| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Articles |
1
Medical Automation Research Center, Box 800168, University of Virginia Health System, Charlottesville, VA 22908.
a Address correspondence to this author at: The University of Virginia, Department of Pathology, Box 800168, Medical Automation Research Center (MARC), Charlottesville, VA 22908. Fax 804-924-5718; e-mail rfelder{at}virginia.edu
 |
Abstract |
|---|
 Top Abstract IntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Methods: Timing studies were used to quantify the efficiency of the manual processes and to identify areas in the processing of coagulation specimens where bottlenecks and long waiting periods were encountered. We then designed our modular robotic system to eliminate these bottlenecks. Our robotic modular workcell was engineered to allow a choice of specimen introduction manually, by conveyor, or by mobile robot. Additional timing studies were performed during clinical trials of the robotic system.
Results: Prior to automation, the time required for preanalytical processing time was 18–107 min; after automation, it was 45–50 min. Additional improvements in workcell efficiency could be realized when high quality, prelabeled specimens were introduced into the system.
Conclusion: Compared with manual methods, modular automation provides more predictable variation in specimen processing.
|
Introduction |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Coagulation analysis is used to elucidate defects in biochemical pathways associated with hemostasis. Automated coagulation analysis usually requires the availability of specimens that have been centrifuged and sorted. Centrifugation of specimens, sorting, and transportation to the analyzer are essential preanalytical tasks that need to be addressed by laboratory automation with a potential for substantial labor savings and quality improvement.
The justification for the purchase of laboratory automation must be based on demonstrable improvements in efficiency and calculated payback. However, there are few data quantifying process improvement after the installation of laboratory automation using robotics. Therefore, we performed a study to determine the throughput and turnaround times for specimen processing of coagulation specimens in a clinical laboratory before and after the installation of a robotic system developed in our laboratory.
|
Materials and Methods |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Coagulation testing is performed on one of two MDS 180 analyzers, commercially available instruments (Organon Teknika, Durham, NC). Instrument use was alternated on a daily basis. The second analyzer also served as a back-up instrument in case of instrument failure and to provide backup when the primary analyzer was undergoing routine maintenance. Coagulation processing was subdivided into several discrete steps. Each step was timed by watching a time-stamped videotape. A Sony (model CCD-TRV70) 8-mm video camera was mounted in the clinical laboratory to record all of the processing steps except precentrifugation. The camera had a time stamp, in seconds, that was recorded on the tape. The camera was placed at locations in the laboratory that allowed the reviewer simultaneous views of several processing steps.
Based on the data obtained from our time and motion studies, we designed a preliminary version analytical processing robot (10). In developing the prototype, we determined that the essential components of an automated workcell (automated device dedicated to a single analytical area) should include a specimen storage area, a centrifugation device, a robot for specimen manipulation, and an output storage area. Both the input and output storage areas were designed to buffer the influx and output of specimens during the time that technologists were not available to add or remove specimens. To maximize the flexibility of the automated workcell, we designed the automated workcell to accept a variety of specimen input mechanisms. Specimens may be presented individually (e.g., for stat samples), in racks by a technologist, via conveyor as part of a TLA system, or via a mobile robot. Automated centrifuge loading and unloading addressed the analytical time bottleneck that we observed in the manual time and motion studies. Automated specimen bar-code scanning and specimen loading into the instrument streamlined the mundane task of assuring that the specimen bar code was properly aligned in the coagulation analyzer.
We present here the results of our examination of two recent versions
of our preanalytical system, named coagAutoLink, interfaced to either
one or two MDA180 instruments (Organon Teknika; Fig. 1
). We designed the system to include a CRS A255 anthropomorphic
robot arm, a CRS C500C controller, which runs the RAPL-3 language (CRS,
Burlington, Ontario, Canada), a modified Jouan C422 (Jouan, Nantes,
France), and a workcell enclosure. The enclosure was designed to
restrict human access to the robot’s workspace, both to protect the
operator and to prevent objects from being accidentally moved or placed
into the robot’s environment. The enclosure had two electrically
operated door locks, which prevented the robot from moving when the
operator was accessing the sample racks. During our clinical trials,
these safety measures were successful in ensuring operator safety.
|
The first version of the automated workcell, coagAutoLink SPIN, operated with a single MDA180 and a centrifuge. The second version, coagAutoLink 2CST, mechanically and electronically interfaced to dual MDA180s and a Labotix specimen transportation conveyor (Labotix). The coagAutoLink 2CST does not include a centrifuge because specimens are separated before arrival at the coagulation station by the TLA system.
In addition, we implemented and tested delivery of specimens using a mobile robot (RoboCart model 2110; CCRI, Lake Arrowhead, CA). The RoboCart is an autonomous guided vehicle that follows a guide path consisting of reflective tape adhered to the floor. Specimens were contained in input racks that were manually placed on the top of the RoboCart. The RoboCart was then programmed to follow a reflective tape path directly into a specially designed docking station on the coagAutoLink. The arrival of the robot was announced by a sensor that prompted to the CRS robot arm to automatically unload the full specimen racks. Racks of tubes that had been analyzed previously were loaded onto the RoboCart for delivery to the specimen storage area of the laboratory.
For development and optimization of the control code for the coagAutoLink 2CST, we developed a discrete-event model of the laboratory sample flow, single and dual MDA180 behavior, and conveyor and robot activities. The simulation was developed and validated according to methods used previously in our laboratory (11). This discrete-event model was used to simulate a variety of design concepts and to maximize system throughput. After the simulation showed the desired performance characteristics, the actual robot control software was generated directly from the simulated robot control software.
The timing and throughput studies were designed to determine the performance of coagAutoLink and to compare its performance with a manual process. Before installation of the coagAutoLink, we identified each of the steps that are performed for manual processing of coagulation samples at the University of Virginia Health System Core Laboratory. These steps included: (a) receiving-centrifugation-loading (sample reception, including accessioning; labeling; transportation to the centrifuge; and loading of the centrifuge); (b) centrifugation-idle (includes idle time after the centrifugation cycle has completed); (c) sample transport to MDA (loading of spun samples into MDA racks and taking the racks to the coagulation testing area); and (d) sample placement into the MDA180 (includes time the sample waits in the coagulation testing area for medical technologist to load it into the MDA). Each of these steps was observed, and timing measurements were taken.
|
Results |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
). The next-largest variation of time was during centrifugation.
All specimens were spun for a fixed time of 10 min, but the specimens
were rarely removed from the centrifuge as soon as the centrifuge
stopped because technologists were occupied with other tasks.
Processing times at this step ranged from 10 min (when specimens were
removed immediately from the centrifuge) to 28 min (18 min spent idle
in the centrifuge; mean, 15.8 ± 5.5 min; n = 27).
|
Routine coagulation specimens were removed from the centrifuge and placed into a MDA rack by a processing clerk. Racks then had to be placed on the MDA instrument for analysis by a medical technologist. These racks waited a range of 1 s to 11 min for loading by the medical technologist (mean, 1.9 ± 3.1 min; n = 28). One day a week a large volume of specimens was sent to the laboratory from the dialysis unit. Because of specimen queues on dialysis days, the largest variation of time was encountered in the pre-centrifuge processing step. The minimum time on those days with a high number of specimens was 38.5 min, and the maximum was 106.0 min (mean, 81.2 ± 18.1 min; n = 44). The average total processing time on a typical day was 48.6 min, and on a busy day it was 100.3 min.
Stat processing of coagulation specimens had a much faster throughput but showed similar bottlenecks. The average receiving/centrifuge loading time was 3.2 ± 3.7 min (n = 29). Processing times during centrifugation varied from 10.0 min when specimens were removed from the centrifuge immediately to 23.7 min (mean, 12.9 ± 3.4 min; n = 55). The specimens are placed in the centrifuge by processing clerks but are removed by the medical technologist in the coagulation area. Even with a stat specimen, the removal from the centrifuge was sometimes delayed for >13 min after centrifugation was complete. Once the stat specimens had been removed from the centrifuge, final processing usually took <1 min to check for problems, load into MDA sample racks, and be placed on the analyzer.
As shown in Fig. 2
, large variations existed in the amount of time
spent in the manual process. In particular, the receiving step varied
from 8 min to >1 h. Clearly, reducing process time variability was a
reasonable goal of coagAutoLink, which was subsequently installed in
the University of Virginia Core Clinical Laboratory.
timing studies after automation
After installation of our coagAutoLink automation module, we
repeated our timing studies. Three timelines in Fig. 3
summarize the data from timing measurements made after the
installation of coagAutoLink. Fig. 3
depicts the "average" times
for steps in the manual process so that it may be compared with the
average time in the automated process. The coagAutoLink processing and
centrifugation steps are shorter than their manual counterparts, and
have predictable ranges. The process still required manual sample
receiving and labeling. Each of the timelines is labeled to indicate
the time required to aspirate the specimen, i.e., the time that the MDA
actually begins analysis. The average times to complete the manual
process, the process performed with coagAutoLink using manually labeled
samples, and fully automated coagAutoLink (prelabeled samples) were 50,
48.5, and 26.5 min, respectively.
|
In Fig. 4
, four scenarios are compared: manually processed routine
samples, manually processed stat samples, coagAutoLink SPIN as observed
after installation, and projected values for coagAutoLink SPIN when
prelabeled samples are used. Our data demonstrate that manual processes
have a larger variation in processing time compared with the automated
process. In the manual process, the longest wait periods were
experienced during the removal of specimens from the centrifuge. In the
routine area, the mean wait time from when the centrifuge stopped until
the specimens were removed was 5.8 min, and in the stat coagulation
centrifuge area, the mean wait time was 2.9 min. In both the stat and
routine centrifuge areas, efficiency was lost when specimens waited for
the next processing step to begin. Often technologists were preoccupied
with other more pressing tasks and thus were not available at the
precise time the centrifuge was finished. Preanalytical systems such as
coagAutoLink reduce sample idle time, thereby improving overall
throughput.
|
We measured the throughput of coagAutoLink SPIN. The measured maximum throughput of the coagAutoLink SPIN is 99 samples/h. Thus, the throughput of the coagAutoLink exceeds the analytical throughput (assuming two assays per sample) and will not present a throughput bottleneck. In separate experiments, prototype system testing of coagAutoLink 2CST has shown that sustained throughput will be ~135 samples/h. This increased throughput is a result of the lack of need for a centrifuge in the 2CST model.
|
Discussion |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
We demonstrated in clinical trials of our automated preanalytical device that the automation was sufficiently efficient to keep pace with the analytical hardware. In other words, we matched our automation to the clinical need. The major benefit of the automated preanalytical device was a reduction in the variability in turnaround time for coagulation specimens. In a similar clinical trial, we demonstrated that centrifugation alone was able to reduce the turnaround time and labor required for preanalytical processing (12). When additional components were added to the preanalytical processing (e.g., sample stocking, aliquoting, and sorting), the process became even more efficient. Versatile preanalytical processors that can accommodate a wide variety of specimen sizes and shapes are becoming commercially available (13). There are no reports that describe the return on investment for modular preanalytical automation. This study focused on the ability of our system to reduce the time associated with preanalytical tasks. However, it did not address the issues related specifically to return on investment.
The efficiency of any robotic system, including the coagAutoLink, may
be improved by providing specimens that are correctly filled, labeled,
and transported. Our timing studies determined that substantial delays
were experienced in dealing with improperly labeled specimens. During
one 10-h period, 32 of 193 specimens (17%) were mislabeled or had
incorrectly applied labels and required 
1 min for corrective measures
by a technologist. Preanalytical variables can also cause significant
deviations from expected coagulation results. For example, improper
tube filling can render some specimens virtually unusable for routine
coagulation analysis. Underfilling of tubes will produce a larger
specimen-to-anticoagulant (3.2% citrate) ratio and can cause prolonged
prothrombin time and activated partial thromboplastin time.
Siegel et al. (14) found that partially filled heparinized
coagulation specimens produced shortened activated partial
thromboplastin times. Short-draw coagulation tubes that have their
original labels obscured by multiple institutional labels are
particularly problematic. It is necessary to remove the labels before
one can determine whether the tube has been designed for a short draw
or is simply under filled. Excessive waiting during the preanalytical
phase can cause in vitro neutralization of heparin. In our institution,
we found that an average of 2.52% of the tubes had compromised quality
(1.25% because of incorrect blood-to-anticoagulant ratio and 1.27%
because of hemolysis). A significant improvement in preanalytical
processing efficiency in coagulation testing can be realized simply by
improving specimen quality.
In summary, we characterized the steps involved in routine and stat coagulation processing. We found a wide variation in wait periods for coagulation specimens, which include a delay in loading the centrifuge, time to verify correct specimen container, and time to address improperly labeled tubes. We developed an automated preanalytical processor to reduce these delays. Improvements in task time variability and overall task time were shown.
|
Acknowledgments |
|---|
|
References |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
The following articles in journals at HighWire Press have cited this article:
|
|
 |
|
  P J Twomey, A S Wierzbicki, and T M Reynolds Chemical pathology and the new contract for GPs J. Clin. Pathol., October 1, 2004; 57(10): 1022 - 1024. [Full Text] [PDF]
|
|
|
|
 |
|
 J. W. Holman, T. E. Mifflin, R. A. Felder, and L. M. Demers Evaluation of an Automated Preanalytical Robotic Workstation at Two Academic Health Centers Clin. Chem., March 1, 2002; 48(3): 540 - 548. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
), maximum (Max), and
minimum (Min) times for each task are shown.
, maximum; 
minimum. Note that the range was smaller for the automated system
compared with the manual method.