Data and Destiny: A Conversation with Commissure CEO Michael Mardini
The mantra for practice management in the post-DRA era is data management, and everyone knows how much data a radiology practice can produce. From the multitude of bills (service points) to the vast number of customers—not to mention the deluge of clinical data—radiology practices need tools. Commissure offers tools that assist in three domains: reporting and communications, decision support and utilization management, and data analysis tools. For the purpose of this interview with Commissure CEO Michael Mardini, focused primarily on the utilization management tools and the leverage they provide the radiology practice in the post-DRA environment. Mardini founded voice recognition and reporting company Talk Technology in 1994, which was acquired by Agfa Healthcare in 2001. He founded Commissure in 2004. As the business side of radiology evolves, where do practices need to be in their ability to manage their data? What are the flash points? MARDINI: Right now, outside of the most business savvy clients, they really are not prepared. I didn’t know this coming in. The thing that has amazed me more than anything is the limited access clients have to their own data. They have to buy special tools, and they are hiring programmers to write the reports, so only the savviest of sites can really use their own data. With P4P looming, there is no way, absolutely zero chance, that any of the items or any of the changes focused on utilization management are going to be dealt with unless there are some major changes in the ability of clients to access their own data and to manipulate it. There’s a lot to be done. RadCube goes a long way, but it really starts with accessibility and consumability of the data, opening these closed architecture databases that exist now so that people can actually use their data. Commissure supports IHE-compliant methods of accessing this data that virtually every vendor conforms with. Wikipedia defines commissure as the place where two things are joined. In aggregating tools for reporting, decision support, and practice management, what do you intend to bring to radiology? MARDINI: We look at the round trip. We as radiology professionals, whether you are a radiologist, a technologist, or an administrator, provide a service to the outside world, and the service we provide, at least internally, takes into account a lot of things: what kind of systems we are using in modalities, PACS and RIS, etc, and how we manage all of this. Internally, this is all very important, but the outside world generally cares about two things: how do I get my patient in for the right test in an efficient manner, and when are you going to communicate the results? From a service perspective and for the people we provide service to, those are the two things that matter most and the two things that we are ultimately being measured on. Commissure wants to affect appropriate ordering by providing support and knowledge on the front end at the time of ordering. Next, we want to provide tools for reporting that provide a consistent, concise method of reporting results in an accurate and efficient manner, and tools for communicating those results in an audited, confirmed, and traceable manner. Finally, we aim to provide some means of looking at this data, both for internal use, as well as external use, whether for communications, utilization management, practice management, or clinical research. These are the areas we feel we can ultimately help radiology improve on. Transcending all of this, there is the whole concept of orchestrating workflow in a heterogeneous environment. We don’t necessarily care that we are the worklist, nor do we think there is even a need to provide worklist and workflow functionality in a hospital that has a single PACS and a single RIS to work from. These systems all have worklists. The problem is they are provincial and only know their own work, they only know what is in their own system. So, in a scenario where there are multiple RIS or multiple PACS or multiple visualization tools, Commissure can fill the role of providing a common worklist, common workflow in an environment where, because you have all of those other systems, you would not otherwise be able to do. That is the connection, that’s the “commissure” in Commisure. Our solutions add value to existing IT infrastructure in the department. A department may turn off some feature or functionality in their PACS, like the worklist. But we don’t want to be the RIS or the PACS. These functions are better left to the vendors who have spent many years performing these functions. Our focus is on innovation and on providing service and solutions that are sorely needed, yet not offered in the market. As reimbursement declines, radiology practices are desperate for tools to help them manage their internal data, but frequently complain that the available tools need further customization once delivered. At the risk of receiving a sales pitch, how would you rate RadCube in its utility across a multitude of practice settings? How much work does it need on delivery? MARDINI: It’s a good question because we had struggled early on with what the deliverable was: RadCube Data Analytics, as much as we internally attempted to define it as a product, really is a tool. In order to understand it, you have to explain what it is. We did two things. Extracted data—from the RIS, the PACS, or RadWhere—removed it from its current relational database environment, and built a data warehouse to create what’s know as an online analytical process (OLAP) server, which allows you to analyze and mine the data. Once we created the OLAP server, we organized data and defined these structures called cubes, which are new data relations outside of the SQL relations of data. Defining these new relational points is an art as much as science. We’ve taken some leaps of faith with how we are structuring the cubes based on the types of data collected, how we believe our clients want to visualize the data, and the types of reports they want to produce. Organizing the data in an appropriate manner is only the first step and one that was not trivial. The second step is the visualization tool to build and see the reports in a way the client wants to see them. OLAP servers have been out for a long time, cubes have been out for a long time, but there weren’t any affordable and consumable tools that would allow someone to create the reports, create the views, and publish the views in an efficient manner. We built a front-end application that allows for easy creation and viewing of reports, as well as the ability to publish those reports out to people who need to see them. That being said, the final product that we box ship is a RadCube, based on the RadWhere database. The RadWhere database is everything we get from the RIS, all the reports that we produce out of the reporting system, and, the real key thing, all of the structured data that comes out of the reports from LEXIMER analysis, which is a patent-pending natural language understanding algorithm that extracts structured elements from an unstructured radiology report: all of the clinical findings, all of the follow-up recommendations, all of the codifed data that gets created by LEXIMER out of these unstructured documents. What was the finding? What was the recommendation? Where was the finding? What was the size of the mass? All of that information gets populated into the RadWhere database as elements that can be mined. Savvy clients can build additional cubes, add data to the OLAP server, use RadCube to slice and dice data with additional data. At one of our largest sites, they have all of the RadWhere data and the clinical data, and they wanted to connect pathology data to the cube. So now, instead of running a report that just shows a list of radiologic clinical findings for 50-to-60-year-old male patients who came in for neuro-MR complaining of dizziness and headache, they can now cross reference pathology data. Now they are looking at male 50-year-old patients who came in complaining of dizziness and headache and had benign granuloma in their middle ear. You can add more data and more analysis that potentially gives a big-picture view. For specific radiology data, there is a lot you can do with very little training. We ship RadCube with about fifty canned reports that cover areas such as radiologist productivity, radiologist recommendation rates, utilization management and clinical findings, basic canned starter items for the site to get going. In answer to your question, RadCube provides the ability to view data in ways most people only dream of. Ultimately, a tool like this is really the only way practices are going to be able to produce the reports and data required of them to maintain P4P initiatives and the ever increasing requirements needed for payment and accreditation. You start right out of the box and a one-day training session is really all you need to be an advanced user. As pay for performance emerges, where do practices need to be going today? MARDINI: Better access to data, looking at their ordering patterns, how they are accepting ordering, and how they are managing appropriate ordering is going to be key. That is where it all starts. Capturing the appropriate data at the time of ordering in a structured manner for the right study is what will allow you to do all the other things. As an example, there is a P4P initiative underway right now that addresses reporting in the case of stroke for neuro imaging. Practices receive an additional payment if they report a certain way for this indication. Well, if you can track these indications from the time of ordering and produce a reporting template at the time of interpretation based on the order, you solve the problem in an automated manner. Without tracking the ordering process and tying it to a reporting tool capable of handling the task, this would have to be manually performed. Today, most practices would either miss this opportunity or need to deploy resources and labor to do this. How can practices control their own fate by embracing more robust utilization management models? What are you observing among your most sophisticated customers? MARDINI: There are two markets for this. On the outpatient side, who is the customer? Who is at risk? It’s the payer. The practice actually wants to do more studies, that’s what they get paid for and that is where they make their money. But the payers want to reduce the number of studies. The customer for the radiology practice is the referring clinician, so radiology practices don’t want to put up hurdles either. In fact they will go out of their way and in many cases make phone calls to do the pre-auth for them. There is a ying and yang going on here that is going to have to change to truly deal with how to effectively manage utilization on the outpatient side where there is a third-party payer. It will begin with providing content and information at the time of ordering and some means of being able to analyze outcomes and data on the back end to truly get this done. Right now, the radiology business management companies (RBMs) require everybody to make a phone call. The reality of it is, 10% of the ordering clinicians are responsible for 60% of all of the inappropriate ordering at any given practice, but you don’t know who they are. It’s not enough to simply look at the ordering indications. Looking at the outcomes of these orders combined with indications is what will give you a true picture. For example, if 60% of head CTs have positive outcomes based on analysis of the reports, and you’ve got three doctors with outcomes of 10%, you found outliers. Chances are you found some individuals who don’t know how to effectively utilize head CT because they are not getting clinical yield as compared with their peers. On the other side, you’ve got radiologists. If on average, when a report does not have a clinical finding radiologists are making recommendations for some other follow-up study 5% of the time, and you have an individual ordering follow-up exams 16% of the time, you may have found someone who is hedging a bit, costing someone money for additional studies. You can look at this data, analyze, and educate to affect utilization in a positive way. The inpatient side is a completely different story. On the inpatient side, who is the payer? The hospital. In most cases they are capitated. Every additional MR, every additional CT they do, they are paying for it. In that scenario, on the inpatient side, the answer is much easier. You’ve got the provider and the payer on the same side of the fence, it’s just a matter of rolling a program out and getting the clinicians to use it. What are you observing among your most sophisticated clients in utilization management models? MARDINI: This area is so nascent. They are almost completely in a reactionary mode, reacting to a looming threat of the 1-800 pre-auth call. This said, we have many sites that are trying to be proactive. They are bringing in the payers and asking them to participate in the process, and while this is a long process, I think we are getting a lot of buy in. Payers see a lot of value in not only tracking and approving on the ordering indication and patient history, which they do now through the traditional RBM, but in being able to track outcomes and identify what is actually happening with specific patient populations, for certain studies under certain conditions. Combined with looking at ordering and service providers based on what actually happened with the procedure, that is really the holy grail here. What about in the private practice setting? How can private practices leverage these tools to provide a value-added for payers without alienating their referrers? MARDINI: It all comes down to relative pain. If the referrer is already being required to make the 1-800 call, the people at the front desk are faxing and re-faxing to get the study approved for their patients, they are feeling that pain. Then, if you present them with this web-based process of doing that, and they don’t have to make phone calls and incur all of the costs of doing that extra work, and you tell them it’s just going to be monitored on the back end, they love it. If you have referrers in an area where they haven’t had to make a phone call for a pre-auth, or if the radiology practice has been the one making the phone calls for the pre-authorization, then this is looked at as negative because it is something new that they have not had to do. It has to be presented in a manner that is a zero-penalty experience. If they are already making the phone call for pre-auth, then it’s easy. If not, then they need to know that they can visit a web site and place the order, or make the phone calls and send the faxes, but no longer do nothing. How do payers perceive this method? Is it an alternative to RBMs? MARDINI: The payers care about reducing their cost-per-patient, period. In the end, that is what they are paying the RBM for, which charge fees that amount to anywhere from $10 to $25 per phone call. Would a system such as ours save insurers a huge amount of money in fees? Yes. But the fees they pay the RBMs are dwarfed by the reduction in cases, assuming that is what they are getting. Radiology service providers and referrers generally do not like the RBMs. After all, they are the gatekeeper so it’s understandable. I suspect that payers don’t like them, but they are viewed as necessary given the lack of alternatives. The new method has the potential of providing the savings associated with reducing unnecessary exams, while doing it a manner that does not alienate providers. In addition, there is significant value in tracking outcomes to allow for education to improve future behavior. So, yes, I’d say they will perceive this as an alternative. As the founder of Talk Technology—and now Commissure—you helped to provide tools that have been widely embraced for their ability to speed throughput. What’s the next step in reporting and what is the role for LEXIMER. MARDINI: I didn’t come into this business to do Talk Technology again, or anything that anyone else was doing. We did this because we saw some deficiencies in the market and the opportunity to really address some enduring needs. There are problems that are not being addressed by the existing players, and with good reason. Large vendors with thousands of customers have got to keep the lights on. They have to get the SMA renewals, and sell the incremental upgrades. In that environment, it’s very hard to innovate. It’s the classic story of innovator’s dilemma, and health care IT is especially susceptible to this, because there has been so much consolidation and the barrier to entry is so high. My time in the industry allowed me to get to know many people who not only understood the problems because they lived them first hand, but could provide guidance regarding what should be done. Our medical advisory board is second to none. We have access to engineers with familiarity with what needed to be done, so it was a perfect storm for us. We were able to pool together all of the resources to deliver something special. What’s the next step? The reporting piece for us is a necessary item. We’ve got to do reporting because in the end, the report is the bottom line. You can spend hours looking at the images, but that internist wants to read the report. And not only do they want to read the report, they want a nice, concise, clean report. They want a consistent look and feel. So from the perspective of the reporting component, Commissure will do the production of the report and the delivery of that report in a consistent and efficient manner better than anything else today. We also recognized that being the reporting application on the desktop of the user is valuable real estate. Radiologists are giving us their attention. So, when we looked at what else we could do, it wasn’t just speech. We had their mindshare, they were in our application, so there were two things that we wanted to do. We either wanted to make them faster and more efficient, or help make them smarter. I sat down with a radiologist once, and he was describing something. I asked him, how do you know you’re right? He said, ‘I’m right’. I left it at that. Radiologists sometimes work in a box. They describe things; they make assessments, move on to the next case. They very rarely get any clinical correlation on what they describe. They don’t know that the liver diagnosis they describe 20 different times in a month is actually what they say it is. They may be describing it over and over again, incorrectly, the same way. So providing tools on the desktop for not just reporting, not just for the workflow, but for all of these other things that can be automated as part of the reporting and workflow process—that can track and provide this information at their fingertips, with communications and alerts that are part of their everyday tools—that’s what Commissure wants to be to the radiologist. Everyone is talking about evidence-based medicine to improve care. Nobody knows how to get the data, pull the data, and present it. At least inside of radiology, LEXIMER will pull this evidence from all of this unstructured documentation, organize it, and store it so that it can be used. What about the provocative clinical correlation piece? MARDINI: There are many places to address this with various solutions. As a radiologist, I’d like to know if what I describe is accurate from the perspective of knowing if the pathology or discharge note corroborated with what I described. Today, there is no sure-fire way for me to do this. We’re scratching the surface here, but this is where medicine needs to go. Commissure wants to lead the charge in this direction. We are talking art here. There’s a lot of art in medicine and rightly so, but there needs to be a lot more science on a day-to-day basis, and the only way you can do that is not only to create the data and navigate the data, but you also have to know that data is even available. It then needs to be presented at the appropriate time in a manner in which clinicians can consume it at the time they need to.