Biblioepidemiology: the importance of case definitions (6 January 2026)

I keep a record of the books that I've read. As the old year ends and the new one begins, I look back at my list for the year. (Hey, if President Obama can do it, then so can I.) This brings up an interesting question. I tend to read "in parallel." I always have multiple books going at once. This means some of them take quite a while to finish. It is not rare that I'll start one in the autumn and not finish it until after December 31. So here's the issue: when did I "read" a book? In what year? When I started it? When I finished it? When I read the bulk of it? Or something else.

This is a simple example of the importance of case definitions. I need to decide when I will consider a book read. I could choose any of the above: the year I started it, the year I finished it, or what have you. Some people might be inclined toward one definition, and others to another, and that's OK. You just have to pick one and stay with it. I chose the year I finished a book. So for 2025, that makes 22 books for me. If I had chosen the year I started a book, I might have come up with a different count.

A meaningful example

Now to a substantive example. Back around 2012-2014, when much attention started to be focused (again) on opioid overdose, colleagues started to ask me, "How many people die of drug overdose in (our jurisdiction)?" I'm sure this was a common question everywhere. The trouble is, it cannot be answered, at least not as it stands. Leaving aside the trivial issue of "during what time period?" there are several issues around case definitions that need to be hammered out before such a question can be answered. So my answer always was, "Well, pull up a chair and let's talk."

• What do you mean by "drug?" Do you mean the patient on warfarin who inadvertently takes too high a dose or has a drug interaction, becomes supratherapeutic on their INR, and suffers a fatal intracranial hemorrhage? How about overdose of alcohol? Usually my questioner would reply, "Oh no, no, that's not what I mean." OK, so what do you mean? "Well, you know, like durrrr-rrrrugs." (Always two or more syllables.) Eventually we pin things down to overdoses of opioids.
• What do you mean by "in (our jurisdiction)"? Do you mean people who were residents of (our jurisdiction) when they died? And they might have died outside of (our jurisdiction). Or do you mean people who died while physically within the geographic boundaries of (our jurisdiction)? And they might just have been visiting here.

And so the conversation went. Eventually we'd negotiate a shared case definition for their question, and I could provide an answer. And the answer was always accompanied by a brief restatement of the case definition.

Why this matters

Whenever you count something, you have to be very precise about what exactly you are counting, and how you are counting it. There may be different things to count, and different ways of counting---different indicators. Any of several indicators might give some insight into the phenomenon of interest, but we cannot expect them all to yield the same numbers---because they count different things in different ways. When answering questions like this, we should avoid giving "bare number" answers; if we give a number, we should always briefly describe the definition of what that number represents. The best epidemiological understanding of any phenomenon, the best situational awareness, probably comes from synthesizing all available indicators into a coherent picture.