Case study: Automations impact on productivity and turnaround time

May 1, 2002

Over the past four years, our laboratory has automated many sample handling steps. This article will

In September of 1996, St. Marys Hospital, a 414-bed acute-care hospital, implemented the first core lab in the province of Quebec, a model based on processes rather than professional disciplines. Our core lab, now running approximately 2,400 samples daily, is composed of a reception and sample processing area, where all preanalytical manipulations are performed; an automation section, where more than 80 percent of all specimens are run, (i.e., general chemistry, hematology and coagulation, and immunoassays); and a manual section for all specimens requiring special manipulations, (e.g. electrophoresis for Hb, protein, IFE, immunology tests, etc.) This physical reorganization, coupled with a well-implemented laboratory management information system (LMIS) performing auto-verification, reflex algorithms, delta checks, and auto-reporting, led to the accomplishment of our main objective: a single and seamless process for processing more than 80 percent of our laboratory test volume with minimal human interactions.1

Automation Phase I

By the third quarter of 1998, St. Marys lab had acquired a new generation front-end automation system (Power Processor, Beckman Coulter Inc., Fullerton, CA). Due to certain limitations at that time, this automation system could only accept one size of tubes and consequently, only specimens for chemistry, immunoassays, and immunology were processed on it. Hematology
(CBC) and coagulation (Pt-Ptt) were still processed from the reception area. Furthermore, electronic accessioning (scanning of the barcode) was still performed manually at the reception area for all specimens. Even with these limitations, we registered considerable improvements in turnaround time and efficiency.2

Preanalytical Automation Phase II

We upgraded our system in May 2000 to accept tubes of all sizes and incorporated a new outlet for noncentrifuged samples. Electronic accessioning is now performed directly from the system to the LMIS. Since our outreach program represents 35 percent of our volume, samples already spun at point of draw automatically bypass the centrifuge and go directly to the proper rack on the outlet. All these improvements led to the elimination of many sample-handling processes, and 90 percent of our blood samples are run through a single process on the preanalytical automation system. Only blood gases and blood bank samples are not processed by the system. 

Results

Preanalytical automation led to drastic changes in habits and tasks performed by our technologists. The samples are drawn in the outpatient clinic and at the bedside. The barcode is placed on the sample at the point of draw, and the samples are placed into the rack for the preanalytical automation system. Our procedure has changed. We used to have to wait for a batch of samples to start processing; now we have a continuous process in which samples are processed immediately. Presently, 90 percent of all blood samples sent to our lab are dropped into the automation system, thus eliminating manual sample accessioning and sorting. A few minutes later, when the process is completed (accessioning, sorting, centrifugation, etc.), the technologist picks up the analyzer racks of already sorted tubes from the outlet and manually loads these racks into the proper analyzer. Since the results of 80 percent of our specimens are auto-verified by our LMIS, once completed, the next step is storage. Our technologists manually validate the results for the remaining 20 percent and also perform the occasional rerun or reflex test (Table 1).

Impact of automation on turnaround time

A two-phase preanalytical implementation process has helped us to better understand the real impact of automation on turnaround time (TAT). During the first phase, preanalytical automation significantly improved the TAT of the overall analytical process. Since approximately 60 percent of the volume (chemistry, immunoassays, and immunology) was redirected to the front end automation system, only 40 percent (mainly hematology and coagulation) was treated in the processing area, leading to a faster process reflected in the TAT for routine and stat hematology. 

When Phase II was implemented, the number of manual steps decreased further. In the case of hematology, the number of steps decreased by 50 percent, from six to three. Once again, the TAT of the overall process significantly improved. Table 2 summarizes the evolution of TAT before and through the different phases of automation. However, a closer look at the variability (SD) may lead us to better realize the importance and the impact of preanalytical automation. Indeed, if we correlate the decrease in the number of manual steps (i.e., automatically performed by the preanalytical automation system) with the variability (SD) of the TAT, we observe the following (Graph 1, see May 2002 issue of Medical Laboratory Observer): 

For chemistry (routine and stat), as the number of manual steps decreases, variability statistically significantly decreases as well. For hematology, during Phase I, even though TAT improved significantly, the variability did not decrease statistically significantly. However, in Phase II, as the number of manual steps decreased, TAT as well as variability statistically decreased significantly Correlation between the decrease of manual steps and variability is well illustrated in stat chemistry (Graph 2, see May 2002 issue of Medical Laboratory Observer).

What about coagulation?
(Pt-Ptt)++

For coagulation, the process is different. Prior to automation, every sample was centrifuged for 15 minutes. When we implemented Phase II, we had the opportunity to include Pt-Ptt into the automated process. However, since our preanalytical system was equipped with only one centrifuge, we would have had to increase the centrifugation time for all samples from four minutes to six minutes. (For Pt-Ptt, studies between 15 minutes manual centrifugation and six minutes automated centrifugation showed good correlation. For chemistry and immunoassays, 10 minutes of manual centrifugation versus four minutes of automated centrifugation showed no clinically significant differences in test results.) Since our daily volume of Pt-Ptts (130) represented not even 10 percent of the total number of specimens to be centrifuged, it was not worth delaying 1,500 specimens for such a small volume. Instead, we opted to introduce a hyper centrifuge (Stat Spin) that centrifuged coagulation samples in two minutes.3 For coagulation, the process went as follows:

Phase I, no change in process. In Phase II (a), all coagulation specimens were dropped into the inlet of the automation system, accessioned, but sorted out from the noncentrifuged outlet to then be processed manually, but still centrifuged for 15 minutes. With this change in process, a small decrease in manual steps slightly improved the TAT (from 36 minutes to 30 minutes) and the variability (nine minutes to seven minutes), but not statistically significant. In Phase II (b), the process is exactly the same as in Phase II (a) with the exception of a change in technology (i.e., centrifugation time decreased from 15 minutes to two minutes). This led to a highly statistically significant improvement in TAT (from 30 minutes to 17 minutes); however, variability remained unchanged.

What does automation really do?

Automation decreases the number of manual steps, but what really happens? At this point, I would like to introduce a new term: transition time. We define transition time as the time elapsed between each step of a process. For example: A technologist brings a rack of tubes for centrifugation. When he/she reaches the centrifuge, he/she realizes that the centrifuge still has six minutes to run from a previous load. During that time, the tubes sit idle and the technologist performs other chores. Does he/she come back exactly six minutes later? Most likely not. Then the next step is to load the centrifuge for a 10-minute spin. Once again, during that time the technologist will perform other chores and probably wont come back in exactly 10 minutes. One of my colleagues did some time studies on the post-centrifuge issue, and the transition time varied from two minutes to nine minutes per load. This is where automation performs best. In our automation system, from the time the tube is accessed (in the inlet) to the time it is loaded into the centrifuge, it takes approximately one minute. As for the post-centrifuge issue, as soon as the centrifuge stops, tubes are unloaded, then decapped always with the same reliable and reproducible transition time. This explains the drastic decrease in the variability of the TAT which led to fewer outliers, fewer phone calls to the lab (from more than 25 calls per day to less than five calls per day), and a better service to our physicians as reflected in the improvement of the TAT at the 90th percentile (Table 2).

Graph 1 clearly shows that unless the number of manual steps is decreased, i.e., the time of the transition steps through automation is stabilized, the variability will not improve. We believe that our automation system improved our TAT at the level at which it should be addressed: decreasing variability. To monitor the variability of the process, we observe the TAT for 90 percent of the samples so we can monitor the longest times to the physicians. 

Impact on productivity

It is very difficult to adequately benchmark laboratory productivity based on cost efficiency since we all fulfill different needs, operate in different environments, and account for expenses differently. Furthermore, being in different countries where medical and laboratory practices are different and where currency fluctuation can further bias the results adds to the complexity of proper benchmarking. Being part of a cross-laboratory comparison program and looking at the production of units per worked hour remains a fairly good indicator of productivity. St. Marys Hospital laboratories, accredited by the College of American Pathologists (CAP), are using the CAP Laboratory Management Index Program (LMIP), as well as the workload measurement system in place in the province of Quebec using a weighted unit (Graph 3, see May 2002 issue of Medical Laboratory Observer). Both programs, though quite different, demonstrate very similar results. By LMIP and Quebec standards, automation improved our productivity by 30 percent and 26 percent respectively. In both cases, our labs productivity is better than 95 percent of our peers in the studies.

As economic factors continue to drive the healthcare industry, for the past three years, our laboratory maintained excellent records in productivity and is regarded by many as one of the most efficient laboratories in North America. Automation played a crucial role, not only financially, but also in providing better service to our community such an important factor when we all strive to increase our market share. While staffing remains a problem, we have not been faced with this issue. Performing less mundane tasks led to higher job satisfaction and lower turnover rates. Cross-training of our staff coupled with automation turned out to be not only a good retention strategy, but has also placed us in the advantageous position of being able to efficiently manage a large amount of work with minimal staffing.

Our next step is to add an aliquotter and connect our chemistry and immunoassay equipment (~60 percent of our volume of specimens) directly to our preanalytical system. By progressively automating more steps in the process, our ultimate goal remains unchanged: to improve our quality while remaining one of the most efficient laboratories. 

Ralph Dadoun is vice president, corporate and support services, St. Marys Hospital Center, Montreal.

References

1. Dadoun R. Impact on Human Resources: Core Lab vs. Modular Robotics. Clinical Laboratory Mgmt Review (Journal of
CLMA). July/August 1998; 12(4): 248-255.

2. Dadoun R. Implementing Pre-analytical Automation: The right volume, the right workflow. Medical Laboratory Observer
(MLO). January 2000;23(1):32-36.

3. Nelson, S. et al. Rapid Preparation of Plasma for Stat Coagulation Testing. Archives Pathol Lab Med 1994; 118: 175-176.

© 2002 Nelson Publishing, Inc. All rights reserved.

Photo 115879330 © Karenr | Dreamstime.com
Photo 124723736 © Steveheap | Dreamstime.com
Photo 129549884 © Katarzyna Bialasiewicz | Dreamstime.com
Photo 159136554 © Alberto Masnovo | Dreamstime.com