WEBVTT Kind: captions; language: en-us NOTE Treffsikkerhet: 83% (H?Y) 00:00:01.100 --> 00:00:07.300 Okay, so we're going to talk about study designs and there are more designs than we can 00:00:07.300 --> 00:00:13.600 discuss. So this is just an introduction. This is a little bit of everything. So to start with, 00:00:13.600 --> 00:00:18.000 we've got three topics that I would like to cover actually. There is variables. We talked a little 00:00:18.000 --> 00:00:23.800 bit about what our variables, what type of variables do we have. Then we talk about study designs, 00:00:23.800 --> 00:00:31.650 the most commonly ones that we use and what our levels of evidence and why is it important? NOTE Treffsikkerhet: 91% (H?Y) 00:00:31.650 --> 00:00:39.300 So lets have a chat about variables. A variable is an object, event, idea, feeling, you name it or any 00:00:39.300 --> 00:00:45.400 other type of category that you are trying to measure. Now, sometimes you will struggle measuring 00:00:45.400 --> 00:00:51.200 things like happiness while usually then you come up with an ordinal scale or something, but you 00:00:51.200 --> 00:00:55.450 know that certain things are easier to measure than others, but whatever you're trying to measure 00:00:55.450 --> 00:01:01.750 that is a variable. Now, we've got different variables and you probably have talked about that at NOTE Treffsikkerhet: 83% (H?Y) 00:01:01.750 --> 00:01:08.000 Stats already, but we've got independent and dependent variables. An independent variable is a 00:01:08.000 --> 00:01:15.200 variable that stands alone and it isn't changed by other variables. The variable represents inputs 00:01:15.200 --> 00:01:24.200 or causes and in the experimental setting, I as an experimenter control that variable. Now the other 00:01:24.200 --> 00:01:29.300 ones are the dependent variables and the dependent variable is something that depends on other 00:01:29.300 --> 00:01:32.050 factors. That's why it's called dependent variable. NOTE Treffsikkerhet: 86% (H?Y) 00:01:32.050 --> 00:01:37.100 It represents usually the outcome, what you are trying to measure of the variation you are 00:01:37.100 --> 00:01:38.600 studying. NOTE Treffsikkerhet: 87% (H?Y) 00:01:38.600 --> 00:01:45.850 Now, if I give you an example, have a look. So if we are talking about a study to determine 00:01:45.850 --> 00:01:54.100 whether how long is student sleeps effects test scores, then the independent variable is the length 00:01:54.100 --> 00:02:01.500 of time spent sleeping. The depending one is my test score because I will not influence the length 00:02:01.500 --> 00:02:09.800 of sleeping but I will look at the test scores, the outcome. NOTE Treffsikkerhet: 91% (H?Y) 00:02:09.900 --> 00:02:15.500 And these are just stupid examples for just to get our head around. If we talk about compare brands 00:02:15.500 --> 00:02:22.700 of paper handkerchiefs to see which holds the most liquid. Well then again, the independent one is 00:02:22.700 --> 00:02:28.500 the brand of paper towel that's affect the dependent one is the amount of water that I can pour 00:02:28.500 --> 00:02:34.000 into that handkerchief and that is being absorbed. Is this clear, things like this? NOTE Treffsikkerhet: 86% (H?Y) 00:02:34.000 --> 00:02:42.700 Otherwise just interrupt, please do. I'll give you more examples. An experiment to determine how far 00:02:42.700 --> 00:02:49.800 people can see into the infrared part of the spectrum. Okay, again, then the independent is the 00:02:49.800 --> 00:02:55.900 wavelength. We don't influence that, it's affect. The dependent one is whether the light 00:02:55.900 --> 00:03:00.900 is observable whether we can see it, the response, the outcome. NOTE Treffsikkerhet: 91% (H?Y) 00:03:00.900 --> 00:03:09.100 If I ask something like does caffeine affects your appetite, then the independent one is again the 00:03:09.100 --> 00:03:16.000 amount of caffeine, that's just a fact, that's there, but I will measure how hungry you are. How I do 00:03:16.000 --> 00:03:23.500 that, that's a different problem. Now, another one just to keep our heads 00:03:23.500 --> 00:03:28.000 around because this is important that you understand the difference, if you want to determine whether 00:03:28.000 --> 00:03:30.150 a chemical is essential for NOTE Treffsikkerhet: 79% (H?Y) 00:03:30.150 --> 00:03:37.650 rat nutrition. You want to determine that so you come up with an experiment and what you do is then 00:03:37.650 --> 00:03:42.400 the presence absence of the chemical is, again, the independent one, that's the fact. 00:03:42.400 --> 00:03:47.700 The dependent one you're going to measure the rat's health. 00:03:47.700 --> 00:03:53.500 Now you determined in the same experiment, suppose you determine that the substance is 00:03:53.500 --> 00:04:00.000 necessary for proper nutrition. Then you do afollow-up experiment and you want to know so how much 00:04:00.000 --> 00:04:00.300 do I need NOTE Treffsikkerhet: 83% (H?Y) 00:04:00.300 --> 00:04:08.500 of this chemical. In that case the independent variable is the amount of the chemical, not 00:04:08.500 --> 00:04:14.900 just presence or absence, but the amount and dependent one is still the health of my rat. 00:04:15.300 --> 00:04:22.100 Now, you've got models and experiments tests to determine the effects what that the independent 00:04:22.100 --> 00:04:26.900 variables have on the dependent variables. Yeah. We talked about that. Now, we also talked about 00:04:26.900 --> 00:04:30.250 that sometimes you've got a problem with confounding and we NOTE Treffsikkerhet: 85% (H?Y) 00:04:30.250 --> 00:04:37.700 make assumptions that are wrong, that we think that the growing of my tree is affecting 00:04:37.700 --> 00:04:44.700 eyes melt or that because the Dutch we eat a lot of cheese and that's 00:04:44.700 --> 00:04:50.800 independent one and growth would be the dependent one. This is confounding. This is not true 00:04:50.800 --> 00:04:57.300 because you think there's a relation but it's actually not. If this would be fat that we've 00:04:57.300 --> 00:05:00.299 overweight. Yes, then. NOTE Treffsikkerhet: 82% (H?Y) 00:05:00.299 --> 00:05:07.000 With cheese, there is a link with that one. But growth length, no. Now you've got different type 00:05:07.000 --> 00:05:11.500 of measures or different type of variables more than just independent and dependent, but let's have 00:05:11.500 --> 00:05:17.500 another look. I've got two Cactus. I don't know what the plural is in English actually. 00:05:17.500 --> 00:05:23.250 I've got two of these plants and I want to know the effects of coffee on my Cactus everyday. 00:05:23.250 --> 00:05:30.250 And my hypothesis is if I only would give my Cactus more coffee, it will grow better. NOTE Treffsikkerhet: 89% (H?Y) 00:05:30.250 --> 00:05:36.900 Okay, so then we've got some variables like temperature and light and I've got my coffee. Now, 00:05:36.900 --> 00:05:44.600 growth is the outcome. Now, if I need to decide what type of variables have I got here, then I can 00:05:44.600 --> 00:05:51.800 tell that so which is the independent variable and a look at this then I hope you can see that my 00:05:51.800 --> 00:05:56.250 coffee is the independent one. That's what I am doing. NOTE Treffsikkerhet: 91% (H?Y) 00:05:56.250 --> 00:06:06.900 Then the dependent variable is that one, growth. But what are then temperature and light? Well, 00:06:06.900 --> 00:06:13.700 factors like or variables like temperature and light, that is what we call control variables. 00:06:13.700 --> 00:06:20.500 They're actually constant. You don't want to have them change because they may have an influence on 00:06:20.500 --> 00:06:26.299 the experiment. So you deliberately control these variables, you hold them NOTE Treffsikkerhet: 80% (H?Y) 00:06:26.299 --> 00:06:32.600 constant during the time of that experiment. So you've got independent ones, dependent and you've got 00:06:32.600 --> 00:06:34.700 control variables. NOTE Treffsikkerhet: 89% (H?Y) 00:06:34.700 --> 00:06:43.600 It's no part of an experiment. But for instance, if you look at health outcomes then 00:06:43.600 --> 00:06:49.500 there are many variables that could have an influence and you want for instance to say ''well it's 00:06:49.500 --> 00:06:56.300 different effects in men and women. So let's only include women. '' So you keep things constant. 00:06:56.300 --> 00:07:04.200 Better examples are actually things like temperature or we will run into more examples. 00:07:04.700 --> 00:07:09.600 Three types of variables that we need to talk, a little bit about this. Again, I think you have seen this before, 00:07:09.600 --> 00:07:16.800 but there is such a thing as discrete and continuous data. Discrete data can only take certain 00:07:16.800 --> 00:07:22.800 values, whereas continuous data can take any value within a range usually. Now examples are, you 00:07:22.800 --> 00:07:28.000 can have one or two students, you cannot have half a student. So that is an example of discrete 00:07:28.000 --> 00:07:34.650 data. Same with your dices. If you roll them, you can never have 3.5., It's impossible. NOTE Treffsikkerhet: 91% (H?Y) 00:07:34.650 --> 00:07:41.200 Continuous data examples are your height or timing variables. They are ongoing. They can get 00:07:41.200 --> 00:07:49.400 any value that is possible. There's also such a thing as levels of measurement that you 00:07:49.400 --> 00:07:56.799 need to know and there are in this case we've got four different levels. We've got nominal, ordinal 00:07:56.799 --> 00:08:04.550 interval, and ratio level. And all of these levels are different from one another. 00:08:04.550 --> 00:08:11.150 If you look at the top one, nominal variables are about categories. 00:08:11.150 --> 00:08:18.800 Attributes are only named. If I look at ordinal, then there's a ranking, you can order them. 00:08:18.800 --> 00:08:25.700 Interval, the distance has got a meaning and ratio is an absolute zero. Now, let's look 00:08:25.700 --> 00:08:31.000 at examples. Otherwise, this is just a little bit difficult to discuss. If you talk about something 00:08:31.000 --> 00:08:35.600 like eye color, can be blue, green, but it can't be orange, NOTE Treffsikkerhet: 85% (H?Y) 00:08:35.600 --> 00:08:39.500 but you've got the categories there. If you only have got two categories, for 00:08:39.500 --> 00:08:46.450 instance, male and female you call that nominal variable also dichotomous. Dichotomous variable, 00:08:46.450 --> 00:08:56.100 is a nominal variable, but with only two categories. So categories only like professions, bakery guy 00:08:56.100 --> 00:09:01.800 or some grocery guy. Those are professions. You can't sum them up. You can't do anything with 00:09:01.800 --> 00:09:04.800 it. It's just a category, a group. NOTE Treffsikkerhet: 91% (H?Y) 00:09:04.800 --> 00:09:12.500 Ordinal on the other hand, there is a ranking and if you go to a restaurant, sometimes you see the 00:09:12.500 --> 00:09:22.000 spicy thing or this one, you see these spicy hot, hotter, hottest. And this is ordinal. So this little 00:09:22.000 --> 00:09:29.100 bit hot, it's very hot... But there is order in these attributes. If that order would be the 00:09:29.100 --> 00:09:34.400 distance is meaningful, you talk about interval. So for instance temperature, NOTE Treffsikkerhet: 73% (MEDIUM) 00:09:34.400 --> 00:09:40.700 every grade, there is a distance between each grade and the distance is same, the distance between 1 to 2 00:09:40.700 --> 00:09:48.700 or 300 301 grades. The distance is the same. So the distance has got a meaning and then 00:09:48.700 --> 00:09:54.400 if you say there is also an absolute zero, then we talk about ratios variables. And one is you can 00:09:54.400 --> 00:09:59.650 have savings and you can have zero savings. Then you have got a problem I think. 00:09:59.650 --> 00:10:04.750 Okay. This is just the same thing but a different slide again, so we've got nominal, NOTE Treffsikkerhet: 84% (H?Y) 00:10:04.750 --> 00:10:10.100 you've got named variables, there's nothing else you can do with it. If you can 00:10:10.100 --> 00:10:19.300 order them, you talk about ordinal variables. If the ordinal variable has got an interval that has got 00:10:19.300 --> 00:10:27.900 a meaning you talk about interval variable. And ratio if there is a 0 present. And you can 00:10:27.900 --> 00:10:35.150 see every time it's a higher level of measurement and therefore we've got this this pyramid. NOTE Treffsikkerhet: 91% (H?Y) 00:10:35.150 --> 00:10:42.100 The weakest ones are actually the nominal variables, ordinal. It's also like what you can do with 00:10:42.100 --> 00:10:47.200 stats. There's not much you can do with nominal. You can make a beautiful picture, a graph with 00:10:47.200 --> 00:10:53.100 something, but you can't calculate or determine any stats with that. The higher you are, the more 00:10:53.100 --> 00:10:57.050 possibilities you've got for statistical analysis as well. NOTE Treffsikkerhet: 87% (H?Y) 00:10:57.050 --> 00:11:05.650 Okay, so nominal again, professions, sports, gender, marital status, all nominal. Ordinal, your grades, 00:11:05.650 --> 00:11:13.400 A B C D E Etc. That's an ordinal scale. Ordinal variables, if I ask you how satisfied are you with 00:11:13.400 --> 00:11:19.900 online teaching, ordinal scale again. Interval, we gave examples like temperature --whether it's Celsius 00:11:19.900 --> 00:11:27.400 and Fahrenheit, doesn't matter--, calendar years... And then if we go to ratio, height, age, weight, NOTE Treffsikkerhet: 89% (H?Y) 00:11:27.400 --> 00:11:37.750 typically examples of a ratio level variable. Is this all clear or did I lose anybody? NOTE Treffsikkerhet: 88% (H?Y) 00:11:37.750 --> 00:11:45.300 If I don't hear you, I just keep talking. Okay, study designs, because that's what we are want to 00:11:45.300 --> 00:11:51.000 talk about. But you need to understand variables in order to understand study designs. So if we talk 00:11:51.000 --> 00:11:57.000 about study designs, then there is a lot out there. I'll give you first a little bit the broad 00:11:57.000 --> 00:12:03.200 terminology. So we talked about experimental studies versus observational studies, 00:12:03.200 --> 00:12:08.300 andexperimental is I can influence it and we had a few of these slides already, but you NOTE Treffsikkerhet: 84% (H?Y) 00:12:08.300 --> 00:12:14.400 influence the event, effect, investigate the effects of the intervention. And that is usually when 00:12:14.400 --> 00:12:20.500 you have a clinical trial or animal studies or anything like it. Observational study is, you don't 00:12:20.500 --> 00:12:27.500 have too much control over the events. You just inventorize. You just want to know, well, what are 00:12:27.500 --> 00:12:33.800 the prevalence, what are the the numbers of people, what are the risk factors... So something like a 00:12:33.800 --> 00:12:38.250 survey or a prevalent study, is an observation NOTE Treffsikkerhet: 91% (H?Y) 00:12:38.250 --> 00:12:43.300 I'll study, you do not intervene. You do not interfere. You do not have an intervention or something 00:12:43.300 --> 00:12:49.900 going on. So a little bit like this. Experimental, I am doing something. Observational, I'm not doing 00:12:49.900 --> 00:12:52.750 something, I just observe the data. NOTE Treffsikkerhet: 76% (H?Y) 00:12:52.750 --> 00:12:59.200 Now we also talk about cross-sectional studies. 00:12:59.200 --> 00:13:07.500 And that is actually the opposite of longitudinal study. So longitudinal studies is you follow people, a 00:13:07.500 --> 00:13:14.900 sample of individuals or maybe one over time. And cross-sectional is one moment, one specific moment 00:13:14.900 --> 00:13:20.550 in time and you go for prevalence studies for instance when you want to know the relation between 00:13:20.550 --> 00:13:22.250 certain variables, but it's one NOTE Treffsikkerhet: 86% (H?Y) 00:13:22.250 --> 00:13:28.300 specific point in time, there's no follow-up. So if you look at this, the top one here we've got, 00:13:28.300 --> 00:13:32.900 this is a longitudinal because you follow that person forever. And here these are all 00:13:32.900 --> 00:13:43.000 cross-sectional moments in time. I'm going to give you some examples. So this is a way 00:13:43.000 --> 00:13:49.000 of classifying study designs. So you can see on the left, my descriptive studies 00:13:49.000 --> 00:13:52.349 and on the right you've got these analytical studies. NOTE Treffsikkerhet: 91% (H?Y) 00:13:52.349 --> 00:13:59.300 Now, let's first move to the descriptive *bark* studies and ignore the dog. So destcriptive 00:13:59.300 --> 00:14:06.300 studies are *bark* *bark* One moment. Stop! Okay, I'm trying to train a dog in the meantime. Descriptive studies, 00:14:06.300 --> 00:14:12.800 case reports are typically descriptive. So it is about a description *bark* of a patient and usually 00:14:12.800 --> 00:14:19.050 you use that for unusual diseases, where you've got small groups or maybe only *bark* one special group. 00:14:19.050 --> 00:14:22.250 Case series is usually you've got a few NOTE Treffsikkerhet: 77% (H?Y) 00:14:22.250 --> 00:14:29.400 subjects has the same problems, similar cases, but still not enough to *bark* really do a cross-sectional 00:14:29.400 --> 00:14:35.800 study *bark* or something like that. *bark* I'm going to intervene here. Give me one moment. NOTE Treffsikkerhet: 91% (H?Y) 00:14:35.800 --> 00:14:38.000 *R.I.P doggie* 00:14:38.000 --> 00:14:40.500 Okay. NOTE Treffsikkerhet: 81% (H?Y) 00:14:41.500 --> 00:14:48.700 Whereas case series is, as I said, a number of cases in a row and that is can be a small group of 00:14:48.700 --> 00:14:55.300 patients that have a similar treatment or but it's still actually something like N is 5 00:14:55.300 --> 00:15:01.600 or maybe N is 10 or 20. But you just report about one group of kind of similar diseases and 00:15:01.600 --> 00:15:07.400 condition. It's a little bit better and higher level than case report, but still very little 00:15:07.400 --> 00:15:11.500 statistics you can do and usually it is really purely descriptive. NOTE Treffsikkerhet: 79% (H?Y) 00:15:11.500 --> 00:15:17.000 Now if we go further than we've got a little bit larger groups, then we talked about for instance 00:15:17.000 --> 00:15:22.900 cross-sectional studies and that those are typically like prevalence studies. They can be much 00:15:22.900 --> 00:15:29.400 bigger but it is still a snapshot, it is one moment in time and you check that certain state. What is 00:15:29.400 --> 00:15:37.300 the prevalence of Dyslexia in seven year olds, in Oslo, first graders, so you will have the first 00:15:37.300 --> 00:15:42.200 graders, you measure one time in one particular point in time. NOTE Treffsikkerhet: 83% (H?Y) 00:15:42.700 --> 00:15:48.200 It is usually also that's why people like it. It's usually simple to do because there's no 00:15:48.200 --> 00:15:53.700 follow-up, no intervention, but that could be risk of selection bias. You really need to be very 00:15:53.700 --> 00:16:00.150 careful with prevalence studies. You need to make sure you've got a good representative sample, 00:16:00.150 --> 00:16:05.000 because that's usually what you do, you want to know the prevalence or something and you take 00:16:05.000 --> 00:16:07.150 a sample to do it. NOTE Treffsikkerhet: 86% (H?Y) 00:16:07.150 --> 00:16:13.400 But if your sample is too small, of course you've got a problem, if you have selection bias because 00:16:13.400 --> 00:16:18.000 you selected already, it's not represented for the whole general population or something, you've 00:16:18.000 --> 00:16:25.200 got a problem. Now ecological studies are usually studies that look for association between exposure 00:16:25.200 --> 00:16:31.700 and outcome. They're often can also be between different countries and very often people will use 00:16:31.700 --> 00:16:37.450 register. So the data's maybe already be collected and all you do -well, all you do still big job to do- NOTE Treffsikkerhet: 74% (MEDIUM) 00:16:37.450 --> 00:16:44.800 but you ask for all the data and you combine them and compare them Etc. Every country has 00:16:44.800 --> 00:16:50.600 different national surveys. Usually they are in databases, we've got in the Netherlands. 00:16:50.600 --> 00:16:54.900 I worked for a couple of years for instance at the Cancer registration, the National Cancer 00:16:54.900 --> 00:17:02.350 registration, which means you've got a lot of data available and you can do your stats as needed. 00:17:02.350 --> 00:17:07.400 You look at risk factors, Etc. All the things. If your sample is big enough, NOTE Treffsikkerhet: 78% (H?Y) 00:17:07.400 --> 00:17:15.400 you can also look for confounding factors for relations between variables, for variables that you 00:17:15.400 --> 00:17:19.599 actually did, were not aware of, but you need big numbers then. NOTE Treffsikkerhet: 80% (H?Y) 00:17:19.599 --> 00:17:24.700 The opposite of a big number is again, this single case study. If I look a little bit more in 00:17:24.700 --> 00:17:31.300 detail, it's an experimental clinical study, is single case, very simple, we select one 00:17:31.300 --> 00:17:37.900 case and this would be an example: I've got here a tracking adult literacy acquisition with 00:17:37.900 --> 00:17:44.600 functional MRI, a single case study. Well, I will give you some time, these are abstract and 00:17:44.600 --> 00:17:49.250 I could ask you for what kind study design. Well, it doesn't matter, NOTE Treffsikkerhet: 84% (H?Y) 00:17:49.250 --> 00:17:55.600 all the texture around it. You just see it's about one adult. Meaning, this must be a single 00:17:55.600 --> 00:18:03.400 case study. Nothing else it can be. Now, if we look at case series, then you've got a group of patient 00:18:03.400 --> 00:18:09.700 at the similar diagnosis, but still a rather small group and you follow them over time. Usually a 00:18:09.700 --> 00:18:14.900 unique condition, but there's no control group. 00:18:14.900 --> 00:18:19.300 So no statistical validity or anything. This is what you do; you follow NOTE Treffsikkerhet: 91% (H?Y) 00:18:19.300 --> 00:18:24.200 a number of patients because you think they've got similarities there, interesting. You describe them 00:18:24.200 --> 00:18:31.300 and you report, that's all you do. Now. If you go to PubMed or anywhere, there are a lot of 00:18:31.300 --> 00:18:37.100 case series. So I'll give you another example, this is Guillain-Barre syndrome, patient that recent 00:18:37.100 --> 00:18:44.200 history of Zika Etc. And you can see it is case series of 19 patients. And if I look at the 00:18:44.200 --> 00:18:49.350 abstract, then I can immediately see 19 patients. NOTE Treffsikkerhet: 76% (H?Y) 00:18:49.350 --> 00:18:58.000 It's a descriptive study. No intervention. They just describe this disease, the zika virus. Now that 00:18:58.000 --> 00:19:04.100 was more on the left. That was the descriptive studies. On the right, we've got more complex studies, 00:19:04.100 --> 00:19:09.750 and we've got the analytical studies. And if we started observational on that side, you can see, 00:19:09.750 --> 00:19:15.900 there are several study designs as well. Now, cohort study is one of the nicest study designs, but 00:19:15.900 --> 00:19:19.300 you need a lot of patience or actually participants. NOTE Treffsikkerhet: 90% (H?Y) 00:19:19.300 --> 00:19:27.200 Since it's a group of people with a shared characteristics, all SNE students in 2021, that is a 00:19:27.200 --> 00:19:34.900 cohort. And I could compare this cohort with previous cohorts. Those are studies. And what you do is 00:19:34.900 --> 00:19:41.300 you are looking for exposure, or risk factors, like, well, again to be boring, take covid, you'll 00:19:41.300 --> 00:19:46.100 had online education. And we want to know, well, what is the difference between online teaching and 00:19:46.100 --> 00:19:49.250 outcome versus the on-site teaching. That could NOTE Treffsikkerhet: 91% (H?Y) 00:19:49.250 --> 00:19:56.500 be a cohort study if you've got enough students involved, it needs to be bigger numbers. Usually, 00:19:56.500 --> 00:20:02.800 it is forward, so prospective, typically prospective studies that I follow you now, I decide 00:20:02.800 --> 00:20:10.800 Okay, I'm going to follow all these students from now on till 2023. Sometimes you also look back, you call 00:20:10.800 --> 00:20:19.050 that a historical cohort and very often you measure, you've got exposures that you study similtenous NOTE Treffsikkerhet: 91% (H?Y) 00:20:19.050 --> 00:20:25.700 you want to know... So you've got age and sex and smoking, diabetes and you want to know how do 00:20:25.700 --> 00:20:31.700 they influence each other. So it's not just one thing that you will follow, especially in huge Health 00:20:31.700 --> 00:20:38.700 cohorts, you will see that there are long list of variables people are interested in. NOTE Treffsikkerhet: 84% (H?Y) 00:20:38.700 --> 00:20:45.400 Now, this is again, how a cohort study could look like. So, you see here, my overweight group. I've 00:20:45.400 --> 00:20:52.800 got a group of people that have normal weight. I follow them in future, in time, on time 00:20:52.800 --> 00:21:01.200 and I compare the risk factors. Now. This is an example. Effectiveness of influenza vaccine 00:21:01.200 --> 00:21:09.199 in the community dwelling elderly. Here is my abstract and you can see it is the effectiveness of NOTE Treffsikkerhet: 74% (MEDIUM) 00:21:09.199 --> 00:21:17.400 influenza vaccine in seniors over long-term. Now that sounds like a cohort. You can also see here, it's a 00:21:17.400 --> 00:21:24.400 pretty complex cohort. They have got 18 different cohorts of community-dwelling elderly members, 00:21:24.400 --> 00:21:32.900 and of one certain organization. They also have got different years that they follow them, that 00:21:32.900 --> 00:21:39.150 makes that you've got in total over 700,000 so-called person NOTE Treffsikkerhet: 84% (H?Y) 00:21:39.150 --> 00:21:46.300 seasons of observation. So that is we've got an X number of persons over different number of years. 00:21:46.300 --> 00:21:54.500 If every year is counted for every person. We've got over 700,000, almost 714,000 person seasons. 00:21:54.500 --> 00:22:04.100 Now, that means 10 seasons they were following many of these patients... 00:22:04.100 --> 00:22:09.300 How does it look, that outcome, how does that look? On the left NOTE Treffsikkerhet: 84% (H?Y) 00:22:09.300 --> 00:22:17.500 I've got my 18 cohorts. There is no influenza. The other group has got the influenza vaccination. 00:22:17.500 --> 00:22:24.800 I follow them up for 10 seasons and then I've got over 700,000 00:22:24.800 --> 00:22:32.100 seasons and I compare risk factors. This is concurrent cohort studies because they are next to each 00:22:32.100 --> 00:22:38.400 other. It's more than just one cohort. Very nice study, extremely good study, but we probably are not 00:22:38.400 --> 00:22:39.150 capable of doing NOTE Treffsikkerhet: 81% (H?Y) 00:22:39.150 --> 00:22:47.000 something like that. And for sure, not during your master thesis. But it is a common study design. 00:22:47.000 --> 00:22:51.900 So all the issue, if you read research, you need to understand what kind of designs are being used and 00:22:51.900 --> 00:22:58.300 what are the pros and cons. Now, this is an analytical observational study and we call that 00:22:58.300 --> 00:23:04.550 one case control study. So, first, you start with cases. Those are participants with the outcome, 00:23:04.550 --> 00:23:09.150 ADHD, dysplasia, you name it, and then you've got NOTE Treffsikkerhet: 89% (H?Y) 00:23:09.150 --> 00:23:16.300 controls, participants that I have not got that outcome. So they are typically developing 00:23:16.300 --> 00:23:22.600 or they don't have the disease. Then next you retrieve history of exposure in each group. Now, this 00:23:22.600 --> 00:23:27.500 is backward direction, this study. So it goes back to the past. There is a risk of 00:23:27.500 --> 00:23:34.100 selection bias. Recruitment is always tricky. You can mess up your total 00:23:34.100 --> 00:23:39.150 study design there, you need to check for that. There is a risk of confounding, of course. NOTE Treffsikkerhet: 78% (H?Y) 00:23:39.150 --> 00:23:44.800 But it is cheaper and less consuming for internet cohort studies be cool because cohort studies 00:23:44.800 --> 00:23:50.600 need big big numbers. Ideal for rare diseases, so you've got a few people that developed especially 00:23:50.600 --> 00:23:58.000 special disease and you want to know why is the other group with similar characteristics, why did they 00:23:58.000 --> 00:24:05.449 not develop that disease? What is different in their history? Sometimes they use matched case control, 00:24:05.449 --> 00:24:09.250 meaning they want for instance to control the NOTE Treffsikkerhet: 91% (H?Y) 00:24:09.250 --> 00:24:16.900 distribution of age, sex or socioeconomic status between groups, the control group and the study 00:24:16.900 --> 00:24:18.050 group. NOTE Treffsikkerhet: 91% (H?Y) 00:24:18.050 --> 00:24:24.700 Again, this it looks like, it can be anything, we've got ADHD or cancer. 00:24:24.700 --> 00:24:30.200 We've got a group with a disease or a condition and a group without, you look back in history's to 00:24:30.200 --> 00:24:36.100 see what happened here. Maybe these, if this is lung cancer, maybe they smoked a lot. Maybe that 00:24:36.100 --> 00:24:42.000 would be a risk factor. So you're going to check in the past what was different 00:24:42.000 --> 00:24:43.200 there. NOTE Treffsikkerhet: 73% (MEDIUM) 00:24:43.200 --> 00:24:49.700 Now, if I look at here, the case control study. I've got an example here again, tobacco use in 00:24:49.700 --> 00:24:57.000 risk of myocardial infarction in 52 countries in bloody blah, a case control study. What 00:24:57.000 --> 00:25:02.800 did they do? Here is the abstract but I'm going to help you again a little bit. And then, you can 00:25:02.800 --> 00:25:09.850 see that for instance, to assess the risk associated with tobacco. That was what they wanted to do. 00:25:09.850 --> 00:25:13.050 It was about tobacco and NOTE Treffsikkerhet: 84% (H?Y) 00:25:13.050 --> 00:25:20.100 they were looking at acute problems, acute myocardial infarction. They had a huge group there over 00:25:20.100 --> 00:25:27.200 many many countries. So this is again, one of high evidence very nice study. The next thing they 00:25:27.200 --> 00:25:34.600 did is they want to look at the relation between risks of that disease and former smoking, Etc. 00:25:34.600 --> 00:25:42.100 Okay. This is their conclusion. But how does that look in this picture? There are my patients 00:25:42.100 --> 00:25:43.450 with an infarct. NOTE Treffsikkerhet: 66% (MEDIUM) 00:25:43.450 --> 00:25:50.500 There are the ones that do not have that infarct. I go back in the past and I'll check who did use 00:25:50.500 --> 00:25:58.000 tobacco and what kind of tobacco did they use. So that is a case control study. NOTE Treffsikkerhet: 85% (H?Y) 00:25:58.100 --> 00:26:06.900 Back to this one. So we are still with the analytical and we are about to talk about the 00:26:06.900 --> 00:26:15.000 experimental. So if I talk about experimental studies, then I usually talk about randomized control 00:26:15.000 --> 00:26:21.700 trials or in short you will very often see RCTs, not randomized control trials. So they did 00:26:21.700 --> 00:26:27.000 not randomize people but it's still having different comparison groups. We will talk about shortly 00:26:27.000 --> 00:26:28.650 about factorial designs and NOTE Treffsikkerhet: 87% (H?Y) 00:26:28.650 --> 00:26:37.000 cluster randomized designs. Any questions here? Am I too fast? Am I too slow? NOTE Treffsikkerhet: 91% (H?Y) 00:26:38.300 --> 00:26:46.000 Let's see. Just hang in ten more 00:26:46.000 --> 00:26:53.700 minutes and I'll let you go for a break. So this is an RCT. So randomized control trials 00:26:53.700 --> 00:26:59.900 are actually the highest level of studies that you can perform and it is really useful. 00:26:59.900 --> 00:27:07.100 What you do is you've got a group of participants that are randomly assigned to two or more arms, 00:27:07.100 --> 00:27:09.000 an arm is a group. NOTE Treffsikkerhet: 79% (H?Y) 00:27:09.000 --> 00:27:14.800 And one is getting an intervention. The other ones could be comparable groups and 00:27:14.800 --> 00:27:22.100 control refers to that you've got a concurrent control or comparator group, and the control group 00:27:22.100 --> 00:27:30.000 usually receives no therapy or a placebo therapy. Meaning, I'll give you something like a pill 00:27:30.000 --> 00:27:35.200 but it's not working or Sham activity, which they do for instance when you give 00:27:35.200 --> 00:27:38.950 neurostimulation in physiotherapy or in swallowing exercise NOTE Treffsikkerhet: 85% (H?Y) 00:27:38.950 --> 00:27:46.800 you put neurostimulation on particular areas of your skin, it will stimulate, but if it's Sham 00:27:46.800 --> 00:27:54.800 neurostimulation, you feel something sensory going on but it doesn't really result in 00:27:54.800 --> 00:27:58.100 contraction of any muscles. NOTE Treffsikkerhet: 80% (H?Y) 00:27:59.000 --> 00:28:06.949 Now, some terms, when you talk about RCTs, allocation concealment, that means random allocation 00:28:06.949 --> 00:28:13.000 and that is implemented without any bias. So it should be at random. You've got groups in 00:28:13.000 --> 00:28:21.700 at random, you assign them to one of the arms. Blinding, also referred as masking is you refer to 00:28:21.700 --> 00:28:27.449 make individuals involved in RCTs, but they are not aware NOTE Treffsikkerhet: 80% (H?Y) 00:28:27.449 --> 00:28:34.300 of the intervention, meaning that could be the participants who does not know what this is Placebo or 00:28:34.300 --> 00:28:41.800 not. The researcher, who may not have a cue to see which group 00:28:41.800 --> 00:28:48.700 these participants belong to. Healthcare, practitioners Etc. 00:28:48.700 --> 00:28:54.600 So it can be blinded from the side of the patient. But also on the side of the in the clinician who's 00:28:54.600 --> 00:28:57.400 providing the intervention or even or NOTE Treffsikkerhet: 82% (H?Y) 00:28:57.400 --> 00:29:03.400 data analyst, who is doing the statistical analysis, blinding can be at different levels. Then we've 00:29:03.400 --> 00:29:08.100 got things like intention-to-treat analysis that anybody who's enrolled in a certain group should 00:29:08.100 --> 00:29:13.200 actually intend to finish that certain group will also not change group 00:29:13.200 --> 00:29:20.200 or loss to follow-up and compliance because very often a problem is the degree to which you 00:29:20.200 --> 00:29:26.700 can keep your participants at here to the prescribed interventions. If you've got a beautiful RCT, 00:29:26.700 --> 00:29:27.300 beautiful desing, NOTE Treffsikkerhet: 87% (H?Y) 00:29:27.300 --> 00:29:33.700 but compliance of participants is low, you've got a problem because then you can't 00:29:33.700 --> 00:29:38.000 measure the outcome of your intervention because people were not interested or maybe your 00:29:38.000 --> 00:29:44.000 intervention was painful or maybe it was time consuming. So there's always this patient burden that 00:29:44.000 --> 00:29:50.100 you also should keep in mind when you design your studies. Also, when you design for instance surveys, 00:29:50.100 --> 00:29:55.500 and that's of course for RCTs, you can't just keep adding questions because 00:29:55.500 --> 00:29:57.450 sometimes the survey takes so much NOTE Treffsikkerhet: 81% (H?Y) 00:29:57.450 --> 00:30:02.100 time, you will lose your participants. They just stop. They're not willing to spend two hours on 00:30:02.100 --> 00:30:09.900 your survey. But back to my RCT. Cancer patient, random treatment, getting a placebo 00:30:09.900 --> 00:30:13.800 and real treatment, follow-up, comparing results. NOTE Treffsikkerhet: 79% (H?Y) 00:30:13.800 --> 00:30:20.200 Here we go. This is a real one. This actually one that we perform. Surface Electrical Stimulation in 00:30:20.200 --> 00:30:26.450 Dyspchagic Parkinson Patients A Randomized Control Trial. What we did was... So you've got electrical stimulation. 00:30:26.450 --> 00:30:32.400 Those are electrodes you place in the neck and it should help swallowing and especially we did that 00:30:32.400 --> 00:30:38.300 in Parkinson patients, patients with Parkinson will have voice problems but also swallowing 00:30:38.300 --> 00:30:44.150 problems. So, that was the whole idea. So what we did is a randomized control trial. We had NOTE Treffsikkerhet: 87% (H?Y) 00:30:44.150 --> 00:30:49.200 Surface Electrical Stimulation, surface means we didn't put needles in these poor people. We just 00:30:49.200 --> 00:30:55.750 put electrodes on top of the skin and we did that in the neck of patients with dysphagia problems. 00:30:55.750 --> 00:31:04.000 We had three groups, meaning three arms. One was traditional speech language pathology 00:31:04.000 --> 00:31:10.000 treatment, then we are to stimulation. We had a high stimulation and a low 00:31:10.000 --> 00:31:14.100 stimulation. High stimulation, then you get contraction, motor contraction. NOTE Treffsikkerhet: 83% (H?Y) 00:31:14.100 --> 00:31:20.200 Lower level of simulation is more sensory stimulation. So, three 00:31:20.200 --> 00:31:26.850 groups. Apart from the conclusion, It's more interesting to look at this picture. 00:31:26.850 --> 00:31:35.200 So on the left, I've got my Parkinson's disease patients. We had one random group allocation, one 00:31:35.200 --> 00:31:43.400 was traditional therapy, this was the motor level therapy and the sensory stimulation therapy group. 00:31:44.200 --> 00:31:49.000 And then you follow them. They all had something 00:31:49.000 --> 00:31:55.200 like 15 sessions over three weeks. We compare the outcome and we were looking for differences 00:31:55.200 --> 00:32:00.450 between the groups. This is an RCT with something like over a hundred and twenty people involved. 00:32:00.450 --> 00:32:06.600 It is labor intensive. You need a lot of resources and a lot of time, it's complex. NOTE Treffsikkerhet: 91% (H?Y) 00:32:06.600 --> 00:32:13.600 I'm checking the time. I am going to stop right here for now. NOTE Treffsikkerhet: 85% (H?Y) 00:32:14.000 --> 00:32:17.600 And I will stop recording. NOTE Treffsikkerhet: 83% (H?Y) 00:32:18.700 --> 00:32:26.900 So interventional, if you look at intervention experimental study, so we've got the rcts. We've got 00:32:26.900 --> 00:32:32.900 also Interventional studies without concurrent controls, and we've got the before after pre-post 00:32:32.900 --> 00:32:41.700 interventions, which are single arm, no comparator ones. Now if I look at a few of these in 00:32:41.700 --> 00:32:46.300 particular, if we will look at 00:32:46.300 --> 00:32:48.450 for instance in experimental design NOTE Treffsikkerhet: 79% (H?Y) 00:32:48.450 --> 00:32:55.100 that you may run into is victorial study designs. So, what you've got is two or more 00:32:55.100 --> 00:33:01.300 interventions and they are available for a particular condition. And you do not know whether an 00:33:01.300 --> 00:33:07.900 intervention itself has good effects or the combination is better. So, most 00:33:07.900 --> 00:33:13.800 simple factorial design is a 2x2 factorial design. So suppose you've got two 00:33:13.800 --> 00:33:18.350 interventions, you've got intervention A and B, and participants NOTE Treffsikkerhet: 90% (H?Y) 00:33:18.350 --> 00:33:24.700 are randomly assigned to one or more intervention combinations. Because there is an A alone, 00:33:24.700 --> 00:33:32.500 B alone, A and B, or nothing, that is no treatment. Now in a picture that looks like this. So here you 00:33:32.500 --> 00:33:40.000 can see treatment A, treatment B. And the whole group is randomly assigned to these four conditions 00:33:40.000 --> 00:33:46.100 and that way you try to understand so what is better, should you have a pill and physiotherapy, 00:33:46.100 --> 00:33:48.350 or just physiotherapy, NOTE Treffsikkerhet: 91% (H?Y) 00:33:48.350 --> 00:33:55.800 or don't bother at all and don't do anything. So control is very often used for no treatment at all. 00:33:55.800 --> 00:34:02.850 But people use the term not consistently. Then we've got 00:34:02.850 --> 00:34:09.500 another type of intervention studies is experimental the crossover study designs, and that is where 00:34:09.500 --> 00:34:16.300 you've got a group of participants that is being divided into groups. And then you cross them over, 00:34:16.300 --> 00:34:18.350 first you get a first treatment, NOTE Treffsikkerhet: 83% (H?Y) 00:34:18.350 --> 00:34:24.850 then there is a washout period and then you make them cross over to the other treatment. 00:34:24.850 --> 00:34:30.600 And as a result, participants are their own control. So you can see, well, were they better 00:34:30.600 --> 00:34:37.699 functioning under treatment A or actually under treatment B. And this is, you need fewer subjects 00:34:37.699 --> 00:34:43.900 that way. So, I've got here, I've got my group of overall participants are randomized into 00:34:43.900 --> 00:34:48.400 treatment A or B. Whatever happens, happens. Then a washout here. NOTE Treffsikkerhet: 72% (MEDIUM) 00:34:48.400 --> 00:34:54.199 Suppose it simple. This is a pill A and that's a pill B. You give it, you check how 00:34:54.199 --> 00:35:00.600 they are functioning. Then you give a period of no pills. So the affect washes away and then you 00:35:00.600 --> 00:35:08.800 swap pills. That is the whole idea behind it. Wash out period means a period that any effects from 00:35:08.800 --> 00:35:12.800 the previous treatment is kind of faded away. NOTE Treffsikkerhet: 85% (H?Y) 00:35:13.100 --> 00:35:20.800 Okay. Now then we've got something that's being called, that's being named a Interventional or an 00:35:20.800 --> 00:35:25.300 intervention of cluster randomized trial. 00:35:25.300 --> 00:35:35.000 Emphasis is on cluster. So, the RCTs we discussed previously, you randomize the participants, but in 00:35:35.000 --> 00:35:41.100 a cluster randomized trial, the intervention that you apply to groups, it's not the 00:35:41.100 --> 00:35:42.500 individuals you randomize NOTE Treffsikkerhet: 74% (MEDIUM) 00:35:42.500 --> 00:35:50.800 but for instance the social units or groups, for instance you randomize schools. 00:35:50.800 --> 00:35:56.900 School A to a certain treatment and school B is going to do the other 00:35:56.900 --> 00:36:04.850 treatment, but you randomize the unit to a certain intervention, not the particular individual 00:36:04.850 --> 00:36:11.300 within that unit. So not the child, but the school. Or you could have for instance private practician 00:36:12.450 --> 00:36:18.800 that are being randomized, and then you've all the participants from one practitioner are 00:36:18.800 --> 00:36:24.200 getting the same treatment. So you don't randomize at patient level or subject level, 00:36:24.200 --> 00:36:31.900 but at a higher level and that is useful for instance, sometimes if you've got three treatments and 00:36:31.900 --> 00:36:38.800 you are in a clinical setting, sometimes it's impossible that people are blinded to treatments or 00:36:38.800 --> 00:36:42.550 patients see that whatever they NOTE Treffsikkerhet: 83% (H?Y) 00:36:42.550 --> 00:36:48.100 get is different from something else, or it's too complex to have three different treatments with 00:36:48.100 --> 00:36:54.700 patients that are just being on the same ward. For that reason they say, 00:36:54.700 --> 00:37:01.100 okay, we randomize per unit and each unit will have a certain 00:37:01.100 --> 00:37:06.900 treatment and you can understand the cluster randomized trials are big big big RCTs. 00:37:06.900 --> 00:37:12.450 We're talking about many different subjects involved. 00:37:14.500 --> 00:37:21.800 Now, we talked already a long time about systematic review and that is based on, that is secondary 00:37:21.800 --> 00:37:29.100 research because you do not perform an original study. It is based on treatments from interventions 00:37:29.100 --> 00:37:35.400 from others or literature from others and you summarize it using guidelines in case of a systematic 00:37:35.400 --> 00:37:42.300 review you will use the guidelines from Prisma. Why? Because those are internationally 00:37:42.300 --> 00:37:43.950 considered to be the best guidelines NOTE Treffsikkerhet: 81% (H?Y) 00:37:43.950 --> 00:37:51.700 for reviews. And we've looked into that and we said why do we do reviews. Because 00:37:51.700 --> 00:37:57.800 there are simply too many articles. What we do is you go to PubMed. You find 00:37:57.800 --> 00:38:04.700 yourself a mesh term or subject heading, here education. You click on review or systematic review and 00:38:04.700 --> 00:38:11.100 you can see there is quite a few hits there, over 50,000 and you can also see that the number of 00:38:11.100 --> 00:38:13.950 reviews are increasing, the number of Publications. NOTE Treffsikkerhet: 80% (H?Y) 00:38:13.950 --> 00:38:19.300 In the 70s maybe they read every single article. I've no idea how they survived, 00:38:19.300 --> 00:38:24.000 but now thank God, we've got reviews. These are just low numbers because this is an old picture. 00:38:24.000 --> 00:38:30.800 It was just at the beginning of the year. All these study designs 00:38:30.800 --> 00:38:38.800 were just examples, nothing else but examples, so depending on the purpose of your study and that is 00:38:38.800 --> 00:38:43.950 also for your research proposal and that is for your master thesis later on, you will select NOTE Treffsikkerhet: 91% (H?Y) 00:38:43.950 --> 00:38:49.800 a certain study design and some study design. You just want to stay away from because there's no 00:38:49.800 --> 00:38:58.200 way you can get a cohort organized or a RCT, but you can see also the different type of studies like 00:38:58.200 --> 00:39:04.800 again, we talked about prevalence, you use a cross-sectional survey, so something one moment of 00:39:04.800 --> 00:39:12.050 measurement, we talk about collecting qualitative data, interviews, 00:39:12.050 --> 00:39:14.200 cognitive interviews and all of that, NOTE Treffsikkerhet: 89% (H?Y) 00:39:14.200 --> 00:39:20.600 effectiveness, the control trials, but again, I think RCTs and cohorts are little bit out of 00:39:20.600 --> 00:39:28.200 your reach. But if you want to report in your master about good evidence, you would love to have 00:39:28.200 --> 00:39:35.649 data from cohorts and RCTs. You want to report on the highest level only. But take-home message 00:39:35.649 --> 00:39:42.300 is different research questions have different study designs. That is good. So you don't want all 00:39:42.300 --> 00:39:43.850 of you do a survey, NOTE Treffsikkerhet: 80% (H?Y) 00:39:43.850 --> 00:39:50.900 it depends on what do you want. An advantage for instance if we look at here again, with serveys 00:39:50.900 --> 00:39:57.500 you can have much bigger numbers of participants, you can do a national survey. One of my 00:39:57.500 --> 00:40:03.000 masters students did a national survey because it's the same survey, you just 00:40:03.000 --> 00:40:09.400 need to have access to recruitment. If you do interviews, that will very likely be a much smaller 00:40:09.400 --> 00:40:13.750 group. You can, however, the advantage of an interview is of course that you can NOTE Treffsikkerhet: 76% (H?Y) 00:40:13.750 --> 00:40:19.500 while performing an interview, you can adjust your questions if needed. But sometimes you don't want 00:40:19.500 --> 00:40:24.600 that and you want a very structured interview to get the same questions for 00:40:24.600 --> 00:40:31.450 each participant. It depends on what you are looking for. But the purpose is of your research. 00:40:31.450 --> 00:40:39.300 Now, we know by now pros and cons for, different study designs. There are many websites where you can 00:40:39.300 --> 00:40:43.950 have look at different clinical studies and these are only examples, NOTE Treffsikkerhet: 81% (H?Y) 00:40:43.950 --> 00:40:49.400 you will come across when you read literature. You will think what on Earth is this, than 00:40:49.400 --> 00:40:56.900 have a look, read through and just know that people are creative, people combined study designs and 00:40:56.900 --> 00:41:02.700 it has consequences for things like statistical analysis, but just that you know and are familiar with 00:41:02.700 --> 00:41:12.000 the most basic ones. Now, very briefly on levels of evidence, or do I need to allow anybody to ask a 00:41:12.000 --> 00:41:13.600 question? NOTE Treffsikkerhet: 91% (H?Y) 00:41:14.500 --> 00:41:22.900 Okay, then I just continue. Level of evidence. Now. There are different systems to report on level 00:41:22.900 --> 00:41:30.300 of evidence. And the most simple one is the ABC level of Siwek. I must say, at the beginning I was 00:41:30.300 --> 00:41:34.700 quite pleased with that one. I actually never use it anymore because it is a little bit 00:41:34.700 --> 00:41:42.000 over-simplification. So he says, level a is RCT. Yeah, sure. That is fantastic level. He says level C, 00:41:42.000 --> 00:41:44.950 I know I skip level B, but level C he says that's consensus or NOTE Treffsikkerhet: 88% (H?Y) 00:41:44.950 --> 00:41:51.700 expert opinion. Nine out of ten, you you exclude consensus or expert opinions for instance in 00:41:51.700 --> 00:41:57.800 reviews. This is just one person telling his or her opinion. You can tell of course there is a 00:41:57.800 --> 00:42:04.200 huge difference between RCT outcome versus one expert only, but if you use the system, the 00:42:04.200 --> 00:42:11.000 problem is, you get all these non randomized clinical trials. So many, many are 00:42:11.000 --> 00:42:14.850 ending up as a level B. It is a very easy to use this, NOTE Treffsikkerhet: 77% (H?Y) 00:42:14.850 --> 00:42:20.700 I must agree. I myself have used it as well at the beginning but that are better systems. 00:42:20.700 --> 00:42:27.000 So RCTs, these are well-designed, meaning, if it's poorly designed that 00:42:27.000 --> 00:42:34.400 actually also excluded. But this is one way of doing it. This the ABC by Siwic. And another way of 00:42:34.400 --> 00:42:39.400 doing it and that is more sophisticated. That is the one that we've seen before from the national 00:42:39.400 --> 00:42:45.200 health and medical research Council NHMRC, that is a spelling error. NOTE Treffsikkerhet: 78% (H?Y) 00:42:45.200 --> 00:42:51.500 And they come up with these levels. I make it a little bit nicer. We've seen this before, again, 00:42:51.500 --> 00:42:57.500 single cases, even below case series, then we've got the case series, comparative studies, two types. 00:42:57.500 --> 00:43:04.100 This is historical. This is cohort or case control, the pseudo RCT, so it's almost a RCT. Maybe it's the 00:43:04.100 --> 00:43:08.400 blinding not too good, maybe there's something with the randomization. Then we've got the real RCTs. 00:43:08.400 --> 00:43:14.200 Then you've got systematic reviews of RCTs on top of that. NOTE Treffsikkerhet: 78% (H?Y) 00:43:14.200 --> 00:43:21.800 And this these are examples. So case series is group reports of patients with similar diagnosis. That's 00:43:21.800 --> 00:43:30.000 one group. Comparative studies is these are without concurrent control so it is a 00:43:30.000 --> 00:43:36.000 historic control group, but not a group at the same time and that makes it weaker compared to this 00:43:36.000 --> 00:43:42.400 one where you've got two or more groups at the same time. So one is getting a 00:43:42.400 --> 00:43:44.900 play intervention and the other one will not NOTE Treffsikkerhet: 85% (H?Y) 00:43:44.900 --> 00:43:50.800 get the play intervention, but you measure at the same time. And the advantage of doing that is that 00:43:50.800 --> 00:43:58.800 for instance with historic components or groups very often there are slight differences in who 00:43:58.800 --> 00:44:07.300 was giving the treatment, with the assessments or with protocols. There's always risk of bias. It's much more neat, 00:44:07.300 --> 00:44:14.850 it's much more professional iff you can do concurrent at the same time two groups that NOTE Treffsikkerhet: 87% (H?Y) 00:44:14.850 --> 00:44:21.500 needs to be comparable in subject characteristics like the degree of severity or severity of 00:44:21.500 --> 00:44:29.400 problems or age and gender, any variable that could influence the intervention. Then we've got the 00:44:29.400 --> 00:44:36.600 pseudo ones so that it's almost RCTs but for instance, the randomization is not really 00:44:36.600 --> 00:44:41.600 at random. Maybe it was just that they say in a certain order. One, two, three. One, two, three, 00:44:41.600 --> 00:44:44.850 three arms, each group goes to one. First one goes in one, NOTE Treffsikkerhet: 78% (H?Y) 00:44:44.850 --> 00:44:50.600 the second patient in group 2 and 3 to group 3, and then the 4th to group one. Again, that is 00:44:50.600 --> 00:44:57.100 not real RCT. Actually, not really a randomized, but you can understand it's very near to 00:44:57.100 --> 00:45:03.500 randomized control trials. Systematic reviews are of course, the good thing about systematic 00:45:03.500 --> 00:45:11.300 reviews of RCTs is when you can perform meta-analysis and we've seen that meta-analysis combines 00:45:11.300 --> 00:45:14.850 the output of individual RCTs. Well, what's nicer NOTE Treffsikkerhet: 84% (H?Y) 00:45:14.850 --> 00:45:20.700 than having an overview of any intervention in a certain area. All these are RCTs, 00:45:20.700 --> 00:45:26.700 highest level of evidence and then you combine them. Of course, there are big, big problems or 00:45:26.700 --> 00:45:34.100 risks, challenges. For instance you generalize, meaning, there can be differences 00:45:34.100 --> 00:45:42.000 in target populations, differences in interventions, differences in maybe statistical analysis that 00:45:42.000 --> 00:45:44.950 these RCTs used. So there are limits NOTE Treffsikkerhet: 72% (MEDIUM) 00:45:44.950 --> 00:45:51.200 to combining when it's too different, you simply cannot perform 00:45:51.200 --> 00:45:58.200 meta-analysis or you need to interpret the outcome of your meta-analysis really, really careful. 00:45:58.200 --> 00:46:06.000 Okay. Now, on top of all of that we already said that is where clinical guideline comes. If you 00:46:06.000 --> 00:46:10.400 write a clinical guideline that is a massive, massive job. You've got people that write all books 00:46:10.400 --> 00:46:14.800 about it. So usually what you do, if you write a clinical guideline, you NOTE Treffsikkerhet: 81% (H?Y) 00:46:14.800 --> 00:46:24.100 first meet with experts that represent all important professions that are linked in your area. If 00:46:24.100 --> 00:46:29.800 you talk about children in ADHD, you want to have maybe a psychologist in there. You want to 00:46:29.800 --> 00:46:34.649 have someone who understands stats, you want the teachers involved, any 00:46:34.649 --> 00:46:41.500 clinician or educationalist or parents, they can be involved. Then first a group like that decides 00:46:41.500 --> 00:46:44.899 what are the main topics we want to address in the guideline. NOTE Treffsikkerhet: 82% (H?Y) 00:46:44.899 --> 00:46:53.100 Intervention, diagnosis, prognostic factors. So you need to decide and then very often step 00:46:53.100 --> 00:46:59.300 number two for clinical guidelines, you perform systematic 00:46:59.300 --> 00:47:05.300 reviews to retrieve the highest level of evidence. What do we know about all these questions that 00:47:05.300 --> 00:47:12.100 have been formulated by an expert panel, then you summarize all of that into a clinical guideline. 00:47:12.100 --> 00:47:14.750 I make it sound like a simple step. Believe me it NOTE Treffsikkerhet: 81% (H?Y) 00:47:14.750 --> 00:47:20.400 takes several years for to write complete full good clinical guidelines because there is also feedbacks. 00:47:20.400 --> 00:47:27.800 You write it, it goes next to the expert, goes back to the field, to the clinicians, 00:47:27.800 --> 00:47:35.400 to the educationalists. You ask for feedback. You revise it Etc. And also clinical guidelines are not 00:47:35.400 --> 00:47:42.300 static. They are non-stop. You need to update them. You need to improve them and add what do we 00:47:42.300 --> 00:47:44.800 know now, what did you learn, what new research NOTE Treffsikkerhet: 89% (H?Y) 00:47:44.800 --> 00:47:51.300 is out there. So once guidelines out there you need to keep continuing to update them, which is also 00:47:51.300 --> 00:47:57.900 a big big responsibility and what you see sometimes is that guidelines are published and then people 00:47:57.900 --> 00:48:02.900 are so glad that is published but they don't update them ever and that's a 00:48:02.900 --> 00:48:07.750 problem because then you are actually using all outdated information. NOTE Treffsikkerhet: 90% (H?Y) 00:48:07.750 --> 00:48:16.100 Okay, now NHMRC, they also have got a few statement guidelines and they say evidence based in 00:48:16.100 --> 00:48:21.500 terms of number, is the first two number of studies, if you've got guidelines you talk about number 00:48:21.500 --> 00:48:26.400 of studies, the level of evidence and the quality of studies and that is you need to address is 00:48:26.400 --> 00:48:33.500 there risk of bias, you want to know about consistency between studies. If one study says your 00:48:33.500 --> 00:48:38.300 therapy is fantastic and the other study says, well, I can't see anything, I don't see anything... NOTE Treffsikkerhet: 89% (H?Y) 00:48:38.300 --> 00:48:44.900 You need to know are there 10 studies confirming the results or are the outcome all 00:48:44.900 --> 00:48:52.200 over the place. So consistency is important. You want to know about the clinical impact of the 00:48:52.200 --> 00:48:59.600 proposed recommendation. Does it make a difference in clinics? It should have a clinical impact. 00:48:59.600 --> 00:49:05.200 Otherwise, why do you write the recommendation if it's actually not important? And you need 00:49:05.200 --> 00:49:08.200 to know can you generalize it. If you are looking NOTE Treffsikkerhet: 73% (MEDIUM) 00:49:08.200 --> 00:49:14.400 at the literature and you find certain statements, is that only for that single study? Or can you 00:49:14.400 --> 00:49:20.800 generalize it? Is it only for ADHD or can you generalize it to ASD? And is it only for children? Maybe also 00:49:20.800 --> 00:49:27.800 for parents? So generalizability is always an issue that keeps coming back. NOTE Treffsikkerhet: 88% (H?Y) 00:49:28.000 --> 00:49:33.600 And then you need to have a look at, it says Healthcare context because they are more written 00:49:33.600 --> 00:49:38.300 from Healthcare context, but actually you can replace it by educational context or whatever you 00:49:38.300 --> 00:49:44.500 want. Once you've got your guidelines, they are evidence-based, they should be consistent, clinical 00:49:44.500 --> 00:49:52.300 impact, generalizability. And of course, applicability. Can you actually apply your recommendations 00:49:52.300 --> 00:49:57.500 in education or is it so detailed and a nightmare that is just impossible NOTE Treffsikkerhet: 85% (H?Y) 00:49:57.500 --> 00:50:04.900 to meet all these suggestions, all these recommendations. That together, make something a good 00:50:04.900 --> 00:50:12.400 guideline. If you give people recommendations that you cannot follow in clinics or in education, 00:50:12.400 --> 00:50:19.250 then yeah, no one will you will follow up on that. Now, that was 00:50:19.250 --> 00:50:27.100 way too fast because I'm speeding up. Variables, study designs, and level of evidence. 00:50:27.100 --> 00:50:27.450 Now, I'm going to stop here, NOTE Treffsikkerhet: 75% (MEDIUM) 00:50:27.450 --> 00:50:32.800 because I want to show a few examples. Okay, I'm going to stop the recording.