Transcript for:
Understanding the ROC Curve in Biostats

- Asalamu alaikom, welcome back  to another biostats ATP video,   this time around we’ll be talking about the  “receiving operating characteristic” curve,   also known as the ROC curve. Introduction: This curve helps us understand   how well a test can distinguish  between two groups. In other words,   how well can the test see if the patient does or  does not have the disease. The Y axis is the True   Positive rate otherwise known as Sensitivity.  The X axis is the FP rate, otherwise known as   the complement or “the opposite” of Specificity  or 1-Specificity if you will. In other words,   x axis represents true positive rate (TPR) while  Y-axis represents false positive rate (FPR) ROC Curve: The curve is   defined by its specificity and sensitivity and  associated cutoffs. In other words, how well   it can identify those with a disease compared  to those without a disease based on different   cutoffs is what determines the shape of the curve. For instance, let’s take a group of people we   tested for condition X with a cutoff of 100  units. Suppose after all the calculations,   we ended up with a sensitivity (SN) of  80% and specificity (SP) of 85%. However,   what if we want to see what would happen to  SN and SP if we raised or lowered the cutoff,   for example, when we raised the cutoff by 20  units, the SN dropped to 70% while SP became 90%. What if we lowered the cutoff  by 20 points? Then SN became 90%   and SP dropped to 70%. I’m just making up the  numbers here to keep it simple but notice how   we’re sort of getting a nice curve here. If you  rinse and repeat this with many different cutoffs,   you’ll get a curve that shows you what the SN and  SP would look like based on the chosen cutoffs.   This makes it easier to understand what cutoffs  would be most helpful to apply in each situation   without having to manually write down all the  possible cutoffs and their associated SN and SP. Notice how the lower left corner suggests a  SN of 0% and SP of 100%, this would be like   raising the cutoff so high that you won’t get any  FP. Hmm, still not really that practical, huh?  On the other end of the spectrum, the  top right corner suggests a SN of 100%   and SP of 0%, this would be like having a  very low cutoff, sure you wouldn’t get any FN,   but this isn’t all that practical either. The top left corner suggests a test with   100% SN and 100% SP. That is where we  would ideally want every test to be.  Please note that sometimes, the curve is not  perfectly symmetrical. This is just a testament   to how real life doesn’t always provide you with  perfect curves, but beggars can’t be choosers, eh? ROC Curve Applications: Moving on, how can we apply this concept?   Well, sometimes a test is needed to screen or  confirm diseases. If you are using the test   as a screening tool, your job is to look  for the test with the highest sensitivity,   in other words, the highest point on the curve  within acceptable detriment to specificity.   That is to say, we can’t just take the  highest point overall because this would   leave us with a SP that is very close to 0,  and thus the test would have much less value.   A good example of a high point that doesn’t  completely wipe SP would be like this one for   example. But, if it’s a confirmatory test, you’re  looking for the left most point, so you would pick   this point without completely wiping out SN. Golden rule when using tests in medicine:   if you are SCREENING focus on sensitivity,  but if you are DIAGNOSING focus on specificity  As you can see, the ROC curve is immensely  useful in helping us visualize the different   cutoffs of a test and allows us to make  decisions as to which test or which   cutoffs of a particular test would be better  suited to being a more SN tool or SP tool. When it comes to comparing two different tests,  eyeballing the ROC curves becomes very handy.   Since the top left corner suggests an ideal  test, we want curves to be closer to that corner.   Another way you could go around this  is to check which curve has more area   under it. Every curve will start in the  bottom left and end at the top right. So how   much it approaches the top left corner will also  determine how much area it covers under it. So,   for example, when looking at these two curves,  it’s clear that A is the better performing test. With this paradigm in your mind, we’d like to add  a few more concepts to help tie things together:  Accuracy is another important term and relates  to the validity of the measurement. I like to   think of it as the “trueness” of a measurement to  help remind me of the formula which is TP + TN/   TP + FP + TN + FN (pretty much everything goes in  the denominator, while only the true ones go into   the numerator). A test is considered more accurate  the closer it approaches the top left corner, and   less accurate as it approaches the diagonal line. We wanted to just mention a few points here to   help tie in a few more concepts  in relation to the ROC curve.  Last but not least, since the ROC curve can  reveal to us the SN and SP of a particular cutoff,   we can actually find out TP, TN, FP, and  FN. However, we also need a bit more given,   specifically, we need the amount of people  we are testing and how many of them have   the disease, as known as, the prevalence. Using this given, we can find out the TP,   TN, FP, and FN. Turns out ROC curves  can do a lot of cool things, huh? - In summary ROC curve is built using  2 axes: sensitivity and 1-specificity,   otherwise known as false positive rate - It allows you to choose the optimal   cutoff for a test by looking at the ROC curve  of that test. That way you can have an idea   about the SN and SP of that cutoff - ROC also allows you to compare   between different tests, by calculating the AUC,   to determine which one is better - Lastly, remember this relationship:  SN – Screening for a disease SP – Diagnosing/conforming a diagnosis With that, we hope you benefited from this  video! Consider liking and subscribing,   and as always, thanks for watching!