Chapter 22: STATISTICS

STATISTICS OVERVIEW

NOTE: Statistics questions are quick and easy “give me” points on the exam if you know where to put in the numbers and how to use the formulas in this overview. Please MEMORIZE the table below. Later in the chapter, the “validity hierarchy” of different types of research studies is discussed. This was added to PBR based on repeated requests from past PBR members. The boards could ask you which type of study design has higher validity, or you may be asked to choose the best type of study for a particular problem being researched.

CALCULATIONS OVERVIEW

TP & TN = True Positives & True Negatives. FP & FN = False Positives & False Negatives. PPV & NPV = Positive Predictive Value & Negative Predictive Value.

PEARLS/MNEMONICS: Note in the table above that everything you need to calculate is ALWAYS within either one column or one row. Also, letters are bolded as a memory aid. N-P-N-P (seNsitivity, sPecificity, Npv, and Ppv).

STATISTICS TERMINOLOGY RELATED TO DIAGNOSTIC TESTS

The simplest form of a diagnostic test is one like a pregnancy test that yields a positive or negative result. No test is perfect, and a number of statistical concepts describe how accurate a test is. A test is sensitive when it is very good at detecting all the cases of a condition. It is specific when it correctly identifies nearly all the non-cases (people without the condition). A test with a sensitivity of 100% is always positive for cases, and one with a specificity of 100% is always negative for non-cases.

In our example of a pregnancy test, a 100% sensitive test would be positive for every single pregnant person; a 100% specific test would be negative for every person who is not pregnant.

Note that there is a tradeoff between sensitivity and specificity. Any test could be made 100% sensitive just by designing it so that it always reads positive, or 100% specific by making it always read negative. An ideal test would be 100% sensitive and specific, but in the real world we have to compromise.

A common strategy for detecting disease in low-risk populations is to first use a very sensitive test even though it may be relatively non-specific. This will catch nearly all the cases, even though it may also catch a lot of non-cases. Since it’s being applied to a large number of people, most of whom don’t have the condition, a screening test should be cheap, non-invasive, and safe. Then, a very specific confirmatory test is used to rule out most of the false positives caught by the screening test.