Cost analysis of immunochemistry instrumentation.

How to calculate the cost of replacing a manual, labor-intensive technology with one that is automated and may require fewer employees.

We describe in this article a rational approach to establishing current and future costs associated with replacing manual immunoassays with a new generation of random-access immunochemistry analyzers. We build on an earlier article we published (February 1995 MLO, p. 32) in which we discussed the cost analysis and justification for replacing a routine chemistry analyzer. Although the concepts are the same, this time the mechanics are somewhat modified.

Replacing one chemistry analyzer with another makes establishing current and future costs straightforward. For the most part, all the supplies for a given instrument come from one, two, or (at most) three vendors, so the sum of six months of invoices from those vendors divided by six months of reportable results yields the current cost per reportable result. Due to recent changes in the area of immunochemistry automation, however, it is unlikely the costs of all the tests you put on your new, random-access immuno wizbang will come from just one or two vendors. We also simplified our previous analysis by assuming that we were staffed at a minimum level, and there would be no change in labor costs. This assumption is almost certainly invalid in the present example because we will be replacing a technology that is primarily manual and labor intensive with one that is automated.

As before, we recommend postponing cost analysis until you have narrowed down your choices of instruments to (at most) three and ranked them in order of preference. There is so much competition in the field, vendors must be price competitive or risk losing market share.

While reviewing contracts, you may find your current vendor is able to control the contract or reagent prices to favor use of your current system. For example, if you want to run a TSH on the new system, the current vendor may be able to raise the insulin assay price high enough to make the new system less attractive, if that system does not yet have an acceptable insulin test. In this situation, you must build the increased cost of the insulin assay into the estimated costs of using the new system before an acceptable insulin reagent is available for use with the new system, or before the contract with the current vendor expires. You will have to do a lot of estimating because the new generation of immunochemistry instruments does not yet have a full menu of tests. You likely will not be able to automate all of your immunochemistry tests at once.

Establishing current costs

To establish current costs, let us first establish the test menu. This is unlike replacing an existing analyzer in one regard - when you do this in your lab, you probably will assemble a test mix from several different technologies. For example, you may have half a dozen radioimmunoassays [TABULAR DATA FOR TABLE 1 OMITTED] to replace along with three or four tests from a batch immuno-chemistry analyzer and a couple of manual ELISA procedures. We established both our initial test menu and the additions to our menu projected over the next two years (see Table 1), as well as our current volume of reportable results and associated costs.

It is important to note that we established our costs in a way consistent with our previous report. That is, we sorted through all of our invoices to obtain the actual costs for each assay. Obtaining the cost of consumables may require more effort since these typically are purchased in bulk. For this group of assays, consumables include pipette tips and disposable tubes that are not included in the purchase of the kit. To be accurate, items such as re-pipettors, test tube racks, and micropipettors also should be included in the cost of each assay. These costs are relatively insignificant when spread over a large number of assays; however, if you are projecting that you will need a new gamma counter if demand for a certain test is not reduced, you will want to include the estimated cost of the gamma counter when comparing costs. Table 1 shows that the cost of consumables varies with the assay as well as with the method used to perform it.

It is important to note that we did not divide the purchase price of the kit by the number of tests in it. A 100-test [TABULAR DATA FOR TABLE 2 OMITTED] kit does not contain 100 reportable results. As you can see in Table 2, variables such as batch size and duplicate testing have dramatic effects on the cost of reportable tests. In this example, it is apparent that triplicate testing is more expensive than duplicate. But of equal importance is the number of tests in a run. If given the choice, reagent utilization would be more efficient if we were able to perform testing only on large numbers of patients at a time. Unfortunately, service demands often preclude this option. On a related note, if you find the need to increase the frequency of testing (and therefore decrease batch size) to meet new service demands, the cost-per-reportable test will increase. Things like repeat testing, expired reagents, and lost runs will add further to the costs; don't even try to make the calculation - just go to the invoices.

Projecting costs for the new system

Since the immunochemistry instrument of choice will be relatively new, you may have some difficulty projecting your own reagent and consumable consumption independent of the vendor. Ask current users about the frequency of calibration and performing quality control. There may be variables unique to your lab that make it difficult to find an independent source. We recommend you discuss your reportable test volumes and the intricacies of how you plan to operate the system with vendors and ask that they estimate costs for reagents and consumables and assume some of the risk in developing estimates. You do reach a certain level of comfort by establishing these figures independently, but it is desirable to have your vendor specify in writing the projected use with a commitment to review all costs at the first anniversary of the contract so appropriate financial adjustments can be made. This approach also will reduce the danger of somebody providing an unrealistically low quote. To know whether projections reflect reality accurately, you must be aware of the cost of things such as calibrators, how often you will have to replace a lamp or syringe, and who provides the pipette tips for the sample dispenser.

Table 3 shows projected costs of reagents and consumables for our proposed test menu on our new immunowizbang. We began by dividing our annual test volumes by 12 to determine an average monthly volume of reportable tests, and to make it easier to predict monthly calibration and QC rates. In reviewing and comparing the dynamic ranges of the new reagents with the old, we established a repeat rate of 10%. In conversations with the vendor, we determined that we would require 18 tests per month for calibration (6 calibrators in triplicate) and 180 tests for QC. Although the vendor may have FDA approval for running QC once a day, our policy dictates two levels on each shift of operation (2 x 3 x 30 = 180). We rounded the figure up to 200 tests.

Next, we established how many kits will be required monthly. With this information we were able to calculate the cost per test for reagents. Once again, by working with the vendor we established that, on average, we will spend $0.21 per test on consumables. By adding the costs of reagents and consumables per test, we established our cost per test. To come up with a cost per reportable test, we simply multiplied this figure by the number of tests performed, and divided by the number of reportable results. To make the next step simpler, we calculated an average cost per reportable test by dividing the sum of this column by the total number of reportable results. Because we are dealing with an average cost from this point on, it is important to note that a change in test mix will effect a change in the average cost and thus alter the final analysis.

We can see that there is a discrepancy between our current costs of $1.18 and our projected costs of $2.38. Even though we chose to take a reagent rental approach (no available capital) and this cost included two instruments, if we were to stop here, we would not be able to make a defensible case to administration.

At this point we must perform a more comprehensive cost analysis. Table 5 compares the costs of using the current versus the new system over a five-year period. Because we are proposing to phase in testing over three years, we calculated the costs of using both the current system as well as the proposed system for each of the three years of the phase-in.

We looked at the tests run using the current system and decided that, during the first year, we could cut back from three technologists to one by using the automated system. (Note that these are real FTEs and not a soft-labor reduction.) But such a precipitous reduction is an easy trap to fall into, especially when pushed by overanxious vendors. Our approach was quite straightforward: We decided that the bulk of this testing was performed by three [TABULAR DATA FOR TABLE 3 OMITTED] FTEs on the day shift. The remainder was absorbed by afternoon, midnight, and weekend crews. If we were to automate, we would require only one day-shift operator to do the testing. The other shifts would handle the rest, as before. But while automation might make things easier for the afternoon and midnight staffs that have other responsibilities, it was not possible to reduce their staffing levels. When you are told that a new system is going to reduce labor by 50%, do not assume that translates into requiring 50% fewer staff members to perform a given set of tasks. This example is, of course, oversimplified; we ignored vacation and sick days, among other things. The important fact to remember is that labor reduction is not a saving unless you reduce payroll.

The materials costs in Table 4 came from Tables 1 and 3 without adjustments for inflation or increased test volume. As we expanded the test menu, we increased the number of FTEs required to do the work with the current system but left the number unchanged with the proposed system. We are assuming that, using the automated instrumentation, the increase in testing will be handled by the one day-shift operator. Note, too, that we will not reduce our indirect labor until the third year when all of the testing will have been moved. Again we used the logic that we require the same amount of labor to uncap, aliquot, refrigerate, or freeze regardless of whether there are 10 samples a day or 100. When none of this has to be done, that labor no longer will be required.

Cost comparison

In the five-year cost projections in Table 5, we reported the cost of using the current system as revenue when comparing five-year costs of both systems. We used the test volumes from Table 1 and the test costs (materials) from Table 4 for the first three years. For years 4 and 5 we calculated test costs based on a 4%-per-year rate of inflation. We also ignored the effect of depreciation due to minimal start-up costs. The initial evaluation looked good, but we felt there may be room for improvement. We looked at tests we sent to a reference lab and identified the five most frequently requested tests. By performing these tests once a week in house, we could maintain service, and even though we had to add one FTE, we still could realize additional savings. We incorporated these savings into the analysis shown in Table 5.

Table 4

Total cost comparison over next three years

                           Current systems         Proposed system
                       1996    1997    1998    1996    1997    1998

Materials ($)          1.18    1.38    1.61    2.38    2.44    2.74
Direct labor FTEs      3       3.5     4.5     1       1       1
Direct labor ($)       2.23    2.24    2.77    0.74    0.64    0.62
Indirect labor FTEs    0.5     0.5     0.5     0.5     0.5     0
Indirect labor ($)     0.19    0.16    0.16    0.19    0.16    0.00
Cost/reportable ($)    3.60    3.78    4.54    3.31    3.24    3.36

[TABULAR DATA FOR TABLE 5 OMITTED]

As you can see, the projection looked very good. It showed us breaking even in less than a year; over five years we predicted a cost savings of nearly half a million dollars. Now it is time to review and confirm all the data, and just as importantly all the assumptions. When you have the spreadsheets built, it is easy and worthwhile to go back and test various assumptions. Make absolutely certain you were conservative every step of the way. For example, we assumed no cost increases in year 2; what if there were some?

As we suggested before, a payback in fewer than three years is not likely to meet with much resistance from administration. Just remember that it is better to be too conservative than to be too optimistic.

Joseph D. Artiss is associate professor of pathology at Wayne State University and technical director of the chemistry laboratories at DMC University Laboratories, both in Detroit, Mich. Frank Butros is administrator for the division of laboratory medicine and the division of pathology at the M.D. Anderson Cancer Center in Houston.