| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
1 ARUP Laboratories, Inc., 500 Chipeta Way, Salt Lake City, UT 84108.
2 Department of Pathology, University of Utah, Salt Lake City, UT 84132.
aAuthor for correspondence. Fax 801-584-5207; e-mail hawkercd{at}aruplab.com.
 |
Abstract |
|---|
 Top Abstract IntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Methods: Project management software was used to track the project. Turnaround times of our top 500 tests before and after automation were measured. We compared the rate of hiring of employees and the billed unit per employee ratio before and after automation by use of linear regression analysis. Finally, we analyzed the financial contribution of the project through an analysis of return on investment.
Results: Since implementation, the volume of work transported and sorted has grown to >15 000 new tubes and >25 000 total tubes per day. Median turnaround time has decreased by an estimated 7 h, and turnaround time at the 95th percentile has decreased by 12 h. Lost specimens have decreased by 58%. A comparison of pre- and post-implementation hiring rates of employees estimated a savings of 33.6 employees, whereas a similar comparison of ratios of billed units per employee estimated a savings of 49.1 employees. Using the higher figure, we estimated that the $4.02 million cost of the project would be paid off 
4.9 years subsequent to placing the system into daily use.
Conclusions: The overall automation project implemented in our laboratory has contributed considerably to improvement of key performance measures and has met our original project objectives.
|
Introduction |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
This report describes the implementation of these different elements of our automation initiative and especially the automated transport and sorting system, which was placed into daily use in November 1998. We describe the impact of these elements on improvement of turnaround time (TAT), quality of service, and productivity as well as descriptive information about our experience in implementing these changes. Finally, we detail the capital costs incurred in the implementation of our automation initiative and present an analysis of return on investment (ROI) based on data from the most recent fiscal year.
|
Materials and Methods |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
The automated transport and sorting system from MDS AutoLabTM Systems (Toronto, Canada) is described in the report appearing in this issue of the journal (1). Initially, it consisted of 360 feet of linear conveyor or track, 5 high-speed sorting machines, and almost 20 robotic devices of various types to perform functions such as merging, diverting, barcode reading, and loading of the sorters. Specimen processors, sitting at one of 30 different workstations, would perform order entry and processing of specimens and complete this process by affixing a barcoded label to each tube. They would place the tubes on one of three "feeder" tracks that transported the tubes to merge onto the main track. The main track consisted of two parallel tracks with 180-degree crossover devices near either end, which created the equivalent of a clockwise loop. The track’s primary function was to transport tubes to high-speed sorters. Three of the four sorters, with a total of 90 lanes or sort groups, sorted the tubes for the laboratory sections into high-volume tests or into work groups of closely related tests. A fourth sorter was used to sort tubes for archival storage after completion of testing. An additional sorter, not attached to the conveyor system, was used as a backup in the event of down time on one of the other sorters and for subsorting (detail sorting).
Four measures were used to compare pre-automation baselines to post-automation performance. These were (a) TAT;(b) monthly lost specimens per 100 000 specimens as recorded by the our Specimen Processing section; (c) quarterly billed units per full time equivalent (FTE) for those laboratory sections receiving the majority of their specimens via the automation system; and (d) the rate of increase in FTEs each quarter for those same laboratory sections.
TATs were measured on ARUP’s top 512 different test codes ranked by volume (representing 70% of the total laboratory testing volume) for three separate 12-month intervals. These intervals were October 1, 1997, through September 30, 1998 (a pre-automation baseline period) and the calendar years of 1999 and 2001. Not every test in the top 512 was transported by the automation system because of specimen type or temperature, but >80% were transported, which was sufficient to assess the impact of automation on this estimate of "total"laboratory TAT. In addition, to make a valid comparison of the pre- and post-automation TAT, only tests that were performed during all three periods were included. The actual TAT of each patient test was measured by subtracting the date and time in the Cerner Pathnet laboratory information system at which order entry was completed from the 2date and time at which the result was verified. We then did a frequency distribution of these TATs in 4-h intervals and graphed cumulative percentages of total test volumes for each 12-month period as a function of TAT in hours.
We attempted to estimate the potential FTEs who were not hired as a result of automation by using measures (c) and (d). We used linear regression analysis (2) on 2 years of pre-automation history to estimate what the post-automation measures might have been without automation. Comparisons of these estimates without automation to the actual post-automation results were then used to estimate the FTE savings, which were used in the analysis of ROI and the payback time on the investment. Eleven different laboratory sections receive the majority of their work via the automation system, and our annual growth rate has averaged 20% for the past 15 years. Therefore, these two regression analyses were used to estimate the number of employees who might have been hired in those sections to keep up with that growth but were not hired because of the impact of the automation initiative. Measure (c) estimated these labor savings by comparing the pre-automation rate of improvement in productivity (billed units per FTE) to the post-automation period. This method was used to correct for productivity improvements implemented by the laboratory sections that were unrelated to this automation initiative. Measure (d) estimated the labor savings by comparing total FTEs in the 11 laboratory sections in the two pre-automation years to the post-automation period. The rate of increase of FTEs observed in the two pre-automation years used in this report was the same as the rate observed over the 3 years that preceded that baseline period.
|
Results |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Since implementation, the average daily volume of new specimens loaded by specimen processors for transport and sorting on this system has nearly doubled, from 8000 to 15 000 at the time of preparation of this report. Because 80% of all tubes are loaded on the system one or more additional times to be delivered to other stops, for either testing or storage, the average number of total tubes transported and sorted each day currently ranges from 25 000 to 30 000. Only 2% of new tubes placed on the system have a second route stop at one of the primary sorters before going to the storage sorter. An example would be a urine test in which the tube’s first stop is for the ordered test (e.g., urine oxalate or urine copper) and the tube’s second stop is for a urine creatinine measurement, which is generally performed on each urine specimen. The storage sorter would then be the third route stop for these tubes.
In general, the implementation of the system and its performance went very well. We experienced problems with hardware and software such as would be expected with the implementation of a complex automation system. Most of these issues were resolved satisfactorily over the succeeding weeks or months. Over the first year of performance, however, we did experience substantially more down time with the sorters than we had anticipated. AutoLab recognized these issues with the sorters with all of their installations and developed a new sorter design to eliminate those issues and improve performance.
In May 2000, several major improvements were made to the AutoLab transport and sorting system, including the installation of five new automated sorters to replace the original sorters. The new sorters had substantially fewer mechanical parts and sensors and used gravity to direct tubes to the front of the sort lanes instead of friction belts. They have been extremely reliable and have had much less down time. The total annual support by our biomedical engineering staff for the automation system has been reduced from two FTEs to one FTE as a result of the new sorters. Other improvements included the installation of a new software module in the AutoLab Process eXpert (APXTM), called APX Alarms. This software enables our biomedical engineers to instantly localize a fault by its location and nature. Two system status light towers were also installed to quickly notify any nearby employee of a system fault. Previously we had light towers only on the five sorters and the reject lane. In general, the support of our excellent biomedical engineering team has assured continual performance of the system with only minimal down time.
The impact of our overall automation initiative [i.e., ESP, the AutoLab system, the Automated Core Laboratory, and the other parts of the project (1)] on TAT is illustrated in Fig. 1
, which plots cumulative percentages of total tests as a function of TAT for three 12-month periods. The line for 1999 (the first full year of the system’s operation) is to the left of the line for the pre-automation period all along the length of the line, and the line for 2001 is to the left of the line for 1999, demonstrating even faster TAT. Fig. 1
suggests an improvement in median (50th percentile) TAT of 5.5 h when the pre-automation period is compared with 1999 and 7 h when compared with 2001. The change in TAT at the 95th percentile estimated in Fig. 1
was 10 h from the pre-automation period to 1999 and 12 h to 2001. The total volumes of the tests whose TATs are graphed in Fig. 1
were 1 689 698 in the 1997–1998 pre-automation interval, 1 954 320 in 1999, and 3 072 466 in 2001.
|
The automation initiative has also had a substantial impact on operational quality, of which we judged the best measure to be the incidence of specimens lost during handling within the laboratory facility, but not including specimens lost en route to the laboratory. The definition we have used for a lost specimen is any specimen that we were unable to locate to complete all originally ordered tests. Fig. 2
, which includes the time period in 1998 before the implementation of the automation system, shows the monthly incidence of lost specimens per 100 000 total specimens. Over 9 measured months of 1998 preceding automation, the incidence of lost specimens was 4.04 per 100 000 specimens. This rate was considered the pre-automation incidence of lost specimens.
|
In the first 2 months there was an increase in lost specimens, which we then learned was related to a software bug in the robotic communication. Some specimens were prematurely discarded without all testing completed. Since March 1999, when the software corrections were implemented, the lost specimen rate has averaged 1.68 per 100 000 over a consecutive 34-month period, a 58% reduction compared with the pre-automation period. During 6 of those 34 months, we had less than one lost specimen per 100 000, and 1 month, we had no lost specimens of 305 907 total specimens.
The impact of the total automation initiative on the productivity of the 11 laboratory sections that receive the majority of their specimens via the automated system is illustrated in Fig. 3
. The ratio of total billed units reported per calendar quarter per FTE for these 11 laboratory sections is plotted beginning 2 years before the implementation of the automated system through the end of our most recent fiscal year (ending June 30, 2001). Independent of our automation initiative, all laboratory sections have had an annual productivity improvement objective of 10% per year. The actual achievement of the 11 laboratories against that objective in the two pre-automation years (represented in Fig. 3
by a dashed line determined by linear regression analysis) averaged 4.4% per year. However, since implementation of the automated system, the productivity of these laboratory sections has increased by 12.9% per year to 5798 billed units per FTE. The regression line based on the 4.4% observed rate of increase before automation predicted a ratio of 4766 units per FTE for the quarter at the end of the last fiscal year.
|
To estimate how many FTEs we might have employed in those 11 laboratory sections had we not implemented automation, we divided the total billed units for the last quarter (1 314 904) reported by those 11 laboratory sections by the ratio of 4766 units per FTE estimated by linear regression. We believe that this estimate of 275.9 FTEs represents the number that would have been on staff had those 2laboratories only continued to improve their productivity at the same 4.4% annual rate of improvement observed before automation. Because only 226.8 FTEs were actually employed at the end of the fiscal year, this was an estimated savings of 49.1 FTEs that we believe can be attributed to the impact of automation.
A second means of estimating the FTEs we have not had to hire as a result of automation is illustrated in Fig. 4
, which shows the actual FTEs by calendar quarter in the 11 laboratory sections. Linear regression analysis for the eight quarters preceding automation estimated that those laboratories might have employed 260.4 FTEs had they continued to increase their total FTEs at the same rate as before automation. Compared with their actual FTE total of 226.8, this suggests that 33.6 employees were not hired as a result of automation. Also shown in Fig. 4
are total billed units for each quarter for these 11 laboratories. The slope of the billed units line increased sharply in the last 1.5 years. We believe that the regression estimate of FTEs that would have been hired based on the pre-automation rate may be too low a number to have performed that work without automation. We believe that the regression line (dashed line in Fig. 4
) should have an upward change in slope during the past 1.5 years to better match the workload. If true, this would mean that the estimate of FTEs we have saved with automation is much higher than the 33.6 FTEs estimated in Fig. 4
and perhaps even higher than the 49.1 FTEs estimated in Fig. 3
, possibly as high as 60–70 total FTEs saved.
|
The FTE data in Figs. 3
and 4
include 10 sorter technicians who staff the automated sorters. The sorter technicians remove tubes and rack them in their sorted manner, load tubes back on the track that are being rerouted to a second destination or to storage, and segregate tubes that must be handled manually. Because the automated sorters and sorter technicians have replaced the manual sorting and the manual status changes to "in-lab" status previously performed by laboratory personnel, it was necessary to include them in any estimate of net FTE savings.
Shown in Table 1
is a summary of all capital expenses related to the total automation initiative. It includes the actual expenditures compared with our budget estimates made before project initiation. The primary expenditures that exceeded our original estimates were the development of ESP and the renovation of our facilities. Table 1
also includes upgrades made to ESP and the automation system subsequent to the original project definition through May 2000, or 1.5 years after implementation. The total figure of $4 021 908 was used as the amount to be paid back.
|
Shown in Table 2
is a summary of our estimate of pay back or ROI. The data are presented on a fiscal year basis. The elapsed time from the November 17, 1998, implementation date to the end of that fiscal year (June 30, 1999) was 0.625 years. Each additional column represents a subsequent fiscal year. The most recently completed fiscal year, which ended June 30, 2001, represents a total of 2.625 years of actual data. The italicized columns are based on estimated labor savings and expenditures for future fiscal years. In this analysis we used the estimate of 49.1 FTEs saved after 2.625 years as derived from the billed unit per FTE ratio method described in Fig. 3
. This amount is intermediate between the two estimates related to Fig. 4
: the estimate of 33.6 FTEs based on historical hiring rates and an estimate of 60–70 FTEs based on relating the predicted hiring rate to the actual workload, which sharply increased starting 1 year after automation. The FTE savings estimated by Fig. 3
for each fiscal year are year-end savings. To derive an average FTE savings for each fiscal year we divided each year’s incremental savings by 2 and added that to the year-end FTE savings for the previous year. These numbers, shown on the next line down in Table 2
, were the actual FTE savings used for the ROI analysis. Also shown in Table 2
are all other costs related to this initiative: the maintenance contract on the automation system, biomedical engineering support, savings in supplies and labor attributed to our standardized tube, and 8% interest on the capital investment. Estimates of future labor savings are speculative, but are based on our budgeted growth rates and are supported by our achievement of savings estimates in the first 2.625 years that matched or exceeded the estimates we made before project initiation. This analysis estimates that we will pay off the full $4.021 million approximately 4.9 years subsequent to implementation.
|
|
Discussion |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
We have attributed the improvements in total TAT to several factors. The first factor is that, as mentioned in the description of the automation system in Part 1 (1), tubes are transported to sorters and brought to in-lab status on arrival at the sorter. Previously, after entry of the order in Specimen Processing, the tubes were manually sorted according to their laboratory section and were placed in a refrigerator or freezer until section employees carried them back to their work areas. Even then, tubes may not have been brought to in-lab status until just before testing. Thus, automation has substantially shortened the elapsed time from the completion of order entry until tubes were in the laboratory sections and able to be listed on worksheets by eliminating or markedly reducing (a) sorting in Specimen Processing, (b) wait time for tubes to be picked up, (c) the manual process of bringing the tubes to in-lab status by reading the barcode of each tube, and (d) the wait time in the laboratory sections for testing to be initiated.
The second factor is that our TAT improvement is clearly volume related. As total laboratory volumes have grown, automation has enabled more specimens to quickly flow to the laboratory sections and be tested. This is illustrated by a dramatic improvement in median TAT, which is evident in Fig. 1
. However, in most laboratories, TAT is also greatly affected by the handling of specimens that have multiple tests to be performed in different laboratory sections. This "running around to find shared specimens" usually severely impacts both TAT and productivity in every laboratory. Although it is not as discernible in Fig. 1
, an even greater improvement in TAT was measured at the 95th percentile, which suggests that automation combined with our ESP tracking capabilities has dramatically reduced this "running around". We also believe that had we not automated, our considerable volume growth would have caused our TAT to deteriorate. Our observed TAT improvement is thus even more dramatic.
Other possible contributing elements to our improvement in TAT are the availability of specimens for repeat testing in our centralized storage system (with higher reliability and better timeliness) and an increase in scheduled run times. Nevertheless, it is our opinion that the first two elements discussed above are the most important contributors to our TAT improvement.
There is a clear correlation of the reduction in lost specimens as a percentage of total specimens with the implementation of our automation initiative (reengineering, ESP, and the automated system). Because only 80–85% of total workload is placed on the automation system, these lower lost specimen rates are even more impressive. It is possible that the majority of our lost specimens are manually handled specimens. However, we have not specifically studied this question.
Although an estimated 50 laboratories in North America have undertaken total laboratory automation (TLA) projects, fewer than one-third of these have published on their experience, and only a handful have published details of actual FTE savings. Some have reported layoffs of existing staff as a means of paying for their project’s capital expense and have provided the numbers of eliminated positions (Bauer S: Design of the Beth Israel Medical Center TLA system. Oral presentation at the conference "Total Laboratory Automation", May 17, 1997, at Beth Israel Medical Center, New York, NY; Graves S: The modular approach. Oral presentation at the AACC conference "Laboratory Automation: Smart Strategies and Practical Applications", November 3, 1999, in Philadelphia, PA; Abbott P: Key issues in staffing an autolab. Oral presentation at the AACC conference "Laboratory Automation: Smart Strategies and Practical Applications", November 3, 1999, in Philadelphia, PA) (3). In our laboratory, layoffs were never anticipated. For 18 years we have had an annual growth rate of 20% per year in specimen counts, and before this project, we believed we would continue to grow at this same rate or higher for the foreseeable future. This led us to conclude that the only means for estimating our productivity improvement and an ROI for this project was to estimate how many employees we might have hired had we not implemented automation and then measure the difference between that estimate and our actual FTEs.
We used linear regression analysis of two different measures for the 2-year period that preceded automation as the basis for this analysis. One measure was the rate of increase in actual FTEs in the laboratory sections supported by the automation system. The other measure was the rate of increase in productivity, expressed as billed units per FTE. In both cases, the impact of automation on these measures was substantial. As shown in Fig. 4
, there was an immediate decline in the slope of the line for the actual number of FTEs employed in those laboratory sections. Even in the most recent six quarters, when the workload has increased even more dramatically, the growth in FTEs has stayed well below that predicted by regression analysis. In Fig. 3
, all of the 10 data points representing ratios of billed units per FTE since automation was implemented are above the regression line, and the points for the last six quarters have been substantially increased. It is possible that some of the increase in productivity is attributable to specific improvements (such as newer analyzers and better procedures) implemented by these laboratory sections and are not related to our automation initiative. However, because the data are inclusive for all tests performed by all 11 of the laboratory sections, including tests not transported by the automation system, we feel confident that the major contributor to this improved productivity is the overall automation initiative.
A pay back or ROI of 4.9 years is not exemplary among typical capital projects in clinical laboratories. However, it should be noted that we have chosen to consolidate several separate projects into a single ROI analysis, although these projects could have been independently implemented and justified. These include the reengineering of our Specimen Processing department, the renovation of our facilities, the development and implementation of ESP, and the installation of the automation system. We chose to consolidate these expenses into a single analysis because the integrated effect of these different elements was likely greater than the sum returns of the separate projects. Moreover, in contrast to many other TLA endeavors, our automation initiative was based primarily on the need to keep up with our anticipated growth by substantially increasing our capacity to handle new work. We were also optimistic that we would see dramatic improvements in TAT and quality, which we did. In the view of our management team, the financial pay back was least important as an objective. The fact that we will experience a reasonable ROI is therefore considered a bonus. In addition, we note that had we hired the additional 49.1 FTEs estimated in Fig. 3
, we might have incurred additional capital expense for the facilities to house these employees in the laboratory.
TLA projects can have a major impact on a laboratory’s performance and productivity if the project is well conceived, planned, and implemented. We believe this project is an example of that statement. In the coming years we will be expanding this system to keep up with an anticipated high growth rate. In addition, we expect to undertake the design and implementation of a robotic laboratory with direct-from-track sampling of patient specimens. Such projects have only been undertaken in laboratories performing mostly routine testing, but are now feasible for the more specialized testing that we perform.
|
Acknowledgments |
|---|
|
Footnotes |
|---|
|
References |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
The following articles in journals at HighWire Press have cited this article:
|
|
 |
|
  C. D. Hawker, S. B. Garr, L. T. Hamilton, J. R. Penrose, E. R. Ashwood, and R. L. Weiss Automated Transport and Sorting System in a Large Reference Laboratory: Part 1. Evaluation of Needs and Alternatives and Development of a Plan Clin. Chem., October 1, 2002; 48(10): 1751 - 1760. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
) for each quarter for the 11 laboratory sections receiving the majority of their work via the automation system.
) shows the total billed units performed by these 11 laboratory sections.