Test sensitivity and specificity made easy

Today we are going to cover test sensitivity and specificity in a non-boring way! In this video from our Epidemiology Essentials course, you will learn exactly what they are and why they matter. Furthermore, we’ll discuss what happens to a test’s sensitivity and specificity if thresholds for a positive test are increased or decreased. Understanding these variables is critical for the correct interpretation of almost any clinical trial. Make sure to watch this video before reading your next paper!

As always, please share your thoughts in the comments section below and don't forget to sign up for a free trial account!

[00:00:00] Hey, everyone. It's Franz from Medmastery. Today, we're going to talk about test sensitivity and specificity. I can hear you moan already and I agree that this stuff can be taught in a really boring way. But take my word for it, this video is not going to bore you. So, just give it a try. In clinical medicine and in public health, we need tests in order to separate healthy from diseased individuals. Most of the time, these tests assess a continuous variable like blood pressure, which is measured in mmHg for example.

[00:00:30] Diseased and healthy individuals usually have different distributions of that variable. So, we choose a threshold above which individuals are classified as diseased and below which they're classified as healthy or non-diseased. Where we choose the threshold makes a huge difference. If the threshold is too low, many healthy individuals will be falsely classified as diseased. If it's too high, many diseased individuals will be falsely classified as healthy. If a test is able to classify a

[00:01:00] large proportion of diseased and non-diseased correctly, it is said to have a high validity. Test validity has two major components, one is sensitivity and the other one is specificity. These are quality criteria for the test. Let's look at an example for better clarity. Let's pick a population of 1000 people and let's say that 200 of them have the disease of interest. So, this means that the prevalence is 20%. Now, let's also say

[00:01:30] that 160 out of the 200 diseased people test positive, whereas 40 are missed by the test. On the other hand, 720 of non-diseased are correctly classified as negative by the test, whereas 80 are falsely classified as positive. The sensitivity of that test is calculated as the number of diseased that are correctly classified, divided by all diseased individuals. So, 160, divided by 200, times 100 and that equals 80%.

[00:02:00] In other words, the sensitivity is the proportion of diseased individuals correctly classified and that's 80% in this case. The specificity is calculated as the number of non-diseased individuals correctly classified. So, 720, divided by all non-diseased individuals of 800, times 100 and that equals 90%. So, the specificity is the proportion of non-diseased correctly classified. Validity is usually determined when a test is

[00:02:30] newly introduced and when that's done, it's compared to a gold standard. So, a lab test assessing the presence of Helicobacter pylori could be compared to a gold standard of let's say biopsy or a test of coronary artery disease could be compared to the gold standard of coronary angiography. Once the new test is used in the real world, we don't know who's diseased and non-diseased at the outset, otherwise, we wouldn't be doing the test. So, in the real world, we end up

[00:03:00] with positives or negatives and we have to do something with them. So, sensitivity and specificity give us an indication as to how much trust we can put into these tests. So, diseased individuals who are tested positive are called true positives. And non-diseased individuals who are tested negative are called true negatives. Whereas diseased who are tested negative are called false negatives and non-diseased who are tested positive are called false positives. Ideally, we would like

[00:03:30] everyone to fall into the true positive or true negative groups but no test is perfect, so we'll end up with people in the false negative and false positive groups. Now, what's the problem with these groups? Well, someone who is falsely labeled as positive will be sent for further testing. This will subject them to the risk of these potentially invasive tests, which will consume dollars, create fear in the side of the patients and their relatives. Also, the label associated with the false test might stick for a long

[00:04:00] time. Think about your own patients. Once falsely diagnosed with hypertension, for example, they might not get rid of that diagnosis forever because it's copied from one letter or patient report to the next. On the other hand, someone who's falsely labeled as negative, who has a potentially treatable disease might be sent home and die or become much sicker because nothing's done about the disease. So, when we choose the threshold of the test, we have to weigh the relative importance of problems associated with false positives and false negatives.