IPPCR 2019 Issues in Randomization Part 2 of 4

>> Paul Wakim: Hello. I'm Paul Wakim. I am chief of the Biostatistics and Clinical Epidemiology Service at the NIH Clinical Center. And this is part two of a four-part segment, and we're going to talk in part two about the -- how to randomize. And we're going to continue that, also, in part three.

So, let's start with how to randomize. Actually, we are going to start with how not to randomize. And so, you can randomize by birthdate. You can say odd and even. Or you can say the last digit of the medical record number or the odd and even hospital rooms numbers. These are examples of how not to randomize.

You're going to say, "Well, why not? It sounds pretty objective, pretty random, and pretty --" well, maybe not. You don't know. That's the problem. Is you don't know what other things go along with these kind of birth rates, birthdates, and digits of the medical record. But I could like -- for example, for the hospital,

the room number, let's say, for example, the odd rooms are in they have windows. And the even rooms don't have windows in the hospital. And so, you do by odd and even, and let's say that for some reason having a window makes recovery faster or better. And so, you're not randomly assigning them.

So, again, there could be confounding -- unknown confounding factors that influence recovery or outcome, or whatever you're measuring, your primary outcome measure. So, avoid these kinds of randomization schemes. So, I just want to point out here at this point we're going to go through these five randomization methods, simple randomization, permuted block, stratified, cluster, and adaptive randomization.