Filly's 9th Law
There are a few important things to know about statistics: Positive and negative predictive values.
I practiced radiology in one of the most prestigious hospitals in the USA for 41 years. It was literally loaded with academics. You will always hear academics, when speaking about statistics in diagnosis, focus on the “sensitivity and specificity.” In the 41 years I practiced at UCSF, I kept an accurate count of how many times a clinician walked into my reading room and asked what the “sensitivity and specificity” of diagnostic ultrasound was for the diagnosis I was asked to render on his/her patient. The grand total was ONE TIME. Got that – 41 years/ONE TIME!
However, the question that virtually every clinician asked me when he or she came into my reading room was some variation of, “Does Mrs. Smith have a gallstone? Does Ms. Jones have an ectopic pregnancy?” Etc. Etc. And they often followed my answer by asking “how sure I was.” The clinician wants to know whether or not you were able to confirm that their patient has the diagnosis that they suspected clinically. They are, in essence, asking for your positive predictive value.
It is extremely important to have a solid grasp on this issue. This is especially important when you are the “final” examination before the patient goes to the operating room for definitive treatment. In the instance of the question, “Does Mrs. Smith have a gallstone?” and your answer is “Yes,” your opinion is based on the positive predictive value. Seeing a high-amplitude reflector within the gallbladder followed by an acoustic shadow has nearly a 100% positive predictive value. Ergo, your confidence is high and the clinician will almost certainly not ask for confirmatory testing. He/she didn’t ask you to state your positive predictive value, but the fact is that the patient will likely end up in surgery if they are symptomatic.
Conversely, when the clinician asks, “Does Ms. Jones have an ectopic pregnancy?” and you answer is “No” because you are confident you saw an intrauterine pregnancy you are relying on nearly a 100% negative predictive value.
Clinicians are not interested in the entire population of patients with gallstones. They don’t want to know what percentage of the total population of gallstone patients you could correctly identify (sensitivity). They want to know “Does Mrs. Smith have a gallstone?”