Computer generated transcript
Warning!
The following transcript was generated automatically from the content and has not been checked or corrected manually.
Hi, it's me. People from surgery should be arriving soon. I'm just going to hit record. And I'm going to do that for all 14 and then come back because this is my beautifully yours. Uh, you do. I'm just wondering whether these guys have had the the little session. I think they've had it yet. I'm pretty sure they haven't had a session. I mean, uh, we had some, like, session for, like, two hours today. What's his face? What's what's his name? Uh, no. The nice guy, Uh, Richie? No, the surgeon. What's his name? Jame Jame is really nice. Yeah, he is. He's gonna give you He's going to give you all the keys to I C three. OK, so he does. I think every year he does a little. I don't know if you guys have it booked in, but he does a little session, and then he basically will give you, like, title's for your I say three. Or you can ask him like I think it's a little bit further on, but like you can bring your drafts and your graphs and like you can present what you have. And they'll tell you if It's good or not. What? Really? For performative like summer service for the summer. Okay. Thank you. I'll get that in mind. Have you had, uh, you know, So we literally had a two hour session going through some sort of I don't know what kind of data you had, but for us, we have some cataract audit data. Sure. Um, and which went through that? To be honest, it wasn't that helpful. I think a lot of us are gonna fused. What is the, uh, this Friday or the Friday after? Um, this Friday. Okay. So Okay. So in five days, yeah. Yeah, but it's a group, right? Yeah. Yeah, yeah, that's fine. Um, is anyone else coming or I don't know, I saw more people in the chat. I mean, in the main lecture. I'm not sure. Well, I mean, if you guys have any questions, I mean, I can't I don't Really? Yeah. Or feel free to share screen or something. If you have anything. I don't know if you can do that, actually, but for us just to give some context, uh, we've been given, like, national audit data. Yeah, the same the same contracts know we had bowel cancer. Okay. Like for the official one, It's probably going to be the national. What is it will be the same one that they get given for the formative. Yeah, national bowel, like bowel cancer outcomes audit. And it's so it's really annoying because audit data doesn't really translate well into, like, answering a good question. Yeah. Yeah, that's exactly our problem. Yeah, Yeah, it's a bit of like, just you have to fake a little bit, like, make it seem like it's like, you know, world changing and everything when it really isn't. But, like, they're not going to give you our CT data. So you have to make it seem interesting if that makes sense. Yeah, it's like, Yeah, no, go ahead. OK? Um yeah. So they told us that, um, we need to come up with something novel if we got to get a high mark. Um, but like you said, it's not the stuff we've been. The data we've been given, For instance, things like number of surgeons in the trust that did this operation, um, a number of, um, cataracts performed a number of complications. All these stuff like it doesn't sound very noble. So I was wondering how how do you go about finding a noble hypothesis? I think that's the hardest part so far. At the end of the after, you have to consider that considering, like, if you if you want to find something novel in a piece of audit data is gonna be very hard. And realistically, the stuff that is novel everyone's going to do. Um I think the main thing I would say Whatever you're investigating, make sure it has a real life clinical impact or like link it to impact. For example, if you're saying like I don't know like radiotherapy, does it reduce the rate of, um, does it does it improve your cancer resection margins? You need to link that to the real world rather than just like you need to contextualize as much as possible. Basically, what impact does have I Why do you want to find this out basically like, for example, like okay, if we do radiotherapy, um, if we increase the about radiotherapy, quite surgery and you basically find lots of articles to tie into, like why this state it was important to the real world, is that's what they like, really care about And then the how the the story is told. Basically, yeah, the they do love a good story like it's it doesn't have to be like a lot of people. I don't I don't remember our formative. I think we had. We did have bowel cancer, but it was like a shorter, wasn't it, like a shorter data set or something like that, like it's not a lot of people try to ask, like, really like random questions or try to make, like, take it way too much out of context and then with the thing with the amount of data that you have, then you can't really explain it or you need like, a really long report. And you have, like, a word limit of like 2000 something, um, so, like you can keep it simple as long as you're like explanation and your analysis like, makes sense like it. I don't know how to explain it like I did. I did not like number of surgeries and mortality as like a primary correlation and then, like a number of surgeries and another outcome, and it's like if you look at Autodata like That's probably the first two that you can like, maybe pare and like, it actually turned out fine. Like I got a 68% which is relatively good for surgery BSC. But it's like I kept it really simple, like you don't have to go do like, world changing questions as long as you like, can fit a good analysis and, like clean charts like don't overpopulate things, um, a clean hypotheses And then you can, like, throw in nice arguments for the discussion. Like it I know they like they make it seem like you need to do this big thing, but just as long as you know what question you're asking and like, how to, like, explain it in a nice and like, wow, we've discovered something amazing here, even though you didn't I don't know if that makes sense, but like don't go and try and make weird correlations that you know don't exist. Like don't don't make shit up. Right? OK, so So your advice is to be novel in your discussion, as opposed to, like, try to Yeah, hasn't, like, make sure you can approach it from, like, a nice angle and maybe maybe your angles to looking at. The question can be innovative, but the question itself doesn't have to be innovative. Does that make sense? Yeah, but you should. I I really advise you to go to this thing with Jamie because he's he'll answer most of, like, he's so patient. We spent, like, four hours last time with him. Thing is, this, uh, I think they might change. Who's your lead for the, uh, the three? Um, I think I don't know. I think it's, uh, fine, but And they should have a feedback session. Like, for your some motive where you can like a check in. I think that's what they called it. Like you can make sure you have, Like, a question and some graphs, uh, for that so that you can, like, show them and ask them. Like, I think, you know, I'm going to do this and conclude this or whatever. Does this sound reasonable? And either they say yes or no, or they they should. All right. How, like, how has it been so far? Uh, I see a three, uh, like, uh, like all of them. Uh, I see one or two are really boring. So, like, uh, I like doing I see a to Did you Did you make a nice video? If there's not enough time to Well, I find that there's not enough time to fully um, do what you want to especially, um, for us before I see to We have to do, um Shafiee. Yeah, And that that a alcohol week? Pretty much, Yeah. The turnaround is really like, I don't think the things produced are a very good quality, to be honest, but it's like you just don't have time to do better. That's that's fine. Um, any a question about I see a three? So they said, uh, well, Jamie said not to do correlation, but explain association or something. I don't quite understand it. Um, he said, Didn't do coronation do association. Yeah, it's kind of like you had a big thing with us about like if you're just looking for correlation like don't look aimless of Correlation, basically, and he was like he went on the ground about the R number, and how the R number isn't is a lie is a lie basically, and then you need to apply to the real world and stuff like that. So I think the main thing to take away from it is, um, like some viste askey with regards to, like, correlation stuff like there's an obviously, if there's an obvious thing that makes sense should be associated and go with someone like that, Rather something like How many pigeons does it take is that I mean, like, something random, just like someone clinically makes sense. Because the way you have to think about is people who mark that I see a our surgeons, they don't whilst they do look at obviously the individual structure, what is the end goal? The end goal is that they can take them away from it where they're like, OK, actually, clinical conclusion. Yeah, right. And he says, association of correlation, because you can't like audit data is retrospective. So, like, Wait, you're gonna draw associate. You gonna draw like associations without yeah, like you. You you can't tell if you know the death the 10 deaths in the hospital were because of the 10, um, surgeries or 10 durations of stay like you just know that they both happened at the same time. But you don't know that one caused the other. So, like, yes, you may do like a course like you may plot a correlation and get our number, but like that, our number like you're more trying to see if there's a trend rather than like, Oh, this causes this All right, I get basic data like audit data will not let you. It's not a study like it's not elected with a purpose of yeah, of comparing things to each other. That makes sense. Yeah, so, basically, just don't do a starts test to find an honor, but you can't do it, but don't make it the focus of your results or your discussion like you can present the are if you want, like, no one's gonna I had art like in my descriptions. I have the r and I have the significance, but it doesn't just because the our is under or bigger than 0.5 or one or whatever doesn't mean I'm gonna be like, whoa, This is like a massive correlation that I found because it it might not be Yeah, and there are They're not going to be very like none of the audit comparisons are going to give you a good are. Yeah, I would say just, uh no, no, go ahead. Yeah, The main thing for me, the most difficult thing that this I see A was finding a suitable hypothesis after that because obviously binding hypothesis involves going through the data, cleaning up, doing all this stuff. Um, it does take loads of energy and effort to do. Once you've got that set, I'll say, that's your priority to get that set. And then once you're set with that, like you're you're flying like you can just go ahead with it because the rest of it is just like basically we call it. Yeah, I completely agree that that's the problem I think everyone's having. Yeah, but really take your time with it. Yeah. Sometimes, like, is there any, like, specific tips how we can go about it Feels like there's so much information. Yeah. I mean, I can tell you what I did so basically like, they'll give you the XL, and it's like, takes you a while to understand, like, definitely use your group chat or whatever to like, decipher things, and you're welcome to ask them as well. If there's like acronyms and everything, Um, and just have a look. See, Like what is usable? What isn't like there will be values on there that you don't really want to mess with. Um, and there's data missing and stuff like that. Um, so understand what each column is first, like, actually, understand it. Don't just, like, read the title and assume that the date is good. Um, and then right up, maybe like three or four different questions that you think could work and quickly, like you can quickly do it, Like on, uh, like Graf, pad prism or something where you just, like put in the columns of data a plot them, like, have a look, uh, and just kind of work through three or four and then the one that you think looks best. Then you use That's what I did or the or, just like write down a couple questions and just think like, Can I explain this, or does this make sense? Like if it's not correlated, if it's if it the comparison is not significantly different, can I still make a discussion about it? Can I still have a few points to discuss about this? So maybe like a draft document a few questions, a few points about what you could say and then just see which one you like best. But I would I would try and draft more than one. Like, don't just go for the first one because then you might halfway. You might realize it's not really that good. Yeah, people will repeat like a lot of people do the same similar question, but obviously different graphs and, like different discussion's. Can I ask, um, basically So obviously we had I say one and we're getting feedback from I see a too soon. Are there any bits of feedback from those I c. A s? You think we should use kind of to inform our I see a three? Or do you feel like they don't really add? I don't remember. I don't think they really add. I don't think that relevant, To be honest with you, um, they're relevant learning kind of. But yeah, and the relevant Uh, yeah, it's It's Yeah, it's more for your own learning like you're developing. You're like scientific writing your appraisal, which you are going to use in all of the I. C. A s. But like the format or the structure itself doesn't really help. Um, but it's kind of like, I don't know, maybe if you didn't do so well, you're motivated to do better. That that was the one that worked for me also. Did you? Did you Were you given, like, a cataract audit for your bowel cancer? Yours is bowel cancer, Okay, but it's probably going to be an audit, whatever it is. Okay, because they Yeah, it's just, like, take data. Is it, like, different, different methods, thunder and cataracts? What we call them, you have Basically, it's separated the data into trusts, and then we've kind of got, like, cataract surgery outcomes and kind of stuff like that. And then who performed the surgery as well? Like I said, it was it someone consultant grade was someone experienced surgeon, less experienced? See, that's really interesting. Like that could give you so many questions. Yeah, So many points of discussion. Like is like, does experience or hours like, uh, of surgeries or surgeries completed or supervision, or does it really not matter? Or is it the size of the trust, the money the trust has, like you have so many ways to and some people, Was it? Did you use some outside data? Must see. I didn't know. No, I would say just avoid outside data data. Yeah, some people went out and like, make it hard because I keep it because basically, with hypotheses, I went through loads of hypotheses. And then I was looking for some, like, Crazy Link. And then I go back to the first ones I thought of that was very simple, and I was a hold on. This makes a lot of sense. That's exactly what I can. Then just go for it. Like, don't don't like, think twice about it. That's it's so true like, yeah, people love to, like, outdo themselves in imperial. But, like, honestly, I did not. I did not put in that much work and, like, I just actually try and do what they tell you to do, even though it sounds a bit abstract, but like it will work out. I think if you keep it like, keep it interesting, but keep it simple, You know what I mean? Like, some people went and tried to find, like the financing data for the trust's or like geographical data, like some people did I don't know, like distances from major trust or something like that. But it's like, Why? Why are you going out of your way? Like you only have two weeks or whatever, so don't Don't do that. Um, could you? But I think, yeah, because you'll repeat the advice on like, how we, um, look for look for novelty. Like, do we? How do we look for something that's novel? Is it like, for example, going on, um, problem med and looking at what's already out there and like seeing how we could use the data to So, uh, what I did that helped was you can. So the audit that we got like, I just went to the actual published audit because it was from 2020 and they like it, had a big document, like summarizing all the results already for you and like you can see, like if something's below usual or above average, or something like that, and that might give you an idea of Oh, I want to look into this like, oh, why is this mortality, like in these trusts a bit lower and, like, what else could I look at? That might be associated with it, if that makes sense and you can Yeah, I don't I don't want to give you specific, cause I'm not sure you'll get the same audit data. So, like, I don't want to confuse you. So So you mean, for example, like you can look at other audits and other audit analyses If you want to get an idea of how they how they study audits. But I can. But But you're saying that primarily, we should just look within the audience. C was, like, super like it was like, really, like, apparent. Like, what might be negative, like the negative, Like, you know, the mortality, for example. Like it can be. Yeah, it can be negative or or positive. Like you can also look at why something like maybe these Eastern trusts are doing really well. Like, why is that or why do some, like, why surgeries performed by consultants have, uh, smaller rate of complications and, like, I mean, that's kind of the first question everyone's going to do. But for your formative, I would pick something like that and just see how it goes. Okay, Um, so basically, just like a look within the date. Just generate questions within the data. And yeah, just just use, like, the published audits and pub Mitt as a as an adjunct driver like that. And then the main drive mean? Yeah, as as an, uh, as just, uh if you're really stuck and you don't know how to pick apart an audit, then just go online and have a look like don't actually use data from outside. Just have a look at how they've analyzed. Yeah. And you can, um I'm not sure who Harry has. Like, the best had the best grades, so you can probably ask him for his i c A. I'm happy to share mine as well. Uh, it was a 68 so I'm not sure that's no. That would be amazing. I think it's really good. I would love to see it. Yeah, I'll send just message me. I'll send you. OK. Okay. Thanks. Um uh, actually, Katherine, if you could just, like, leave your email and then like, yeah, just email you asking, and then you could just send one big one. Um, I wanted to ask because I think I'm quite I'm quite scared of starts. Um, do you have advice when it comes to the actual starts Bit, uh, for me, I agree with you. Same thing. I'm I hate stats. I'm not a big stat with the new codes and stuff so similar to the to the theme of simplicity is key. They recommended graph, pad, prism A lot. I think a lot of people use that. And it was the most generally the most simple. It was. It was like it was like a bad version of excel, but very usable. You get really nice graphs. It's really easy. Yeah, everything's like there for you. You just It takes, like, maybe like, an hour to play through. Just understand what's going on. But, like, it's like a lot better than XL. A lot better than S P s s. Yeah, as an if you if you know how to work with more complex things that that will give you, like, even nicer graphs you're welcome to, but graph pad like just download it like there's tons of, um, tutorials online. But for this, you don't even need, like, a regression or like you don't need to go that far. It's literally like basic basic stuff like you just check the distribution, and then you know that little table of, Like, what? What test to use. And you can just do that. Non Parametric. Yeah, literally. I didn't go. I I think I had a correlation chart, and then I had a bar chart comparison. I had Those two was fine. So people have a plot, a funnel plot as well. Oh, really? Just graph pad makes your funnel plots. Okay. I was wondering, Are they Are they really, like, particular about the way graphs look? Yes. Okay. Do you have any advice for that bit? Um, I think that Eduardo will tell you exactly how he wants. Like the figure to look, I I don't know. I don't remember how, but like, definitely triple check your table title's descriptions and just make it like, clean. I don't know if that makes sense, but like, you know, 12 colors, like big size, you know, stuff like that. Yeah. It's like you should be able to look at it and and, like, not have to read the description. Oh, brilliant. OK, right. But I think they go through. They definitely go through figure like Eduardo loves to talk. So, like, Yeah, yeah, I remember when we went to the I c a to video session. I remember, like realizing how much there was, like, some sort of creativity interpretation of the video that he had in his mind. Oh, yeah, Yeah. You told me stories like, Oh, someone once did this. And this Did you tell me about our video? Our video got, like, remember, we got really? We got really lots of praise for a video, but I did. I think I did better in the formative than the summit of, But that's fine. Um, yeah, he does. I don't think he'll be grading the ice. Oh, is he grading there, say three. But he does love his little storytime and go along with it like you'll get into it. I guess it's very subjective, Like whoever is grading you is just gonna do whatever they feel like doing. So, um, what's the difference between the lodge and and the small And when? When we, um right, the legend, Uh, the population again. I don't know what the big one was. A large, large and and small in. Yeah, I remember. It was briefly covered in the video source. Uh, I'll just go back home. Yeah, So the big n equals the size of the population, and the small line equals the sample size. Uh, so again, it's the total population. Everything included small and is whatever you're investigating, uh, or male female? Yeah, because also, like, for example, for r i c a like like they gave us the raw data, but some like, and you get the audit and it's like, you know, some trusts only reported 50% of the data some trusts reported not like 99. Adherence as in like, they reported 99% of the things they had to. So you could choose like it was up to you whether you wanted to use everything, whether maybe you wanted to use trusts above this threshold. So it's like no one's gonna have the same end like it's It's what you define it as your population like, you don't have to use all of the data, not your data. Your clinic of the data is basically well, excluding the kind of subsets of data I don't apply to you or don't kind of fit your data. Yeah, not for your data. That sounds like you're skewing it. But basically like that you can't physically use your analysis that makes sense. And that comes into your methodology. You begin to break it down. Obviously. Why did they exclude these trust and stuff? Yeah, for example of cataract surgery and some trusts, like reported 500 like reported results for 500 surgeries and then one trust reported results for two surgeries. Then you might want to exclude that cause it might be an outlier, because it's it's not going to have as much impact like things like that you can think about. I think a good piece of advice I got for write ups was trying to write your right up, as if you want the people who read it to be able to do the same thing you've done with the data so they'll be able to follow. Exactly. It's like, kind of like a recipe, because then it was a lot clearer and, like, easy to follow. It's just a general rule of thumb to go by. Even first I see a and for your project right up and stuff. Thanks. All right, guys. Uh, more questions. Yeah. Question about, uh, you mentioned earlier where you said something like, um, you went on, um, published data, and you found, um, in your data said you found, like, hospitals, which you're above average and below average and things like that. And I was wondering, where do you find that information? Mhm. I don't know. Because I didn't do it. I'm just with most what? In terms of sorry. Can you repeat again? Yeah. So I remember. Katrina said, um, but she, um, found for the date you said that you guys were given? Yeah, she went on. Um, um, like an analysis website or something. And she found hospitals. Which, uh, yeah, uh, average or something like that. Maybe. Uh, yeah. As in, like, I just I just found the actual published audit. Yeah. So the one they give you is probably not going to be like the current ongoing data. It's going to be reported from last year or however, many years ago, so you can go online and find the report of that data. If there is one which there should be like a specific website. They used to just google the name of the of the audit Yeah, yeah. Oh, my God. Portugal just scored. And I missed it. Uh, that's fine. Um, any other questions, guys? You can ask anything, by the way. Like if you have a specific question drug, break it down, I have another question. So say after you find your hypothesis and you did your statistical analysis or something and stuff. How do you then go about creating an argument for your, um, whatever purposes that you presented is Do we, like, just go through multiple other published data? Essentially, the, uh, at the very start. You should before you do the hypothesis, you should kind of before you, like, start to analyze the prosthesis. You should kind of find out. Why. Do you want to look at this hypothesis, Uh, at the start, rather than at the end of all your work. That makes sense. Yeah. Try and establish. I can aim your study like you said, Like a clinical aim, for example, um, I did think of Senator Katrina. Man was like a number of cases per trust versus, like, uh, outcomes, for example, mortality, unplanned return to theater. And then I kind of be down saying, uh, for us to have more cases. Might have, like, better, more, uh, mortality rate. So that makes sense. Therefore, move together all the trust. You know what I mean? Yeah, you You can have a look. Yeah, you can. I mean, you have. I think there's a limit for citations, but you have, like, 10 or something. Um, and you also have to write a lace summary, don't you? Yeah. Yeah, that was awful. But it's it's easy enough. Just just make it, like, make someone nonmedical, read it and then just, um But I I did quote a lot of, like, studies that had we're looking at the same analysis. So lots of, like, uh, surgeon volume and outcome analyses, which I mean, you'll find some stuff done before. So whatever your question is, google it before and see if you have, uh, some based, like some evidence to back up your discussion and all these analyses, like our CTS or they will be They will usually be like, in a specific center, or like in a country or something, and they can be from audit. But it can also be like a retrospective study. Yeah, there'll be. There'll be some good. Some good stuff. Um, sorry. I wanted to ask the question. So basically, when you you formulate your hypothesis from the audit, do you need to be able to use the audit data to, like, show how you prove your hypothesis, or is it like, out of, uh, project of I see a scope? What do you mean? Like suppose you you fallen happens. It happens is like, uh, like you said. So because as an increase in, like cases and how mortality changes with it, you'll be able to use, like, your audit data to say how you would, like, go through a method and everything to prove your hypothesis or yeah, uh, I wouldn't say Yeah. I mean, prove or disprove, right? Cause you're gonna you're gonna you're gonna make up what you think is the relationship. But you do have to you have to do an analysis of the data to either prove or disprove. Yes. Okay. So you use everything from the or did it serve? Not from like, you know, it has to be from the audit. You can't. You can't go and use other data. Okay. Okay. All right. Thank you. No worries. Okay, guys, I've got to go. But feel free to email me. Uh, if you have any questions, um, and yeah, I think that's it. Thank you. No worries. Thank you. Thank you. Uh, if you have any more questions, you can just let me know or if you, uh oh. I had, like, one quick question. Yeah. So, um, Eduardo said that, um, we can look at, um the different qualities of each trust that involved in this audit. Um, I try to do that, like, doesn't find out how they performed the surgery, All that stuff, but literally. Like, the website doesn't have much information. Yeah, this is some advice that you got last year as well. And how do you go about finding a job? There was a few things similar to that farm. Exactly. Worth. Yeah. That is like, that is a good way to start, to be fair, like, for example, is, uh, similar to what you're saying. Like there's one trustees, a certain method. Cataract, uh, is another method that would be nice to distinguish in terms of finding that information. I'm not too sure. Uh, I guess new goal is your obviously like he is there for you to, uh you can ask Eduardo himself. You have a question. That's, uh we do that. I'm sure you can do that. Thank you. So it's a nice nice thing to do that. That sounds like quite a good thing to look at. Uh, thank you very much, Mr For, um, I think where the last breakout room in attendance. You can get on with your evening. Um, relax. Thank you so much for coming. And your advice has been incredibly useful. Uh um, all right. I think we'll end it there. Um, please, can you guys fill in the feedback if you haven't already? Um, and thank you so much, Mr. For you're free to go. Thank you. The talks ended. You guys can also go and enjoy your evening, and I will upload the recordings of the, uh, basically, everything be recorded. I'll upload it all tonight. All right.