DIP Episode 363 - Taking the confusion out of confounding

Okay, welcome. My name is Divine. This is episode 363 of the Divine Intervention Podcasts. Until these podcasts, we're going to be talking about a BIOS, this is a BIOS Starts Podcast. We're going to be talking about something called Confounding. We've got a lot of email requests on this. So I've decided I will literally make a podcast on Confounding and then by the grace of God, time permitting, I'll make another podcast on Effect Modification. Again, this should be a short podcast, but I'll try to make some integrations and give you a lot of very rich examples. So you can truly understand what Confounding is. Now before I jump in, if you're taking the USM Listep 2 CK or step 3 exams, or the complex level 2 or 3 exams within the next few weeks, I do offer a course that's taking place next week. On Monday the 24th, from 2 to 4 30 PM Mountain Standard Time, which is 4 to 630 PM Eastern. We offer an MBME Testicking Strategy course. Again, how tons of people take this course? They've done extremely well on the exams. I've had people where they were percentages of bumped up 20, 30%, just after taking the course, or they bump up 30, 40 points on the MBME practice exams, or they've done really well on the exams. Again, tons of people have taken this course. They've found it to be really helpful. And then between the 25th and the 28th, that's Tuesday to Friday. I'm offering a 24 hour review course. It's going to be from 7 a.m. to 1 p.m. Mountain Standard Time on each of those days.

That's basically from 9 p.m. to 3 p.m. Eastern. I mean, sorry, 9 a.m. to 3 p.m. Eastern. So if you're interested, again, we're going to be going over surgery, peds, internal medicine, OB-guine, psych, neuro, bio-stats, ethics, professionalism, communications, multi-system processes and disorders. Again, many people have taken these courses. They've done extremely extremely well. And we go over like about 2000 concepts caught across all those subjects. And we'll go over them by way of scenarios. We're not going to be doing stock lectures. No, that's not very useful. So we're going to be doing a lot of scenarios. And then I'll go over them. And there is also time about a set one, six of the total course time is devoted to taking and answering people's questions. So again, if you're interested, just shoot me an email through the website. I'll give you some more information. There's still some spots so you can sign up over the weekend. Okay. So let's jump right into it. So what is confounded? Well, basically, a confounder is something that is associated with an exposure and an outcome. So the thing is many times in bio statistics, when we're doing a study, we look at some exposure or some risk factor and try to look at the outcome that's generated from it. The thing is, sometimes what we think, oh, wow, we see this thing, this exposure, we see this risk factor, what we think sometimes is associated with a risk factor.

I mean, what we see as an outcome, sometimes associated with like an exposure or risk factor ends up not really being associated because there's actually something else that is modifying the results that you get. Let me give you an example. So for example, you can say that, oh, you know, if you carry a lighter, oh, let's say you do some study, oh, carrying lighters increases your risk of getting lung cancer. Oh, that I've done this wonderful study, this divine intervention study, and I noticed that, wow, the more people that carry lighters, they have a high risk of developing lung cancer. Well, if you really think about that, yes, you've seen an association, you've seen what is some kind of correlation like, wow, lighters, lung cancer, lighters, lung cancer, that's a correlation. But that doesn't necessarily mean that that's what's causing those people to have lung cancer. The thing is, in this example, something that can be a confounder is smoking, right? Because if you think about it, the person that smokes is more likely to carry a lighter, and present that smokes is more likely to develop lung cancer. I'll say it again, a present that smokes is more likely to carry a lighter, a present that smokes is more likely to develop lung cancer, right? And let me tell you something, the fact that you carry a lighter around does not mean you're a smoker.

You can carry lighters around and be like a park ranger or you can carry lighters around because you live in a country where there is no electricity or something like that, right? In fact, this example I just gave, I would say it will treat three good things to keep in mind when you're reading an in-beaming question, you're trying to determine if something is a confounder or not. First, the confounder must be associated with the exposure. I'll say that again, the confounder must be associated with the exposure. So again, we said this study, let's not call it the divine intervention study because I would not do that kind of study because it makes no sense, right? But smoking is a sort of carrying lighters. Remember, carrying lighters is the exposure here, getting lung cancer is the outcome, right? Being a smoker is a sort of carrying lighters, right? Think about it, how you're going to light up those cigarettes and they're just going to rub your palms together, make it warm enough and boom, light up a cigarette. No, you need to carry lighters around, right? So smoking is associated with the exposure, right? That's carrying lighters. So that's the first true homework of a confounder. The confounder is going to be associated with the exposure, okay? Now the second true homework of a confounder is that the confounder is also associated with the outcome. The outcome here is lung cancer. Smoking is associated with developing lung cancer, right?

So smoking is associated with developing lung cancer, right? So again, the confounder is associated with the outcome. Now the third thing about a confounder is that the confounder must not be a result or it must not be caused by the exposure. Basically, like the exposure must not be a mechanism behind the confounder. I'll say that again, the exposure must not be a mechanism behind the confounder, right? Let me explain this. Carrying a lighter does not make you a smoker. It's not like, wow, okay, if divine goes and buys a lighter, boom, automatically becomes a smoker. No, the fact that I carry a lighter does not mechanistically explain then one becoming a smoker, right? So you need those three things to tell you that you're dealing with a confounder, right? Like for example, for example, people can say, oh, you know, let's say you have a study that, wow, you know, if you drink a moderate amount of alcohol, this is a study I'm sure many healthcare professionals have heard cited in the literature, right? Oh, you know, consuming a moderate amount of alcohol actually lowers your risk of cardiovascular disease. Well, the thing is, if you think about it, if you look at it from a biochemical perspective, right, when you consume moderate amounts of alcohol, that actually raises your HDL. If your HDL goes up, then your risk of cardiovascular disease is going to go down.

So because there's like a mechanism that exists between your consuming moderate amounts of alcohol and raising HDL, then HDL cannot be a confounder because the exposure in this case alcohol is mechanistically related to raising your HDL. There's literally a biochemical mechanism that relates those two things. We're not going to go into that mechanism because this is a bio-statistics podcast, right? Or say, for example, if you look at the relationship between a beast, let's say you do a study and he says, oh, wow, people that are obese are at risk, wow, you notice that, well, put the obese at high risk of cardiovascular disease, okay? Well, if you look at age, age could potentially be a confounder, right? Because on average, people that are older, right? People that are older tend to be more obese. That's just the truth of life, right? You know, they get married, they're eating good food or they make more money, so they're able to eat out more and all that stuff, right? And also, being older is actually not having cardiovascular disease, right? Most people that have cardiovascular disease are old people, right? So age is an example of a confounder. Notice it's associated with the exposure obesity. It's associated with the outcome cardiovascular disease. But notice that obesity is not a mechanism behind your age, right? It's not like, for example, we say this is how old you are based on your BMI. Like you look, you read a person's BMI like, oh, wow, your BMI is 50.

So that means you have 50 years old. No, that's ridiculous, right? So BMI, obesity, like it has no mechanism behind behind each. So age can be a confounder in that kind of study, right? Another example I can give, really the best way to understand confounding is with examples, right? Or they say, oh, you know, the birth order of male children is associated with having preter willing, right? The birth order of male children is a zero preter willing. So what do I mean by that? Oh, like if you have four male children, the fourth male child has a high risk of preter willing compared to the first male child, right? So you'd be like, wow, I found a great awesome study. No, you've not found a great awesome study. You just feel to see the confounder, right? The confounder in this case is paternal age, right? Remember actually paternal age is one big risk factor for person getting a preter willing, right? Like, increasing paternal paternal age, right? So you'd be like, okay, divine. Well, how does paternal age confounder look at it this way? First, let's define it with a three criteria. First is paternal age associated with the birth order of male children. Absolutely, yes, right? Think about it. By the time you're having your fourth male child, you're probably going to be an older man than when you were having your first male child, right? Okay. Now, second criteria is this confounder associated with the outcome. The answer to that is yes. The answer to that is yes, right?

If you an older dad tends to have more kids with preter willing, right? But let me ask you this, is there a mechanistic association between the number of birth children and the age of the father? No, absolutely not. Again, you cannot look at the number of birth children, like the number of male children you have and say, oh, wow, that means the father must be old. Like, wow, he has six male kids. He must be 15. No, there are some people in this world that have six male kids and they are not even 20 years old yet, right? So again, that meets the criteria. Oh, wow, okay, we're likely dealing with some kind of confounder, right? Another good example you can think about is like, oh, wow, okay. You do some study and you notice that, wow, you know, there's an increased risk of getting the flu if you take the flu vaccine. Wow, like, you know, if you take the flu vaccine, you have an increased risk of getting the flu. Well, one potential confounder there can be being a healthcare worker, right? Being a healthcare worker, right? So being a healthcare worker can be a confounder. Again, let's look at our three criteria. The thing about me and I wish more people had this orientation in life be very standardized about the way you do things. Like, to be honest with you, this is one of the things I teach in my test taking strategies class, standardization. You've got to be standardized. You see many people, they take exams in very non-standard fashion, right?

They're very emotional about the exams, right? But if you, the key thing you should do is ask yourself, what are some reliable principles behind doing well on tests, okay? Then use those principles every time, regardless of how you feel. If you do that, then you're going to get good results from taking that test, right? You're going to get consistent results, right? Like, for example, again, many people have heard me talk about Kawaii Leonard. I'm not a clippers fan, but Kawaii Leonard is very standardized in the way he plays basketball, the way he shoots, the way he does things, or he's very standardized, right? So like, for example, let's look at our criteria. We've used this criteria over and over again, right? Well, think about it. It's being a healthcare worker associated with the exposure of getting the flu vaccine absolutely, right? Most healthcare workers get the flu vaccine, right? Your chance of getting a flu vaccine is much higher if you're a healthcare worker, okay? That's criteria one. Criteria two. It's being a healthcare worker associated with getting the flu absolutely. Why? Because if you're a healthcare worker, chance is almost like the people you interact with for most parts of your life are people that are sick. So since you have like a richer exposure to sick people, you have a higher risk of developing the flu, okay? Now, let's look at the third criteria.

Is there a mechanistic relationship between getting the flu vaccine and being the healthcare worker? Like, does getting the flu vaccine make you a healthcare worker? The answer to that is, no, right? There are many people that I've got in the flu, you can't say, oh, you've got in the flu vaccine, oh, that means you must be a healthcare worker. No, you can't, you can't do that, right? There are millions of people all over the world that I've got in the flu vaccine that are not healthcare workers, right? So being a healthcare worker, is a confounder in that circumstance, right? So the thing is our friends at the MBA means, again, they recognize that people just memorize stuff, right? But the thing about the USML is just about being good at application, not just understanding, right? Again, there's this gradation of knowledge, understanding and wisdom, right? You know, you need to acquire knowledge, that's true. We need to understand and understand that knowledge, that's true. But then you need to apply that knowledge, that's wisdom. Wisdom is just basically the prudent application of knowledge. Although many times you cannot prudently apply knowledge of not understood. So you see some people, you know, like two to people one or one, and you see some people, one, you don't have the right knowledge, they're using the wrong resources, right? Or some people, they have the right resources, they just don't have the right to understand it, okay? Right?

Or some people, they have the right knowledge, right, understanding, just cannot apply. Again, that's why my, with my review course, I try to do this thing where we use examples, right? I use clinical vignettes for pretty much everything so that you can just see how those things could be applied. That's why you see even with my podcasts. I love to use rich, rich, rich examples, right? So let's talk about how this could be applied in the context of an NV Me question, right? So say, for example, we look at this whole thing of, oh, you know, current lighters is associated with an increased risk of developing lung cancer, okay? And you, you know, you do this also awesome study, you notice that wow, the relative risk is like 10. You're like, wow, okay, we're going to get a publication in, in a nature or, you know, the BMG or something, I'm going to show people my also awesome findings, right? But then your PI says, wait a second, how about we try to divide up this study? Let's, let's, let's, it's stratify this study, right? That's actually one of the best ways to deal with confounded, right? One of the best ways to deal with confounded. You're like, okay, you know what, let's, let's hold up before you start publishing in nature. Let's kind of break this thing down. Let's look at people that are smokers and people that are non-smokers.

So we take a bunch of smokers, we look at the lighter, like the, the lighters they use and, you know, a number of lighters and the ancestral lung cancer. And they will look at non-smokers, a number of lighters they have, the ancestral lung cancer. And you notice that wow, okay, that thing that you see, you notice that wow, by stratifying, you then see the effect, you actually notice that wow, within the group of people that are non-smokers, the risk of lung cancer is pretty equal amongst those people, regardless of the number of lighters they use. And then, wow, in the smoking group, wow, they have a much higher risk. And the risk is roughly the same between that group of smokers. That's what stratification is, right? Stratification is a great way to see that confounding has happened in a study, right? It is a very good way to see that confounding has happened in a study. Why? Because again, when you separate those people, you then begin to see that wow, oh, I see, this thing I thought that actually existed, it's actually this thing that's controlling those states, right? So say, for example, if you look at that birth order of male children and pretty willy, right? And you say, okay, you know what? Let's break this thing down by paternal age. Let's look at dads that are 20 years old. Let's look at dads that are 25 years old. Let's look at dads that are 30 years old. Let's look at dads that are 35. That's that are 40.

You notice that wow, okay, if you look at the risk, if you look at 20 year old dads, right? That have one child, two, one male child, two male children, three male children, four male children, five male children, you'll see that wow, the risk of preterwillis syndrome is actually equal amongst dads of this age group, right? It's actually the same. So you're stratified by the confounder paternal age. So you look at wow, oh, wow, you know, in a group of 20 year old dads, if you have one male child, it doesn't matter. Two male children, three male children, four male children, all of them. This again, within that same group of 20 year old dads, right? They all have the same risk of preterwillis, but then if you then compare those groups, you stratified. Let's say or let's even take this further. Again, I want to make this painfully obvious to you. You look at 40 year old dads, right? 40 year old dads that have one male child, two male children, three male children, you notice that wow, within this group of 40 year old dads, if you have one male child, the risk of getting preterwillis in that one male child is the same as the risk of very similar to the risk of getting preterwillis. If you have five male children as a 40 year old dad, right? So again, that's one very good way to see that confounding has happened in a study, right? Just by stratification, stratification, stratification.

You'll notice that wow, that risk that you thought you saw gets pretty much disappears when you begin to do your analysis by the age group, by the by the by the confounder, I mean, I've used age a lot as a confounder by the confounder that exists, right? Or say for example, you know, you look at the risk of, you know, people that get the flu vaccine and them getting the flu, right? You then stratify, okay, you know what? Let's do this analysis in a bunch of people that are healthcare workers and in a bunch of people that are not healthcare workers, let's kind of separate them that way, separate the data, kind of piece it out better. You notice that the risk of flu vaccine, people getting the flu vaccine, getting the flu, right? It's going to be very similar within the cater or category of people that are healthcare workers and then you also notice that wow, within the people that are not healthcare workers, they have similar risk, right? Of even if they've gotten the flu vaccine, they have similar risk of getting the flu, right? But if you compare, you then see that wow, okay, you notice that wow, people that are healthcare workers they seem to be getting the flu more than build and are not healthcare workers, again, by just teasing the data apart by means of stratification, you can really see what is going on, right? So again, the MBA means they can easily give you a question, right?

They can easily give you a question where you notice that they give you like a population of like a thousand people, right? Oh, number of lighters, you know, used associated with increased risk of lung cancer, right? And you notice that wow, okay, the relative risk is 10. But then, and many times they like to make these things those are passage questions, right? Where they give you like a drug out or something. And then you notice that they didn't decide to do some kind of subgroup analysis in the question. And that's a group analysis. You notice that wow, okay, if you take everybody together, the relative risk is like 10, right? You know, lighters and getting lung cancer, lighters, you know, number of lighters is the risk factor or exposure, getting lung cancer is the outcome. And then you notice that wow, okay, you look at the people that are non-smokers and number of lighters and risk of developing lung cancer, you notice that wow, the relative risk just for that subgroup drops to like one. But then, you then notice that wow, okay, for the the smoking group, the relative risk is like 20, right? When you notice that wow, these relative risks change when you do a subgroup analysis, that tells you that you are likely dealing with a confounder in that circumstance, right? Or if you notice that wow, the risk of people, because again, there are many ways they can present this, right?

Like you see in the study that wow, the number of people that have, you know, the number of lighters you use, increase your risk of lung cancer, like, oh, if you use one lighter, you have a 10% risk of developing lung cancer, two lighters, you have a 20% risk of developing lung cancer, three lighters, 50% risk of developing lung cancer, whatever, right? But then you notice that wow, okay, let me do the subgroup analysis and then you look at non-smokers, I noticed that in non-smokers, if you use a one lighter, your risk of developing lung cancer is 10%, use two lighters, your risk of developing lung cancer is 10%, three lighters, risk of developing lung cancer is 10%, you notice that wow, the percent risk is the same across the number of, like, it doesn't matter how many lighters you use within that group of just non-smokers, then that tells you that the number of lighters you use has no association with you being a, being getting lung cancer, it's smoking that is the big thing, right? So again, there are two is they can present that, you can notice that in the subgroup analysis, for each different subgroup, the relative risk changes, right?

So like I said, let's say you're studying everybody, the relative risk is 10, again, I'm repeating, repeating, repeating, repeating, but I really want to get this into your head, the relative risk is 10 for everybody, number of lighters and risk of developing lung cancer, but then you do the subgroup analysis, now you notice that wow, for one group, the relative risk just drops precipitously, for the other group, the relative risk just rises precipitously, that tells you you're dealing with a confounder, right? That's one way they can present it, or within the same group, so let's say we're just looking at, oh wow, let's just look at non-smokers, you notice that the number of lighters they use seems to have no association, like the risk of getting lung cancer is similar, regardless of the number of lighters you use within one group of people, right? Then that also tells you that you're dealing with a confounder, again, I'm explaining this, but I really want you to get it down, if you get it down, then it's going to really make your life easy, and notice I already talked about this, how can you analyze a study to see or to a limited effect of confounding, right?

I think you can do as a post-test thing, again, it's stratification, but at this some things you can do on the front end, are they pre-test because again, the MBA means, again, they just want to make sure you understand, they want to make sure you understand, so one thing the MBA means can do is they can see which of the following pre-test interventions would reduce the risk of confounding in the study, or they can say which of the following post-test interventions can reduce your risk of confounding in this study, right? So guess what, do you know what they will do? They will give you different measures that can all reduce confounding, but some of them will be pre-test interventions, some of them will be post-test interventions, stratification is something you do after the test has been concluded, after you've got in your data, right? It's a post-test intervention, right? But what are some pre-test interventions, you can do, a pre-test intervention you can do is something called restriction, right? It's something called restriction, restriction is where you're like, you know what? Just to make sure that I don't have confounding, I'm going to design a very narrow scope of people that are going to come into the study in the first place, so you can say, you know what, you know, I do want confounding to be a problem, so for this study I'm doing, I'm only going to limit it to people that are between the ages of 40 to 45, right?

And they must be male, they must have this demographic, they must have this, this, this, this, this, you're putting all this criteria in place, right? So that you just narrow down the criteria, so that you are very selective about who comes into your study, so that you don't get confounding. The only problem though with that kind of approach is that one is going to vastly reduce the number of people coming into your study, right? Because you've put so many criteria, right? So it's almost like, wow, they have very few people that meet this criteria, it's just like, for example, if you say, oh, you know what, to be admitted to my medical school, your college GPA must be four, you must get a score above a 525 on the MCAT, you must have published 20 papers, you must have done, you must have been a road scholar and a four-bright scholar already, right? The more criteria you have, well, the fewer people that are going to get into that medical school of yours, right? Just too many criteria, right? So restriction is a good way, it can help you get rid of confounding, but it can put some problems in your study, again, it can reduce the number of people that are coming into the study, right? Because again, you put too many criteria, it can make it very hard for you to find subjects for your study, right? It can make it very hard for you to find subjects for your study. And also get this, it limits your study because your study becomes less generalizable, right?

Your study is so narrow in scope, it's so narrow in the demographic of the participants that you're studying, right? That your study does not become super generalizable because if you're studying people who think the age is of 40 to 45, it's going to be really hard to apply the results of that study, the people over 70 or people in their 20s, right? The thing is, you may be like divine, why are you spending time going over why restriction as a study, as a pre-test intervention may not be a great idea. I'm going over these limitations because they are things that could be tested on an exam. Again, the NBA means they are very fond, especially with these drug-at-questions, they are very fond of asking questions where they say, oh, which of the following characteristics of this study limits the general flexibility of the study results or this or that, right? Again, just understanding these factors, you can pick them out of a line or pretty easily on exams. A pre-test intervention you can use to limit confounding is restriction, right? Another pre-test intervention you can use is matching, right? It's something called matching. What is matching? Matching is basically, this is, I feel like maybe the best way to explain matching is just what happens in a case control study, right? You look at a bunch of cases, you look for a bunch of controls, but you try to make those people as similar as possible, right?

So you're like, oh, cases that meet these criteria, controls that meet these criteria, right? So by doing that, again, you're like getting people as closely matched, basically, getting people that resemble each other very well. The only thing that they are not very similar in is in some risk factor or something you're trying to examine, right? But again, this is why case control studies, it has very similar problems to the restriction business, right? Because again, just finding the right pair, right? The right pair, right? So finding the right person to pair up, like the right control to pair up with your case, or finding the right case to pair up with your control may be really hard. So many of the limitations you find by using that restriction method, right? It's something you also find in this matching method, right? But also another pre-test intervention for limiting confounding in study or x-effects, right? It's just randomization, right? Randomization, if you randomly assign people to groups, chances are that randomization process will kind of abit reach out or get rid of some of these little discrete factors that may be a difference between groups that can, again, introduce confounding in your study, right? So again, I mean, sure, many of you, this idea has been rammed into your head so many times that, oh, you know what? Correlation does not imply causation. Correlation does not imply causation.

Let me tell you this, there isn't that many bio statistician see that correlation does not imply causation. It's because you don't know of every confounder that exists because that correlation you may be seeing, right? Maybe caused by a confounder that you presently do not see, right? So again, I know some of you may be like, wow, divine, just spend 30 minutes roughly talking about confounding. Again, if you understand it deeply like this, then when you present a question to your exam, it'll be a piece of cake for you to deal with, right? Again, remember, there are three criteria you must meet to be a confounder. One, the confounder must be associated with the exposure or the risk factor we're dealing with. Two, the confounder must be associated with the outcome that we're examining. And three, right? The confounder must not be mechanistically caused by the exposure, right? It must not be mechanistically caused by the exposure. If you keep these factors in mind, any confounding question you get, you're going to understand it, right? So by the grace of God in a future podcast, I'm going to make a podcast something on effect, effect modification, and then I mean try to compare and contrast both of them. So as I do at the end of every podcast, I do offer one and one tutoring for me in exams. Again, step one, step two, seek is step three, complex level one through three. The only thing I don't do for is OMM, but I do for pretty much any other thing.

And then I also offer longitudinal tutoring, right? So if you're, I've done this with quite a number of people, you know, I'll tutor you for like all your medical exams. But then at the same time, by me having tutoring for those medical exams, I'm also heavily preparing you for your upcoming USMEL exam. Also, I would also say that, you know, I, you know, obviously I tutor for regular medical exams, 30-year clerkship shelf exams. And again, I do offer these review courses for step two, seek is step three, complex level two and three. Again, the 24-hour review course and the NV Me testing strategy scores. And then I do have these podcasts on Apple podcasts, Google podcasts and on Spotify. So if you subscribe, you'll see the most recent 150 podcasts. If you want everything from episode one, you want to make sure you go to the website, divininterventionpodcasts.com. If you actually sign up for Word Press and subscribe to me on, you know, subscribe to website divininterventionpodcast.com. You will get an email notification whenever I drop a new podcast. And also have a You Tube channel in its core divin intervention, USMEL podcast and videos. You subscribe. Again, that's where I post the videos that I make. And then many people have gotten more meals than I can remember. Like, oh, wow, Divine, I really love your life lessons podcasts. They made a big difference for my life, right? So like, you know, where I talk about a life lesson at the end of a podcast.

And I then decided to start a new website. It's called divininterventionlifelessense.com. You know, so today I believe I made episode 53 today. Again, it's a lot of Bible based teaching. Many of you listen to this podcast, not my Christian. Just Bible based teaching that addresses just a common problem of humanity. So if you're interested in that, just go on to divininterventionlifelessense.com. I do have the podcast on Apple podcasts. Again, it's called the Divine Intervention Life lessons podcasts. And again, I try to make like one or two podcasts a week. I try to make two, but you know, things, you know, sometimes you just just get really busy. But I guess one thing I want to maybe go ahead and wrap up with is just understanding the value of time. So I want to throw in a small life lesson here. Understand the value of time. I mean, there's this part of the Bible that says that we should redeem the time because these are evil. Many people don't respect time at all, right? The thing is you'd be surprised at how much more you can get out of your day if you just put time stamps behind the things you do. Even your leisure, if you plan out how much sleep you want to get, you will get more sleep, right? But many people, you know, they don't want to go with any plan, right? They just say, yes, okay, how long do you want to study for? You know, as long as I'm able, I'm really willing to do whatever is required. No, that is that is not a helpful answer.

Like whenever I'm tutoring people, right? Many times I if I'm making a study plan for people to ask them, okay, how many hours are you going to be able to give consistently every day? And if they tell me, well, to find, you know, I can I can put in as many hours as required. I'm like, that's not the answer I'm looking for. How many hours can you put in, right? If I know that you have eight hours, then I can make a plan that can help you get the most possible ring out every ounce of nutrition from those eight hours, right? So respect time, right? Respect time. The thing is, again, kind of shared this. So if you again, if you want to get more on this, listen to episode 53 on my life lessons podcast. But I shared this that, you know, as a physician by the grace of God, you know, I've worked with people that I'm in the ICU. And when people know they have terminal disease, they have actions adramatically different from people in the world that think that their next 10, 20, 30 years are completely guaranteed, right? So many times in medicine, we have this term, high, high, high, high, high, live a high, high, live a high, live a high, high, just plan out to the end, say, okay, all from eight to 90, I'll get this done. From nine to 10, I'll get this done. From 10 to nine, I'll get this done. Literally just that little intervention, can make you get more out of your day, get more out of your year. So I'm encouraging you today.

Don't let, don't leave life and say, oh, let me see where the wind will take me. No, that's a terrible way to live life, right? Have a plan for your life. Have a plan for your hours, right? Plan your hours, respect time, respect time. Many people, and I guess maybe one thing I should say with this timing of a thing is, it's not enough to just plan out your time. Make sure that the activities you putting your time are useful for your life, right? Many people, they spend their time on things that just have no bearing on your future, right? The thing is, you can see some people, they look busy, but they are busy in the wrong things, right? And then they notice that, wow, my life is not fruitful. I'm not getting anything out of my life. Well, hello, maybe try to use your life for things that are productive, right? So I'll encourage you to just be careful with the way you use time, respect time this year, and you'll get a lot more from your time. So thank you for listening to me. Again, I wish you all the very best, and I will see you in the next podcast. God bless you. Thank you.

1) Associated with the exposure; 2) Associated with the outcome; and 3) Must NOT be mechanistically caused by the exposure (or vice versa).

Age is often cited as a confounder, as older people tend to be both more obese/at higher risk for CVD and are also generally more prone to developing CVD.

It vastly reduces the number of participants in the study, making the results less generalizable (narrower scope).

Stratification. This is a post-test intervention.

Matching cases and controls based on age (or other key variables).

It suggests mediation, meaning the variable is an intermediate step in the causal pathway, not a confounder.

When an observed association between an exposure and an outcome is distorted or misleading because of a third variable (the confounder) that is associated with both.

Stratification is a post-test analysis method (analyzing data after collection), while Restriction is a pre-test design choice (limiting who can participate).

1) It is difficult to find suitable pairs (right control for every case); and 2) Like restriction, it limits the generalizability of the findings.

Randomization, as it distributes all known and unknown confounders evenly across groups.

Alcohol consumption $\rightarrow$ raising HDL levels $\rightarrow$ lowering CVD risk (HDL acts as a mediator).

Paternal age. It is associated with both the exposure (birth order) and the outcome (preeclampsia), but not mechanistically caused by it.

When an observed association between an exposure and an outcome is distorted or misleading because of a third variable (the confounder) that is associated with both.

Stratification is a post-test analysis method (analyzing data after collection), while Restriction is a pre-test design choice (limiting who can participate).

1) It is difficult to find suitable pairs (right control for every case); and 2) Like restriction, it limits the generalizability of the findings.

Randomization, as it distributes all known and unknown confounders evenly across groups.

Alcohol consumption $\rightarrow$ raising HDL levels $\rightarrow$ lowering CVD risk (HDL acts as a mediator).

Paternal age. It is associated with both the exposure (birth order) and the outcome (preeclampsia), but not mechanistically caused by it.