 Well, hello everybody and welcome to today's presentation in my series, Portfolio Projects with Pizzazz. And this presentation today is going to be on FAIRS data and how you can use it for your portfolio projects. I'm so glad to see all of you here in the chat. I'm Monica Wachee and I'm an epidemiologist and a biostatistician and a data scientist. And one of the things that's going to happen today is we're going to talk about not only the FAIRS dashboard from the point of view of an epidemiologist or biostatistician like me, you know, what if you want to do a portfolio project? But we're also going to talk about it from the point of view of this is this dashboard is an application, right? Like it's an app. And I want you to be thinking about how does this app get there? Like the government made it. The FAIRS dashboard is for the adverse event reporting, which I'll get into the post surveillance adverse event reporting for medications and is the US federal database for that. But I've worked with customers who were dealing with the Saudi version of that. So when I had, because I have some customers from Saudi, so I started learning about the Saudi post surveillance reporting, adverse event database, which, which is similar. I mean, you know, how different can they be, right? But in working with that customer, I started learning about the global surveillance data databases for these adverse events. And so I'm going to be talking about exactly what FAIRS is, like what's the data in there from a specific point of view of what's in FAIRS versus what's in these other ones, if you go to other countries, like, you know, in general, then we're going to go over the dashboard, the FAIRS dashboard, I'm going to show you how to query it. I'm going to talk to you about what are the issues with adverse event data. I mean, sort of the big picture I want to tell you in addition to we have to think about applications, like this application I'm going to demonstrate. But it's to also think about how can you use these data or portfolio projects, even if the data are not perfect, they're not like perfect epidemiologic data, they're like, you know, passive surveillance data. And you'll see what I mean when I actually query it, because you'll be like, Oh, look at these results. And then you'll be like, Oh, I don't know exactly about these results, you'll see. All right, I'm looking over in the chat to see if you have any questions. Okay, so first of all, on the right side of the slide, you'll see some links. And I put, well, let me do it again here. Let me just make sure that you've got the link to the slide in the chat here, just so you can get these because it's just easier if you download the slides to get these links. Now, the only one I'm probably going to show you today is this top one, which is this blog post. And what the blog post does is it goes through like an example of what I'm going to demonstrate now, but it's like a really super clear example. And I planned it all and whatever. So if you sort of forget exactly what I do, go to that blog post and that that'll help you walk through what I did. Okay. So what is adverse event data? So adverse event AE. Okay, after drug companies have a drug approved in a country, and it's available and on the market, they are ethically bound to collect data about adverse events or adverse reactions. So those of you involved with clinical trials or or pharmacovigilance or all that. I mean, you know the truth, there's only so many people who can participate in your clinical trials, like maybe hundreds before, or if you're lucky thousands, you know, in different studies before you get FDA approval in any country, the US or Saudi Arabia or anywhere, right? Okay, so that that's not millions of people. And sometimes there are adverse events that are going to happen at a very low rate, like if you have a certain genetic situation or an unusual comorbidity or whatever. And you're just not going to see it until a million people get the drug, you're never going to find it if you don't have this surveillance system. Okay, so it's really important that every country set this up. So I'm proud of Saudi Arabia. They did it. We did it. This one is actually a pretty good one I'm going to show you. And so, but then you have to sort of think about it, like, let's say that I take a drug, you know, like maybe I'm on a hypertension drug and I take this drug, or I'll even tell you a real example. Unfortunately, my mom has breast cancer and so she's on a lot of cancer drugs. Well, she had this problem like last summer where she's like, Monica, I feel really sick. Well, she attributed it to her cancer drugs because they had been changed. But what was really happening is she stopped taking her iron pills and she was anemic. Okay. So it had nothing to do with the cancer drugs. So that's one of the things that hit me when I was querying this database is like, how do they know this effect is associated with like the drug? Like, how do they know? So the good news about the fares database, this public dashboard, is that it's very easy to use and and actually people from the community contribute this stuff. People like my mom, she didn't contribute to this. But if she really thought she was having an adverse event from her cancer drug, she would have said, yeah, it makes me feel lethargic, which was actually her anemia. But the bad news is that the data are hard to use because there's not really a formal process. I mean, there's a semi formal process. But like if my mom had really thought this was an adverse event and reported it, then it would just be wrong, right? Like nobody who makes sure it was right, you know, I shouldn't pick up on my mom. She's got enough going on. All right. So we're going to go to the database. But before we go, I just want to show you this is what it looks like when you get there. Okay. And I'll show you this again. But one of the things you'll notice is that there's not like a menu up here. Right? Like, aren't you kind of like expecting a menu here? So because it doesn't really have, and it doesn't do like, let me walk you through a tutorial and have these tutorial things. So I actually, when I went to, I wanted to get to know this dashboard because I'm working with someone who I thought, hey, maybe this would make good data for their portfolio project, right? And I said, wow, before I recommend it, I better like get into it and see how it works. So I went over there and I was like, oh, this looks good. But what do you do? And so the reason why I'm even persevering on that is because think about it. If you design a dashboard, which is an application, people are going to just show up and they need to know what to do. And you'll know this from putting apps on your smartphone. There's some apps that you download it. Every time you open them, I know Venmo is like that for me. Every time I use Venmo, which I don't use it very often, I open it up and I just do not know what to do. I don't know where the menu is or anything. Well, maybe they changed it since I last looked at it. But that's one of the challenge with dashboards is when the user gets there, they should kind of know what to do. And I kind of didn't know what to do. So that's why I'm holding this discussion. So I'm going to demonstrate it and I'm going to demonstrate the functionality. There are probably other functionalities I didn't figure out, but it really wasn't obvious. Also, I just want to step back and say there are other ways of getting this data. You don't have to go through the dashboard. This is public data. It's not really private. It just says how many counts of a certain adverse event you had for a certain drug, like not you, but people by year or whatever. So it's not really personal data. I mean, what they're serving up here. And so I didn't really get into it, but I found something on GitHub about a way you could fetch the data. But anyway, it's fun to use the dashboard to begin with because then you can start narrowing your decision space about what you're going to actually download. So without further ado, let's go to the dashboard. So here it is. Okay. So just to be clear, this is the blog post, the first one here, where I walk through each of the steps. And I went there to go and get to the dashboard, which is here. And then you click on this and you accept it. You say, I didn't read it. It was really long, but I wasn't going to do anything bad with it. So then I got here. Okay. So let's just look at this for a second. Just a second. So remember this is the US FDA. So if you're thinking, well, I care about a different country, in this one, we're caring about the US. Okay. So people had to report it in the US. They could be from anywhere, but they reported it in the US. So I kind of was like, wow, there's 27 million reports in this whole database. Oh, and by the way, you're going to keep seeing this thing show up. And oh, and so, and see this Q lick sense. So you're already seeing that this dashboard's just a little buggy, right? Like, oh, here, it's refreshing. And this is my, I have to re-accept this. Okay. This thing's a little buggy. Okay. So you'll see that this thing shows up once in a while. Okay. And you'll also see like weird buggy things like that. So there's 27 million here, serious reports. Okay. And all of the definitions, like what is serious? And like, well, we probably know what that is, right? But serious reports, you have to read the definitions of like, how they classified things as serious, or was it the person who was writing it down, you know, submitting it? And then how did you tell them to classify it? Like, that's all you're going to have to look at. And then also here, the, and again, this dashboard, this is just what I see when I show up. So I'm sort of looking at, well, what is all in the dashboard? So we see these reports here. And it says total reports, which is 27 million, expedited, non expedited and direct. And I admit you, I do not know what direct is. What I believe expedited and non expedited are, is when there's like a new drug on the market, and you start seeing a whole bunch of adverse events you didn't see in your studies, and they're the same ones over and over. I think that they do expedited reporting or something. And then direct, what I believe direct means is it's from healthcare, like the healthcare person is doing it, but I'm not sure you'd have to look that up. And so I tried to look this stuff up, but it is really serious, that you have to get in the weeds. And then the question is, do you really need to know, right? So what I usually do, since I'm talking about using this portfolio project, you would need to step back and know something about a certain drug or a certain, you know, topic before you approach this dashboard. I'm just sort of demonstrating how to use it and how to think about getting data out of it that you can use for a portfolio project to like maybe make some inferences, maybe just a descriptive analysis, but make some inferences. Why not? It's real data, you know? And speaking of real data, I was like, okay, well, when did it start? And I went all the way down here. Like a long time ago, like in the 60s. So the FDA was basically invented in the 60s. And so that's why it goes way back here. And so you might be like, well, you know, do I need to filter it? Well, you'll see on my blog post, you'll filter it like for recent drugs. Well, you see on my blog post, the example drug I use, I did a search. And the example drug I use was Ozymbic, okay? So Ozymbic's a new drug. There's not going to be reports in there from a long time ago. So sometimes you don't even need to really worry about the that kind of stuff. What you really have to worry about is what is your hypothesis, okay? So of course, I'm looking at number of reports, but you're probably like, well, what about the adverse events? What about this death and everything? And so I was like, okay, that's what I want to know about. And I really didn't know where to click. And then I realized, well, maybe I should just click on search, right? And that's actually what I would suggest to you is to click on search when you start, okay? I mean, if you want to look at that stuff, I looked at that's fine. So then you search by a product. And I was trying different ones. One of the products I tried was Lysinoprol, which I know is a blood pressure drug. Now, you see, it's like a smart thing. And it's coming up here. And here, you immediately see the problem, right? Like, you can only add five of these. And so let's say I add Lysinoprol, and then Lysinoprol, and this is, these are just two blood pressure drugs. And then, oh, here's amlodipine with Lysinoprol, or Lysartin with Lysinoprol, which is two. Oh, and this, so this is chlorothiazide. Well, which one? Oh, here, Lysinoprol and hydrochlorothiazide. And this is hydrochlorothiazide. I don't know. I don't know. See, that's what I'm saying is like, you're probably going to have to do a lot of post-processing on this. And then my limit has been reached. Okay, so let's go. And I actually don't know a lot about Lysinoprol. I mean, I used to in the olden days. And this data, these data are updated. Okay. So now I want to make sure you can understand what we're seeing here. Okay. What we're seeing here is different from what we were just looking at, even though it looks really similar. Okay, but let me go through it. Remember before there were no nothing here? This was like blank. Now there's stuff here. Okay. There's demographics, which is what we're on, actually. And then reaction group, reaction and listing of cases. Okay. And that's there because we put this filters on. So you really want to start with like the search. Okay. Now Lysinoprol is like kind of old. You know, it's not like Ozympic. So you can go back here to 88. That's when probably it was approved. And you've got all of these. So maybe you don't care about those old ones. Maybe you do. It's up to you. But I'm just going to be simple about it. And then this is kind of cool. You have these cases by these years. And I, you know, here's a total case and whatever. And so of course, you see how when I go over here, you keep seeing these things. And I go over here and I keep seeing these things, right? I was like, well, what is that? Well, what I realize is over here, if you click on this, you get these choices. And obviously I was always into downloading data. So I go download as and here you can have image PDF or data. So data. And then I don't know, I will I'll say no, I don't need table formatting. Okay. So now I'm clicking and viewing my data file. It's an XLSX. I'm just I was playing with it before I started here. Okay. So it comes up here. And I can open it. So it came up on the wrong thing. Okay. So see, I just want to point out, see how it, it calls a category, but see how it goes all the way down to 1988, like we can see all of these cases. That is awesome, right? Here's the problem. When I went over here, see this histogram or whatever. Like I get a bar chart, I guess it's a bar chart here. When I tried to download that as like a PDF, and even when I put fit to page, it like wouldn't give me the whole image. You know what I mean? Like, let's see here, click here to download your PDF. Let's open it. See how it cut off at 2002. So I'm not sure how to get that, but you probably don't even want that to be honest with you. Because if you can get this, you can throw it into R or throw it into SAS or Python. But you're probably noticing like, do I really care about them by year? And the answer is probably not. Probably what you care about is like reaction or reaction group. Okay. So let's go to reaction here. Okay. So this is probably what you care about only it's like separating it by age group. Right? Maybe let's just go to reaction. Yeah. So, I mean, obviously you might care about these cross tabs. But what you'll see if you go to that blog post, what I cared about, I was just, you know, it was so Zempik. So it's like newer drugs. So I was just curious what was coming out, right? And what you'll find is actually, let me just download this just to show you. I didn't really look at it before, but I'm sure we're going to see. So remember, this is Lysinopril, which is a blood pressure drug along with other like or mixtures of Lysinopril. Okay. So this is the total case. So this is the reaction group, right? So this is the it's sorted by reaction. So the total case is so the most common reactions is general disorders and administration site conditions. What is that? Like, what does that even mean? And this is skin and subcutaneous tissue disorders like what? And I mean, I kind of get gastrointestinal disorders. That's usually up there with everything. This respiratory disorders, that's kind of interesting, nervous system disorders, injury poisoning and procedural complications. You know, as you go down here, you start to realize, well, you know, maybe this isn't very useful, like surgical and medical procedures, like that's an adverse event. So it's almost like you end up with more questions and answers when you do that. And when you go to my blog post, you'll see how I was like, okay, one of the things I saw in the Osembic data that I noticed when I downloaded that was that there were a whole lot of reactions like this, like there were a whole lot of reactions that had to do with like off label use, like using it like administration issues. But then it also had these weird categories like sexual abstinence and X alcohol user, like what does that even mean? You know, that's an adverse event. And so what I ended up doing was making classifications. And if you're interested in how to make classifications, make sure you sign up for and come to my event next week at this time, because I'll explain to you, like how I make crosswalks, because that's how I do it. Because what I show you here is that you can make sense of the data. The problem is, the problem is that you have to go through and classify all of them. Okay. And the one I just downloaded from Lucinapril had maybe a few hundred, but there were like over, I think over a thousand of rows of, for the Osembic one. And I had to go through and make a crosswalk. Like I said, if you come next week, I'll show you how. But I was kind of lazy. I said, well, what if I only classify like the top ones, right? So I classified the ones that I could as either use errors, you know, the administration errors or other errors and then unclassified. I went down, I did like, I went down to like where there was less, 40 or less in each row. And I said, oh, these are low. Who cares? Well, unfortunately, that's what, that's 20, almost 25% of the data set of the reports. So I can't get away with not classifying them. So the fairs public dashboard is a great place to start with adverse event research questions. So you can use it as a dashboard for, and as a starting point for deeper analysis, but you do have to know something from the literature about it. But once you figure out what you're doing with it and you can filter your data and get what you need, and you do your crosswalk or whatever, you can do a simple descriptive analysis and come up with something because what you can see from this here is if I were to go through and classify all these, some of this would go into the green and some of this would go into this blue. And that means like at least 10% of the Osembic adverse event reports have to deal with a use error. Like, I don't think anybody knows that. Like actually just making that blog post, I was like, wow, that's amazing. And so, so I've been talking throughout this presentation about applications, specifically this dashboard application, and how actually kind of cool it is that it's out there, the government made it so we could download data, but it's actually kind of difficult to use. And often people like us, you know, data scientists, bioscientists were expected to be able to just figure out these dashboards and get data out of them. They're not like data from cohort studies or cross-sectional studies or like surveillance like BRFSS, which are prospectively gathered and really well documented. Like I said, when I was downloading, like I don't even, I didn't even know how to navigate this dashboard, I'm just guessing and telling you. And so what if you're expected to analyze data from an application? Maybe not this one, but any other one, like maybe a smartphone app that collects health data. Or if you work in academia, like on a research team, and the PI comes to you, principal investigator, and said, oh, we got data from Medicaid or something, or, you know, how are you supposed to know what to do to analyze it? And so I'm holding a workshop, and I call it application basics. And the theme this time is adding analytics to pipelines. If you think about it SAS Viya, and that dashboard, and any analytics platform like Twitter analytics, it's an application. So how do you deal with it? The learning objective of the workshop is to understand data sets from applications well enough to analyze them and produce results. And so we're going to talk about computer applications, the different design approaches, the team structures that they use when they create these applications, and how the data are stored in the applications, like even the dashboard I was just demonstrating. I'll teach you terminology used in application development. And with this knowledge, you'll be able to break through communication barriers to get the answers you need to complete your analysis and be seen as an expert, even if you're given data from a dashboard and not from an epidemiologic research study. So here are the details about the workshop. It's, as I said, application basics, adding analytics to pipelines. And it's next Saturday and Sunday. And each session starts at noon. And it runs about three hours, noon Eastern time. And it's very interactive. You'll get to meet each other and talk. And normally, I price this out, normally price workshops like this where you're on Zoom and you actually interact with people and you get all this information, they're priced at about like $250 to $750 per workshop, especially if they're more than one day. But you're lucky because your cost is free. So if you want to come to this workshop and get the free workshop this weekend and learn more about analyzing data from applications, just register here. I'll put it in the chat. You'll be able to go there and sign up. And then I'll send you an email. And then we can spend more time talking about application basics. Well, I wanted to thank you for coming today. Before I end, let me ask, does anybody have any questions? You can put them in the chat. I'll just wait for a second here in case anybody has any questions. I'll try to answer them. I'm not a super expert on adverse events, surveillance, but obviously, if I've done projects on them, I've looked into these systems. They're just like any other technology system in the sense that as technology improves, these dashboards improve. I see them redoing them. So sometimes my knowledge is a little old. But you can also look in the peer-reviewed literature. They often write like methodologic papers about these. Thank you very much for coming today. I hope you have a good week and make sure to follow our company page, DPS, because then you can find out whenever I'm holding an event like this and come in case it's on a topic you want to hear about. All right. Well, thanks for showing up today, and I hope you have a good Tuesday, and I hope you have a good rest of the week. Thank you for watching this video, which is part of the Public Health Today to Science rebrand program. If you are interested in joining the program, please sign up for a 30-minute Zoom interview using the link in the description.