One way to interpret and analyze a diagnostic test is by using the likelihood ratio (LR), which is basically a ratio of the probability that a result is correct to the probability that the result is incorrect. LRs are generated from the sensitivity and specificity of a given test:

• Positive likelihood ratio = sensitivity/1-specificity

• Negative likelihood ratio = 1-sensitivity/specificity

One of the useful characteristics of LRs is that they do not vary in different populations or groups of patients because they are based on a ratio of sensitivity and specificity. Other useful features are that they can be used directly at the individual patient level, and they allow the clinician to quantitate the probability of a disease for any given patient. Nomograms are useful tools to calculate LRs. One example is Fagan's nomogram. 

An approach to using LRs in practice is to look at the pre-test probability, then the likelihood ratio, then pre-test odds, and then the probability of diagnosis.

For example, lets say you were given a specific case and you wanted to determine if that patient has lupus. In the case, your original pre-test probability for "Patient A" having lupus is high. Let's say it is 75%. An ANA is ordered and the test comes back positive. Using the calculation or Fagan's nomogram, the likelihood ratio increases from 75% to 94%. This strengthens the likelihood that the patient has lupus.

Here's a second example. "Patient B" has a low pre-test probability of lupus. Lets say it is 25%. If the ANA is positive, it increases the probability to 62%. But if the ANA is negative, it decreases the probability to 1%.

As you can see, the likelihood ratio is more valuable in ruling out a disease than ruling it in. 

In summary, a probability for a possible diagnosis can be assigned. A diagnostic test can then be requested based on background knowledge and knowledge of the test's performance. The post-test probability can then be estimated using the pre-test probability and the likelihood ratio calculation. A treatment decision can then be established.

Share This Report

About the Author

Trending Reports From ACR 2016