Smooth operator in the lab: middleware

February 2006
Feature Story

Anne Paxton

From one perspective, it’s just the latest buzzword for a product that’s been around for years. But the buzz about "middleware" is causing mild to moderate tremors in the clinical laboratory world—and even predictions that middleware might up-end traditional thinking about IT needs in the laboratory.

Defined as software that sits strategically between laboratory instruments and the lab information system, middleware uses rule-based decision processing, sometimes known as expert systems, to manage the generation of test results. It might autoverify results, add new orders to test files, insert comments, or take other actions for specific patients of specific physicians.

Suddenly, middleware is igniting enough interest that an entire pre-conference workshop at the 2006 Lab InfoTech Summit in March will be devoted to middleware topics.

Laboratories are starting to appreciate the multitude of functions that middleware can perform, says Gregory R. Vail, CEO of Data Innovations Inc., South Burlington, Vt. His company, which has been shipping middleware exclusively since 1991, now has installed more than 4,000 systems and reports that sales are surging.

"The growth of middleware really comes from the fact that more and more people want to solve problems between their instruments and their LIS," he explains. "The LIS is not meeting their needs, and the instruments may not be to the point IT-wise where they can do what the customers want. Middleware is a customer-driven ’third source’ to solve those problems."

There’s nothing new about instrument interfaces, which laboratories have long purchased to smooth out communications between diagnostic instruments and LISs. The instrument interfaces used to be referred to as "black boxes," says Bruce A. Friedman, MD, professor and director of pathology data systems at the University of Michigan Medical School, who organizes the annual Lab InfoTech Summit. "In the early days of LISs it was very problematic, because a laboratory would buy an instrument from an IVD company, then find out it needed the interface, that software that served as a bridge between the instrument and the LIS." And these interfaces were notorious for being expensive.

Dawning and then Data Innovations started specializing in these interfaces and interface appliances, and they got very good at it, Dr. Friedman says. "They have literally hundreds of interfaces they can take off the shelf, so one of their specialties is interfacing instruments to LISs for small hospitals or multi-specialty clinics and physician office labs."

But the term middleware has come into wide use in the last couple of years as those interfaces have transformed into rules-based engines. Rules, or algorithms, are "if-then" statements that can facilitate some of the laboratory’s processes, and they date back many years, to when Cerner launched a rules module called Discern Expert for its classic LIS.

Interest is higher now, because there are not enough medical technologists to meet staffing needs, and the rules-based engines are seen as a way of supporting lesser-trained laboratory personnel on off-shifts or meeting manpower requirements by operating laboratories more efficiently, says Dr. Friedman.

Because of these trends, more and more IVD manufacturers and LIS vendors are forming partnerships with middleware companies.

Data Innovations, for example, has about 40 partners including the largest diagnostics companies as well as LIS vendors McKesson, Dairyland, and Siemens. "We partner with both sides, and our partner companies actually resell our software," Vail reports. Some agreements are regionalized, while others—DI’s partnership with Ortho is an example—are worldwide.

In hematology, instrument vendors like Sysmex have an extra incentive to develop a middleware product for their line of hematology analyzers because they sell reagents that are diluents, not patentable products, Dr. Friedman notes. "So those vendors don’t make the profit margins like some firms that make reagents for immunoassays or molecular diagnostic tests, and they’re looking at middleware not only as an important adjunct to instruments but as a way of competing in the market, since a lot of hematology analyzers do the same thing."

The partnership and reselling arrangements have become so widespread that it can be difficult to sort out the players, Dr. Friedman says. "It’s complicated, because you have the incumbents—the LIS vendors—plus the IVD vendors coming in with their own products or their logo on third-party products, then the pure middleware vendors like Data Innovations." Some kind of industry shakeout is inevitable, he predicts, but it’s hard to say what it will look like when the smoke clears.

Dawning Technologies, Fort Myers, Fla., the second main third-party supplier of middleware or what its president John Selmyer calls "specialized interface engines," has more than 3,500 client sites in more than 28 countries.

"Hospital laboratories have always been under intense financial pressure, and it’s probably worse today than it’s ever been," Selmyer says. Having to produce error-free results faster and with fewer resources "is a tough prescription to handle, and hospitals have diminished capital resources to go out and replace laboratory systems as frequently as they might like." Middleware is one way to give existing systems a longer useful life, he says.

"We used to think of instrument connections as establishing uni- or bidirectional communication," Selmyer says. "But that was just getting connections running so orders could be sent from the LIS to instruments, and results could be properly reported back."

"Well, these days the systems need to be a lot smarter. Through advances in middleware, we can apply a variety of rules that eliminate the irritating workarounds that people had to deal with, do simple things like add canned comments to data records after certain tests, and conduct more complex tasks like reflex testing."

Dawning has longstanding relationships with many LIS companies and instrument manufacturers, and its partnerships with other companies vary, says Selmyer. "We may provide all or some of the instrument connections for a vendor. In some cases an instrument manufacturer may contact us to simply buy one of our interfaces to include with a placement of one of their instruments, because the client may use Dawning and we know we can provide the kind of connectivity they need."

Another model is Dawning’s relationship with the nucleic acid testing division of Bayer Diagnostics. "The System 340 bDNA Analyzer is a combination of an instrument and a controlling PC system, and Bayer was looking for a general purpose way to make the system connectable to a variety of laboratory systems the client might have. So they wound up integrating our software into the System 340 PC, and it appears to the customer that that’s just another element of the system developed by Bayer that handles interface connections." Dawning’s software handles such embedded applications fairly well, Selmyer says, adding that it was an FDA-approved implementation.

Roche Diagnostics, Indianapolis, has done more than 400 total installations of Middleware Solutions, says Roche product manager Jon Wearly, and its partnership with Data Innovations continues. "To best meet the customers’ needs requires the partnership of Roche and DI working on a solution together," he says. As an IVD manufacturer, Roche is subject to stricter FDA controls than are middleware vendors like Data Innovations.

By and large, developing software is not one of the core competencies of most of the IVD companies, Dr. Friedman notes. "It’s a sort of ’second act’ for them, and for a long time they thought they didn’t need to be in the software business. Now they’re coming to understand they need the software to protect their positions as vendors of instruments and reagents."

"Roche and Beckman Coulter for the most part go to third parties which produce for them the software or interfaces between their instruments and LISs. For example," Dr. Friedman says, "Orchard has a product called Aqueduct, and Beckman Coulter shrink-wraps the Orchard software with its hematology analyzer to provide its hospital clients with an interface that works with multiple LISs."

The term middleware is something of a misnomer, suggests Keith Chaitoff, senior director of global brand marketing for Abbott Diagnostics, Abbott Park, Ill. "A lot of laboratory industry customers and manufacturers have migrated to this term, but in the IT world, middleware is a software product that allows two disparate software systems to talk to each other. It doesn’t add value except as a communication device."

"So if you talk to IT professionals, they’ll give you a blank stare because their ’middleware’ doesn’t add the kind of functionality we’re talking about." Though Abbott’s middleware product is called Decision Manager, the category of products is referred to generically at Abbott as "expert software."

Expert software emerged to fill a gap as labs were asked to do more with fewer resources, and the need grew once patient safety became a hot issue, Chaitoff says. "Since the 1999 Institute of Medicine report stated that 44,000 to 98,000 people die each year due to medical error, there has been a ton of focus on reducing that medical error, and that’s a lot of what decision software is about."

"Prior to expert software coming on the market, really the only option to add some of the functionality it offers were LISs, and this functionality was often limited, expensive, had long lead times, and was disruptive to the lab. And today IT companies that supply LISs are focused on other things that will enhance their business model or revenues, such as expanding EHR [electronic health record] and CPOE [computerized physician order entry] programs."

Chaitoff says some laboratories think they have autoverification, but it’s not at the sophistication level that expert software can provide. "They might have a simple delta check where the doctor orders a glucose and the delta check simply says here is today’s result versus yesterday’s. You could argue that’s autoverification or validation, but once customers understand they can set up much more complex algorithms with different reference ranges for different physicians, different patient populations, different wards, then they see how much more valuable expert software is."

Middleware itself has become more and more "feature rich," Vail says. "The early requests from customers were for simple things. For example, ’My instrument produces numeric results, say, for a pregnancy test, but I really want qualitative results. Can you take numeric results and change them to ’positive’ or ’negative’ before they’re reported?’"

What people want now with rules, he says, can be "incredibly sophisticated."

"They may want to take into account a patient’s age, race, the ordering physician, and what kinds of things the physician wants done with the data."

Vail sees the word middleware as meaning "in the middle of laboratory operations in general," and he points to a number of things DI is doing in the laboratory that have nothing to do with instrument interfacing.

"We have a specimen storage and retrieval package to manage specimens in refrigerators. We’re also coming out this quarter with a new release that includes a module called Maintenance Manager." That module covers everything from the big maintenance notebooks sitting on top of the analyzers to the fire extinguisher check-offs and recorded temperatures for the refrigerators, he says.

"A lot of laboratories are evaluating what the [LIS] vendor has for them as an upgrade path and thinking, There’s not a lot of meat and potatoes here; therefore, changing or upgrading is not worth the hassle or expense, so what else can I do? That’s really opened the door for middleware, which involves a lower financial outlay for acquisition and, more important, is much easier to implement than upgrading or changing an LIS."

DI’s customer base, and perhaps the laboratory IT consumer in general, have "become better at asking for something even if it doesn’t fit the definition of what they think middleware can do," Vail says. DI’s Instrument Manager is an example. "A lot of people would not have thought of some solutions coming from middleware, but now we’re intimately involved in laboratory operations, and a lot of laboratories call us and say, ’Here’s a hurt point—can you help?’"

Middleware can give a boost to hospital laboratories’ outreach programs through "intelligent routing," Vail notes. "There can be some technical difficulties with bringing outreach samples into the laboratory because they haven’t gone through the traditional accessioning process." Middleware can help because while it can manage the entire sample and data flow between the instruments and the LIS, it also can connect simultaneously to any number of outreach programs.

"So when a sample comes in with a bar code, it can be put right on the instrumentation and run, because middleware knows about samples from both in-house and out-of-house sources, and is intelligent enough to send data belonging to the hospital back to the LIS, and outreach data back to the appropriate outreach application," Vail says.

That capability plays into a lot of HIPAA concerns about where data goes and where it resides, he adds.

What Roche’s Wearly likes about rules engines is that several if-then statements can be strung together to complete multiple actions. The newest version of Roche’s Middleware Solutions has a feature called specimen routing, he reports. It allows mid-size hospitals, or smaller ones that can’t afford automation, to determine the route through the laboratory that a specimen would follow so that turnaround time is optimized. For example, the specimen might go through a centrifuge, then to instruments A, B, and C.

"You can write a rule that says if test A is done and the physician is Dr. Smith, then add tests B, C, and D," Wearly says. Or an archiving rule will dedicate an archive space for a sample, "so when the test result for patient A comes out, you can basically wand that sample under a bar-code reader, and it assigns a rack position with a row, a shelf, a refrigerator—even a whole building."

Roche case studies show that the archiving rule can cut more than 60 percent off the time it takes technicians to retrieve samples.

"The other thing is you’re reducing a lot of mundane, repetitive tasks. That doesn’t mean a normal patient result is any less important than a result that’s not normal. But computer-based logic can help you reduce or better utilize the staff time you have. When I talk with laboratories, one point I always stress is, ’Let’s spend 80 percent of your time on the 20 percent of results not within normal ranges—not 80 percent of your time on the 80 percent that are normal."’

With the relatively new popularity of rules engines for handling laboratory tests, some companies are trying to determine whether the rules they develop have intellectual property value. Can a middleware company copyright its rules? Perhaps it can, but Dr. Friedman thinks claims of exclusive rights to certain rules will fail.

"These are relatively simple algorithms, and they’re so basic, any trained medical technologist would know what rules would help you operate more efficiently. They could just reel them off," he says, noting that generic algorithms for hematopathology are posted on the Web. (For example, see www.islh.org/2006/ConcensusGrp/ConcensusRules.asp for a list of 41 hematology rules.)

"If you asked me where the intellectual property is, I’d say it’s in the deploying of the rules and customizing them for the individual laboratory. To me it’s like Linux: It’s open software that you get for free."

"The ’secret sauce,’" he says, "is the company’s coming into your laboratory, looking at your workflow, volume, hospital environment, and outreach testing, then deciding which rules, deployed where, would provide the most efficiency and decrease the most friction in the system. That’s where the talent and the expertise lie."

The complex map of partnerships and reselling arrangements between middleware developers and IVD manufacturers may also mean, in some cases, that there are competitive reasons to keep middleware from doing too much, says Jacques Baudin, general manager and executive vice president for Technidata America Medical Software, Tucson, Ariz.

Technidata, which is a global LIS company with over 700 LIS, 390 middleware, and 6,000 instrument workstation installations worldwide, defines middleware as an information system dedicated to managing laboratory test production. "So we design our IT solutions to address the weaknesses of legacy LISs, especially their lack of flexibility in the area of process and workflow," Baudin says. "At the end of the day, middleware is a productivity issue."

"But we are an IT company. We’ll connect anything. If you are an IVD company, how far should you go? You enhance your own instrumentation by putting middleware on it, but the question then is, do you limit functionality to your instruments, or do you allow consolidation with other IVD manufacturers’ instruments?"

Some companies have been suggesting an open solution while others have indicated their policy is to limit connections by other manufacturers’ instruments. "I’m curious to see what will happen," he says, noting that the IVD manufacturers’ business model for selling more instruments may not result in the best IT solution.

For example, using several closed IVD middleware systems could end up leaving a laboratory with many different quality control packages. "Clearly, for me, if you want the gain from middleware, it has to be an open solution," says Baudin, who plans an address at the March Lab InfoTech Summit on this issue using the catch phrase: "’Open’ Software Solutions: Don’t Get Caught in the Middle."

Emergence of the virtual LIS

Twenty-five years ago, laboratories purchased a multi-purpose LIS that met all its needs, says Dr. Friedman. "But that was a simpler time, and the classic LISs were designed to provide functionality and support for hospital laboratory processes. The large reference laboratories had a lock on the physician office business, and they had huge IT departments to support those office practices."

"Now the classic LIS no longer suffices to meet all these business needs. Added on to laboratory practice are inpatient, outpatient, ambulatory care, outreach as a profit maker, point-of-care testing, home testing, and direct-access testing. So a number of pathology groups are just hunkering down under an aging LIS, not purchasing new systems, because they believe they don’t add that much."

"Essentially they’re buying specialized modules that are now coming to market that provide functionality the LISs don’t."

The first of these were the so-called laboratory portals by Atlas Medical and Labtest, which provide Web-enabled order-entry and result-reporting systems to support outreach, he says. "We call this the ’virtual LIS’ as the next revolution replacing the classic LIS."

But whether LISs could really be replaced is open to debate. "Lots of people have been forecasting the demise of the classic LIS for a long time, and they’re still out there," says Selmyer, who envisions a more gradual drift toward other models.

"The systems are changing; we see a lot of large data-storage systems built to hold all the data that may be generated within the hospital, and a variety of middleware products are used to get the data in and out of big warehouse type databases," Selmyer says.

Frederick L. Kiechle, MD, PhD, former chair of the Department of Clinical Pathology and former medical director of Beaumont Reference Laboratory at William Beaumont Hospital, Royal Oak Mich., believes that LIS vendors have moved too slowly to keep pace with laboratories’ needs, and the LIS platforms are antiquated.

"They might stick a graphic user interface on the front, but in the background they have been running the same old software as back in the old days. Middleware has resolved issues that were created, frankly, by the inefficiency of LIS and HIS systems to respond to the needs of customers in a timely manner."

Nevertheless, Beaumont, which was one of the early adopters of middleware, found it could not do everything. "We used middleware with our very first implementation with Sysmex. The concept was that it would do rules-based logic, then provide automatic verification, but our volume was much too large," Dr. Kiechle says. Driven by outreach business, Beaumont’s volume of CBCs went from 980 a day in 1998 to 2,000 a day in 2004.

"We actually had to have the HIS do the rules because the middleware was not capable of doing them rapidly enough and couldn’t keep up with the volume. Since then, I think all the manufacturers have figured out how to provide enough functionality, and now Sysmex has a separate middleware computer perfectly capable of doing all those rules." As more small hospital laboratories implement modular automation, he adds, "middleware has been crucial to their ability to do so effectively."

"If there was an absolutely perfect laboratory system out there in the country, everyone would purchase it and the secondary players would disappear, but no one has yet arisen from all the fire and brimstone to be the most effective providers of the services a modern laboratory requires," Dr. Kiechle says.

Abbott’s Chaitoff thinks the lab’s IT environment will look different in the future. LIS companies, he says, are focusing on initiatives outside of laboratories, like the electronic health record for which there is a huge federal push. This creates "a great opportunity now for a product like Decision Manager that offers benefits that are specific to the laboratory," he says.

LISs have never been focused on clinical management tools, Chaitoff notes, suggesting that middleware might bring a natural evolution possibly away from the standard LIS. "I would say the companies could have made a bigger impact in some of the functionality that Decision Manager offers, but they have focused elsewhere."

"Especially the older legacy LISs were designed as large administrative systems, to let database managers interface with the HIS, and offer modules around billing, but they weren’t designed to add clinical management tools."

Wearly of Roche doesn’t see any time in the future where either the LIS or middleware disappears. "Both have to have the other to exist. Middleware has a very distinct responsibility and job, and so does the LIS, and there are some overlaps."

"Some LISs do autoverification and delta checks—not all, but quite a few—but where middleware is very beneficial is with LIS continuity. So, for example, if you have unscheduled downtime with your LIS, Roche Middleware Solutions has a process built in so the lab can still accession the sample, run tests, and report results even if the LIS is down."

In Vail’s view, middleware will not become the LIS "as we know it today," but it will largely replace the traditional LIS of today. "It’s already happening; we’re in facilities where there is no traditional LIS anymore," he says.

"The information is coming directly from all kinds of methodologies of order entry—the HIS, or Web portals, or the outreach program—and therefore results have to go to electronic patient records or back to the outreach program—again, not really to the LIS per se. So we already can see the LIS getting eliminated from the equation."

The next stage for expert software is a greater level of customization, Chaitoff predicts. "There’s so much configurability and power behind Decision Manager, once we install it, the customer starts thinking, What other possibilities does it have?"

In the meantime, participation at Lab InfoTech Summit shows a dramatic shift has occurred since Dr. Friedman started conducting the technology summit 23 years ago: Diagnostics manufacturers have turned into information technology exhibitors because of their new engagement with middleware.

"At my meeting last year I had 28 vendors; in March I’ll have 35 and have turned away five," Dr. Friedman says. In the past he has never been able to attract an IVD vendor as an exhibitor. "But now I have four, and I have other vendors coming out of the woodwork."

It’s a turning point for the laboratory industry. Dr. Friedman has long believed that added values for laboratory testing come from information technology. "The whole field," he says "is starting to recognize the criticality of this information."

Related Links

• What labs should look for in middleware