Latest / This Is Why with Dr. Busti / Types of Study Designs in Clinical Research Explained & Made Easy

Transcript

Types of Study Designs in Clinical Research Explained & Made Easy · This Is Why with Dr. Busti

• 0:00
  SPEAKER_00: Hey guys, this is Dr.
• 0:01
  SPEAKER_00: Bustaye, and we're in the Study Design Development and Interpretation Module or Playst, and this lecture will focus on the core basics of study design.
• 0:12
  SPEAKER_00: Now, when we think about the process of doing clinical research, at least in its broadly described process by the NIH, we move through several steps.
• 0:21
  SPEAKER_00: We develop the study concept, we secure funding as needed, we prepare the study protocol, which is really where the study design comes to really have a big influence.
• 0:31
  SPEAKER_00: We protect patients, we we do uh conduct and monitor the study, we close out the study, analyze it, and obviously communicate those findings to our colleagues so that we can apply it.
• 0:42
  SPEAKER_00: But everything really starts at the study concept stage.
• 0:47
  SPEAKER_00: Nailing down the best study design to answer your study hypothesis is absolutely essential.
• 0:54
  SPEAKER_00: In fact, there are now guidelines for protocol development that you would submit to the IRB if you're a researcher, right?
• 1:02
  SPEAKER_00: Before you start.
• 1:03
  SPEAKER_00: Those are things known as the spirit guidelines.
• 1:06
  SPEAKER_00: And doing that not only helps to improve the quality and transparency of the research, but also many journals now require it.
• 1:16
  SPEAKER_00: In fact, you also may be required to prospectively register a clinical trial with a group like clinicaltrials.gov before you also start.
• 1:26
  SPEAKER_00: In other words, understanding study designs, how you publish, what journals, what are the requirements, their pros and their cons, okay, all go into protocol development, but also help facilitate the success of publication.
• 1:45
  SPEAKER_00: As a reader, it also gives you better content to decipher.
• 2:10
  SPEAKER_00: And that changes everything about the patient decisions that we make every day, even from the literature.
• 2:16
  SPEAKER_00: Now, in this module and series, each lecture is part of a larger learning sequence.
• 2:22
  SPEAKER_00: So if you find this helpful, be sure to explore the collection or playlist that you have access to on YouTube, the podcast channel, or at thisiswide.health.
• 2:32
  SPEAKER_00: Now, these topics do and can stand alone by themselves.
• 2:38
  SPEAKER_00: But if you need that general step-by-step process, there are some areas that do build upon each other.
• 2:45
  SPEAKER_00: And with that, this is the module that we're in where we've already covered things like hypothesis testing.
• 2:51
  SPEAKER_00: So if you haven't seen that lecture, you want to go to it, again, go check that out.
• 2:56
  SPEAKER_00: We're in the particular area of study design, and that's where we're going to be anchoring our time.
• 3:03
  SPEAKER_00: And then there's also types of bias and internal and external validity.
• 3:07
  SPEAKER_00: All of these things are essential to be at the beginning stage if you're a researcher in putting a protocol together.
• 3:14
  SPEAKER_00: I don't care what the type of research is or what study design is.
• 3:17
  SPEAKER_00: You should think ahead about these things.
• 3:21
  SPEAKER_00: As a reader, clinician, you should be thinking what should the researcher have done, okay, and or as you're reading the protocol, knowing that no study is perfect, but it guides you into the overall appraisal of the paper so you can apply that information appropriately in its proper context.
• 3:41
  SPEAKER_00: Now, we've talked about the steps, okay, and again, this flow that I'll repeat over and over again in these lecture series on this particular topic of clinical research is to burn in the process that you should go through.
• 3:55
  SPEAKER_00: When you understand the process, then you are oriented to your success.
• 3:59
  SPEAKER_00: But here's where we're at.
• 4:01
  SPEAKER_00: We're at this level of the development of the study concept and preparation of that protocol.
• 4:08
  SPEAKER_00: Remember, that's where those spirit guidelines come in that you may even have to publish the protocol, okay?
• 4:15
  SPEAKER_00: Or you may go out and see if someone else is doing a study on these registries that are available that we'll talk about at the end, and see if somebody is already doing the study that you're doing or you want to do, right?
• 4:29
  SPEAKER_00: Or answering your hypothesis and the questions you have.
• 4:32
  SPEAKER_00: But it also is important as we go through the study design to think about where are you going to put this paper or where are you planning to try to present the information.
• 4:41
  SPEAKER_00: Those journals may have expectations, and I mentioned that at the beginning.
• 4:46
  SPEAKER_00: I'll point back out publishing guidelines.
• 4:49
  SPEAKER_00: So you have pre study guidelines that help you structure things, and then you have the analysis, and then, of course, you have the communication to the journals, and they have expectations that you may be required to document.
• 5:04
  SPEAKER_00: And so here we are over here in the study design step.
• 5:09
  SPEAKER_00: Okay.
• 5:10
  SPEAKER_00: You have already developed what is the main question or hypothesis, so that gives you your null and alternative hypothesis.
• 5:16
  SPEAKER_00: And in the next couple of series, we're going to talk about ways and types of bias and ways to prevent that.
• 5:21
  SPEAKER_00: Then we're tying in the internal and external validity.
• 5:25
  SPEAKER_00: Does the information that is produced in this paper and the results actually are they real?
• 5:30
  SPEAKER_00: Can you extrapolate that information to your general practice?
• 5:34
  SPEAKER_00: So again, for the researcher, this should be done before the study.
• 5:39
  SPEAKER_00: Someone who sees it and experience, and this is where if you're early on in your training, where a mentor can really guide you, if you're the clinician, you're trying to really understand what that researcher should have done.
• 5:50
  SPEAKER_00: Again, with some degree of grace, knowing that not everything is perfect like we want it to be.
• 5:57
  SPEAKER_00: There's things like cost, access to resources, time, all of these things go into that.
• 6:03
  SPEAKER_00: And you'll see that when we go into the study design, some of the pros, cons, issues to consider.
• 6:09
  SPEAKER_00: Okay.
• 6:10
  SPEAKER_00: So not everything can be a randomized controlled trial like we want it to be.
• 6:14
  SPEAKER_00: It's just not feasible.
• 6:17
  SPEAKER_00: We'll come back to this, but why are we showing you this study design hierarchy?
• 6:22
  SPEAKER_00: Well, this is a traditional pyramid of you know, basically the amount of the degree of systematic error or the risk of bias.
• 6:31
  SPEAKER_00: So at the top, Cochrane reviews of well-designed randomized controlled trials, man, they're they're the best, right?
• 6:39
  SPEAKER_00: Because Cochrane follows a very structured, systematic, rigorous process for how they analyze the quality of a study, the bias in particular studies and a group of studies, and then do a quantitative analysis.
• 6:55
  SPEAKER_00: All of that is important to reducing the overall risk of bias and trying to minimize the systematic error that's in there.
• 7:05
  SPEAKER_00: So as we go down that, you can see that cohort studies, as we get into observational studies, case reports, case series, those, you know, the risk of bias starts to really increase, right?
• 7:16
  SPEAKER_00: Down into where you even get editorial or opinions of experts, right?
• 7:21
  SPEAKER_00: We're all biased by what we know, perceive, what we've experienced.
• 7:25
  SPEAKER_00: Um, and so that is the general hierarchy that we're and we're gonna be kind of going through the structure of many of these study designs.
• 7:34
  SPEAKER_00: So let's start out with the most basic, descriptive, or what we would consider non-analytic.
• 7:42
  SPEAKER_00: We're not doing uh statistical analysis on these kinds of study designs, and they're they're fairly straightforward.
• 7:48
  SPEAKER_00: You're documenting an experience.
• 7:51
  SPEAKER_00: They begin to search for some of these initial explanations that maybe nobody has ever seen before.
• 7:58
  SPEAKER_00: It's not we don't, it's an observation, it's new, right?
• 8:02
  SPEAKER_00: It can be hypothesis generating that leads to, right, and gives you the reinforcement, it sets up the stage and the support for doing additional analysis.
• 8:14
  SPEAKER_00: And those include things like case reports, case series, cross-sectional studies that are survey-based, okay, so non-analytic, qualitative studies, which we'll break down, compare that to quantitative studies.
• 8:31
  SPEAKER_00: Okay, this is a growing area of research and study design.
• 8:36
  SPEAKER_00: Very much harder to do, but very important because it's talking about that human experience.
• 8:40
  SPEAKER_00: And then big data epidemiology, if it's more descriptive, big data epidemiology tends to be more observational type of studies in most cases where there's some quantitative assessment or analytic and analysis being done.
• 8:56
  SPEAKER_00: So, what do descriptive studies do and not do?
• 9:02
  SPEAKER_00: Okay, well, descriptive studies act as a stepping stone.
• 9:06
  SPEAKER_00: I've said that all right, they they can be uh helping us to generate a hypothesis, they can be supportive background information that lead to the reason another study needs to be done because we're subputting people's even information, individuals and their uh their bodies and their biology subject to risk of exposure, all right, and being involved in studies.
• 9:30
  SPEAKER_00: And so we don't want to just do things to do them for no good benefit.
• 9:34
  SPEAKER_00: So sometimes we have to lay the groundwork.
• 9:36
  SPEAKER_00: And those studies are fine.
• 9:37
  SPEAKER_00: It's if it's the best available evidence, it's the best available evidence.
• 9:40
  SPEAKER_00: But they do not, guys, provide an explanation for the cause of a disease, or they cannot provide evidence to evaluate the efficacy, and I would even argue effectiveness, right?
• 9:55
  SPEAKER_00: So that pragmatic piece of the interventions that we do, they cannot prove causality, and that is really, really important.
• 10:06
  SPEAKER_00: You you people will talk about something was studied and published, and it's like, but was it able, was it designed to prove causality?
• 10:15
  SPEAKER_00: Was it an explanatory study design?
• 10:19
  SPEAKER_00: Now, descriptive studies, okay.
• 10:21
  SPEAKER_00: So here we are, and this is a table that we put together and is available at you know thisiswhy.health for you to download for free.
• 10:31
  SPEAKER_00: Um it is summarizing the types of studies or publication, right?
• 10:37
  SPEAKER_00: And then the statement here in the middle, okay, the statement here in the middle, which is that guideline statement that helps you in the way you report it in the literature.
• 10:48
  SPEAKER_00: Again, many journals will require you to follow that structure.
• 10:54
  SPEAKER_00: Some of them make you sign a statement that you follow that structure.
• 10:58
  SPEAKER_00: Some of you have to document that in the actual paper, but it becomes obvious, right?
• 11:04
  SPEAKER_00: And so I've also provided you for those of you that are wanting to publish, there's the links there for those tools.
• 11:11
  SPEAKER_00: If you're reviewing a paper and you're really diving deep into the methodology that was done, knowing these publication guidelines are also helpful to make sure that the paper that you're reading was done correctly.
• 11:25
  SPEAKER_00: So here we are down here at kind of case reports, really is that bubble that we're in, right?
• 11:31
  SPEAKER_00: And Care is a UK-based organization.
• 11:37
  SPEAKER_00: They do a lot of different checklists, but that one can be a guide.
• 11:41
  SPEAKER_00: Now, this is different than some of the tools, which we'll talk about here in a second, that help us to add to elements of a case report, okay, and I'll point those out here.
• 11:52
  SPEAKER_00: So, for example, if you have adverse drug event, which is a lot of times a case report, you made an observation, someone got a drug, something happened, this is the you know the thing that resulted.
• 12:04
  SPEAKER_00: You're trying to determine is this plausible, right?
• 12:08
  SPEAKER_00: And so there are tools that have been published out there, and this one in particular, which is still one of the most highly cited, recommended by journals, been around for a long time since the 80s, right?
• 12:21
  SPEAKER_00: But at the Neuronho ADR probability scale, it's 10 questions, and you know, you get to the more points, the more definitive that there likely is something going on there.
• 12:33
  SPEAKER_00: Okay, I've published case reports before, and I used this particular tool, and it was very helpful in not only getting successful publication, but really helped me to wrap my mind around the likelihood for doing that.
• 12:48
  SPEAKER_00: There are other tools that are out there that some journals may also recommend, some even further divided into areas of specialty or practice.
• 12:59
  SPEAKER_00: Okay, the World Health Organization has one Liverpool ADR causality assessment tool, less common, but the LCAT can be also one of those tools that you might see.
• 13:10
  SPEAKER_00: So I wanted to make sure that you you had that.
• 13:14
  SPEAKER_00: What about the appraisal process?
• 13:17
  SPEAKER_00: Your review, you're the reader, you're now critically appraising.
• 13:21
  SPEAKER_00: And remember, I I don't like the word critical because many times we always look out the negatives of everything in the published in the literature.
• 13:28
  SPEAKER_00: When if you've ever published papers and done research, it is not easy.
• 13:33
  SPEAKER_00: So be again, show some grace, but again, you want to make sure that you appropriately evaluate all the parts of a paper, and that's what CATs are designed for.
• 13:44
  SPEAKER_00: Okay, and so over here where we have case reports and case series, JBIGlobal.global is one of the tools that helps us in assessing the proper publication and basically the parts of a case report and case series.
• 14:05
  SPEAKER_00: Qualitative studies down here again, CASP, again, another US UK-based group.
• 14:12
  SPEAKER_00: It's a tool, and it may seem kind of elementary for some people.
• 14:18
  SPEAKER_00: Um I don't care who you are, I don't care how much experience you have.
• 14:22
  SPEAKER_00: We all start to develop bad habits, sometimes we skip things, right?
• 14:26
  SPEAKER_00: And if you have a systematic reproducible approach, these tools will help you with that if you don't have one, but doing that helps you to reduce the chance of missing something really important.
• 14:40
  SPEAKER_00: Okay, now this leads us to the next sort of major category.
• 14:45
  SPEAKER_00: So we just finished up descriptives, non-analytic study designs.
• 14:50
  SPEAKER_00: Now that we're getting into explanatory or analytic study designs, which are really broken down into those that are experimental in nature and intervention is being applied by the researcher, or there's the observational studies, which there's no action being made by the researcher, okay?
• 15:11
  SPEAKER_00: No action or intervention.
• 15:13
  SPEAKER_00: You're making observations in a more structured format, and you're going to do some analysis, but that is a key difference.
• 15:23
  SPEAKER_00: So you need to keep yourself oriented in the space, if you will, of this framework for study design.
• 15:31
  SPEAKER_00: So we have explanatory, okay, or analytic, and then we have descriptive, non-analytic.
• 15:38
  SPEAKER_00: Within the explanatory analytic, we have these experimental with an intervention being made, observational, no intervention being made.
• 15:47
  SPEAKER_00: Okay, that's fundamentally the basics.
• 15:50
  SPEAKER_00: Okay.
• 15:50
  SPEAKER_00: We are going to break this down even further.
• 15:54
  SPEAKER_00: And just so you know, there continues to evolve study designs.
• 16:00
  SPEAKER_00: Okay.
• 16:01
  SPEAKER_00: Most of those evol the study designs that are evolving have to do with our now technological advances.
• 16:08
  SPEAKER_00: There are laws that have been put into place that give researchers more flexibility for, especially with large database acquisition.
• 16:17
  SPEAKER_00: There's now AI, where AI is getting involved in even decentralized randomizations and other parts of protocol development, even management, monitoring of data in real time in a much more efficient and effective way than we did in the past.
• 16:36
  SPEAKER_00: And so, with that, the evolution of different types of study designs have occurred over the past two decades.
• 16:43
  SPEAKER_00: Okay.
• 16:44
  SPEAKER_00: The classic ones, many of which are listed here, let's go over on the right side, so observational, so no intervention is being made.
• 16:52
  SPEAKER_00: We're looking for possible associations that can lead us to another level of study.
• 16:58
  SPEAKER_00: Case-controlled studies, okay, and that's where we already have the outcome and we're going backwards in time, typically retrospective.
• 17:05
  SPEAKER_00: Cohort studies, which we have exposure to something, again, not by intervention of the researcher.
• 17:10
  SPEAKER_00: They just were exposed.
• 17:12
  SPEAKER_00: We're following them forward in time.
• 17:13
  SPEAKER_00: We're going to break that down.
• 17:14
  SPEAKER_00: Cross-sectional studies that are analytic, okay.
• 17:18
  SPEAKER_00: Correlational studies, and then we get into that big data epidemiology, okay.
• 17:24
  SPEAKER_00: Now that we have more access and technology and the tools, real-world evidence, RWE, right, and then real world data studies.
• 17:35
  SPEAKER_00: Most of these are observational.
• 17:38
  SPEAKER_00: There are some that can be and fall into the pragmatic clinical trial.
• 17:44
  SPEAKER_00: So you'll see some fluctuation.
• 17:46
  SPEAKER_00: Again, that goes back to the actual design.
• 17:48
  SPEAKER_00: You have to pay attention to that.
• 17:50
  SPEAKER_00: And then the last one is the development of this target trial emulation or TTE, which is a hypothetical randomized controlled trial that is using most commonly real-world data.
• 18:04
  SPEAKER_00: Okay.
• 18:06
  SPEAKER_00: So those are all observational.
• 18:08
  SPEAKER_00: We're going to break those down here in just a minute and cover their pros and their cons when you do them, when you don't do them, what do they answer for you, what they don't answer for you.
• 18:17
  SPEAKER_00: Okay?
• 18:17
  SPEAKER_00: Next is over on the experimental, that's your classic randomized and non-randomized controlled trial.
• 18:23
  SPEAKER_00: It also includes crossover studies, okay, especially those that are obviously randomized.
• 18:28
  SPEAKER_00: And then the evolution that seems to be gaining steam are these adaptive platform master protocol trials.
• 18:37
  SPEAKER_00: Well, randomized controlled trial is a master protocol.
• 18:40
  SPEAKER_00: All studies and clinical trials will have a study protocol, but some of them adapt.
• 18:47
  SPEAKER_00: They have strategic known points in their protocol where they're allowing for prospectively adjustments and adaptations to different things that they already anticipate that are going to happen.
• 19:01
  SPEAKER_00: And why would we consider that?
• 19:03
  SPEAKER_00: Well, that goes back to the ability to apply it to real world.
• 19:08
  SPEAKER_00: The pragmatic piece, the understanding the effectiveness of things in our real clinical practice.
• 19:15
  SPEAKER_00: Because back in the day you would do a study and you get this really clean data, but remember, we don't practice medicine, nursing, pharmacy, healthcare inside of a clinical trial where people get treated differently and get more monitoring, have more access, get follow-up phone calls, compliance, adherence, all these things are being monitored.
• 19:37
  SPEAKER_00: That's not real-world clinical practice.
• 19:39
  SPEAKER_00: And so this real world practice perspective is evolving.
• 19:44
  SPEAKER_00: And even decentralized clinical trials, which are now, you know, especially with tools and technology, we can do that easier.
• 19:52
  SPEAKER_00: Okay?
• 19:53
  SPEAKER_00: Now, within randomized trials, there are these explanatory clinical trials, which tend to be on this side here, short-term surrogate, single center, very per protocol, very tight and clean.
• 20:08
  SPEAKER_00: Okay.
• 20:09
  SPEAKER_00: These are per protocol, so you get more of the efficacy understanding of it, true intervention.
• 20:17
  SPEAKER_00: Right?
• 20:17
  SPEAKER_00: But that doesn't mean that that's what you see in the real world.
• 20:20
  SPEAKER_00: That's where pragmatic clinical trials have really started to increase, okay?
• 20:27
  SPEAKER_00: Because clinically we want to know what's really going on.
• 20:30
  SPEAKER_00: Well, it's long-term, more patient resource focused, okay, tend to be larger, multi-centered, and then include that intention to treat ITT, intention to treat analysis.
• 20:43
  SPEAKER_00: And it again tends to look at that effectiveness or the overall applicability of the data to the regular population or public.
• 20:53
  SPEAKER_00: Okay, so this is our framework that descriptive studies, non-analytic, explanatory, analytic.
• 21:02
  SPEAKER_00: Okay, it's important.
• 21:03
  SPEAKER_00: Make sure you study that.
• 21:04
  SPEAKER_00: You can keep your head kind of wrapped around that.
• 21:08
  SPEAKER_00: It becomes very helpful as you then think through, especially at the protocol development stage, what is going to be the most feasible for you.
• 21:16
  SPEAKER_00: I remember as a resident, right, early on in my academic faculty positions, I didn't have the resources, I didn't have the funding to do randomized controlled trials.
• 21:26
  SPEAKER_00: So I had to go to a case-controlled environment or do a cohort study type of thing.
• 21:32
  SPEAKER_00: So, you know, you start out doing things that you're able to do that lead and to give you the support to go get the funding.
• 21:38
  SPEAKER_00: Okay.
• 21:39
  SPEAKER_00: Now, let's break down those study designs.
• 21:42
  SPEAKER_00: So case controlled.
• 21:44
  SPEAKER_00: Okay, so you have it in your mind.
• 21:45
  SPEAKER_00: So as you're reading the paper, they may not just come out and flat out say it, but you want to be able to read it and you should be able to put it, the skeleton together.
• 21:54
  SPEAKER_00: So here is a structure where the onset of the study is stated right here.
• 22:00
  SPEAKER_00: Okay, so this is on your x-axis is time.
• 22:03
  SPEAKER_00: Okay.
• 22:04
  SPEAKER_00: So we have the cases, patients who actually have the diagnosis or the problem or the illness that you're interested in.
• 22:13
  SPEAKER_00: And then you pick out controls.
• 22:16
  SPEAKER_00: So you have cases, and then you have controls without that thing.
• 22:19
  SPEAKER_00: And then what's ended up happening is you're going backwards in time.
• 22:24
  SPEAKER_00: So you're retrospectively going backwards and you're looking for an exposure in the situation.
• 22:30
  SPEAKER_00: So the thing has already happened.
• 22:33
  SPEAKER_00: You're not really excessively harming anybody other than their information because you're going backwards and looking in data, charts, databases.
• 22:43
  SPEAKER_00: Okay?
• 22:44
  SPEAKER_00: And so this is typically where odds ratios are calculated in this type of study design versus a relative risk, tends to be more prospective cohort analysis.
• 22:55
  SPEAKER_00: But you're looking for exposure.
• 22:57
  SPEAKER_00: So let's say the cases over here are lung cancer.
• 23:01
  SPEAKER_00: And you're going backwards in time, and the cases controls over here have patients no lung cancer, but they look similar to the patients who do.
• 23:09
  SPEAKER_00: And they go backwards in time, and it's and they're really kind of looking at things that they were exposed to.
• 23:14
  SPEAKER_00: Smoking, you may be environmental causes, different life factors, right?
• 23:20
  SPEAKER_00: And they're looking for that, and they're looking at the odds of developing that outcome or that case based on that.
• 23:29
  SPEAKER_00: So the ideal scenario for this, though the sort of advantages, if you have a rare disease, those are those are first of all hard to find, but if you know who they are, you can have them all at the end because they're already diagnosed, and then you can go backwards in time.
• 23:45
  SPEAKER_00: So they're ideal for that.
• 23:47
  SPEAKER_00: Whereas in cohort studies, it's not ideal for a rare disease.
• 23:53
  SPEAKER_00: All right, it's less time requirement.
• 23:55
  SPEAKER_00: It can measure multiple exposures or risk factors.
• 23:59
  SPEAKER_00: Okay, that's helpful.
• 24:00
  SPEAKER_00: Helps to establish causation, the usually lower costs, and the ethical concerns are much lower because you're not giving the patient anything, you're just looking at their information.
• 24:12
  SPEAKER_00: So it's privacy protection at that point.
• 24:14
  SPEAKER_00: It's still a risk.
• 24:16
  SPEAKER_00: IRB is still gonna want to know it, but that is the risk.
• 24:19
  SPEAKER_00: Okay.
• 24:20
  SPEAKER_00: Now, what is the issue?
• 24:21
  SPEAKER_00: Well, it's if you're asking people and going back and trying to gather information, well, then that's prone to recall bias.
• 24:29
  SPEAKER_00: Okay, it's not gonna be good for a diagnostic test because the cases already have it and the controls don't.
• 24:36
  SPEAKER_00: Um, it is difficult to find sometimes a control group that matches up, you know, in that area.
• 24:42
  SPEAKER_00: Not impossible, but sometimes can be difficult depending on your geographic location.
• 24:46
  SPEAKER_00: But it can also not determine any prevalence or incidence.
• 24:51
  SPEAKER_00: Okay.
• 24:52
  SPEAKER_00: We'll see other study designs here in a minute where you can do those.
• 24:57
  SPEAKER_00: So let's now talk about the cohort study.
• 24:59
  SPEAKER_00: So just think of it as almost kind of backwards, right?
• 25:02
  SPEAKER_00: You start out with a cohort of patients, and then you push them into those that were exposed to something, smoking, right?
• 25:13
  SPEAKER_00: Eating red meat, living in some geographic area, working in some type of environment of exposure, war, maybe there were some chemicals that you were exposed to, and then you follow these individuals forward in time.
• 25:27
  SPEAKER_00: No intervention is made.
• 25:29
  SPEAKER_00: Remember, we're still in the we're still in this observational area.
• 25:35
  SPEAKER_00: We're not making an intervention on these patients, just like in the case controlled.
• 25:39
  SPEAKER_00: We didn't make an intervention.
• 25:41
  SPEAKER_00: We're looking at someone who's exposed versus non-exposed, we follow them forward in time and we look for that measured outcome.
• 25:50
  SPEAKER_00: Okay, so you can collect data prospectively, but you can also do a retrospective cohort study.
• 25:57
  SPEAKER_00: That means you go back in time, you have a group of people from a database, you look for an exposure, then you follow their data, which was already collected, and you follow it in a prospective format forward to look at what did they end up developing.
• 26:15
  SPEAKER_00: Did they get that outcome or did they not get that outcome from that exposure?
• 26:19
  SPEAKER_00: So you can do a retrospective cohort study, but most commonly we think about them in a prospective format.
• 26:27
  SPEAKER_00: And they they're the things that when we do the analysis, because remember, this is part of the analytic group, we are generating relative risks, okay, versus an odds ratio.
• 26:38
  SPEAKER_00: And so cohort studies, their advantages are helping to determine that natural incidence or natural history of the disease.
• 26:47
  SPEAKER_00: You follow it forward in time and see what happens.
• 26:51
  SPEAKER_00: Okay.
• 26:52
  SPEAKER_00: They can be helpful when uh evaluating multiple outcomes after a single exposure.
• 26:58
  SPEAKER_00: Okay, you just define that at front, and they can be prospective or retrospective in nature, which I've already talked about.
• 27:06
  SPEAKER_00: The disadvantage is that you don't know who's gonna have, if you're looking for a rare disease as an outcome, well, you don't know who's gonna develop it.
• 27:15
  SPEAKER_00: You could spend years, decades.
• 27:18
  SPEAKER_00: There have been cohort studies that have been followed for decades, and I mean that's a lot of time, money, and energy to study something that you're probably not gonna find anyway.
• 27:27
  SPEAKER_00: Okay, and there's more likely to have some degree of bias than compared to an RCT because data typically is either being asked, recalled, you know, or collected.
• 27:38
  SPEAKER_00: It may not be complete, so there's risk of bias in those situations.
• 27:42
  SPEAKER_00: Uh selection bias is built into the cohort itself because you're looking for these people that were exposed and trying to match them to non-exposed.
• 27:52
  SPEAKER_00: Loss to follow-up, you know, they just don't stop answering the surveys and following through with meeting up for their rechecks.
• 28:00
  SPEAKER_00: And then, depending on the size, can be very costly and time consuming.
• 28:06
  SPEAKER_00: Okay?
• 28:06
  SPEAKER_00: Now, what about a cross-sectional?
• 28:08
  SPEAKER_00: Again, we're another observational analytic design.
• 28:12
  SPEAKER_00: We can't prove causation, but this cross-section is a moment snapshot in time, single snapshot in time.
• 28:21
  SPEAKER_00: And this is very helpful when you're trying to re get the prevalence.
• 28:27
  SPEAKER_00: What is going on right now in the country, the county, the world with this particular disease.
• 28:35
  SPEAKER_00: Okay, so that's where a cross-sectional study can be helpful.
• 28:40
  SPEAKER_00: What it cannot do is it cannot determine the incidence because you have no time interval.
• 28:47
  SPEAKER_00: You need that time interval to determine the incidence.
• 28:50
  SPEAKER_00: So it can be helpful at determining the prevalence of a problem, but not the incidence.
• 28:56
  SPEAKER_00: It's pretty easy, it's fast.
• 28:59
  SPEAKER_00: There's no loss to follow up because it's a single moment in time.
• 29:03
  SPEAKER_00: Okay.
• 29:04
  SPEAKER_00: And so we use it a lot of times in assessing uh healthcare needs or determining the what's going on about a particular disease.
• 29:12
  SPEAKER_00: Okay.
• 29:14
  SPEAKER_00: So far, all of those study designs and structures, again, can help us to look for an association, but none of them can prove that causality.
• 29:25
  SPEAKER_00: Okay?
• 29:27
  SPEAKER_00: And sometimes it may not be feasible.
• 29:29
  SPEAKER_00: I mean, think about someone going into a pulseless VTAC or V fib and defibrillating them.
• 29:36
  SPEAKER_00: I mean, you're going to randomize somebody to a group that's not going to get defibrillated and see if, I mean, that's right, it's not feasible in that case to be able to prove causality.
• 29:49
  SPEAKER_00: Okay.
• 29:50
  SPEAKER_00: What about crossover studies?
• 29:51
  SPEAKER_00: Because this was one of those in the randomized controlled trials, you know, so if you just stopped right here in the line, that would be a typical uh randomized controlled, you know, group where patients were giving an intervention versus the comparative group, no intervention.
• 30:09
  SPEAKER_00: We follow them for it in time, we measure the outcome.
• 30:12
  SPEAKER_00: But in crossover studies, all patients will serve as their own control because they eventually cross over.
• 30:20
  SPEAKER_00: But the important thing in a study design here is the period of the washout from the intervention that the patients were in at the first group.
• 30:28
  SPEAKER_00: Now, depending on the intervention, right, it could be that what is the half-life or the expected effect of that intervention?
• 30:37
  SPEAKER_00: So some drugs have half-lives of weeks, sometimes years.
• 30:42
  SPEAKER_00: Some drugs have half-lives of seconds.
• 30:46
  SPEAKER_00: So your washout period is influenced by that intervention.
• 30:50
  SPEAKER_00: Because what you don't want to do is to have the intervention group was still having the effects cross over to the placebo group or arm, where you do nothing, but they're still under the influence of the drug, which then causes a bias of your results.
• 31:07
  SPEAKER_00: So the so the period of washout it depends on the intervention.
• 31:12
  SPEAKER_00: This can obviously prolong the study, cause more time, right, that's needed.
• 31:19
  SPEAKER_00: Sometimes it raises the cost because of the duration, and sometimes you run the risk of losing these patients.
• 31:26
  SPEAKER_00: But when the patients serve as their own controls, they control for their own confounders.
• 31:34
  SPEAKER_00: That's huge.
• 31:36
  SPEAKER_00: That really helps to contribute to the power of the study.
• 31:40
  SPEAKER_00: So you can actually enroll fewer people in this type of study design and maintain 80% or more power, okay, right, which is really the risk of just not making a type 2 error, right?
• 31:56
  SPEAKER_00: Because it's 1 minus beta, so the chance of not making a type 2 error.
• 32:00
  SPEAKER_00: We don't need as many people, but it is harder on us and takes more time to do that.
• 32:07
  SPEAKER_00: And if that washout period is a long time, based on the intervention, that can be sometimes problematic.
• 32:14
  SPEAKER_00: Okay?
• 32:15
  SPEAKER_00: So we have our randomized control trial, we have our crossover study, randomized control trials, not crossover.
• 32:21
  SPEAKER_00: The patients are not serving as their own control.
• 32:25
  SPEAKER_00: And typically we want these patients to be a randomized.
• 32:29
  SPEAKER_00: But you can do a non-randomized approach.
• 32:34
  SPEAKER_00: It's not ideal.
• 32:36
  SPEAKER_00: The risk of bias goes up, and we'll see examples as I do these lectures.
• 32:41
  SPEAKER_00: I try to give you real-world examples of the influence of these particular types of design issues and the lack of control of confounders of bias on the impact of results.
• 32:52
  SPEAKER_00: Okay, so this stuff is very relevant, but this is considered the gold standard.
• 32:58
  SPEAKER_00: Okay, a well-designed, appropriately developed RCT can help you to answer both per protocol efficacy, but also, if done right, can even help you to answer the pragmatic approach, the real world applicability of the data when you do intention to treat analysis.
• 33:20
  SPEAKER_00: And remember, attention to treat doesn't usually look as good as the per protocol analysis, okay?
• 33:26
  SPEAKER_00: But it's the real world.
• 33:28
  SPEAKER_00: But they are time consuming, they are labor-intensive, and very expensive, and blinding can could be sometimes compromised.
• 33:37
  SPEAKER_00: Where we single-blinded studies, double blindness is the intervention.
• 33:42
  SPEAKER_00: Um is there a placebo matching, so it's double dummy, right?
• 33:47
  SPEAKER_00: All these things go into controlling the overall bias.
• 33:52
  SPEAKER_00: Surveys, questionnaires, or obviously the the big advantage is that patients can be anonymous.
• 33:59
  SPEAKER_00: Many times it's a one-time data collection, so there's no follow-up issues or and loss, the loss of patients.
• 34:07
  SPEAKER_00: Um, there's many uh reliable tools out there for gathering data.
• 34:12
  SPEAKER_00: They're pretty quick overall, but they can be time consuming depending on the extent and the number of people that you're trying to acquire.
• 34:20
  SPEAKER_00: Okay, they can so that can lead to expense uh being expensive.
• 34:24
  SPEAKER_00: And they also, depending on the questions you're asking, can be subject to bias.
• 34:28
  SPEAKER_00: So, one of them being the most common is the recall bias from what you did, depending on how far back you're asking the patient to go to.
• 34:38
  SPEAKER_00: Now, qualitative studies, which kind of goes back to that descriptive group, right?
• 34:45
  SPEAKER_00: Non-analytic, this is helping us to understand the patient's experience with an intervention or treatment or process.
• 34:55
  SPEAKER_00: That human element, okay.
• 34:58
  SPEAKER_00: Um, the biggest problem with that is it's subject to observer bias, including also the the even the patient's own bias, if they especially if they know and the way you the researcher asks the questions.
• 35:13
  SPEAKER_00: Did you really like being on that medicine, or was it really hard?
• 35:18
  SPEAKER_00: You know, or you know, versus just tell me, you know, can you explain to me how you felt being on the medication?
• 35:25
  SPEAKER_00: Right?
• 35:26
  SPEAKER_00: So your tone and maybe even the nonverbal body language that you express can influence that patient or subject, and they may want to stay in the study because maybe they're getting paid.
• 35:40
  SPEAKER_00: So they want to give you the right answer or what they perceive that you're looking for is the right answer.
• 35:46
  SPEAKER_00: We don't want all that, right?
• 35:47
  SPEAKER_00: So those are the those are where the risks come and the disadvantages.
• 35:49
  SPEAKER_00: These are difficult studies to do that they have very meticulous processes to try to control for that, which makes it a little bit more uh difficult and time consuming.
• 36:00
  SPEAKER_00: But man, they can help you to really get into that real-world practical application of what you're going to do with your patient.
• 36:08
  SPEAKER_00: So when you take the analytic, non-analytic, what we're really kind of talking about is the ability to quantify the results, measure them, and put them in the statistical metrics that we talk about.
• 36:21
  SPEAKER_00: Odds ratios, relative risk, p-values, 95% confidence rules, all things that in our series on clinical research that we'll address.
• 36:30
  SPEAKER_00: But qualitative research versus quantitative, we've been mostly thinking in the analytic as that quantitative analysis, right?
• 36:37
  SPEAKER_00: We're measuring it, we're quantifying it, we're putting it into measurable numbers that mean something to us.
• 36:45
  SPEAKER_00: Okay?
• 36:46
  SPEAKER_00: Now they're not exact, but they help us to wrap our minds around it.
• 36:50
  SPEAKER_00: That qualitative again is more about that human experience and the reasons or opinions, motivations of the patient and the participants.
• 36:59
  SPEAKER_00: Okay, it's usually done as unstructured interview, people being observed, and then we summarize them and put their information in a descriptive format.
• 37:10
  SPEAKER_00: Okay, versus uh quantitative studies or research are statistical analysis.
• 37:16
  SPEAKER_00: Again, I just want to make sure that the terminology, as you kind of interact with these subjects, and if you're using this content to help you study for board exams, your own, you know, your own coursework, right?
• 37:30
  SPEAKER_00: These are the terms that the writers of these questions really like to throw at people because and it creates a little confusion at times, but you just get oriented.
• 37:42
  SPEAKER_00: So I'm trying to make sure that you kind of understand the framework, and I'm I'm moving around the framework from different angles so that you see all the walls.
• 37:53
  SPEAKER_00: Now, we talked about the study hierarchy, and I'll bring that back up here in a minute, but you've been hearing about bias, and that's going to lead to our conversation on another lecture on internal and external validity.
• 38:05
  SPEAKER_00: But this is really looking at the why do we care about these study designs?
• 38:09
  SPEAKER_00: What are the why do we go through the pros and cons, the advantages, disadvantages?
• 38:14
  SPEAKER_00: Because it has to do with bias, systematic error that's being introduced that determines what the truth of the answer is in the results that we get compared to the actual truth.
• 38:27
  SPEAKER_00: Right?
• 38:27
  SPEAKER_00: And so as you go down these study designs from experimental to observational and descriptive, you see that the risk of potential bias, systematic error, starts to rise rapidly, right?
• 38:41
  SPEAKER_00: Gets higher and higher and higher.
• 38:43
  SPEAKER_00: That's the same concept over on the pyramid.
• 38:47
  SPEAKER_00: But remember, we're looking for, if we want to prove causality, you have to do an experimental study design.
• 38:53
  SPEAKER_00: If you're looking for correlations associations, then then you get away with some of the observable observational studies that are analytic.
• 39:00
  SPEAKER_00: If you're looking for hypothesis generating, you can do that with some correlations or observational data, of course.
• 39:07
  SPEAKER_00: Yes, absolutely.
• 39:08
  SPEAKER_00: But remember, many times we think of if we're just starting out, the the most of the hypothesis sometimes come from a case or a series.
• 39:16
  SPEAKER_00: And then qualitative studies are about the human reasoning behind the experience and what we do, why we do what we do and how we feel about it.
• 39:27
  SPEAKER_00: So this again structure goes back to the study design hierarchy, and while not everything fits in there perfectly, right, you see these transition points.
• 39:39
  SPEAKER_00: But again, the Cochrane review, systematic reviews up here, you can have a terrible systematic review that's a high risk for bias.
• 39:47
  SPEAKER_00: That's why Cochrane does the risk bias assessment.
• 39:51
  SPEAKER_00: And most good systematic reviews, meta-analysis will do a risk bias assessment and the quality of a study assessment and publish that.
• 40:01
  SPEAKER_00: They're telling you junk in is junk out.
• 40:04
  SPEAKER_00: We did the analysis right, but the information is terrible because the studies weren't done right.
• 40:11
  SPEAKER_00: And so the reason we're pushing for these and why we're putting on, and I'm doing these topics for you and other programs around the country is that the idea is to for us to produce good information that's meaningful.
• 40:25
  SPEAKER_00: Just to publish something to say you put it in a journal, please stop wasting time, right?
• 40:30
  SPEAKER_00: We got enough journals with a bunch of junk in them and people claiming publications.
• 40:36
  SPEAKER_00: I get the pressure.
• 40:37
  SPEAKER_00: I was in academia for many years, went through promotion, dossier, and all that stuff.
• 40:44
  SPEAKER_00: I know the pressure that it takes.
• 40:47
  SPEAKER_00: I remember doing all that.
• 40:48
  SPEAKER_00: But the the downside to that is that sometimes you will compromise to find results that are significant, interesting, so you get that publication, so you can get promoted.
• 40:58
  SPEAKER_00: Really, but it didn't do anything to meaningful to clinical practice sometimes.
• 41:03
  SPEAKER_00: So when you're thinking about study design, this is just a general catch-all.
• 41:07
  SPEAKER_00: Think about analytic, non-analytic.
• 41:08
  SPEAKER_00: Is it prospective?
• 41:09
  SPEAKER_00: Is it retrospective?
• 41:11
  SPEAKER_00: What are the risks when it's retrospective?
• 41:14
  SPEAKER_00: Randomized, non-randomized, decentralized, is AI involved now?
• 41:18
  SPEAKER_00: Okay, are you using some other tool to monitor and track people?
• 41:23
  SPEAKER_00: Nothing wrong with that.
• 41:24
  SPEAKER_00: Makes it more feasible.
• 41:26
  SPEAKER_00: Is the interventions, are they blinded?
• 41:28
  SPEAKER_00: Is it just the patient being blinded or is the researcher also being blinded?
• 41:34
  SPEAKER_00: Okay, double dummy.
• 41:35
  SPEAKER_00: Do I know what you're taking because I see you take it?
• 41:38
  SPEAKER_00: Pill versus a shot, right?
• 41:41
  SPEAKER_00: Capsule versus a pill.
• 41:43
  SPEAKER_00: Okay, so you create placebo, you these double dummies so that you avoid that patient being able to figure it out, or the provider.
• 41:51
  SPEAKER_00: Active versus placebo.
• 41:52
  SPEAKER_00: I mean, most people would like to go head to head with the active treatment.
• 41:56
  SPEAKER_00: It's harder to win, but it's a it's the nice thing to do.
• 42:00
  SPEAKER_00: Single center versus multi-center.
• 42:02
  SPEAKER_00: What happens at one location may not be what happens at another location.
• 42:06
  SPEAKER_00: So you want to know in the real world, since we all don't practice at the same place, what is the real meaningful thing that we see happening?
• 42:17
  SPEAKER_00: Right?
• 42:17
  SPEAKER_00: So multi-center gives us that more realistic perspective.
• 42:22
  SPEAKER_00: Um, if you are doing the role of AI, it's really important to understand that what AI is contributing to, what parts it has control over, because we know that AI is not set necessarily perfect.
• 42:32
  SPEAKER_00: Great tool, but it's not perfect, right?
• 42:35
  SPEAKER_00: The influence of the sponsor and conflicts of interest by the investigators.
• 42:41
  SPEAKER_00: Now, core evidence, okay?
• 42:42
  SPEAKER_00: So this is an application piece.
• 42:44
  SPEAKER_00: Does controlling for confounders in clinical trials really matter?
• 42:48
  SPEAKER_00: Well, I mean, I can give you a lot of examples, but the answer is obviously yes, right?
• 42:53
  SPEAKER_00: Just wasting my time doing some of this.
• 42:56
  SPEAKER_00: We're study design and controlling for confounders and looking for risk of bias, which we're going to cover in that if you stay on the series in the next sort of topic, it matters.
• 43:07
  SPEAKER_00: So this is an assessment looking at the influence and controlling for things like socioeconomic status, SES, and the impact.
• 43:17
  SPEAKER_00: And this was a study that was looking at hormone replacement therapy, HRT.
• 43:22
  SPEAKER_00: Okay.
• 43:23
  SPEAKER_00: And so on the left side you see these are the groups before not adjusting, right?
• 43:29
  SPEAKER_00: So not adjusted for socioeconomic status, and then those that were adjusted.
• 43:35
  SPEAKER_00: Okay, you come over here, you see all that force plot, okay, kind of look.
• 43:40
  SPEAKER_00: And so we're talking about it whether it's relative risk cross ratio, but remember that means that the 95% confidence interval cannot pass through and include 1.0.
• 43:51
  SPEAKER_00: That means there's no difference.
• 43:53
  SPEAKER_00: The null would be true.
• 43:55
  SPEAKER_00: Okay, so the 95% confidence interval is the Line, okay, so you see how the length of the line here.
• 44:03
  SPEAKER_00: If it crosses over, touches that line of no difference, it is statistically not different.
• 44:10
  SPEAKER_00: Well, you can see that these two results up here, current hormone therapy use, past hormone therapy use, if you don't factor in the associated economic status, it actually shows a statistical difference.
• 44:30
  SPEAKER_00: When you control for those or adjust for it, that goes away.
• 44:35
  SPEAKER_00: And all of the data shifts over.
• 44:40
  SPEAKER_00: So that's a pretty profound impact.
• 44:42
  SPEAKER_00: Well, that goes back to the study design and your ability to control that.
• 44:48
  SPEAKER_00: If something is retrospective in nature, you can't fix what already has happened.
• 44:53
  SPEAKER_00: That's where that prospective studies, which are harder, more time consuming sometimes to do, can be very better, it can be helpful and more robust at reducing the risk of bias.
• 45:06
  SPEAKER_00: Okay, now moving back to reporting.
• 45:09
  SPEAKER_00: So if you're doing the study design, you're preparing to uh start this, you need to think where you're gonna publish it, and do I need to register it?
• 45:19
  SPEAKER_00: And you'll see as a reader, a good again, good journals, good papers will say this is our registry number.
• 45:28
  SPEAKER_00: And our protocol was published ahead of time.
• 45:32
  SPEAKER_00: Okay, and so there's groups out there for interventional studies, obviously randomized controlled trials, but also observational studies.
• 45:40
  SPEAKER_00: Clinicaltrials.gov is probably one of the most famous.
• 45:44
  SPEAKER_00: Okay.
• 45:45
  SPEAKER_00: Then there's also the international standard standard randomized control trial number.
• 45:50
  SPEAKER_00: Okay, it's a little bit more international, obviously.
• 45:52
  SPEAKER_00: And then even systematic reviews and meta-analysis have a registry.
• 45:56
  SPEAKER_00: And this helps you to know if somebody else is doing the work that you're doing.
• 46:00
  SPEAKER_00: Are they answering the question already?
• 46:02
  SPEAKER_00: That may eliminate your need to do the study.
• 46:05
  SPEAKER_00: Or maybe you will do it and see if the two of you align up.
• 46:09
  SPEAKER_00: Did you use the right studies?
• 46:10
  SPEAKER_00: Did you use the same protocol?
• 46:12
  SPEAKER_00: Right?
• 46:12
  SPEAKER_00: Well, again, protocols are many times now required to be published online that you're doing the study, and this is where this is how we're going to do it.
• 46:21
  SPEAKER_00: And that's the the spirit guidelines, right?
• 46:24
  SPEAKER_00: The standard protocol items, recommendations for interventional trials, spirit.
• 46:30
  SPEAKER_00: Okay.
• 46:30
  SPEAKER_00: This one is also many times updated along with consort.
• 46:34
  SPEAKER_00: Now, consort is the publication guidelines.
• 46:38
  SPEAKER_00: What does a journal expect in the information as it's being written out so that when I, as a clinician, you read it from one journal to another, we're reading the same things in the same format.
• 46:52
  SPEAKER_00: You don't have to figure it all out every time you open up a journal.
• 46:56
  SPEAKER_00: That would be so frustrating.
• 46:59
  SPEAKER_00: And so this actually doing this ahead of time helps you to get more complete of that protocol so you don't make mistakes.
• 47:07
  SPEAKER_00: It's more transparent to the public.
• 47:09
  SPEAKER_00: So if you don't publish the study, why not?
• 47:14
  SPEAKER_00: All right, what are you hiding?
• 47:15
  SPEAKER_00: It also typically increases the quality of those studies.
• 47:20
  SPEAKER_00: And again, coming back to the reporting guidelines, you have consort, okay, we have Prisma for systematic reviews, agree for clinical practice, okay, case reports, care.
• 47:32
  SPEAKER_00: If you're doing diagnostic studies, it's the start E.
• 47:36
  SPEAKER_00: Okay.
• 47:37
  SPEAKER_00: All of these become very important.
• 47:38
  SPEAKER_00: Qualitative research, okay, the SRQR.
• 47:42
  SPEAKER_00: All of these can guide you again and give you the reference link there.
• 47:46
  SPEAKER_00: Again, these are this table is available at thisiswhy.health.
• 47:51
  SPEAKER_00: But consorts, which is probably one of the most famous, and and again, since we are most interested in RCTs, this one has been updated.
• 47:59
  SPEAKER_00: It does get updated periodically.
• 48:00
  SPEAKER_00: It's now up to 30 items that are required in a checklist.
• 48:05
  SPEAKER_00: And again, many journals, depending on where you're trying to publish, will require you to document and show those 30 items because, again, it helps the peer review process for determining publication.
• 48:19
  SPEAKER_00: It also increases your chance of likely being published.
• 48:22
  SPEAKER_00: Um, and in addition to that, it improves what we as the readers get to read and have if we have questions, then we can look for that.
• 48:29
  SPEAKER_00: Now, core concepts.
• 48:31
  SPEAKER_00: Okay, only appropriately designed randomized controlled trials can determine that causality.
• 48:37
  SPEAKER_00: Remember, it's in the analytic side, that explanatory study design.
• 48:42
  SPEAKER_00: Well done Cochrane reviews, systematic reviews and meta-analysis that are basically a good randomized controlled trial, they sit at the top of the pyramid.
• 48:52
  SPEAKER_00: Okay, the top.
• 48:54
  SPEAKER_00: But if they're they could you could Cochrane review could still be done perfect, excellent systematic approach, but they'll tell you this is junk.
• 49:04
  SPEAKER_00: And that means it's really junk.
• 49:07
  SPEAKER_00: I mean, because they follow such a good process that if they determine it's junk in and it's junk out, then it's really junk.
• 49:14
  SPEAKER_00: But we need to know that.
• 49:16
  SPEAKER_00: That's good to know.
• 49:18
  SPEAKER_00: Okay.
• 49:18
  SPEAKER_00: Now, randomization obviously helps us with bias.
• 49:22
  SPEAKER_00: Failure to control for confounders can lead to inappropriate conclusions.
• 49:28
  SPEAKER_00: It also impacts our internal validity in that example that I showed you.
• 49:33
  SPEAKER_00: And then qualitative research is really these unstructured interviews and observations that are trying to be summarized descriptively to explain human experience rather than quantifying those outcomes.
• 49:47
  SPEAKER_00: So, guys, thank you so much for watching.
• 49:49
  SPEAKER_00: And if you enjoyed this uh series on This Is Why, continue to explore the collection and playlist where you'll walk through other evidence-based medicine topics, clinical research, and a concept that is step-by-step if you need it that way.
• 50:04
  SPEAKER_00: And for those of you who are preparing for clinical practice or board style learning, like maybe you're taking a test, your many board exams cover this material.
• 50:13
  SPEAKER_00: Well, that's why I developed high-yield med reviews.
• 50:15
  SPEAKER_00: Okay, we have cue banks over there, we have practice questions, exams.
• 50:19
  SPEAKER_00: They help you to even with this kind of content, learn how to apply it.
• 50:25
  SPEAKER_00: We have programs throughout the country that use it.
• 50:27
  SPEAKER_00: If your program is interested in that, we will work with you to do that.
• 50:31
  SPEAKER_00: This is why uh members at ThisisY.
• 50:38
  SPEAKER_00: So, thank you so much for joining me.
• 50:39
  SPEAKER_00: We'll see you on the next episode.
• 50:42
  SPEAKER_00: Take care, bye.