Screening Tests, Part One

One of the most common tests I performed at the lab was the "Beta Strep Test," a throat swab test for Group A Streptococcus (GAS), the causative agent of "Scarlet Fever" or "Strep throat." It is a simple test really. You swab a person's throat, put the swab in some solution, take the swab out after a minute, drop in a stick, then read the stick for positive or negative. It's so simple that anyone can do it, like a pregnancy test. However, like all screening tests, it is subject to the rules of sensitivity and specificity. That is, you'll sometimes get positives that are not really positives (false-positives) and negatives that are not really negative (false-negatives). False positives are not really a big deal with this test because the patient just gets prescribed simple antibiotics and is on his/her way. False negatives are a problem because, if left untreated, GAS infections can cause some damage down the road.

Let's talk about sensitivity and specificity, throw in some prevalence, then look at positive and negative predictive value... I hope you have your thinking caps on for this one... All after the jump...

He'll be jumping off the "screen." Get it?

CONSIDER THE FOLLOWING
Take a look at the following 2-by-2 table:

Don't worry... I'll explain A, B, C, and D next.

Where A is the number of people WITH the disease who test POSITIVE (true positives). B is the number of people WITHOUT the disease who test POSITIVE (false-positives). C is the number of people WITH the disease who test negative (false-negatives). And D is the number of people WITHOUT the disease who test negative (true negatives).

It's not the only time we'll use 2-by-2 tables in this night school. In fact, they will come in very handy when talking about case-control studies and outbreak investigations. But, for screening tests, we are only interested in certain aspects of the table.

SENSITIVITY
Sensitivity is the ability of a test to truly identify a positive when a patient really has the disease. If the sensitivity is low, there will be more false-negatives. This can be a problem with pregnancy tests, for example. You'll have more women who don't get timely prenatal care if they really are pregnant but test negative. This is also a problem with HIV testing because you'll miss the opportunity to give antiviral medication if you are infected but test negative. In our 2-by-2 table above, sensitivity is A divided by A+C. A is the number of people with the disease who test positive (true positives), while A+C is the total number of people with the disease. For example, if you have 100 people with the disease, and 90 of them test positive, then the sensitivity (90/100) is said to be 90%, and there will be 10 false-negatives (C).

SPECIFICITY
Simply stated, specificity is the ability of a test to truly identify a negative when a patient really is negative for the disease. If the specificity is low, then there will be more false-positives. This can be a bit of a problem if, for example, you're trying to determine whether to give someone an intervention (medication, vaccine, etc.) based on their positive/negative status. A false-positive pregnancy test may delay radiographic imaging. A false-positive HIV test may lead to unnecessary antiviral medication being prescribed. In our 2-by-2 table above, specificity is D divided by B+D. D is the number of people without the disease who test negative (true negatives) divided by B+D, all the people without the disease. For example, if you test 100 people who are disease-free, and 80 of them test negative, then the specificity (80/100) is said to be 80%, and there will be 20 false-positives (B).

POSITIVE PREDICTIVE VALUE
A lot of health care providers are not too worried with sensitivity and specificity. Instead, they are worried about the positive predictive value (PPV) of a test. The PPV answers the question: Of all the people that test positive, how many of them are truly sick? (Specificity, on the other hand, answers: Of all the people that are sick, what proportion will test positive?) PPV is important to providers because they want to have that knowledge in order to act on a positive test.

PPV can be determined from our 2-by-2 table by taking A and dividing it by A+B. A is the number of people sick who test positive (true positives) divided by all the people who tested positive (A+B). For example, 110 people test positive, but only 90 of them are really sick. Your PPV is 90/110 or 81.8%. This tells the provider that about 82% of people who test posi