Feature Story

Believe it or not— do-it-yourself lab automation

August 2002
Karen Titus

Yes, it’s true that William Neeley, MD, has discovered an inexpensive way to automate his laboratory. Indeed, despite a nearly nonexistent capital equipment budget, he’s managing to streamline the pre- and postanalytical phases as well as the analytical process into one seamless workflow at Detroit Medical Center, where he’s the medical director of laboratories. He’s pared down sample splitting, scaled new heights of efficiency, and connected with his LIS in a newfangled way.

When we tell you how he did it, you’re going to want to toss this article aside. But we’re going to tell you anyway:

Dr. Neeley, with the help of a few equally clever companions and the backing of his chairman and lab operations VP, built his own automated system. He designed his own parts using AutoCAD—he milled them and drilled them and turned them on his own lathe, in his own workshop. He and his colleagues designed and wrote their own software.

Have we lost you yet?

It’s not so hard for Dr. Neeley. After all, his master’s degree in pathology consisted of extensive coursework in electrical engineering. He could also draw on his experience from the years he spent in Silicon Valley, designing automated systems and writing software. Not only does he know what he needs in the lab, he knows how to put it there.

Still with us? Good. Because even if you can’t replicate what Dr. Neeley’s doing—indeed, he doesn’t think you should—you may still learn a thing or two about how automation should work (but probably doesn’t) in your own laboratory.

Dr. Neeley, who spoke on low-cost, do-it-yourself lab automation at the Dark Report’s 2002 Executive War College in May, is the first to admit his approach is a tad funky. "Hang on to your chairs," he said as he opened the first of his two talks on the subject. "This is a very, very different approach. What we do, basically, just isn’t done in most laboratories."

Part visionary, part cheerleader, part eccentric inventor—and, of course, all clinical pathologist—Dr. Neeley speaks of his work with a charmingly frank passion. He unabashedly declares that he loves lab automation; in the next breath, he’ll castigate it for its perceived failures. And in between, he spins an absorbing tale, blending how-to advice with philosophical riffs on the true meaning of automation to explain how he and his colleagues at DMC are bringing lab automation into the 21st century.

Dr. Neeley started his tenure at DMC nearly a year-and-a-half ago. Going in, he knew what he had to work with: virtually no money for new capital equipment and an LIS that’s off-limits to the lab. It’s managed by an IT department that oversees the entire medical center—six hospitals with an overall capacity of 1,400-plus beds. From IT’s standpoint, "the lab is only a small cog in a big wheel," said Dr. Neeley, though it’s easy to argue that the laboratory, which turns around some 8.5 million tests a year, is hardly a small cog. Indeed, among the many entities it supports are an acute care hospital, a children’s hospital, a rehabilitation institute, and a large bone marrow transplant program. Only 0.8 percent of its tests are send-outs. "So we have to really tap our resources, and we have to work a little harder to be creative, to figure out how to create a modern laboratory under these very, very challenging conditions," he told his War College audience.

They also have to battle entrenched ideas about what constitutes automation.

"Total lab automation is a misnomer," Dr. Neeley insisted. Most people see it as a complete system, but it’s turned out instead to be a simplistic, expensive sorting and delivery method. "There are some significant features missing," he said. No surprise, then, that it hasn’t delivered the expected labor savings, shorter turnaround times, and reasonable ROIs. "Overall, total lab automation has been a pretty big disappointment in the United States." Chief among the failings, he contended, are the underuse of autoverification (and its lack of sophistication) in postanalytical systems, and the absence of systems integration between the preanalytical, analytical, and postanalytical phases.

He also suggested he’s not alone in his disappointment, recalling a conversation he had earlier this year with the COO of a company that Dr. Neeley declined to name, but is, as he put it, one of the two companies that do more than $1 billion worth of lab reference business annually. The COO told Dr. Neeley he had "absolutely no use" for lab automation. "I was very surprised," Dr. Neeley said. "Because here’s a man who’s risen through the ranks, who’s become COO of a billion-dollar corporation, and he’s completely turned off by lab automation."

"This raises a red flag," he continued. "We need to look at this, and we need to ask, Why is someone with his knowledge and ability and closeness to the lab industry so negative about lab automation?"

High cost, early failures, and lack of systems integration are among the main reasons for automation’s limited success, he said. At the same time, laboratories need to be more discerning when purchasing automated systems.

"Right now, vendors are controlling the process," he said. "Laboratorians have not taken an active interest and tried to understand [automation] in depth." Instead, he says, they’re trying to choose from an abundance of automated offerings with only a vague sense of what they truly need. "Each one of you has to ask, In my laboratory, if I spend $100,000 on this component, is it really going to yield $100,000-plus in good to my laboratory and to my productivity? What’s best for my laboratory?"

In Dr. Neeley’s case, asking such questions has given rise to unusual, do-it-yourself solutions.

He first plunged in eight years ago when he was working at a laboratory in California, where he introduced sample tracking. The lab’s programmers tweaked the LIS so the lab staff could, among other things, keep a complete record of every patient sample as it worked its way through the laboratory. "You could tell, in real-time, where Mrs. Smith’s lavender tube was located," Dr. Neeley said. If it was located in hematology, they could discern everywhere it had been until its arrival in hematology. Or, if it had completed its trip through the laboratory, they could locate it in storage.

He and his colleagues also built a track and, in keeping with accepted practices, began splitting samples up front. "Everybody knows that the first thing you do when you walk into the laboratory is to split your samples four or five ways," he said. "So that’s what we did." The result: a backup of samples.

Then they took a closer look.

Since they were intimately involved in creating the automated system, they enjoyed two luxuries most labs don’t have: the freedom to think up new ideas and the freedom to act on them. Said Dr. Neeley: "We went in and said, ’This doesn’t make any sense. What if we don’t split all samples at the beginning?’"

Instead, they looked at splitting samples later—after the primary tubes had already been to the chemistry and immunochemistry analyzers. Perhaps, they theorized, this would be a better place to begin splitting samples. "We suddenly realized that once our track had delivered samples to our large, random-access analyzers, 80 percent of the samples didn’t need to be split." Since their software tracked samples so closely, they discovered another bonus: They didn’t necessarily need to split samples even after a primary tube was delivered to a third workstation, after leaving chemistry and immunochemistry. "If we had to repeat something or a physician happened to add a test, this wasn’t a problem, because we knew exactly where that primary tube was, in real-time," Dr. Neeley said. "Then, if it had to go to two additional places, we’d split it there." As a result, he said, sample splitting dropped tremendously. So did labor, turnaround time, and errors.

We know what you’re thinking. Bully for Dr. Neeley. But we can’t write our own software or configure our own aliquoter.

That’s OK. Dr. Neeley doesn’t want you to. He just wants to mess with your mind a little, persuade you to ask yourself a few more questions whenever you tackle automation in your own lab.

"Do you need a $100,000 to $200,000 sample splitter up front?" he asked. "Or, if you are going to put in a sample splitter, do you put it on the back end of your automation? You know what my answer would be." That brings up another question: If you’re splitting so few samples, do you need an expensive aliquoter? Says Dr. Neeley: "I would like you to take a different look, take your glasses off, clean them, and open your minds up and be willing to look at different ways of doing things."

<Dr. Neeley brought that spirit with him when he arrived in Detroit and began studying the laboratory’s current setup.

"As you can tell, sample flow and processing, to me, is one of the last frontiers. It’s an area that’s wide open," he said. With that in mind, he and his colleagues began attacking the lab’s problems on several fronts. No one solution would do the trick.

The LIS was ripe for revival, Dr. Neeley noted. He has no beef with the vendor of the system, he was quick to say. But he also has no access to the lab information software. "Everything is subcontracted to a for-profit corporation. Everything is managed by third-party individuals. They manage the hospital information system, and they manage the LIS."

Undaunted, he and his colleagues are building a track system that will be able to handle stat and routine samples. They’re interfacing it to the LIS, using a technique Dr. Neeley developed years ago that doesn’t require software upgrades or modifications to the LIS. And they’re integrating the preanalytical, analytical, and postanalytical phases. "You can’t have a weak link in the chain," he said. The best preanalytical system in the world is useless if it’s mismatched with the analytical instrumentation, or if the postanalytical phase is ignored. But if all three components are well balanced, "then you can get efficiency you wouldn’t believe."

At DMC, as in many labs, the laboratory setup is awkward. "Processing is located in one area. Far away from it is chemistry, and farther away from that is hematology. It’s a totally inefficient design," Dr. Neeley said.

He prefers the "funnel" concept: The highest-volume samples and the ones with the greatest urgency should move the shortest distance. "We call that a BFO—a blinding flash of the obvious," he joked. "But how many other labs are set up like this, where the chemistry is far away, so you’re moving massive numbers of samples long distances?"

A laboratory reorganization ensued, with chemistry repositioned next to processing, and hematology to follow.

Dr. Neeley and his colleagues are installing the track themselves. ("Wear leather gloves if you do this yourself," he advised.) The track is made by the Swedish manufacturer FlexLink AB, which provides many of the tracks used in Silicon Valley clean rooms, Dr. Neeley noted. "It’s self-lubricating, requires no maintenance, and is very quiet." Seven instruments sample directly from the track, and there are eight additional transfer stations. The complete track cost less than $50,000.

As for other costs: "I had to pay big-city rates to take down some walls so we’d have room for our instruments and track," he said. "And I had to pay electricians to come in and put in new wiring for the instruments." Those two outlays came to $40,000.

You can’t help it, can you? You want to say, Right. I’ll just set up a power saw next to my immuno-assay analyzer. Uh-huh.

It would be easy to dismiss Dr. Neeley’s efforts as interesting yet impractical, especially when he says things like, "I couldn’t live without my band saw, because I need it to rough cut my metal," or when he proclaims, "You need to know how to use a lathe and a mill." But that might be a mistake. His approach is both more simple and more complex than what most laboratorians would pursue; it’s likely some useful ideas lie between the two extremes.

Here, for example, is Dr. Neeley’s take on the postanalytic phase.

At DMC, Dr. Neeley and his colleagues have built connections between the LIS and the various instruments in the laboratory to "eavesdrop" on the exchange of data between the LIS and the instrumentation. He also uses a self-designed program to access the data flowing between the LIS and terminals throughout the lab. This enables him to access test information without tampering with the LIS.

"Now, what we do is different than anybody else," he said. "We have our set of algorithms, and we capture the results. Our algorithms aren’t limited by the programming limits of the LIS vendor. When we automatically sign on the LIS, we look at the patient’s files and see what results are there. Then we directly compare what’s in the file versus what came in from the instrument." That’s what any good technologist does, Dr. Neeley noted. At DMC, it’s done automatically.

The system also automatically notes discrepancies between results as well as results that violate the lab’s rules or algorithms. It prints the unacceptable results, identifies the problem, and, below that, tells the technologist what the relevant laboratory policy is for addressing the problem. "We call that real-time quality assurance," Dr. Neeley says. "Your 1,800th sample gets the same level of scrutiny as your very first sample. And the laboratory’s response to a given problem is exactly the same each time."

By adjusting the software and the instrument operation, the lab is now also able to report results from its AxSym analyzer directly back to the LIS. The intelligent system captures the results and releases those that pass a high level of scrutiny, including delta checking—about 97 percent fall into that category. The system is to be used throughout the lab.

The benefits are immeasurable. Technologists are too valuable to spend their time moving samples or releasing results in bureaucratic, mind-numbing fashion, Dr. Neeley says. With the intelligent system, they are now free to focus on only the most serious problems—a far better use of their time and skills.

Dr. Neeley continues to fine-tune the system. Following up on a request from one member of the professional staff, the software automatically identifies hCGs greater than 5 on male patients. "We capture the hCG value, the accession number, and the name and gender of the patient, and we automatically notify the doctor via e-mail. No tech has to lift a finger," he says.

Similar opportunities abound in a personalized LIS, he says. At another institution, he programmed the system to flag potassiums that might be falsely elevated due to hemolysis of the sample. In another case, he tweaked the software to identify cases of falsely low cholesterols due to the presence of large amounts of ascorbic acid. "What we look for are ridiculously low uric acids in combination with ridiculously low cholesterols," he explained. In a third instance, the lab was able to pinpoint high glucoses, normal BUNs, and creatinine levels of 4 or 5 (by the Jaffe reaction). "Guess what?" he said. "Diabetic ketoacidosis. We’d spot that one every time. Then we’d repeat the creatinine by enzymatic means and send back a normal creatinine level, instead of sending back a 4 or 5 to the doctor and expecting them to sort it out."

"This is modern laboratory medicine," he continued. "This is what we were trained to do. This is being good chemists and good laboratorians. But typical systems don’t allow us to do things like this."

That’s because in the past 10 to 15 years those in the laboratory "have turned our backs on automation," he said. "As a result, other people stepped in. We’ve also turned our backs on our LIS systems, and nonlaboratorians have stepped in to fill the gap. As a result, we don’t control our own destiny—65 percent of the modern laboratory is IT and automation. So most of us have no control over 65 percent of the most critical parts of our entire business."

Dr. Neeley, obviously, is a man who refuses to surrender to technology, no matter how limiting it may be. But no laboratorian, even those without his specialized skills and knowledge, can afford to give in without a fight, he said.

"I’m not trying to sell you on the idea that this is the only way to do automation," he said. "This is the best way I figured out how to achieve a cutting-edge lab in the 21st century. Is it for everyone? No, it’s not for everyone."

But, he argues, improved automation is for everyone. The technologist shortage will "force everyone into automation or out of business," he insists. "If we’re going to be competitive, if we want to control our own destinies, we’re going to have to have input into these processes. If we wait on others to do this for us, we’re not getting all we can out of our laboratories."

Karen Titus is CAP TODAY contributing editor and co-managing editor.