College of American Pathologists
Printable Version

  Flexing connectivity muscle with middleware


CAP Today


March 2008
Feature Story

Anne Paxton

On its way to becoming the linchpin of a laboratory's information management, middleware often enters the scene as a kind of IT duct tape.

Initially, many laboratories say, they turned to middleware to troubleshoot something that wasn't working quite right in their test ordering and results reporting. They wanted to avoid a full-scale operational retooling. But frequently they find that what began as a patch or a workaround for a system glitch ended up fixing a range of other problems.

"Every time you get a new instrument, there's something that you need that nobody thought of before," says Marilyn Paulson MT(ASCP), laboratory information manager for Florida-based Bay Pines VA Medical Center, one of the largest Veterans Affairs hospitals in the country. "And middleware can give you the flexibility to be able to customize a solution."

The laboratory at Glan Clwyd NHS Trust, a hospital in the United Kingdom's National Health Service near the Welsh town of Bodelwyddan, is a case in point. "I suppose in your terms we'd be a rural hospital," says Martin Gibson, biomedical scientist/computer manager, but Glan Clwyd is the site of the cancer center for the three main hospital groupings across North Wales. The possibility of middleware arose in 2004 because of concerns about monitoring near-patient testing on a different family of analyzers from that of the main laboratory, which has Siemens Advia 120s.

"We noticed there were occasionally some discrepancies, and we felt it would be a good idea to have Technidata middleware brought in so we could compare the results from the two machines more directly."

Soon, the technology platforms used will be uniform. "So our real original reason for going to Technidata sort of ceases to exist for the moment," Gibson says. But Glan Clwyd has a new objective: It's now running all of its validations on Technidata middleware in preparation for purchase of an all-Wales replacement lab information system. "We don't know who the supplier will be, but once we move all our validations onto Technidata, whichever vendor is chosen, it will be much easier from the laboratory point of view to change over."

Glan Clwyd is hoping other efficiencies can be gained from its middleware. "We wanted to take advantage of the way Technidata works and the fact that we could look at extra tasks that we weren't really able to incorporate within the present LIS. It limits the number of tests you can fit on a request, and we'd already hit that limit but there were extra tests that newer technology makes available. Because Technidata is test-based, there is no set limit."

Stepping up from a plain-Jane "green screen" was a bonus of Technidata, Gibson says. "You could see previous results against current results in a table, but you can't display them graphically. Technidata is Windows-based and takes advantage of current technology, and it's a lot better-looking."

The U.S. Department of Veterans Affairs decided a number of years ago to stop writing its own interfaces and procure middleware vendors to get new instrumentation onboard more quickly. So the nation's 152 VA hospitals are permitted to choose Dawning Technologies or Data Innovations as their middleware, Paulson says.

Bay Pines performs 3.4 million tests per year with a staff of about 105 FTEs. The Vista software used for the LIS has been around for 25 years and has a number of limitations-including a latest version that is 10 years old, with only patch-type solutions to update it.

Dawning Technologies middleware was first used at Bay Pines to interface data from the instruments via an HL7 with the Vista LIS about 15 years ago. Now, each time a new instrument comes in, middleware is used to interface it, and in fact, "without middleware we'd have no way right now to interface anything." Vista, Paulson reports, is thinking of eliminating some of the middleware and doing the interfacing itself in new software, but she questions whether that's possible. Middleware is critical to the functioning of the laboratory's fully automated chemistry and hematology, she adds. "Middleware makes our automation work."

However, translation is not all that Dawning's middleware provides, she says. "You're able to make rules to apply to the data. For example, if you have specimen integrity problems, say with a hemolyzed sample, there are a set of rules for how to handle them depending on whether they are mild or two-plus or three-plus hemolyzed." In addition, Paulson says, the rules give them the ability to get the data ready for autoverification, which the VA does not yet support. "Even though we do not have autoverification yet, the rules minimize the time required for review once the data get to the host [Vista]."

Paulson especially likes the middleware's customizability. "Every now and then I want something that actually requires programming, and Dawning has been very flexible in adding functionalities that didn't exist before. For example, about six months ago for our chemistry analyzers we switched to an ASTM interface so the middleware could send comments to our system. Before, those would have to be input manually by a tech."

All of Bay Pines' interfaced instruments and systems such as the automation lines, both in the main lab and in remote sites like its outpatient lab in Fort Myers, are set up, configured, modified, and diagnosed from the central location. This eliminates the need to travel to the remote labs and clinics to set up new equipment or troubleshoot it. JavaLins, which have Dawning's JResultNet software loaded on them, are used to connect to remote locations via the network, and they allow for continued productivity in the event of network downtime.

The middleware's ability to complement and enhance Bay Pines' existing systems helps extend their useful lifespan, and that's important for the VA, Paulson notes. "At the VA, adapting to system changes is not something we can do fast, so Dawning will do it for us. We do the testing, and our functionality is greatly enhanced."

Once the laboratory at Porter Health in Valparaiso, Ind., started adding Data Innovations middleware nearly four years ago, there was no going back, says laboratory IS coordinator Mark Wellons, MT(AAB). "We kind of went crazy," he admits. The laboratory put middleware on two of its Sysmex XE 2100 hematology analyzers, three CA-1500 coagulation analyzers, its Iris Diagnostics IQs for urinalysis, six versions of Dade Behring's Dimension analyzer, its Beckman Access 2 immunoassay analyzer, and several other instruments. "We even have it on the tiny little Biosite triage meters," he says.

One of the appeals was greater self-sufficiency. "Before I ever used the advantages it was touted to have, based purely on connectivity, for me middleware was a win from the beginning," Wellons says. With Porter Health's LIS, "the instruments had to go into a serial term server, and we have big bulky ones that the vendor had to program for us and nobody understood them."

"With DI, I could use small, dumb, cheap serial term servers that I could program myself. If the serial box becomes fried because we encounter a lightning strike or other electrical problem-and I've seen several-for about $500 I can buy a new one, program it, and replace it in 30 minutes. I don't have to schedule time with the vendor or my IS department to assist with the recovery, and 16 instruments don't go down all at once. So connectivity, as unglamorous as it is, is a wonderful, wonderful thing."

Middleware has changed his role as IS coordinator, he says, to one of essentially "coming up with creative solutions to problems I didn't know I was going to have." For one, he underestimated the impact the middleware would have on users.

Wellons wanted to find a way to promote uniformity among all shifts and facilities. With middleware, he says, "we can almost force the test results to be consistently the same."

"For example, if I have a reference range high result, it has its own color that comes up for the techs, and they have a defined way of handling it. Since it has autotracking capability, I can prove whether the tech did or did not handle it correctly. And the techs were really nervous about that at first-that it was sort of staring over their shoulder. But more times than not it can help them out.

"They've adjusted and tend to like the system, but that was the biggest thing, that it took a longer time for techs to give up control. We all tend to be control freaks; nobody works in the laboratory because they are laissez faire, or lazy and not detail-oriented. Without the great techs and supervisors getting behind autovalidation, it would never work."

The new version of the ODBC drivers now being installed will make it possible for Wellons to customize his own Crystal reports. "The pathologists want to know what activities are going on, the director wants to know what it's costing, how many results are validated, how many FTEs are needed, and I'll be able to drill down and make automated reports to send to all the department heads."

With the ease of rule-writing in DI's middleware, he adds, he has been able to partner with a lot of the department heads and get them involved in the process.

Porter Health's LIS has an autoverification/ autovalidation feature, "but it's very straitjacketed," Wellons says. "We have a number of doctors who don't want to be called for alert values on creatinine. So we can write a rule and the result doesn't come up as an alert; instead, it flags it as a dialysis patient."

When an alert does come up, his staff insists it be in bright red on the screen where results are shown. "You can sort the columns so all the alerts and stats can float to the top, so they can get on the important work first, then continue with the rest of the stuff."

So even though Wellons has written rules conservatively, geared to worst-case scenarios and not what happens 99 percent of the time, he has calculated that the laboratory is now autovalidating 79.7 percent of its hematology tests. "And any way you add that up, it's saving a significant amount of time."

Intuitive graphical user interfacing is an added advantage. "Lots of systems are so 'worky,' but this one is intuitive and it makes it a lot faster to train new techs. With the interface and the rules, when a case of flu wipes out three chemistry techs and replacements have to float there to cover, for every tech on every shift-whether they're an MLT just out of school or an MT who hasn't worked in that department in 10 years-the system points out what the rule is and what they should do about it. It makes all my green troops experts."

The net effect of rules and autovalidating has been to reduce turnaround time by about 20 percent, Wellons estimates, even though he believes the lab's TAT to have traditionally been quite low. To monitor TAT and other measures, he is able to use the DI data-collection management tool for "down and dirty" raw data. But with the ODBC capability, "I'm able to cobble reports together in comprehensible language that doesn't look like computerese, so effective managers can take a look at daily, weekly, or monthly measures and see the trends. And the managers want this because they really don't have time to chew through a 3,000-page report. They want a format they can use."

Much greater flexibility and responsiveness have resulted because of the increase in productivity and efficiency, he says. "We are able to pull or push tests to different facilities if their test volume falls below a certain number or we decide we need to pull that test back and centralize it. Then if the ER at our second largest facility wants to pick up a certain test, we give them back the capability of doing that test. But in the meantime, if one of their main instruments goes down, we need to be able to push those tubes to the other facility to have that work continue uninterrupted. So when people run in the door with the samples, they are ready to test."

The cows, pigs, sheep, cats, and dogs tested by the College of Veterinary Medicine at Cornell University, Ithaca, NY, are the unwitting beneficiaries of middleware.

Cornell's veterinary diagnostic laboratory, one of the largest veterinary labs in the country, does large and small animal testing. It acquired middleware, in part, because of the diversity of its specimens. "Obviously we're dealing with multiple species and we see many more bacteria and viruses [than a hospital for humans]," says Tammy Klock, systems manager. "Some diseases we test for are specific for a species; then there are general tests you can do that run on almost all species, such as thyroid testing."

The problem, she says, is the amount of research and testing needed to develop multiple reference ranges. The laboratory also has volume issues. "The number of cases we get is kind of mind-boggling. You might get 1,000 samples from a dairy herd that you have to process, accession, label-it's a huge amount of labor."

On top of that are the substantial potential hazards. For example, because anthrax virus tends to be common in dairy farm soil, Cornell's veterinary laboratory might culture anthrax from a sample one or two times a year. Because of that, she says, "there is actually a lot of Homeland Security money filtered down to veterinary diagnostic laboratories, which can be a kind of first-line of defense."

When Klock arrived at the laboratory in 1996 as a systems manager, the lab was still using a home-grown system for test ordering and results reporting. When she and others started looking into a new information management system, they also looked into middleware. "We weren't keen on having to write 15 different interfaces for each instrument, so we found a middleware solution from Dawning to do all the necessary coding for us."

In fact, bringing everyone up to speed on the new system proved to be far more complicated than getting the middleware to work smoothly. Now, "it's all done invisibly. The system takes a request in one format, Dawning massages that format into what the instrument is looking for, and does the same for the LIS when the result comes back," she says.

Even though the laboratory doesn't use its middleware for autovalidation, Dawning has become critical to the lab, Klock says. "If something were to go down and the requests didn't go to the instrument, you'd have animals over in the hospital getting ready for surgery or in surgery, and the tests wouldn't be getting done. So it's become an indispensable part of operations."

In the case of BryanLGH Medical Center in Lincoln, Neb., which does more than 1 million billable tests per year, "we had the ability to build rules within our LIS, but the functionality was cumbersome," says LIS manager Nathan Shelbourn. "We have a 'roll and scroll' method for online charting of interfaced instruments-an old prompt-based system-but we really wanted to have a front end for users that was a GUI [graphical user interface] screen for charting results. And we wanted an open middleware system so we'd have the same look and feel throughout the lab."

With Data Innovations middleware, the improvements in the user interface have been striking. "We auto-file about 85 percent of automated chemistry results. But everything we stop has some comment color code telling the technologist why it failed, and they can review that. Our rules might display a comment saying to check specimen integrity, or to make a dilution if there is high linearity. It's definitely had a major impact on our efficiency," Shelbourn says. The lab switched vendors when it acquired new chemistry analyzers, and because of autoverification, he says, the technologists had an easier time learning to use the new systems.

The only downside, he's found, is "if you do ad hoc reporting and data mining out of your LIS, some of your data is now in another system, where before you had it lodged all in one place in your LIS. For example, time stamps showing when you get an instrument and result available-you have to get that data from each system and then join it together."

One piece of its middleware that the medical center has not quite implemented is specimen archival. "We're in the process of launching it," Shelbourn says. "The big advantage will be that it tells what's been ordered and resulted. If you go to archive a specimen and there are pending tests, it will say don't put it away yet."

Five years down the line, he expects the middleware will be implemented throughout the laboratory. "There are areas of the laboratory that are not very well suited for it, but we plan to do autoverification on any of the testing that's feasible." Wellons, of Porter Health, insists middleware is as effective as the effort put into it. "It took longer than I expected to pull this off, but I really appreciate the results; it was definitely worth it." He advises other labs considering the use of middleware to have realistic expectations. "It's not going to come straight out of the box and do magic things," he says, "but if you work and partner with the techs and the LIS people, you can create the magic."

Anne Paxton is a writer in Seattle.