Wrighton-Smith, Peter MEng, PhD; Sneed, Laurie AA; Humphrey, Frances MSN; Tao, Xuguang MD, PhD; Bernacki, Edward MD, MPH
* Summarize the challenges of current approaches to TB surveillance in healthcare workers and the potential advantages of interferon-gamma release assays (IGRAs).
* Review the findings of the new analysis of the cost and adherence impact of switching from tuberculin skin test (TST) to IGRA surveillance.
* Discuss key factors for healthcare institutions to consider in choosing the approach to TB surveillance, including the areas of potential savings with IGRA.
Tuberculosis (TB) surveillance of health care workers using the tuberculin skin test (TST) both upon hiring and thereafter at regular intervals (based on the TB risk of the institution) is a cornerstone of TB surveillance and control in health care institutions.1 The drawbacks of the TST for this purpose are well known and include variable specificity, particularly in previously BCG-vaccinated persons, suboptimal sensitivity, the requirement for a repeat visit within 48 to 72 hours to read the test, the requirement for specially trained personnel to administer and read the test, and high inter- and intrareader variability.2
Over the past few years, a new class of blood tests, called interferon-γ release assays (IGRAs), has been approved by the Food and Drug Administration (FDA) as aids in the diagnosis of TB infection. Two assay formats are available: the T-SPOT.TB test (Oxford Immunotec Inc, Marlborough, Mass) and the QuantiFERON-TB Gold In Tube test (Cellestis Ltd, Chadstone, Victoria, Australia). Recently, Centers for Disease Control and Prevention (CDC) guidelines3 were issued on the use of IGRAs. The guidelines identify, among the recommendations made, two situations of potential relevance to employee health in which an IGRA is preferred over the TST: groups who historically have low rates of return to have TSTs read (as IGRAs, being blood tests, do not require a return visit to obtain results) and groups who have received the BCG vaccine (IGRA results are unaffected by prior BCG vaccination). More generally, the guidelines state that an IGRA may be used in place of a TST in all situations in which the TST is currently used, including employee health, although they point out that the use of the TST is still acceptable practice. The CDC recommends that before implementing IGRAs, each institution should evaluate the availability, overall cost, and benefits of IGRAs for their own setting.
As part of a large health care institution screening almost 18,000 employees annually, many of whom are foreign born, the authors saw that the replacement of the TST with an IGRA would have a number of programmatic advantages, such as the elimination of the need for a return visit to gain test results, the elimination of the requirement to have trained TST placers and readers, the ability to administer tests 5 days per week (TST placement on Thursday is currently problematic because of the need to have a weekend read), and the elimination of two-step testing for new hires. In addition, as the IGRAs have been shown to have high specificity that is unaffected by BCG vaccination, the use of IGRAs could potentially reduce the number of false-positive results obtained in foreign-born health care workers and thereby prevent unnecessary chest radiographs and chemoprophylaxis with attendant risks. However, the authors' concern was whether these benefits would actually accrue in practice and whether the costs of adopting an IGRA would outweigh the benefits accrued.
The SWITCH (Screening health care Workers with IGRA vs TST: impact on Costs and adherence to testing) study was therefore conceived to systematically identify and then measure all the costs of screening the health care worker population using either a TST or an IGRA and, in so doing, to build a model that could be used by other institutions across the United States to judge the cost impact of making the switch to an IGRA. As part of the study, the authors also gathered evidence on the impact of using an IGRA on adherence to screening, as the aim of all employee health TB surveillance programs must be to ensure the highest completion rate possible. The study was designed to examine costs and adherence for annual screens (serial testing of current employees) and new hire screening (baseline testing of new recruits), both separately and in aggregate.
The study was conducted by the employee health department at the Johns Hopkins Healthcare System and the Johns Hopkins Medical School based in Baltimore, Maryland. Ethics approval for the study was obtained from the Johns Hopkins Institutional Review Board (study number NA 00038494).
The study combined a number of elements to build up the total costs and adherence results for the screening program. First, all material and labor costs involved with TB screening were gathered. Second, a time–motion study was conducted to measure, over 393 patient encounters, the time (and therefore labor costs) associated with all steps involved in administering a TB surveillance program using either TST or IGRA. Third, a cohort of 743 health care workers (including both new hires and annual screens) was tested in parallel with the TST and IGRA to gather data on relative positivity rates of the two tests. Alongside this testing, a questionnaire was administered to all participants which solicited their views on TST or IGRA screening. Finally, a series of decision trees was constructed to unite this information and allow the calculation of the total costs and adherence rates of screening using a TST or IGRA in annual screens and new hires based on retrospective epidemiological statistics from the employee health database of all employee health TB screens conducted over a 1-year period. The following sections summarize these steps; detailed further information can be found in the Supplemental Digital Content (http://links.lww.com/JOM/A95).
The study examined costs from the perspective of total costs to the institution, including both material costs and costs of labor expended in conducting/attending screening. Out of pocket costs to the employee were excluded from the analysis.
Material and Labor Costs
All materials costs for all reagents and consumables used in TB screening were obtained from institutional finance records. These were calculated on an “as consumed basis” to account for wastage (eg, it is not always possible to obtain 10 tests from each PPD vial). As well as materials costs, average labor costs for all those involved with TB screening (both administering and receiving the test) were obtained from the relevant HR departments and converted into an hourly rate taking into account working hours and the cost of employee benefits. A weighted-average labor cost for all employees undergoing testing was computed on the basis of the proportions of each job category among the population eligible for TB screening at the institution. All material and labor costs are shown in Fig. 1.
To compute the labor costs of maintaining a screening program (both in terms of employee health personnel labor costs to administer the program and employee labor costs caused by time away from work to attend screening), it is necessary to first measure the time taken to perform all steps in an employee health TB screening program. To do this 393 individual randomly selected patient encounters encompassing each step (eg, time to place and read a TST, time to enter results into the employee health (EH) database, time to conduct phlebotomy for the IGRA, time to follow-up TST nonreturners) were examined, and the total time taken for employee health personnel to complete their task in the encounter and the total time away from work for the attendee caused by the encounter were measured. Averages for each step were calculated and then computed into dollar costs for the employee health department using the relevant labor rate for the person who would perform that task, and into dollar costs for the employees in terms of time off work using the weighted-average labor rate for all employees at the institution. Times and associated costs of each step in the TB screening process are shown in Fig. 1.
Parallel Testing of IGRA and TST
To gather data of test positivity rates to feed into the cost modeling, those undergoing TB testing over a 3-month period (May to July 2010) were offered an IGRA test alongside the TST with the aim to enroll at least 250 each of new hire and annual screen participants. Participation was also offered to those with a prior documented positive TST result who would be getting a symptom screen instead of a TST. Invitation of participants was random, but participants' subsequent participation was voluntary. Participants were also required to complete a questionnaire, administered verbally, containing questions relating to their attitudes and experiences with the TST and IGRA. The TST was applied and read as normal by employee health staff. The IGRA testing was performed using the T-SPOT.TB test by Oxford Diagnostic Laboratories (ODL), a national TB testing laboratory. Prior to this study, the institution had previously chosen to adopt this IGRA in a targeted population by sending out to ODL because of the ease of accessibility of the test; specifically, blood collection and preanalytical steps are easy (one draws a routine heparin blood tube and packages it in prepaid shipping containers provided free of charge by ODL for collection at the end of each day) and that testing is available 5 days per week. Test results are reported according to the test's FDA-approved package insert as positive (≥8 spots), negative (≤4 spots), borderline (5 to 7 spots), or invalid/unusable (no result because of phlebotomy error or test failure). Phlebotomy for the test was undertaken by employee health personnel.
Decision Tree Model Structure and Other Model Inputs
Five separate decision trees were constructed as part of the analysis: two for the TST reflecting the screening pathways for annual screens and new hires (Supplemental Fig. 1a and d, respectively; http://links.lww.com/JOM/A95) and three for the IGRA reflecting the screening pathways for annual screens (Supplemental Fig. 1b and c, http://links.lww.com/JOM/A95) and new hires (Supplemental Fig. 1e, http://links.lww.com/JOM/A95). The screening procedures modeled reflect those outlined in the CDC guidelines1; specifically that two-step testing is performed on new hires, any TST non-returns are followed up to ensure they return and complete retesting, there is annual retraining of staff administering or reading TSTs, and symptom screens are conducted on those with prior positive results (rather than readministering the TST to known prior TST positives). In addition, two deviations from CDC procedure were also modeled: first, what the impact would be on costs and adherence if no follow-up of TST nonreturners was performed and second, a specific decision tree was constructed to reflect the scenario in which in the first year of adopting an IGRA for annual screens, all prior TST positives were screened with an IGRA so as to set a new baseline (Supplemental Fig. 1b, http://links.lww.com/JOM/A95). Although the CDC does not recommend testing those in whom prior TB infection has been identified, the authors, like in many employee health departments, suspected that a large proportion of prior TST positives were false-positives because of prior BCG vaccination, and wanted to examine whether using an IGRA would establish a different baseline in these people.
As the CDC guidelines state that an IGRA can be used in direct substitution for a TST, the model did not attempt to take into account any accuracy differences between the tests. Consequently, a positive and negative test result with either TST or IGRA were treated in the same way. For simplicity, it was assumed that no cases of active TB would arise in the screening process. This assumption reflects the epidemiology at Johns Hopkins, where only one case of active TB in routine TB screening of health care workers (both new hires and annual screens) was identified over the past 20 years (>400,000 screenings). The rate of TST conversions was taken from annual employee screening statistics. In the absence of serial testing data of the performance of the IGRA at our institution, the IGRA conversion rate was assumed to be equal to the TST conversion rate, although this was a key variable examined in the sensitivity analysis.
The costs for each branch in the decision trees were derived from the preceding investigation into material and labor costs associated with each step. The probabilities of each branch were derived from the relative positivity rates found in the parallel testing of TST and IGRA, and also from the employee health database, which records the outcomes of all TB screens conducted annually. In this case, the data from the 17,813 screens conducted in July 2009 to June 2010 was used as the basis of the models. Relative positivity rates of TST versus IGRA were used in preference to taking just the absolute positivity rates in case the enrolled study population was biased versus the overall yearly testing cohort. The derived costs and probabilities are shown next to each decision tree (Supplemental Fig. 1a–e, http://links.lww.com/JOM/A95). The decision tree models allowed the computation of total costs of screening with the TST and IGRA as well as adherence rates (the numbers ending in a decision tree state consistent with an acceptable completion of the screening pathway). The cost of the IGRA test was a floating variable in the models; goal-seeking was used to compute the IGRA test cost at which the IGRA pathway became cost saving versus the TST pathway.
Several aspects of the decision trees reflect the configuration of TB screening at Johns Hopkins. First, new hires have the first of a two-step skin test before beginning work. Hence, there is no cost to employee time off from work for this test, as it takes place before the person actually starts working. In addition, this process ensures that there is 100% compliance for the first step of new hire testing as without a valid first step, the recruit cannot begin to work. The second step TST is placed during the first week of employment. Second, the screening of all health care workers (with some small exceptions who have more frequent testing) is annual, consistent with the institution being deemed medium risk under the CDC classification.1 All new hire testing takes places at the employee health department (a separate building to which all employees have to travel), but almost half of the annual screens are done in locations convenient for employees (eg, a location adjacent to the central cafeteria). This configuration has an impact on the convenience of testing to employees, their time off work to attend testing, and on employee health resources consumed in testing. To enable the evaluation of whether our approach is the most cost-effective, separate models were built to examine the changes in outcomes if all annual screens were performed locally to employees or whether screening required them to travel to a remote location.
It was also an aim of the study to explore the cost effects of how assiduously follow-up of TST nonreturners is conducted to ensure that they fully complete screening. To establish a common benchmark for comparison purposes, two scenarios were modeled, one in which no follow-up is performed (cheaper, but lower compliance), and one in which resources are consumed on following up such that 95% of those nonreturners fully complete screening (the time spent and success rates of following up with nonreturners was examined in the time–motion component of the study). Nonreturn rates for the TST were taken from the annual employee health TB screening statistics.
A separate model was also built to assess different possible practices surrounding the treatment of borderline T-SPOT.TB results. Both because a borderline result is a valid test result (that can be acted on from a clinical management perspective) and to preserve the generality of the model to both IGRAs (QuantiFERON does not have a borderline zone), the base case assumed that borderline results would not necessarily require retesting. The alternative (full retesting of all borderline results) was assessed using a separate model.
To account for errors in point estimates of input probabilities and costs arising from the time–motion and TST/IGRA sample data, as well as to generalize the model outcomes for other institutions with different costs, a sensitivity analysis was also performed by varying all 48 model inputs by a fixed value of ±20% the base amount to show those variables that most impact the model outcomes (see Supplemental Digital Content, http://links.lww.com/JOM/A95, for more detail).
All analysis was performed using Microsoft Office Excel 2007 (Microsoft Corporation, Redmond, WA). Model cost results were expressed both per person and to show the total costs of screening to the institution on the basis of 17,813 screens annually.
Parallel Testing of TST and IGRA
A total of 750 people agreed to take part in the parallel testing part of the study and fully completed all study documentation; 473 were undergoing annual screening and 270 were new hires. An additional seven persons enrolled as they wished to get retested with the IGRA following a known (not necessarily recent) prior exposure; these results are not shown. Comparative data of the performance of the TST and IGRA are shown in Table 1.
From the 473 annual screens, the overall positivity rate for the IGRA was less than half that of the TST (50 vs 119 positives). Of the 113 known prior TST positives (of whom 69 were foreign born, almost all with prior BCG vaccination) who ordinarily receive a symptom screen instead of a repeat TST, two thirds (75/113) were found to be negative by IGRA. The institution's typical TST nonreturn rate for annual screens (without follow-up) is 20%; in this cohort it was just over 10%. This compares with the IGRA where only 0.4% (2/473) of results were unavailable. Of the 312 subjects with known negative prior TST status and results in the parallel testing, six new TST positives were identified (of which five would be considered conversions having an induration increase of ≥10 mm since the last documented reading) versus 10 with the IGRA.
The positivity rate of IGRA was also found to be lower than the TST in the cohort of 270 new hires (23/270 vs 39/270). As with the annual screens, the rate of unavailable IGRA results was very low (1/270). Interestingly, out of the 270 screens, the IGRA gave negative baseline results in 240 versus only 171 with the TST.
In response to the questionnaire, a significant proportion of study participants (22.3%) stated that they had “no confidence” in the TST result (rising to 30.7% in those that were foreign born), compared with 6.5% (and 7.3%, respectively) who stated that they had “no confidence” in the IGRA result. Foreign-born study participants unsurprisingly reported a greater number of adverse reactions to TST than native-born participants (23.4% vs 12.5%). Despite the fact that 18.5% of study participants felt a blood draw was less desirable than the placement of a TST, in a direct preference test, 62.3% of study participants preferred an IGRA, whereas 6.5% preferred a TST. Among foreign-born participants, 66.5% preferred an IGRA, whereas 5.0% preferred a TST. (A full summary of the questionnaire results is shown in Supplemental Table 2, http://links.lww.com/JOM/A95).
The total cost to the institution of just placing and reading a TST was $54.09 per test in annual screening and, because of the need for two-step testing, was $81.38 for new hires (Fig. 1). Given that the material cost for the TST is $2.83, the remainder of the costs relate to the employee health staff time to administer the test and the costs of staff time off work to receive the test. For those annual screens who fail to return for test reading, the costs rise significantly to $88.15 because of the requirement for employee health staff to expend time in follow-up to get staff to return for a second placement, and the wasted costs associated with the placement of the first unread TST.
Adding in the other elements of a TST screening program (eg, follow-up of TST positives, symptom screens, etc), Table 2 shows that the total overall cost of conducting the TST-based screening program is $1,303,855 per year (or $73.20 per person). New hire testing is considerably more expensive than annual screening ($90.80/person vs $63.42/person). This is largely because of the need to perform two-step testing (higher employee health staff costs) and the higher nonreturn rate observed in the new hire population that requires additional employee health time to follow-up. The overall adherence rate was 98.54% with the TST, 99.10% and 97.53% in annual screens and new hires, respectively. These result from the modeling assumption that the employee health department spends the time necessary to ensure that 95% of nonreturners (20% of annual screens and 50% of new hires for the second step) return for screening. Without this follow-up, the overall cost of screening with the TST drops to $52.61 per person (see Supplemental Table 3a, http://links.lww.com/JOM/A95), a saving of more than $20 per person as a result of the elimination of employee health time, employee time off work, and materials costs consumed in follow up and retesting initial TST nonreturners. However, this comes at the penalty of an adherence rate of only 70.8%.
Switching to an IGRA would be cost saving versus the TST if the IGRA cost is $54.83 or less per test (Table 2). Taking only new hires, which is the more expensive of the two screening programs to administer with the TST, switching to an IGRA for only new hires would be cost saving at $81.16 or less per test. Screening with an IGRA has a very high adherence rate (99.99% overall) due to there being no need for the employee to return to have the test read, and due to the low numbers of unavailable test results.
Figure 2 shows the outputs of the sensitivity analysis; the 10 variables that have most impact on the model are shown. None of the other 38 variables had an impact on the test cost of more than $0.75 in either direction at which the IGRA was cost saving. The most important variable identified in the sensitivity analysis was the average labor cost of the employee base, because of its effect on the costs associated with time off work. The base value was $49.38 per hour. If this is increased by 20% to $59.26 per hour, then the IGRA becomes cost saving at the higher cost of $61.16 per test. This is because the TST costs are affected more than the IGRA costs by changes in this input as the TST requires a larger number of appointments, and hence higher costs of time off work, to complete screening versus the IGRA. Equally, if this labor rate is reduced to $39.50 per hour, then the IGRA becomes cost saving only at the lower cost of $48.50 per test. The next eight most important sensitivities were all linked to the employee time off work to undergo various elements of screening or staff time of the employee health department to administer screening. With the exception of the TST material cost, no changes in other material costs or any of the input probabilities were significant enough to reach the top 10. Under all the sensitivities studied, the IGRA was always cost saving at a cost of $48.50 per test.
When the model was changed to examine the costs of screening all employees remotely versus locally, the total costs of screening remotely with the TST were $1,407,967 ($79.04 per person). This fell to $1,114,447 ($62.56 per person) if all screening could be completed locally (Supplemental Tables 3b and c, http://links.lww.com/JOM/A95). There would also likely be an improvement in compliance because of the greater convenience of screening, but in the absence of data, this was outside the scope of the model. Neither scenario greatly affected the cost at which an IGRA was cost saving relative to the TST (as the same benefits of lower time off work from local screening accrue also to the IGRA, albeit in a smaller way). The IGRA was cost saving at $52.38 per test for local screening versus $54.70 per test for remote screening.
If all borderline IGRA results are retested, the screening costs using IGRA increase by 1.7% to $1,326,803 (Supplemental Table 3d, http://links.lww.com/JOM/A95). In this scenario, the IGRA was cost saving at $53.54 per test.
To the authors' knowledge, this is the first study in the United States to systematically analyze, including using time–motion methods, the costs of administering a TB surveillance program for health care workers using TST and compare that with the costs of switching to an IGRA on the basis of first-hand investigation of test performance and labor costs. It is also the first study showing the performance of the T-SPOT.TB IGRA in general health care worker screening cohorts.
The result of this study shows that when the time consumed in administering a TST program is accurately measured and costed, TST screening is expensive for a sizeable institution, both in absolute terms (exceeding $1 million annually) and on a per person basis ($73.20 per person). Given that the TST material cost is relatively cheap, this high cost reflects the significant burden on staff time required to both administer and receive the TST.
In an earlier study by the CDC of the costs of maintaining a TST program in compliance with the CDC guidelines for screening employees in hospitals and health departments, Lambert and colleagues4 reported that the total cost of administering a program in hospitals varied between $57 and $501 per person (values rebased into 2011 from the original 1998 costs5). Of the four hospitals studied, three were of smaller size in which the fixed costs of administering the program were large in relation to the number of employees receiving testing. For the large hospital in that study, the reported costs were $57 to $78 per person screened using the TST, which is comparable with this study's findings of $52.61 to $73.20 per person screened.
The CDC does not recommend retesting those in whom a latent TB infection has already been confirmed, implying that prior TST positives should not receive an IGRA. As it was suspected that many of the prior known TST positives in the authors' institution are false-positives (eg, due to prior BCG vaccination), we chose to retest these individuals with the IGRA. It is interesting to observe that of 113 known prior positives enrolled in the study, more than 50% (75) were found to be negative by IGRA, which would potentially allow them to have their baselines reset and reenter annual screening. This costs marginally more in the first year ($2.28 more per person undergoing annual screening), but could allow a more accurate baseline for all staff to be established, thereby aiding TB surveillance in the institution.
The views of those undergoing testing are often overlooked, but may be an important factor in their willingness to comply with testing. We attempted to understand employee attitudes and preferences to the TST and IGRA through the use of a questionnaire administered to each study participant that showed, despite an aversion by a minority to venepuncture, a preference toward IGRA testing. It is worth recognizing that the questionnaire was only administered to those who chose to participate in the study, so these results may be biased toward those, for example, who dislike the TST or are not averse to phlebotomy, and may consequently not accurately reflect the views of the entire population tested at hire and annually. Like many other institutions, our policies require all new hires to visit the employee health department before starting work to complete both the placement and read of their first-step TST. In discussions with new hires, it was clear that the requirement to see the employee health department on four occasions sometimes imposed a substantial financial burden on potential employees in terms of travel time, childcare costs, and other factors. Although it was outside the scope of this study, it may be worth exploring whether switching to an IGRA makes enrollment procedures more acceptable to potential recruits, or the on-boarding process for new hires quicker.
This study's results show that a costly element of the institution's TST program is the follow-up of those who do not return to have their TST read. There is thus a trade-off between adherence and cost. In this study, the difference between an adherence rate of 70.8% and 98.5% with the TST was $366,793 per year, or $20.59 per person. It is noteworthy that of the TST nonreturners, almost 10% were deemed infected by the IGRA, showing the potential TB risk to the institution of not following up TST nonreturns. As the IGRA does not require a repeat visit and has such low failure rates, the adherence rate calculated was 99.99%. Even if persons with an initially nonreportable result are not followed-up, as these are relatively rare occurrences, the adherence rate drops only to 99.74%. A limitation of the model is that the adherence/compliance rates only express the compliance rate of those who initiate screening in some way. It does not reflect those who do not even attend for any kind of screening in the first place. However, in practice the initiation of screening is usually enforced through administrative controls; for example, at this institution new hires cannot start work unless they have at least completed the first step of their two-step testing.
To assess the factors that had the greatest effect on model outcomes, and to check how sensitive the conclusions of the study would be, should the model inputs from only sampling a subset of the total testing universe have been inaccurately estimated, a sensitivity analysis was performed to evaluate a ±20% change in all the model inputs. This showed that nine of the top 10 most significant costs are staff time in one variation or another. The programmatic advantages of IGRAs (one visit, no need to follow-up nonreturners, no annual retraining costs) remove a lot of unnecessary consumption of staff time, which is why the model shows that costs savings are possible even with IGRA material costs that are an order of magnitude higher than the TST material cost. The sensitivity analysis shows that these programmatic advantages dwarf any cost changes resulting from difference in performance of TST versus IGRA. More generally, the cost saving potential of the IGRA over the TST was relatively insensitive to model inputs; the IGRA is always cost saving at a cost of $48.50 or less per test under any of the sensitivities analyzed.
Given the logistical challenges imposed by the TST, the authors' institution has chosen to perform a significant proportion of its annual screens at a location convenient for employees. To do this, 215 persons across the institution have been trained to place and read TSTs. This increases yearly fixed costs as these persons require annual retraining in PPD placement and reading to comply with CDC guidelines. But it is felt that the reduced time off work for employees has been worth this extra investment. This model has made it possible for the first time to systematically understand whether this decision is, in fact, cost-effective. If the institution was to insist on all annual screens taking place in the employee health department, costs for completing screening would rise to $79.04 per person versus the current cost of $73.20 per person, showing that it is cost-effective to run testing in a way that reduces time off work to employees despite the higher training costs. If the institution was to go further and conduct all screenings at a location convenient for employees (ie, locally), this study's model shows that further significant savings would be possible for both TST and IGRA. The practical challenges of doing this with the TST are significant. However, the programmatic advantages of IGRAs suggest that, when implementing an IGRA, hospitals consider whether workflow can be redesigned to administer the IGRA close to the employee, rather than requiring unnecessary trips to a remote location for screening.
The two IGRAs differ in that the T-SPOT.TB test has a borderline zone whereas the QuantiFERON-TB Gold In Tube test does not. The CDC states that incorporation of a borderline category for the T-SPOT.TB increases test accuracy by classifying results near the cut point (at which small variations might affect the interpretation) as neither positive nor negative, which makes test conversions from negative to positive more likely to represent a newly acquired infection.3 However, the potential concern of a borderline zone is that the need to retest these samples significantly increases costs. This model shows this concern to be unfounded as the effect of retesting all borderline cases only increased the overall costs of the IGRA screening program by 1.7%.
The present study has several limitations. First, the models were built on the basis of positivity rates extrapolated from the cohort that was enrolled for parallel testing. This was a fraction of the total tested population, as it was impractical to concurrently test all employees with both TST and IGRA. This, and the fact that subjects self-selected for inclusion, meant that there was a high likelihood that the TST and IGRA positivity rates identified in the cohort were not perfectly reflective of the rates in all employees, due to enrollment bias. Indeed, the data suggest that this cohort was overrepresented with, for example, known prior TST positives and foreign-born persons, which predisposed this cohort to higher TST (and IGRA) positivity levels than observed for employees as a whole. To overcome this enrollment bias, actual TST positivity rates from employee health records (captured for the entire screening population every year) were used in the model and the expected positivity rate for the IGRA was calculated using the relative positive rates of IGRA versus TST found in the parallel testing cohort. The sensitivity analysis showed that the study outcomes were, in any case, insensitive to changes in assumed IGRA versus TST positivity rates (changes of IGRA positivity of ±20%, had less than ±1%, or $0.50 per test, effect on the cost-neutrality level).
Another limitation of the present study is that it did not consider the relative accuracy and effectiveness of the TST versus the IGRA strategy in preventing downstream active TB cases. Interferon-γ release assays are preferred over the TST, particularly in BCG-vaccinated populations, as their greater accuracy should enable the more accurate identification of those at risk of developing active TB disease3; thus, screening using an IGRA should yield some effectiveness benefits in terms of cases of active TB disease prevented. This potential benefit was excluded from the modeling for a number a reasons; first, the numbers of active TB cases that have occurred in health care workers over the past 20 years in this institution is extremely low and therefore effectiveness gains could only be extremely small, and second, it adds considerable complexity to the modeling and necessarily increases the use of assumptions to obtain results, which we preferred to avoid in favor of a model based largely on measurable inputs.
A further limitation of the study is that, in the absence of serial testing data for the IGRA, the IGRA conversion rate was assumed to be equal to the TST conversion rate. There are two important issues behind the measurement of this rate, the first of which is highlighted by Gandra and colleagues.6 In any year of a TB screening program using the TST, there will be a number of people who are deemed infected (the known prior TST positives) and a number of people deemed as newly infected this year (convertors). In the first year of the switch to an IGRA, a new baseline will be set for all employees as to who is infected. This baseline will be different to the previous TST baseline in a large number of cases for the following two reasons. First, given that the IGRA has higher specificity than the TST,7–11 it would be expected that a number of the previous TST positives would now not be deemed infected by the IGRA (as was observed in this study's cohort with 75/113 prior known positives being deemed uninfected by IGRA). Second, given that the evidence on IGRAs shows that they may have higher sensitivity than the TST (and that even when sensitivities have been equal there is discordance as to who positive by either test),12–18 it would be expected that some of the previous TST negatives would now be deemed infected by the IGRA, as was also found in this study's cohort (10/312 previous TST negatives were found to be positive by IGRA vs 6/312 for the TST). It is our belief that these are not conversions, and hence should not be investigated as such, as they are just as likely to reflect discordance between IGRAs and the TST rather than actual new infections occurring within the last year. Nonetheless, this scenario, in which a new baseline is set with an IGRA (including offering INH to all newly identified positives) was explicitly modeled with the fifth decision tree (Supplemental Fig. 1b, http://links.lww.com/JOM/A95) which showed that the costs do increase in the first year of IGRA adoption, but only very modestly (+2.9%, Supplemental Fig. 1b, http://links.lww.com/JOM/A95).
Once the new baseline has been established as to who is deemed infected after the first year of switching to an IGRA, the question is then what the subsequent conversion rate of the IGRA will be. In this study, it was assumed that the steady-state (ie, year 2 after switch and beyond) conversion rate of the IGRA would be the same as the TST. Given that TST must satisfy a “double test” to be deemed a conversion (ie, both a change from a negative to a positive test result and an increase in induration of ≥10 mm), whereas an IGRA only needs to move from negative to positive, one might expect conversions with IGRA to be more common than with the TST.3 Should this be the case, the sensitivity analysis shows that the model outcomes are actually insensitive to changes in this input. For example, even if the IGRA conversion rate was not +20%, but double that of the TST, then the costs of screening the annual screens with an IGRA rises by only 0.4% (data not shown). However, from a programmatic perspective, further studies looking at the conversion rates of IGRAs (from a previously negative IGRA baseline) and an examination of the practicality and parameters of a “double test” with IGRAs would be extremely helpful.
In summary, the results show that administering a TST-screening program at this institution in accordance with CDC guidelines is expensive at $73.20 per person, largely because of the significant time burdens both on employee health staff and the employees themselves to administer and receive screening. Primarily because of programmatic benefits, switching to an IGRA results in better adherence and is cost saving versus the TST provided that the IGRA cost is $54.83 or less per test. Whichever screening strategy is chosen by an institution, this analysis suggests that analyzing and changing workflow practices to minimize time off work can result in significant reductions in the total cost of a program.
The authors thank David Baron, MBA, who assisted in the data gathering process.
1. Jensen PA, Lambert LA, Iademarco MF, Ridzon R; Centers for Disease Control and Prevention. Guidelines for preventing the transmission of Mycobacterium tuberculosis in health-care settings, 2005. MMWR Recomm Rep. 2005;54:1–141.
2. Huebner RE, Schein MF, Bass JB. The tuberculin skin test. Clin Infect Dis. 1993;17:968–975.
3. Mazurek GH, Jereb J, Vernon A, LoBue P, Goldberg S, Castro K; IGRA Expert Committee; Centers for Disease Control and Prevention. Updated guidelines for using Interferon Gamma Release Assays to detect Mycobacterium tuberculosis infection—United States, 2010. MMWR Recomm Rep. 2010;59:1–25.
4. Lambert L, Rajbhandary S, Qualls N, et al. Cost of implementing and maintaining a tuberculin skin test program in hospitals and health departments. Infect Control Hosp Epidemiol. 2003;24:814–820.
6. Gandra S, Scott WS, Somaraju V, Wang H, Wilton S, Feigenbaum M. Questionable effectiveness of the QuantiFERON-TB Gold Test (Cellestis) as a screening tool in healthcare workers. Infect Control Hosp Epidemiol. 2010;31:1279–1285.
7. Diel R, Goletti D, Ferrara G, et al. Interferon-gamma release assays for the diagnosis of latent Mycobacterium tuberculosis infection: a systematic review and meta-analysis. Eur Respir J. 2011;37:88–89.
8. Bienek DR, Chang CK. Evaluation of an interferon-gamma release assay, T-SPOT.TB, in a population with a low prevalence of tuberculosis. Int J Tuberc Lung Dis. 2009;13:1416–1421.
9. Wang SH, Powell DA, Nagaraja HN, Morris JD, Schlesinger LS, Turner J. Evaluation of a modified interferon-gamma release assay for the diagnosis of latent tuberculosis infection in adult and paediatric populations that enables delayed processing. Scand J Infect Dis. 2010;42:845–1850.
10. Detjen A, Keil T, Roll S, et al. Interferon-gamma release assays improve the diagnosis of tuberculosis and nontuberculous mycobacterial disease in children in a country with a low incidence of tuberculosis. Clin Infect Dis. 2007;45:322–328.
11. Nienhaus A, Schablon A, Diel R. Interferon-gamma release assay for the diagnosis of latent TB infection—analysis of discordant results, when compared to the tuberculin skin test. PLoS ONE. 2008;3:e2665.
12. Diel R, Loddenkemper R, Nienhaus A. Evidence based comparison of commercial interferon-gamma release assays for detecting active TB: a meta-analysis. Chest. 2010;137:952–968.
13. Kobashi Y, Mouri K, Yagi S, et al. Clinical evaluation for diagnosing active TB disease and transitional change of two commercial blood tests. Scand J Infect Dis. 2008;40:629–634.
14. Soysal A, Torun T, Efe S, Gencer H, Tahaoglu K, Bakir M. Evaluation of cut-off values of interferon-gamma-based assays in the diagnosis of M. tuberculosis infection. Int J Tuberc Lung Dis. 2008;12:50–56.
15. Lee JY, Choi HJ, Park IN, et al. Comparison of two commercial interferon-gamma assays for diagnosing Mycobacterium tuberculosis infection. Eur Respir J. 2006;28:24–30.
16. Kang YA, Lee HW, Hwang SS, et al. Usefulness of whole-blood interferon-gamma assay and interferon-gamma enzyme-linked immunospot assay in the diagnosis of active pulmonary tuberculosis. Chest. 2007;132:959–965.
17. Meier T, Eulenbruch HP, Wrighton-Smith P, Enders G, Regnath T. Sensitivity of a new commercial enzyme-linked immunospot assay (T SPOT-TB) for diagnosis of tuberculosis in clinical practice. Eur J Clin Microbiol Infect Dis. 2005;24:529–536.
18. Chee CB, Gan SH, KhinMar KW, et al. Comparison of sensitivies of two commercial gamma interferon release assays for pulmonary tuberculosis. J Clin Microbiol. 2008;46:1935–1940.