Skip to content ↓
Please register to view this content

Paradigm SHYFT to RWE: Fundamentals for Success

*Required Fields

Transcript Collapse

Priya: So, with that, once again, many thanks to all of you for taking the time to join me today. As I mentioned, my name is Priya Sapra, and I have the sincere pleasure of serving as a chief product officer at Shyft Analytics. For those of you who are unfamiliar with Shyft, we’re a technology company focused on the life sciences truly committed to delivering data driven insights across the clinical and commercial continuum. And it is really this commitment that drives me to speak with you all of you today. It’s no secret that we’re in the midst of a paradigm shift within the industry, specifically catalyzed by the emergence of RWE.

Priya: Five years ago, many of us would have, potentially, no idea what the acronym meant, if it even existed much less choosing to spend an hour on the phone with me listening to us speak about it. so, where are we today? Number of data sources under the umbrella of RWE is increasing exponentially.

Priya: Each potential partner I have a conversation with is interested in monetizing their data, whether it’s a payor organization, a medical affiliation, or a patient group. They’re really interested in how to commercialize their information. And why shouldn’t they? They all have information that provides a unique perspective on the patient. In addition, as an industry, we’re already spending millions of dollars on the data. One client recently estimated that they’re spending close to $20 million just on RWE. And the fact is that only a handful of people within the organization are using it. It’s clear that something needs to change and needs to change now. So, what needs to change? In my opinion, we need to set the foundation in place to properly manage and optimally utilize this data.

Priya: And that foundation is built on three fundamentals. First, we need a rigorous data strategy. As I mentioned, the number of data sources is only going to increase. And the only way to leverage all of these data sources is through a common data model. Second, we need to leverage the right acumen. The industry needs to become comfortable with comparative analytics looking at a single metric from multiple vantage points. And most importantly, we need the right people to conduct these comparative analytics and translate them into valuable insights. Third, and perhaps the most essential, we need a robust technology platform, a platform that expands and evolves as the data and market grow and evolve. A platform that allows for the delivery of insights across the clinical commercial continuum of an organization.

Priya: So, let’s examine each of these a bit further. Let me start off by saying there is no single best RWE data source. Time after time, I get asked by our clients, “Priya, what data source should we invest in? Which one do you think is the best?” The honest answer is none of them. The answer is it depends. It depends on the therapeutic area and business question you’re trying to answer. Each data source has its own inherent strengths and weaknesses, depending on the data collection methodology, the geographic coverage, and the ultimate content of the data. As an industry, we need to become adept at leveraging each data source for its advantage and juxtapose one source against the other to get a complete picture of the market and achieve what I call the data equilibrium.

Priya: All right. So, hopefully, we all agree that we’re going to have to play nice with many, many different RWE data sources. However, remember, at present, each data source comes with its own data structure and vocabulary, as well as its own set of analytic tools for the end user. Within each organization, we have factions of loyalists, the data source A people, the data source B people. And they each speak a language that the other does not understand. Now, given that the mandate that we need to become multilingual will add a bit of an impact. So, how do we get past it? Many of us agree that a common data model is imperative. So, let’s all learn to speak the same language.

Priya: Our friends at the OHDSI Group, Observational Health Data Science and Informatics Group, have developed the OMOP common data model, which provides us one foundation for us to do so. What remains undeniable is the need to translate all of the data into a single taxonomy, so that we’ll be able to take advantage of multiple RWE data sources and no longer be beholden to any single one. So, now that we have this common language, how do we really transform it into powerful pros and meaningful insights? The answer is comparative analytics. As we discussed earlier, each data source provides its own perspective. We need to examine all of those data perspectives to find the truth.

Priya: In my opinion, the days of one data source informing one metric are over. Each metric needs to be driven by multiple data sources. For example, we, as an industry, have spent years estimating analytics like market share purely from sale data. This is fair, but we’re all familiar with the shortcomings of this data, with due respect to our failed data providers. Therefore, in some of my projects recently with some more provocative clients, we’ve been leveraging different RWE data sources as a means to compare and contrast the sales data and estimate true product usage. Together, we have found that these two types of data provide a balance and a more accurate picture of the market reality, further confirming our belief that comparative analytics are, indeed, the new mandate. So, how do we build these comparative analytics? And how do we translate them into meaningful insights?

Priya: The answer is simple to describe but always one of the hardest to solve. The right people. Well, who are the right people? They are those who understand the data, understand the relevant analytics, but most importantly, also fully understand the business questions facing the end users. Some might call them data scientists that are tailored for the life sciences industry. Ultimately, it is the marriage of data, analytics, and business questions that create the best insights. Now, the question that many of you may have is where do you find the right people? And the honest answer is I don’t know. If I think about our team, we have created the right people. They came with an inherent, data centric mindset and analytic inclination. However, as they became more familiar with the industry, they transformed themselves into the right people.

Priya: So, now, if you’ve followed along with me, we have all of the data translated and the analytics prepared and the insights generated. So, how do we get this intelligence into the hands of the end users. Technology. Technology is the only way to sustain the ever-increasing number of data sources to officially conduct powerful comparative analytics and to deliver RWE insights across the organization. We need a platform that can expand and evolve as the data and market expand and evolve. And let me be clear. By platform, I mean product. Services related to RWE are readily available, and they do have their inherent short-term value. However, if we do not invest in a product, we will soon find ourselves unable to keep up with the paradigm shift that we’re experiencing.

Priya: Currently, most of the value of RWE is centered around the clinical domain. I’m sure this doesn’t surprise any of you. Research studies based on RWE are increasing in prominence. At a recent symposium I attended, a member of the FDA stated that not far from now, it would be required that an RWE based study would be part of any submission for a new clinical trial. So, our comfort level with mining this data has to increase and has to increase fast. The common data model gives us a consistent way to look at the debt. Now, we have a way to look across multiple data sources in a consistent way. However, now, we need an analytic solution that allows us to analyze the data in a consistent way. A tool that can be leveraged not only by the few who have the advanced programming skills required to explore the data, but one that will guide a broader group of clinical researchers to analyze the data. So, why is this important. So, based on my understanding, a current, average RWE study takes about six weeks to conduct. Estimates are that a technology based solution could shorten that time to six days. As we said, there’s no argument that the current focal point of RWE is in the clinical domain. However, I think many of us would agree that the future is well beyond clinical to the rest of the commercial enterprise. We’re only starting to uncover all the uses of this data across the organization.

Priya: The commercial side has been conducting business with the same data set and metrics for many years. And it’s going to be challenging to change, at least in the short term. Therefore, my suggestion is this. We need to transform RWE into the commercial vernacular. We need to align RWE with the commercial questions that this part of the organization asks themselves day in and day out and stitch those answers together to tell a story. A story that gives them the intelligence to make business decisions and take action. So, now that I’ve had the opportunity to share our position, my position, on the fundamentals of RWE, I thought it would be important to share with all of you how these fundamentals are manifest within the Shyft RWE solution.

Priya: So, if we start off and think about data strategy, I mentioned it is critical that we have an approach that allows us to examine multiple different data sources with a singular point of view. The Shyft RWE solution uses a common taxonomy across the data sources in the shape of the OMOP common data model, the one I mentioned earlier. And, as I said, each data source, typically, comes with its own underlying data structure. The metaphor I often use is that each data source speaks its own language. Some speak French, some German, some Italian, some Spanish. The Shyft data platform allows for translation of all of those data sources into one language, English, if you will, so that now, all of the data can speak to one another.

Priya: This becomes readily apparent if we search for clinical concept in our search engine. We can search for any term, such as Type 2 diabetes, and we can see how the disease and the corresponding underlying codes are defined within all the multiple clinical vocabulary that governs the original, raw data sources. We can easily see how one concept translates from the original source vocabulary into the OMOP standard vocabulary. And, in addition, we can see how many patients we have within each data source based on the standard vocabulary for a specific clinical concept. Also, it’s important to note that now that you have the data speaking a common language, you’re able to conduct the same series of analytics and conduct the same study across multiple data sets.

Priya: Imagine, if you will, that each of the data sets that you’re purchasing as part of this drop down menu. You can leverage one and design a research study. And then, simply copy and paste the study and run it against a different data source. But now, because all of the data structures are exactly the same, you can compare and contrast the results from one study and one data source to the exact same study built on a different data source. As I mentioned earlier, in my opinion, there is no such thing as the best data source. So, here, by looking at the same study with different data sources, you’re able to juxtapose one against the other and really start to determine where the truth lies. So, now, let’s talk about the use of technology.

Priya: As you know, at present, we talked about how the focal point of RWE is on the clinical side, and really, organizations using it to conduct retrospective observational studies. And I said, as I mentioned, the criticality of this research will continue to rise, particularly with the FDA mandates that are on the rise. The power of the Shyft RWE solution is that it allows you to conduct an entire observational research study end to end within the platform. Let’s quickly walk through a study as an example. Incidents of CHS in patients on SDLT2s versus DPP4s for Type 2 diabetes to try and understand the difference in risks between patients on each of these two drug classes. To start, in addition to selecting a data source, you can easily select the exact data timeframe, look back period, look forward period, and longitudinality of the data.

Priya: Furthermore, you can define very specific and nuanced inclusion/exclusion criteria through a guided user interface. For example, you can use diagnoses, drug utilization, and lab values all in combination to indicate the manifestation of a disease. For example, here, we’ve defined patients with SDLT2 usage and those who have used the class within 60 days of their diagnosis of Type 2 diabetes. We could also very easily create a third group of patients who have used TVB usage to include in this study. Now, easily, we have that third inclusion criteria to augment the study that we’ve created. So, now that you’ve defined the who of the patients that you want to include in your research, you can build a series of data variables to understand the what you want to know about these patients., such as, for example, in this case, presence of cardiovascular events prior to the CHS. Here, we quickly and simply defined CV events as a combination of myocardial infarctions, PAD, etc. We can also include variables simply such as line of therapy variables to determine how long a patient has been on first line therapy.

Priya: In addition, you can drill down into specific sub groups and past patient strata. So, for example, here, we can quickly create age sub groups to investigate further in our analyses. But now, finally, as you’ve gone through this data preparation step, which many of us know is potentially from the hardest up front work, which all can be done in an iterative fashion, we’ve now fully got the data ready for analyses. And you can design, within the solution, design and execute rigorous statistical analyses Anything from descriptive statistics to time to event analyses within the R based validated analytic engine to design and then, execute the analysis and visualize the results.

Priya: So, I pause because it’s important to note that everything that you see here is happening within the confines of the solution. Why is this important? Because, as the volume, variety, and velocity of data is continuing to increase, only a technology platform like this will allow you to go from two to twenty to one hundred data sources. As the number of emerging data sources continues to increase, the platform has the capability to stay in line and expand with it. In addition, technology allows speed. We talked about observational research studies, typically, taking six weeks, at minimum, and now, being able to be completed in six days. All of you are familiar with the current process.

Priya: Typically, there are other researchers who are the brains behind the questions. They work back and forth with SQL programmers to query the data and iterate until they get the exact data set that they want. And this takes a significant amount of time and effort. Then, that same researcher now partners with a PhD statistician to conduct what is, again, the very iterative process of analytics starting with simple statistics to the most complex algorithms. This three person group requires significant time to align and address the question at hand. The Shyft RWE solution, as it exists today, allows that researcher to have the power of all three of these resources at their fingertips reducing the time to execution significantly. This shortened time to result is absolutely necessary within the ongoing RWE paradigm shift.

Priya: Our ability to access and analyze this data has to increase and increase fast. And the RWE solution at Shyft allows you to do that. So, finally, I talked about the need for proper acumen to manage RWE. And this acumen is required to take this data via comparative analytics into insights. For example, here is one of the visuals from the RWE insight suite. It has a traditionally commercially metric market share, one that we would historically leverage traditional sales data sources to calculate, as I mentioned earlier. But now, using different algorithms designed specifically for RWE, we can utilize claims and EMR data to calculate this metric and contract it to the same metric based on sales data. In addition, end users can benefit from visualizations based on RWE that show how patients are being treated across lines of therapy bringing to life brand switching patterns and product discontinuation rates. Analytics such as these are critical to bring into the hands of the end users providing the commercial organization with RWE insights that are translated into the commercial vernacular and drive decision making. Now that I’ve shared my thoughts on RWE and showed you how we leverage them in an RWE solution, I wanted to close with the three fundamental calls to action.

Priya: 1) The increasing data sources and the need to leverage many of them requires a common data model. The translation of these data sources into accurate insights requires comparative analytics and the right people. And finally, the derivation of analytics and the dissemination of information requires a technology solution. Products that will harness the power of RWE and broaden its use across the clinical commercial continuum of an organization. So, with that, that was a lot of the information, the point of view, and some of the highlights from our solution that I wanted to share with you.

Priya: I would love to open it up to questions and get your thoughts and address anything that I may for you. So, please feel free to speak up.

Priya: It looks like we did get a question on what database is the data coming from. So, I assume that this is a question about which data is driving the demonstration that I share with you through the solution. It is, currently, the commercial claims and Medicare data. However, the Shyft RWE platform is completely data agnostic. We’re able to consume any of the variety of RWE data sources that are available. We are capable of converting those data into the OMOP CDM, and then, power both the analytic suite, which you saw, which allows us to drive observational research, as well as drive the insight suite allows more of the data visualization based on RWE. Examples of data sources we work with would include Humedica, GE Centricity, [inaudible], as you saw, Flat Iron, Clipper, so on and forth.

Priya: Another question: So, one of the other questions that came through here is can’t companies just hire service providers to solve this RWE challenge as opposed to buying a whole solution?

Priya: Well, I mentioned my point of view on that. But there’s no doubt. There are a lot of really smart people who can do some amazing analytics based on RWE. And to answer a very specific question, I do see the benefit of some of the service providers that we have out there. But that only answers the 5 to 10 percent of exceptions. It doesn’t address the rules. As a rule, a technology platform is really what’s required to consume all of this information and build analytics and insights that can answer the 90 percent of questions that are facing organizations anywhere from the clinical side of the organization as they think about the epidemiology, HEOR, med affairs, pharmacovigilance groups, to everyone on the commercial side, marketing, market assets, new product planning, corporate development.

Priya: They need a standardized, syndicated way to handle all of this information. And really, a technology platform is the only way to handle it. Any questions from individuals on the phone? Feel free to speak up and unmute yourself. So, another question is what about the terminology , for example, Sno Med, NBC, Metric Post, are these being repaved. So, that’s an excellent question. We won’t take too much credit for this, but the OMOP CDM allows for mappings from all of the different clinical vocabularies, whether they’re disease based, drug based, procedural, so on and so forth to one single vocabulary that is the standard. For example, in diseases, it takes things such as ICD9 and translates it into Sno Med, which is the one standard vocabulary that’s used. The source of the data that exists in the underlying raw data, if you will, is retained. So, we’re able to continually understand how the raw data was mapped into the OMOP data and can leverage those mappings for analysis as well as needed.

Priya: Any other questions from the field out there?

Participant: So, the use of findings from RWE studies are dependent on the scientific soundness of conductive studies. How would scientific soundness be incorporated into the solution you’ve provided?

Priya: So, this is great. So, the question itself is saying – and I apologize if I’m misinterpreting it. It’s that how do we ensure for proper behavior and proper use of the platform. The answer is that the platform itself has baked in business rules that allow or guide the individual to leverage the tools in a proper fashion. For example, the analytic part of the solution ensures that you don’t use a categorical variable to conduct an analysis that can only be done on a continuous variable. However, bad research is bad research. So, yes, the solution could allow individuals to ask questions that may not be as relevant. Now, how do we work around that? We urge each client that we deploy the solution on to put into place data governance and ensure that the right individuals have access to this solution and meet with the guidelines that are manifest for each organization to ensure that, again, those were properly trained and have the authority to actually conduct this analysis after we go forward with it. But the tool itself, though, again, is very smart, cannot take the place for the clinical researcher.

Priya: Any other questions? The next question is how wide and flexible is the range of analytics through statistical methods that the platform provides?

Priya: Another fantastic question. Thank you. Currently, there are 50 plus different statistical methods that are within the solution with that number changing on a day to day basis. Anywhere from all of the different types of descriptive statistics, correlations, regression, [inaudible], time to event analyses, line of therapy analyses are all currently available with the tools. And I said, as the tool continues to evolve, additional methodologies are being incorporated. However, I’ll answer a follow up to that as well. What if I want to do an analysis, and I always use my example of the alpha, which a PhD statistician shared with me recently.

Priya: Does the tool allow you to take the data that you’ve created and pull it outside of the solution, so that you can use something as powerful as SAS to analyze the data?

Priya: And the answer is yes. So, we understand that those 50 different analytic methods may not be sufficient to answer all of the questions. They are, indeed, the ones that researchers are using most often. However, they are not all of them. The tool itself then allows you to do the upfront data preparation steps like defining the cohort and defining the variables and so on and export that data through CSE files, so that it can be used within the power of SAS, MathLab, whatever the application is that’s of interest to you and conduct analytics on top of it.

Priya: Another question is can I add my own methods and functions, i.e., upload them to the analytics platform?

Priya: If we’re talking about specific analytic techniques, for instance proprietary, at present, we don’t allow for that. However, we could consider how that can be done in the future really trying to create a product to meet the needs of the 80 percent and allow as little configuration as possible, but allowing that flexibility so that you would be able to take that data and the example that I just provided to you before to take the data outside of the tool and then, incorporate it into whatever language you have to conduct that analyses. One of the things I didn’t have the opportunity to show you in the variable section, I’m happy, of course, to talk to any of you one on one after this, is that the defined variable section has the capability to do arithmetic, which means that you’re able to use the existing data variables and conduct whatever mathematical functions you’d like to to create a new variable really allowing a great deal of power in defining variables that may not exist within the data. So, it’s not that you’re limited to what is really within the OMOP data, but how do you take different columns within that data structure and combine them, again, mathematically, in a way that provides you things like risk scores. For example, many times how we see the arithmetic function being used. So, though you can’t upload them, you could leverage the arithmetic functionality to create that new variable and that new piece of data to examine.

Priya: The next question is can the use of a common data model from multiple sources, can you demonstrate the integration of multiple sources for analysis? For example, adding Truven to EHR and registry data.

Priya: Right. That’s, of course, the Holy Grail of what we’re all looking for is one patient record across multiple data sources. What we have done, at present, I want to answer the question carefully, what we have done to date is looked at patient cohorts across these multiple data sources. We’ve done it with the claims data, all of the EMR data, and multiple patient registries. We are in the process of developing a methodology that will allow us to get closer to the patient level. But, at present, the solution does not offer that.

Priya: The next question is will the reported presentation be made available.

Priya: That’s a great question. I think, at some point in the future, I don’t know the exact dates and times with that, but I’m sure, at some time in the near future, we’d be able to share the conversation we’re having here with all of you.

Priya: Anymore questions, thoughts from the group? I’d love to hear any of you on sort of what – if I can put a question back to the group, for those of you who are willing to share your thoughts, we would love to hear about what prompted you to join us, join me, join the team here today and any feedback you have for us.

Priya: Another question we have is can you tell us a little bit about the technology infrastructure behind the platform?

Priya: I can. I can get you started. Everything that you’ve seen is proprietary that is built here at Shyft. Everything from the visualization platform to the analytic tool that you saw was all built and developed here at Shyft. There’s, obviously, a big data engine. I’ll leave it at that. That helps us consume the data. And then, we leverage our own business rules engine that we’ve built over the past decade or 15 years that we’ve been in the industry to transform that data and build analytics and insights on top of it. and then, we use the publicly available OMOP conversions and our own home grown OMOP conversion business rules to get it into the proper CDM. I’m happy to answer more questions related to the technology. But I really want to allow the right individual to answer that question, if there is more granularity. I’m happy to address that.

Priya: Another very good question is on the hosting locations. Can global colleagues access the applications?

Priya: That is another great question. Yes. We do, currently, work with global data sources. And we also are providing the insights and analytics back to global users. So, we have the ability to use the global data, allow it to maintain the geography restriction that we all know need to be adhered to, when we’re looking at international data. We’re able to take that data where it resides and use what’s necessary to build insight. But to directly answer the question, yes, we’re able to then take those analytics and insights, whether it’s based on US or international data, and allow access to individuals across the globe.

Participant: Is the database hosted via Amazon?

Priya: Yeah, we do use AWS hosting services for the data. But the data cloud that we have is something we called the Shyft Health Data Cloud which is HIPAA compliant, PII compliant. And it has gone through rigorous, and I’m happy to, again, have a more detailed conversation about all of the validations and verifications that the data cloud has gone through, in order to allow us to securely host all of that information in the cloud. Any other questions, or is anyone willing to share some feedback, thoughts, motivations? I’m happy to hear from any of you on the phone.

Priya: Do we support bringing proprietary clinical study data into the system?

Priya: I think that we’re talking about the use of primary research, such as clinical trial data. Again, it’s absolutely we take extreme value in incorporating this information and juxtaposing it with RWE in many, it seems like, obvious use cases. So, I absolutely support the incorporation of this data. At present, we’re trying to determine the best way to do so to ensure an apples to apples comparison. So, it’s not something that’s offered today but something that we absolutely have considered and is very logical, as we think about the evolution of the product. Any other questions, feedback, motivations?

Participant: Just to confirm, did you indicate that one can download the data?

Priya: Yes. So, absolutely. So, we can leverage the data through the analytics that you saw. You can swap the data source, choose the look back/look forward period, the longitudinality, define the inclusion/exclusion criteria, and define the exact cohort you want. And you can just download the index date and the cohort based information. Or, in addition, you can define variables and add sub groups and download that data set as a CSB and incorporate that into, again, any other analytic tool that you’d like. So, yes, I can confirm that you can download the data.

Priya: The last and final call, if you will. If anyone has any other additional questions, comments, feedback, happy to hear them. We’ll hold on the line for another minute or two here. Well, thank you for the generous compliment. I appreciate it. We put a lot of hard work – it’s a labor of love for many of us here. So, we appreciate the compliment on the interface. So, with that, I don’t see anyone else typing. I think we have one more question coming in. So, we’ll wait for that.

Priya: While we’re waiting to see if there are any other questions, I do want to take the time to say thank you so very much for bearing with me today and taking the time to join. I’d love to have the opportunity to speak with any and all of you one on one to answer some more specific questions that you may have about my point of view or about the solution itself, please don’t hesitate to reach out either to the individual who shared this invite to

Priya: I’m sure we can find any of your requests and make sure they get funneled to the right person. So, I do want to thank you for that. It still seems like we have an individual or two typing. So, we’ll give them another minute here. And to repeat the email address for the general inquiry, it’s info @ So, just to be sure, there’s some question of whether or not it’s info @ shyftanalytics. Probably not. That should work. But for one time only, I’ll share my email address, so that, if you’d like to contact me directly, I’d love to hear from you. And my email address, which I’m confident of is p sapra @

Priya: All right. With that, I think that we have all of the questions. Again, anyone else want to share feedback, comments, follow up directly with me. I’m happy to, again, take those questions and inquiries. Again, thanks, once again, for your time. We’re going to close out the line now. And thank you again. Take care. Have a great day. Bye.