I am excited to welcome Deep Jaiswal, a first year medical student, as a guest author on Manu et Corde. He highlights some awesome learning from the first day of medical school in his first post!
Day one of medical school was dedicated to shadowing a clinician in his or her natural habitat. For me it meant being placed in an emergency department at a community hospital with a really awesome preceptor. This learning was very experiential.
One of the many topics discussed was why certain diagnostic tests were ordered, and what test results meant from a probability standpoint. The clinical shift provided a first hand look at how statistics play a role in clinical decision-making. The clinical significance of the terms pre-test probability, sensitivity, specificity, post-test probability etc… became apparent during the shift, though I am still trying to wrap my head around them.
What does that result mean?
I previously understood test results as definitive. In other words, if a test for a certain disease came back positive, then the patient had the disease or if the test came back negative, the patient did not have the disease. Unfortunately, this approach did not take into consideration the uncertainty attributed to the various diagnostic tests we learned about. For example, a pregnancy test can be erroneous such that if a sufficient number of males were tested, one of the test results will turn out to be positive. I hope you can appreciate this as a false positive and erroneous result.
From the above example, it seems like a natural starting point is the patient’s pre-test probability when deciding whether or not to order a test or assess for a clinical finding. In the case of the pregnancy test, it is
very unlikely – prior to testing – that a male patient would be pregnant. Consequently, there is no point in conducting a pregnancy test.
Okay pregnant male patients aside, pre-test probability plays an important role in setting the baseline probability that a patient has a certain disease. As such, the same test result does not mean the same thing for two different patients. For example:
- We have two patients Ms. A and Mr. B, both have positive test results for disease X.
- Suppose it is determined that Ms. A’s pre-test probability for disease X is 10% and Mr. B’s pre-test probability is 65%.
- It is also known that a positive test result shifts the probability of having disease X up by 30% for patients like Ms. A and Mr. B.
- In essence, a positive test shifts the probability of Ms. A having disease X to 40% and Mr. B having disease X to 95% – two very different post-test probabilities that have different clinical consequences even though the test results were the same in both patients.
Medlow and Lucey (2011) have suggested the following approach to categorizing probabilities, (including pre-test probabilities):
Very Unlikely = probability of disease less than 10%
Unlikely = probability of disease between 10% and 33%
Uncertain = probability of disease between 33% and 66%
Likely = probability of disease between 67% and 90%
Very Likely = probability of disease greater than 90%
If the initial assessment is either Very Unlikely or Very Likely then further testing may not be required. Two situations in which further testing is important: The first is if the diagnosis is very unlikely but the disease is so dangerous that it warrants certainty in ruling out e.g. would you send a patient home if they had a 7% probability of having a subarachnoid hemorrhage? The second situation is if the diagnosis is very likely but the treatment is very dangerous e.g. treating DVT with coumadin if the patient also has liver disease. (Medlow & Lucey, 2011)
While pre-test probabilities set the point of baseline, the degree of uncertainty attributed to different tests and clinical findings helps to determine post-test probabilities. That’s a lot of statistics for one day…
stay tuned for part two (I promise not to use the pregnant male example anymore).
Because I am. I am excited about the huge amount of learning ahead so that I might understand how to estimate this pre-test probability and apply these heuristics to patient care. Good thing there are many years of training left! There is a huge amount to learn and I feel fortunate that my education around evidence-based, rational clinical decision-making started on day one.
Students, what has your experience been in medical school learning about these concepts? Medical educators, how do you teach students these concepts? Leave your thoughts in the comments and please feel free to share this post!
If you haven’t already, make sure you check out these related posts:
- Lauren Westafer’s (@LWestafer) great medical student thoughts on “Thinking About Thinking” and “Metacognition for the Pragmatist”
- “Teaching Clinical Reasoning” by Nadim Lalani (@ERMentor)
- “Thinking about teaching thinking” by Robert Centor (@medrants)
- “Clinical Problem Solving: the Online course” on Manu et Corde
This post was reviewed by Eve Purdy.
 McGee, S. R. (2012) Evidenced-based physical diagnosis, 3rd ed, Elsevier/Saunders ebook pp 9-21
 Melow, A. M., & Lucey, C. R. (2011) A qualitative approach to Bayes’ theorem. Evidenced Based Medicine, 16(6), pp 163-167