IQVIA Holdings Inc. (IQV) Earnings Call Transcript & Summary

Hello, and thank you for joining us today for this webinar, Identifying MedTech's Hidden Complaints and Adverse Events: The Power of Automated Social Media Monitoring, brought to you by Citeline and sponsored by IQVIA. My name is Maire Gerrard, Managing Editor, Custom Content at Citeline, and I will be your host today. We will be running polls during today's webinar to find out your thoughts on the content. For these, you will see a little pop up on the media screen, and we would love to hear from you. Before we go up into the presentation, let's go over some quick housekeeping items. First, if at any time you're having technical difficulties, click on test connection button, and this will take you to a connection checker and FAQs page. If you are experiencing any technical issues, there is a chat now function in order to go through to the support team who can assist you. If you want to enlarge the presentation, just click on the maximize icon or drive the lower right-hand corner of the presentation window to enlarge. [Operator Instructions] Because we know you want to listen to this webinar again, today's session is being recorded and will be available to view on demand in around 48 hours. I would now like to introduce our presenters for today's webinar. First, we have Marie Flanagan, Director of Offering Management at IQVIA's Vigilance Detect at IQVIA. Marie has 17 years of experience in PV operations, strategy and consulting. In her tenure, she has held various leadership positions in PV operational management, the integration of safety technology and services and the strategic expansion efforts of IQVIA's pharmacovigilance department. In her current role, she supports the evolution of safety risk identification technology in response to evolving industry needs. Next, we have Donna Smith, Product Manager for SmartSolve Quality Solutions at IQVIA. Donna has over 30 years of experience in enterprise software and is product manager supporting IQVIA's Quality Solutions, responsible for best practice solutions for complaint management, supplier management and risk management. Finally, we have Sanmugam Aravinthan, Senior Director, Development, Vigilance Detect at IQVIA. Aravinth's main area of focus is on driving the technology department and delivery of a productized solution that enables an optimized approach in detecting adverse events, product quality complaints and other safety risks in large-scale structured and unstructured data. He has over 20 years of industry experience in driving software engineering and systems development with the past 10 years in pharma and life sciences, specifically. And with that, I will hand things over. Marie, the floor is yours.

Thank you, Maire, and thank you all for joining us today. We hope to fit quite a bit in to this webinar today, starting with a few polling questions to get a sense of where you are when it comes to monitoring social media. We very much welcome your feedback. We will collate the results of these 2 polling questions following the final polling question at the end of the presentation and take a look at what you have to say. We'll hear from Donna about some of the challenges associated with complaints handling and the benefits that can be brought to bear with a complaint management system. And you'll hear from me on the process of and benefits of connecting safety and quality technology to all from a social media monitoring, for hidden complaints or adverse events. And last but not least, you'll get to see a demo from Aravinth and gets to see an event on social media automatically populating in case intake. But firstly, let's get straight into some polling questions. So first off, does your organization have dedicated social media channels that people may document complaints or adverse events. Your options are yes, no, or don't know. So if you take a moment to respond, we will appreciate that. And our second question: how does your team currently monitor social media for product complaints or adverse events? Manually; an automated solution; they don't, as in that's something that is currently not performed; or unknown. Great. So let's move along to the evolving environment. I think it's important that we preface our discussion today with a reflection on our industry and take a look at its trajectory and its evolution. We live in unprecedented times. We find ourselves just now emerging from a pandemic, and that's just one of the many reasons we're seeing an increase in the publication of regulations and guidances. And in fact, there has been a 150% increase in regulatory mandates in the last 5 years, with 2,000 new regulations per year and a new or changed regulation occurring every 22 minutes. So gathering relevant information, as you know, is key to developing and maintaining a regulatory strategy. Yet currently, regulatory personnel spend a great deal of time on repetitive tasks like, chasing down changes, which can distract from the more valuable, strategic and proactive work needed. Regulatory staff need to have information that's easily accessible and up to date in a knowledge base enabled by technology. Breaking down information silos that create [indiscernible] in the organization will be critical in our current regulatory environment. And to improve productivity and compliance, information must be accessible, up to date, accurate and shared across the business. And to further compound regulatory challenges organizations are faced with, the safety landscape is also rapidly evolving. And what we've seen and continue to see is the democratization of safety reporting through technology. One example of this is the COVID vaccine distribution, as I'm sure, you'll all well remember, dramatically raised society's awareness to safety reporting. 64% of COVID vaccine events continue to be reported by non-healthcare professionals. That is direct from consumer, which represents a seismic shift in dynamics from traditional safety reporting processes. So the concept of reporting adverse reactions or dissatisfaction with products has become mainstream. And regulators are championing the cause, making that direct to regulator reporting of side effects accessible. And are in toto calling for real-world evidence to face significant role in regulatory decision-making. We also see patients' and the health professional centricity is already demonstrating shifts in the sources and formats of where safety data now exists, notably, on social media with 50 million on average conversations on social platforms a week around the COVID vaccines. We see a rise in audio transactions, emergence of digital assets such as virtual AI agents and wearables. And again, the people doing the reporting are all the consumers. Therefore, these new customer-centric mechanism of reporting needs to be catered for and need to be monitored. And organizations find themselves at a pivotal point of understanding how to harness the growing social data for commercial and clinical insights and for long-term benefit of patient safety. And this, of course, extends to complaints and how organizations can manage complaints in this evolving safety and regulatory landscape. And with that, we'll hear from Donna.

Thank you, Marie. I'm going to begin with a discussion on the current challenges in the quality management and the benefits of SmartSolve's Complaint Management module in addressing those challenges. Let's start with defining why we have a quality management system. Its purpose is to establish and maintain procedures for receiving, reviewing and evaluating complaints by a formally designated unit. Any complaint involving the possible failure of a device, labeling or packaging to meet any of its specifications shall be reviewed, evaluated and investigated. And that's where SmartSolve's Complaint Management comes in. But before we discuss the benefits of Complaint Management, let's take a look at the primary challenges faced by customers. First, inquiries and incidents come from many different unrelated sources, which can cause inconsistency in data and inefficiencies in processing that data. Timely awareness of incidents, which is a key to meeting regulatory reporting requirements and the inconsistency in identification and prioritization of complaints. In addition, as there are many regulatory bodies, all with their own regulations, it's difficult to understand the regulations and keep up with the changes to those regulations. And lastly, it is difficult to assimilate meaningful data, which is pertinent for making improvements to your product and processes. Now we will move on to talk about how Complaint Management addresses these challenges. At a high level, Complaint Management allows the user to streamline the handling of all complaints and regulatory reporting. As mentioned, one of the challenges for users is that data is scattered, often in different unrelated systems. Complaint Management provides the means to gather data more efficiently and effectively by use of single forms to record your incidents quickly and consistently via data entry in the case intake portal, which enables extended demand chain, allowing employees to report incidents in a secure, consistent manner and via secure open integration connections, which allow for intake for multiple sources, PV detect being one of them. Complaint Management also provides a reliable means of handling your complaint process. Recording of incident's awareness date along with automatic notifications, ensures timely regulatory submissions. The Complaint Management workflows are driven by user-defined policies, which are used to assign cases and complaints to the appropriate individuals. These policies can be defined for all records or can be based on specific attributes such as product type and reporting site. Checklists are used to tie procedures and complaint determination criteria to the intake process. And use of the built-in impact assessment can reduce the level of product risk. Another way in which Complaint Management streamlines your case and complaint processing is by simplifying global adverse event reporting by using bidirectional integration with FDA's Electronic Submissions Gateway with the generation of XML files for reporting to the EU and various other regulatory bodies, and by utilizing finished product information and decision trees to ensure a thorough and accurate regulatory reporting. Complaint Management is also a tool for supporting continuous quality improvement by identifying early signs of subpar supply chain components and with the use of complaint data to provide critical feedback into the design, manufacturing, supplier and risk management processes. Now let's take a look at the complete related workflows to see how SmartSolve accommodates the streamlining of processes. Here, you will see the various workflows that make up Complaint Management, beginning with the definition of your finished products. Within each finished product, you'll identify each country in which the product is sold and the regulatory bodies to which incidents are reported. The next step in the process is the creation of the case. As you now know, in addition to cases being created manually within SmartSolve, this process can be automated with data coming from the quality portal from external CRM and field services systems and now internally from PV detect. All automated processes is first to create a case. From there, the case may be reviewed and that is evaluated to determine if the case should become a complaint. This determination is what triggers the complaint workflow. This robust workflow allows the user to perform many tasks, including review of product history, investigation, risk assessment and regulatory evaluation. During regulatory evaluation, the decision trees are run and determination is made if regulatory reports are required and to which regulatory bodies reports need to be submitted. If regulatory reports are required, the regulatory workflow is triggered with the creation of the appropriate regulatory records. Data from finished products, case and complaint are met to the regulatory records. Then updates can be made and additional information added prior to the submission of your reports. In addition to generating XML and PDFs, which can be submitted to the regulatory bodies, Complaint Management supports electronic submission to the FDA Gateway. This includes not only submission of initial and follow-up reports but also receipt of acknowledgments set back from the FDA. Now that you have a better understanding of the benefits of Complaint Management, I will turn it back to Marie to discuss connecting social media and eQMS technology.

Great. Thanks, Donna. All right. So what we hope to demonstrate shortly is the integrated workflow that exists between 2 mature established technologies, being SmartSolve eQMS and IQVIA Vigilance Detect. And you will see how Vigilance Detect automates the proactive identification of potential safety data, a task that would otherwise be manual. And how SmartSolve's complaint module can automatically ingest that data for a review and validation. What you won't see in the demo, but I hope that the point comes across is that behind the scenes, you have a GxP-compliant fully validated solution and that precious time down to all the trail that will demonstrate that you proactively took the time to review social data and, importantly, took action when you found something. But before we get there, what is Vigilant Detect? And I'm aware that many of you today may not be familiar with this technology. It has existed in service of the pharma industry for the last 12 years. In essence, it is a technology-enabled service that identifies potential safety risks. That may include safety adverse events or product complaints. And it works through a combination of automation technology, artificial intelligence, namely Natural Language Processing and IQVIA's IP, which is their safety and product specific terms and patterns, or something you may hear us refer to as ontologies. And this solution is importantly underpinned by human expertise. Plus, I think Aravinth will bring it to life a lot better than I will. So with that, I'll hand over to Aravinth to show us the demo of everything in action.

Thank you, Marie. Hello, everyone. I'm going to share my screen for the demo. Okay. For today's demo, I have chosen to show a particular scenario that may happen in Twitter platform. For that, I have come up with a fictional brand name Gelidium, and I'm going to assume that they have established a Twitter page for their brand, right, which is you're seeing now. And also, I have another test user that I created for this demo, who will be coming and interacting with this brand on their Twitter page. So as you know, once a brand is in Twitter, any Twitter user can come and comment or reply to their tweets. It's normal. There will be tons of content that will be coming from that platform. There's a chance. But all we are interested is if a user is a patient that who is using this brand and he's coming with frustration and typing out some sort of a quality complaint about this brand or some events that he experienced while he was doing this brand is also a scenario that may happen. So our platform -- our integrated solution, with Vigilance Detect and SmartSolve complaint module, so we will be only looking for such event-related conversations that may happen in the Twitter platform. We are not worried about other tweets that are coming. So we are specifically tuned to listen into this type of complaints or events that Twitter user can come and discuss in a brand base. So what I'm showing now is -- as you look, so this Gelidium brand have put up a tweet. So as a user, who could be a patient too, can come in and put up a tweet like that. So he's trying to say, assuming that this brand is a device and it's contaminated device, the surgeon suggested it to remove. And I was bleeding internally and was in pain. So this is something that the brand would be obliged to be looking at, so that is where we come in. So let's go ahead and press the reply so that the user already registered this tweet or the reply for the tweet. So what will happen now is like detect, the system will be continuously monitoring this particular brand page, in order to look for tweets that are risky, potentially carrying events or complaints. So now I'm going to move to Vigilance Detect screen. What you're seeing now is the Vigilance Detect platform. So this is how it looked like. So let me go to this particular tab. So if you look at here, this particular record, let me open that. So what has happened is whenever that particular user was putting up a reply tweet, so Detect sensed it and also like it brought the tweet into the platform. So you could see the original content that the user was typing, and you could see the words like contaminated, bleeding, pain, these are all like highlighted in the platform that the platform knows that this particular tweet is coming with some events that need to be lodged as a complaint later. So we basically capture the whole thing. And if you look at here at the top, so this is the integration part with the SmartSolve complaint module. So we have already sent this to the SmartSolve platform and SmartSolve acknowledged with the case number as well. So in addition to that, we are GxP compliance system, so we completely keep our trail or audit trail of what is happening to this data. If you look at here, so when the Twitter -- the tweet was pulled in, when we detected, there was an event in the tweet, and when that was sent to SmartSolve platform and when did we get the -- I mean, the acknowledgment back. So that's also being mentioned here. In addition to that, we also have MedDRA mappings. So there are some terms, pain, bleeding contaminated. So we were able to map those to MedDRA coding. So this information also can be passed to SmartSolve for their further case intake. So this is all happening within a few seconds. So less than a minute, so we were able to capture this tweet, we were able to sense it, we were able to give a structure for the complaint, and this is already pushed to SmartSolve. So now let me go to the SmartSolve platform now. So I'll copy this case number. So this is a SmartSolve platform. So let me go and open up the case intake section. So I'm going to look for that particular case and opening that intake screen. Okay. So you could see the content is already being populated here. And a case processor can come and fill in the rest of the information that's needed for the case intake. So we have automated certain things so that the case already landed in the platform automatically. So now it's easy for a case processor to come and add any additional information and submit this case. So from someone tweeting in the brand page, so up to its coming and landing into SmartSolve complaint module, is the integration that works within a few minutes or a few seconds. So that's what we have been working on about this integration. That concludes my demo. Thank you so much.

Great. Thanks, Aravinth. So when we point this solution to social media, we see the following. So across 7.7 million [indiscernible] customers, social channels and in the public domain, approximately 74% are managed by the technology. And that is 74% that are safe to remove from the workflow with no human intervention required. This is achieved through really robust safety patterns and has evolved to recognize slang, emojis and character recognition. So 26% of those records still requires some level of manual review. The manual review is required to validate identified events. And in the context of social, you will also have audiovisual or embedded URL files as well. So for those of you who have an Instagram account, for example, I bet you'll have seen somebody upload a TikTok video to Instagram's account. So there are multiple mediums that need to be deconstructed. And you may also have embedded URLs that will take you off that site and revert to another. So this complexity with social media. And very often, that level of complexity requires some level of human intervention to validate what the technology is finding and see if it's relevant. So the benefit for this client -- for social media clients using this solution is felt in the reduced volume of irrelevant data hitting either the safety database or eQMS complaints modules. And yet at the same time, maintaining a record of nonvalid events in a way that you could use this data for trend analysis or to show an auditor. So what we've seen over the last 5 years is that volume is consistently growing on social media, and that's no surprise to anybody. And since 2020, the efficiency gains are commensurate to that growth. But if you look back at 2018 through 2019, you see this drop in efficiencies, and that's a really good example of how unpredictable the social landscape can be. So events can happen that will positively or negatively impact the products profile very suddenly. And tech will be needed to quickly train and adapt to new vernacular surrounding and course of events. There's new social platforms emerge, this new acronyms, slag, viral sensations, you name it. So technology really is evolving very rapidly to keep pace with social media, which is evolving rapidly, and it needs to. But it is important to recognize that it's an evolutionary journey. So what about the concerns? Some concerns, few problems concerning social media, and they do have merit. Undertaking monitoring of this medium comes with cost. It's the need for people to manage volumes and reviews. And the good news is that tech is available today that can support this activity very well. I can handle languages, which is massive consideration when you're dealing with the medium that spans the globe. And it can greatly reduce the burden on humans. There is also a debate around what's found on social media. Will it have any bearing on my product? Will it be of little relevance to the quality or safety profile of the product. What we do know is that for all of that information, there is still valid safety information and product quality, complaint information that has emerged. And it can be used to enhance commercial and product insights. And I know we'll all agree that social media will continue to be an area of growth and diversification. So we need to very carefully consider the way in which we approach it. We can no longer turn a blind eye to it and tech can help, and it can find the data of relevance with very little human efforts. And of course, it may represent complexity and with complexity comes cost, and very often compromise quality. But you can address those concerns by making intelligent connections in your business processes. So what does it take to deliver a successful solution. Here are a few of the components as we see them that are needed to deliver with success. Having mature established technology, something that can provide a seamless integrated workflow. Having the human expertise that knows how to connect that technology is very important. I mean being able to cleverly optimize your processes. NLPs and ontologies or AI of any form, it needs to be customized to the product and the customer's very unique needs. So having the agility to customize the patterns and the searches. And something that's flexible, something that can move at the pace that social media can move at, something that can adapt to an evolving landscape is key. And what do we mean when we talk about established technology? Well, we mean one that has proven performance and has been exposed to millions of users, millions of records and millions of complaints. And lastly, what benefits could be brought to bear? Well, ultimately, it's about keeping the complaint system lean and clean, and all of the other rewards are consequential, of course. So through automating social media monitoring, you could realize efficiency, you could free up professional resource to focus on value-add tasks. Quality can be realized through the minimization of noise and separating those records, needing review from those that do not, and improving safety with early detection and overall compliance. And when we think about social media, we think about the future and how fast this medium advances and the pressure on our industry to stay in lockstep with those advances. Future-proofing is a necessity. We need to consider scale and volume without adding to headcount and improve overall market surveillance and qualitative perspectives and support and embrace this digital journey. So I will round out our presentation with the third and final polling question. That being, what has your biggest challenge been with social media and product complaints or adverse events? Or if you haven't yet engaged with this level, what do you suspect the biggest challenge would be if you were to engage with social media to find product complaints or adverse events? Would you believe that to be strain on resources; unidentified or missed events; disparate processes; inconsistency in quality or all of the above. Thank you.

Well, thank you, everyone, for that great presentation. Before we jump into the Q&A, we had a number of poll questions during the presentation. So Marie, I was wondering if you had any thoughts on any of the results that we might have seen as we were going through?

Yes, I did. Thank you, Marie. Fascinating actually. So the first question, if you remember, it was around dedicated social channels where people come to document any adverse events or complaints they may have had. We asked people if they have those dedicated channels. So 55% of you do or your teams do; 33%, don't; and 11%, don't know. And this would be really what we expect, I guess. I think the days in which we turned off comments on social posts for fear something would be reported and we'd be held accountable to that, have long gone. The trouble with large organizations is you very often don't know what one side of the organization is doing. And particularly, when it comes to marketing or commercial or any entity in the organization that's running maybe brand awareness campaigns, et cetera. Very often, we can have social thieves coming into the organization and not be aware of them. So while we don't have dedicated social channels for reporting of events, that data may be coming into the organization unbeknownst to us. It's not something that we're, I guess, dealing with then having a lot of conversations around lately. So the second question is how do you monitor your social channels? 55% do that manually, 5% have an automated solution, 15% don't monitor and 25% don't know. So I think that we -- I think it's interesting to see that we still do that activity predominantly, manually or choose not to. And I can understand why we don't because of the amount of noise and frankly, nonsense that can appear in social media and the resource constraints that, that represents. And what we talked about was really just trying to put a solution in place so that you can get access to that data. You can rest assured nothing is being missed, and you don't have to have that overbearing resource constraint that goes along with it. With regards to the third poll question here, I'm just pulling up the results, Marie, so bear with me in a second. And I'm not able to see the results in the third one just yet. Oh, yes. Resounding all of the above. So we asked what were the biggest challenges, listing a few of them, strain on resources; identified events; disparate processes. I mean, you name it, there's a whole plethora of issues and challenges they come on social media, and I think you've all considered that that's 100% and all of the above to be challenging. So yes, thank you for participating in the polls. I think it's interesting and it gives us pulse and food for thought as well. So great. I think Marie, if we're good with that, we can move to questions, if any questions have come in from the audience.

Yes, absolutely. And we have had quite a few questions come through. So I will get to those. But just before we do jump into the Q&A, for our audience, you will just see one final poll question, which will enable you to provide your feedback on the webinar today. We'd really appreciate it if you could just take a moment to submit any feedback via that before we jump into the Q&A. But without further ado, I will get around to some of those questions. So the first question that we had through says, I would think having complaints and adverse events coming in from social media would add to the noise in my complaint system. How are you able to keep that noise to a minimum?

Yes, I can take that, Maire. I think that's just what I was referencing. Look, in short, you search and you refine and it's a continual process of doing so. So on social, you're going to pull in a load of noise. And you can use tech to take 75% of that noise away immediately, and we found that to be the case on 7.7 million records. That's really effective in doing so. But it's that remaining 25% represent the work where you can initially review what the tech is finding and you can refine and make that more precise over time. So it's a real balance. You don't want to cast the net wide, especially on something such as social media, but then how can you be assured that you don't miss anything? And that process is achieved through human-assisted machine learning, and that cycle goes on in perpetuity. So it's all part of search, refine and get better and more precise, so that eventually that noise really dwindles down to a minimum. And you can continually customize your search and your MLP to pertain to your product, your portfolio and trends that you're seeing on social media as well. So it's highly customizable. The end result being that you get as little noise as possible hitting your complaints system.

Thank you, Marie. Before we jump into next question, I also wanted to flag to our audience that we've actually been joined by an additional member of the IQVIA team for the Q&A, Kari Miller, Senior Director of Product Management for SmartSolve. So delighted to have Kari on board and potentially we'll hear from you on some of these questions that have come through. And the next question that we had, though, was how do you know that a case in SmartSolve originated from Vigilance Detect? Do the records stay connected?

So yes, this is Donna. And I can answer that. So yes, short answer is yes. The records do stay connected. And if you might notice on the last screen that Aravinth had showed, there were 3 attributes there that we tag when the data comes over into SmartSolve. So the first one -- sorry about that. There you go. So the first one is the case source, which tells where it came from. So it'll say Vigilance Detect. And then the case source number, which is the identifier for the record in Vigilance Detect. So there's always that link back. And then the third field is the case source information, and that field displays the key words that were used when the data was pulled into Vigilance Detect. So on the example that Aravinth showed, you saw bleeding, you saw contaminated and you saw pain, so that information will always be displayed in SmartSolve. And then as well within the filters within SmartSolve, you can go in and you can just filter and say, show me my cases that came from iota.

Thank you, Donna. The next question that we have is, when a case is created from data sent from Vigilance Detect, what happens to the case if it is determined that it is not an adverse event or complaint?

Okay, sure. So I'll take that one as well. So the case will always be retained in SmartSolve, whether or not it's determined that it is an adverse event or a complaint. And then once the case is closed within SmartSolve, you do always have the ability to go back in and reopen that. So if more information came in that you then determined it was an adverse event or needed to create a complaint from it, you could always reopen that.

Thank you, Donna. And then the next question that we have is we already have a global Complaint Handling Unit, a CHU that is effective. How can this help?

Yes. I think -- this is Marie. I can take that. I think just essentially, it supports effective resource utilization. The net benefits to the team being that they can focus on root cause analysis and investigations of unknown failures mood and just got clinical or get close to tactical and technical understanding with the free time that can be generated from removing what is essentially a transactional or administrative task.

Thank you very much. And then the next question that we have is our global business heavily uses dealers, distributors and other third parties, how can this help our organization?

Yes. Along the same vein, it supports process effectiveness. So you can point the solution at something like a global customer service, e-mail repository for example, or dealer portals to search for potential or complaints or cases of any description. It can also just release that complaint entry task from in-country sales or customer services. And with the big bonus being at a global level, it can do some multiple languages. So that's a big win-win for an organization for customers, compliance, et cetera.

And then the next question says context is key. What may be an AEI in another industry could be one in mine and vice versa, how the product is used and the nature of the clinical environment really matters. And then the question is, how does using Vigilance Detect cover that?

Thank you. Yes, I think that's a great question. And like no 2 products, no 2 ecosystems are the same. And I think what's important to know that it starts out as a concentrate process. So as part of the deployment, we train it in keywords and pattern and specific to an organization. So Vigilance Detect test baseline safety terms and patterns that it will catch a very, very broad suite of things. But what needs to happen when we onboard customers that we talk to them, get to understand their products, their portfolio, the unique requirements of their ecosystem and then layer on additional custom or search terms. So it really is a process of consultation to layer on the specifics of the customer. So that like to that question, what's relevant to one may not be relevant to another, and we kind of get that surgery, really precise through that sort of task of customization.

Marie, thank you. And then another question asks, does this, I assume the platform help with knowing global adverse event reporting time lines and content?

This particular one doesn't -- this solution in isolation doesn't. But IQVIA has a right Intel solution. It has dedicated sections on safety, adverse and reporting time lines, et cetera, and that could be used in conjunction with this certainly.

And another question, we have says, pediatric populations and elderly populations related events without an identified patient. It's a question. I'm not sure if anybody wants to take a stab of that.

Yes. I can jump in on that. I think it's really interesting. And -- so pediatric and elderly populations or 2 populations you wouldn't expect to find on social media, first of all. When you think about elderly populations, typically speaking, those reports are going to come in through HDPs they're through a clinical setting and pediatric population, obviously, are not going to be reporting any events through social media, but you do have parents reporting on behalf of pediatrics. And there were some really interesting studies done on Mumsnet, our moms forums a while back that showed a massive level of sophistication in the type of events they reported for pediatrics. So I think that the other layer or the other part of that question is more around you don't have an identified patient. You don't have the consent of the person reporting and also it brings up the fact that there are, from a regulator's perspective, there are 4 minimum elements which must be present in order for something to be a valid adverse event from a GBP perspective. And I know like the difference between medical devices and GBP. But when we talk about validating an event or finding a complaint of relevance, we need to think of it doesn't meet those criteria. So an identifiable patient is typically outpatient. It can be adding reference to a patient. The other elements that you need are identifiable reporter, a suspect product or as an event, whether that's an adverse event complaint or a special situation. The trickiest of those 4 is the identifiable reporter because on social media, a patient, as I said, is our reference to our patients. The one that usually will constitute an invalid report is the fact that it's not coming from an identifiable reporter, because unsorted social handle somebody's Facebook name or their Instagram tag, doesn't constitute an identifiable reporter. And we've worked with clients who will look and see, well, what does constitute an identifiable reporter, and they would take any reporter who will provide follow-up information so that we can perform due diligence and following up with that reporter. So very often, you'll get 3 out of the 4 elements there. And what you have is a whole plethora of nonvalid events that come through social media. The tricky part is that you need to show due diligence. First of all, in having monitored it, but second of all, following up to get as much information as possible, very, very tricky in a social setting. So what we do and what we've seen customers do is a due diligence follow-up where they will respond to that post with a link to a toll-free number, for example, for a medical information call center or a link or webpage where they can report that event directly. So there are, I guess, considerations for everything in between pediatric and elderly populations and how to manage when you 3 out of the 4 criteria or 2 out of the 4 criteria, how do you manage that? And that's something that's done at scale by a number of our customers, and we've sort of learned to read the regulations and follow-up the regulators with regards to their expectations in what they constitute to be an identifiable report or an identifiable patient as well. So it's a very layered discussion. But definitely, if there's a specific question around that and someone wants to follow up off-line with me, please do and we can help you get to the root of that question as well.

And we've also had actually a couple of questions come through regarding Amazon reviews and complaints. So this question asked, I am interested in monthly reporting of Amazon complaints. Is this something that the software offers?

This is Aravinth, I can answer that question. So the system that we have worked with any sort of open-ended text field. So whether it's Amazon, Twitter, data, so all we need to understand is the individual cases, like how we can access the data source. So some of them may be closed. Some of them would be open, and some of them would be access through a client. So once we confirm that individual case and if you get the access to the platform, like I showed example of Twitter. So if you can access Amazon platform in any way, the rest is like works on the text base. So once we receive the data, then we must be able to look for complaints in the data. So it depends on how we get access to Amazon or any other platform. I hope I answered that question. Thank you.

Thank you very much. And then the next question that we have is if a company has multiple products and the product is not properly named in the social media post, how do you align an incoming complaint to the respective products?

Yes, I can take that one. That's sort of comes back to one of the questions I answered earlier, a little bit about the upfront consultation that you do as part of this. So if your company has multiple products, but the product doesn't appear in the way you might write down the name of that product officially, it would be built into those custom search terms. So very ultimate we all see is misspelling. You may have variations, 3 or 4 variations of a product depending on where that product is in the world, depending on the way in which it's used. There may be multiple names. So what we try to do is build a product portfolio, try and customize the terms to that product, but also account from misspellings or how that product might be misinterpreted. In addition, that's just so the technology will catch it and it will catch any iteration of us or any incorrect selling or any iteration of the name, whether that's a generic name a trade name or a local name for the product. What it also does is anything that the technology discounts as not being relevant or not having that product name. That goes into an audit cycle so that it is regularly in randomly reviewed to ensure that if it did discount it, there will be an opportunity to go and see whether or not it made the right decision. So that's part of the backup audit cycle. So essentially, what you're showing your auditors is we make every effort upfront to look for this. We've provided all of the various iterations of this product name so that the technology can ably catch that. But it also shows that you have a very robust order cycle behind the scenes that will catch it periodically and sort of satisfy your regulators to show that you are showing due diligence towards catching complaints on social media or in any form that they appear. So I hope that helps.

Thank you. We probably got time for a few more questions. First of those is, is the system flexible enough to allow events that have been determined not to be complaints or adverse events to be data mined for other useful insights into the products and how they are being used?

Yes. I think that we have a number of customers who look at nonrelevant specifically. So the system and the solution will determine that they are not relevant. They are not relevant to you, your product portfolio. And what they do is they take this data, parse it off and route it out to a what we call a signal detection dashboard, but it can be any name you want. What it does essentially is look at what the technology found to be of relevance, but dismiss it from the complaints of the adverse event workflow because it didn't meet all of the criteria needed to be a valid event. We have customers who look specifically at nonvalid events because they may represent trends. Like I said, they may represent trends in whose reporting. They represent trends in the sentiment of what's been reported, but also like the system organ class. Is there a particularly elevated event in one particular system organ or class, is there particular events that are peaking in particular countries, in particular languages. We have one country that does this at a very large scale to look at what has been the system, the workflow because very little valid event comes through social media. But what does come is value and looking at what else is happening as it relates to the safety of your products. So yes, the answer is yes. And there's great flexibility with those dashboards as well to see them in a way that is helpful to you in your organization. So it's a very flexible setup and it's -- I guess, it's really customizable to the unique needs of our customers.

Thank you. And then the next question that we have is, is there any data on what percentage of the cases can actually be followed effectively in order to gain -- in order to gather adequate complaint data and details around the incident?

Yes. We have a very, very tiny percentage of events that are valid or you get follow-up from. And you can imagine that, that's just the -- that's your situation with social. So around 1% of all events can be followed up on to get additional clinical information. So of those 7.7 million posts that we talked about earlier, around 100,000 of those turned out to be events that we were able to follow up on, the company were able to follow up on and they ended up in a complaint system or in a safety database. So a tiny, tiny percentage, but still 100,000 events is large enough for a regulator to worry about it. So you're not getting a lot of clinical information. You may be able to really actually reroute your reported to a health care professional or give them access to a link where they can provide more information or more depth to their clinical analysis of the event. But really, what you're doing is satisfying the fact that you are proactively monitoring it and hoping that some value will come from that. But like I said, there is a huge value to be found in what the technology dismisses as well. So even if you're not getting follow-up from your reporters because they're reporting through social media, you're still able to assess the trends of what's coming in to tell you a bigger picture. And that's very interesting from a sort of upfront of your safety process to be able to zone in on what you think could potentially represent a problem down the line.

And if I could, Maire, I'd like to just add to this just a little bit here, because there's a lot of questions regarding it's coming into the tech how we share that it's good or bad, go to complaints or go to adverse event reporting or not, et cetera. But with IQVIA with our ecosystem as well, even if it comes through detect, there's still another triage process even within the complaint handling and the adverse event handling system here for IQVIA. So you do have another opportunity to ensure that you can do that triage that uses your speeds within your organization, right, on the product and the event that has been reported to come to a final conclusion on whether this should be a complaint in adverse event or maybe both. So you're going to get an additional opportunity to learn from that and then continue to tune your parameters as you move forward.

It's a great point, Kari.

And then the last question that we have is for which languages is the system available?

Yes. So our -- what we call baseline ontologies are what's currently available like out of the box have been translated into 52 languages. For social media at the moment for our biggest customers, it's been translated into 90 languages. The degree of accuracy varies, of course, depending on the language and depending on exposure of that language. So how many data files have come in for that language. Obviously, the more exposed it becomes, more precise it becomes. So English for very obvious reasons is the most precise, but Japanese and English are sort of head for head in terms of the volumes that we get in for social media, and that speaks to the number of users who are on social media as well. So we can generally tell how accurate language is going to be by doing a proof of concept. I mean it's the best way to sort of try and test the performance and the accuracy using your data, I really would encourage really quick proof of concept to see how effective the technology could be using your own particular data sets that you're after earning from social media. It's the fastest way of getting a degree of certainty as to how effective the solution will be for you.

Well, we have just about run out of time there. So without anything else, I would like to thank all of our speakers today, Maire, Donna, Aravinth and Kari for the great presentation and answers in the Q&A session. And our sponsor IQVIA for making this whole event possible. On behalf of IQVIA and Citeline, we hope that you all have a great remainder of the day, and thank you very much for tuning in today.

Great. Thank you, everyone.