How to Implement Automation in the Lab

Laboratory leaders should know their customers, their workflow, and their goals before beginning to automate

Dr. Charles Hawker has retired as scientific director for automation and special projects at ARUP, where he was employed for 26 years. Dr. Hawker was also professor (adjunct) of pathology in the University of Utah, School of Medicine. He is a past president of the Association of Clinical Scientists and the National Academy of Clinical Biochemistry (NACB). He has received numerous awards including the AACC’s highest award for Outstanding Lifetime Contributions to Clinical Chemistry and Laboratory Medicine. He is the author or co-author of 48 reviewed published papers, 49 abstracts, and 17 invited reviews or book chapters including chapters on automation in three editions of the Tietz Textbook of Clinical Chemistry and Molecular Diagnostics and three editions of the Tietz Fundamentals of Clinical Chemistry and Molecular Diagnostics. At ARUP, he installed several major automation and robotic systems that have made ARUP one of the country’s most automated laboratories and the first US clinical laboratory to achieve Six Sigma quality in any metric.


A: For me, the most important challenge is that the lab’s leadership team understands who their customer is and what their workflow really looks like.

If you’re going to do automation, you’re not automating tests. Tests are already automated on analyzers. What you’re automating is the handling of specimens; you’re trying to replace human labor—handling, sorting, and distributing hundreds of thousands of specimens—to gain some efficiencies and improvements in quality. So, it’s really important that the laboratory understands their workflow—what the peak hours of the day are and how many specimens they get during those peak hours, and then what it looks like over the other hours of the day. In other words, map it over a 24-hour cycle and look at where the specimens go, who touches them, how many times a specimen is handled, and all the different steps to specimen handling. Only when you’ve done that will you really understand what automation might do for you. You might look at that and say you need automation that includes aliquoting or you may [decide] that you don’t need aliquoting because of the nature of the analyzers you’re using. You really have to understand your whole workflow to understand what you’re going to do in terms of automation.

Laboratory leadership teams wanting to implement automation must also do a detailed workflow analysis. There are published methods out there for doing this. With the workflow analysis, you’ll identify what all the problems are in your operation and where you have opportunities to make improvements. You’ll also find out where the bottlenecks are and where you really need to focus the automation.


A: Except for the very smallest labs, the answer is yes. I’m aware of some laboratories that have automated and they’re handling fewer than 1,000 specimens a day, which is a pretty low volume, especially when you spread that volume out over 24 hours. But in the cases that I’ve heard of laboratories doing that it’s because they’re in a very tight labor market, such as a small- to medium-sized community where they’re the only hospital, and they simply can’t find medical technologists even when they’ve tried to recruit throughout their state or their region. If they’ve automated, then they’ve reduced the labor requirements and they can operate that laboratory with fewer staff. I think with the exception of some really small labs, most labs are doing automation or they’re going to need to do some automation if they haven’t yet.


A: Six Sigma is a quality improvement process—a quality improvement strategy if you will—but it’s more than a strategy. There is a quality improvement process that started with Toyota called Lean around 40 years ago, but it wasn’t adopted by many American companies. In the 80’s and 90’s when Americans started waking up and realizing the Japanese cars were much higher quality than the American cars, then the American manufacturers started implementing quality improvement processes such as Lean.

Six Sigma started a little later than Lean but it’s a parallel improvement process that is more statistically based, whereas Lean is a more intuitive improvement process. The definition of Six Sigma is less than or equal to 3.4 errors or defects per million opportunities, where an opportunity is a handling step. For most industries, Six Sigma is the ultimate gold standard because making only 3.4 errors or fewer per million opportunities is extremely hard to do. The typical clinical laboratory operates at roughly four to four and a half Sigma, and some very good clinical labs will get a little bit above Five Sigma. Five Sigma is less than or equal to 210 errors per million opportunities; if I want to improve from Five Sigma to Six Sigma, it’s a 16 and two thirds-fold improvement in error rate. If the laboratory’s already working really hard and is an above average lab and they’re at Five Sigma, now to go to the next level of Six Sigma is mind boggling—it’s almost impossible.

Six Sigma applies to analytical quality as well as pre-analytical or post-analytical. In the analytical area, you might look at having issued wrong reports on a test result, that would probably be the principal analytical error, but there are some others. In the case of pre-analytical errors, you might look at misidentified specimens, mislabeled specimens, and lost specimens. You might even go way back in the analytical chain to physicians ordering the wrong tests. There are a whole bunch of different parameters you could consider measuring, it just depends on where your emphasis is on the quality that you’re trying to improve.

In the lab where I worked for 26 years, which was ARUP Laboratories Salt Lake City, we did achieve Six Sigma quality for one metric, which was lost specimens. We were the first clinical laboratory in the US to achieve Six Sigma quality for anything, and it was only because we had spent more than 20 years at improving processes and then we put automation on top of that. We’re really quite proud of it because it took a heck of a long time and work by a lot of people and that’s because Six Sigma is really hard to do. If somebody loses one tube, you’ve got to not lose the next three million you handle.


A: More and more, vendors are using different variants on something I refer to as machine vision. That is to say, they’re using cameras with the automation to replace the human eye in looking at specimens and looking at the processes to make sure that things are being done correctly. A very basic application is to look at the tube and see if it’s the correct tube for the test. In your LIS, you’ll have the specimen requirement for this test and it may call for a green-topped tube and you simply look at that tube with a camera and the camera will tell you if it’s the tube it’s expecting according to your LIS. If it’s not, the system can set that tube aside into a special lane where somebody has to now investigate it and correct the error or override it. But if you’re depending on people to see if they’ve got the correct tube for every test, sometimes somebody’s going to make a mistake. By using the computer and an automation system and cameras, you can greatly reduce the human errors and just have people looking at problems.

Now it’s going to the next level—there are automation systems that some vendors have where they look at the tube and they can see the top of the packed cells and the serum. Based on the diameter of the tube, they can determine the volume of the serum that they have above the packed cells and use that information to guide a pipette tip that’s going to pipette off that serum and put it into a transfer tube, or guide the aspiration tube that’s going to take a portion of the specimen for a test on an analyzer.

You can also use cameras and even more sophisticated technology to determine whether the specimen is hemolyzed, or if it’s icteric and it’s got the yellowish-green color, or if it’s got excess lipids and cloudiness, in which case you’d need to do an ultracentrifuge on the specimen. I think we’re going to see more of this general technology, what I call machine vision, in which cameras are combined with automation to replace the human eye in doing various inspections that are important to quality.


A: The laboratory has to understand its business, who its customers are, and what its customers’ needs are. If the lab is doing outreach work and serving physicians’ offices, then they have a group of customers who have a certain expectation that they can collect their specimens during the day as patients are visiting the doctor’s offices. Doctors are going to want to have the results the next morning for all the lab work they ordered the day before, which creates a set of objectives or goals for the laboratory and the automation so that work can be run at night and be ready for the doctors the next morning. That’s a simple example, but understanding who your customers are and what [their] needs are is important to putting this all together.

Another tip is when you decide you’re going to do some automation, it’s typical to call all the different vendors and start looking at different systems and narrow it down to two or three or maybe even one. When you talk to these vendors, they will do the workflow analysis for you and they’ll come up with the design of an automation system that they think will meet your workflow. But, I am a strong proponent of trying to do your own workflow analysis independently because if you simply accept what the vendors come up with for you, how can you be sure they’re correct? They obviously are in the business of trying to sell stuff, so if you’re taking their word for it with the workflow analysis, they may be recommending something that you don’t need or they may overlook something you do need. The only way to avoid that is to do your own workflow analysis.

I also recommend visiting labs hat have the automation from that vendor. The vendor may have customers that have had very successful implementation of their automation and those are going to be the ones they’re going to want to take you to visit. If you have the opportunity to talk to the people at that laboratory without the vendor present, you may be able to ask them some frank questions about what problems they ran into with their automation. Or, if possible, try and find some laboratories that had some implementations of the automation that were problematic; find out what problems these people encountered when they were trying to do it, because you don’t want to repeat those problems.

The more you can learn, the better off you’re going be. Even if you are choosing that vendor and there’s no question that’s going to be the right system for you, still it may not be all roses when the system comes in and you’re doing the implementation. You’d like to know ahead of time what some of the problems are so you can come up with a good implementation plan with contingencies. Unless your hospital is building you a whole new facility and you can put in the new system while your old analyzers and systems are still running, which is the exception, you’re going to have some disruption and some analyzers are going to have to be moved or shut down for a while if you’re replacing some old automation with some new automation. You’re going to have to come up with a plan of how you’re going to deliver the service to your customers while all this implementation is going on. So it’s good to talk to other people that have had that experience in order to know how to do that.