Missed and Erroneous Diagnoses Common in Primary Care Visits

John Brush jebrush at ME.COM
Fri Dec 20 19:25:15 UTC 2013


I agree that likelihood ratios are very important, but I am not sure I understand how you would use them without regard to a prior probability.
	Likelihood ratios are dimensionless numbers that are only useful if used as a multiplier of the prior odds. Sure, you can use them in isolation to get some idea of the strength of new information, but to use them without regard to the prior probability (converted to odds), would lead to a fallacy known as base-rate neglect.
	Estimating the prior probability of a single case or event requires a shift in thinking from the frequentist notion of probability to the personal notion of probability. The meaning of probability changes and becomes a measure of your degree of belief or conviction that a patient has a particular disease. 
	When I explain likelihood ratios to students and residents, I usually use an example where I am indifferent about the prior probability. I assign a prior of 0.5, which gives a prior odds of 1. This makes it easy to show how you can multiply the prior odds by a likelihood ratio to get a posterior odds. I then convert the posterior odds back to probability. This exercise gives a student an idea of how the strength of new information should impact on your prior impression, to create a final impression.
John


On Dec 19, 2013, at 4:44 PM, Bimal Jain <bjain at PARTNERS.ORG> wrote:

What is required to minimise diagnostic errors, I believe, Is a deeper understanding of the mechanics of clinical diagnosis. There are two important factors which make diagnosis problematic:
(a) The need to diagnose a disease correctly in a given, individual patient and
(b) The almost infinite range of clinical presentations of a certain disease in different patients, varying from highly typical(high prior probability) to highly atypical(low prior probability).
It is important to appreciate that a prior probability is a statistical concept which symbolises distribution of a disease in a series of patients. It is not therefore a measure of prior evidence for a disease in a given individual patient.
What we require in clinical diagnosis, I suggest, is a parameter which allows us to assess a presentation as evidence for or against a disease in a given patient. This parameter, I believe, is a likelihood ratio, which is a comparison between frequencies of a presentation in patients with and without a suspected disease respectively. It turns out the likelihood ratio of any presentation, even if it is high probability is negligible (Sanson et al Thromb., Haemost. 83: 2000, 199-203. for presentations of pulmonary embolism).
Thsi means any clinical presentation, regardless of whether it is high or low prior probability, does not provide any significant prior evidence for or against a suspected disease. It can only be employed therefore to suspect a disease which is then evaluated by a test. If a test result with likelihood ratio of 10 or higher is observed, the disease is then to be diagnosed definitively ( Jasc hke et al in Users guide to medical literature AMA Press, Chicago 2002, 121-140).
The mistake often made , which leads to diagnostic errors, is to consider a highly atypical presentation(low prior probability) as strong evidence against a suspected disease and rule it out without testing.
Bimal P Jain MD, Pulmonary-Critical Care, Northshore Medical Center (Union), Lynn, MA 01904






Moderator: Lorri Zipperer Lorri at ZPM1.com, Communication co-chair, Society for Improving Diagnosis in Medicine

To unsubscribe from the IMPROVEDX list, click the following link:<br>
<a href="http://list.improvediagnosis.org/scripts/wa-IMPDIAG.exe?SUBED1=IMPROVEDX&A=1" target="_blank">http://list.improvediagnosis.org/scripts/wa-IMPDIAG.exe?SUBED1=IMPROVEDX&A=1</a>
</p>










More information about the Test mailing list