Revisiting annual screening for latent tuberculosis infection in healthcare workers: a cost-effectiveness analysis

Tuberculosis (TB) infection has long been considered a hazard for healthcare workers (HCWs), where occupational factors such as caring for patients with respiratory TB increase risk [1‐3]. Routine screening and treatment of latent TB infection (LTBI) has traditionally been an important element of TB prevention measures in Canadian and US healthcare institutions.

The preferred method of serial screening for LTBI in Canadian HCWs is the tuberculin skin test (TST), with the recommended frequency of testing reflecting the volume of TB patients cared for at the healthcare facility and the risk inherent to specific work activities. Baseline two-step tuberculin testing is recommended for all HCWs upon hiring in Canada [4]. Subsequent annual testing is recommended for those with intermediate-risk duties (e.g. direct patient care) in settings where TB patients are more likely to be encountered. Annual testing is also recommended for all HCWs who perform high-risk duties (e.g. cough-inducing procedures or laboratory procedures with potential M. tuberculosis exposure), regardless of work setting. Other HCWs, who are at lower risk based on setting and/or duties, are retested only after identified exposure where there is concern about possible transmission [4]. US guidelines are similar, although they suggest that an interferon-gamma release assay (IGRA) may replace the TST [5].

Recent data call into question the use of routine serial testing for the ‘typical’ North American HCW involved in most patient care activities which may entail exposure to M. tuberculosis. This is because false-positive test conversions become more frequent than true-positives as the risk of true infection falls, regardless of the specific test used. The true risk now faced by most North American HCWs is small, as TB incidence in the US is now at an all-time low. This is concordant with reported estimated conversion rates of 0.8–0.9% over 6–18 months of follow-up [6, 7]. Several reports have also suggested more frequent false-positive ‘conversions’ with the IGRAs [6, 8‐11], although one study reported lower apparent conversion rates with the T-SPOT.TB test, particularly when borderline results were excluded [7].

Surveillance data in the US (1995–2007) suggest that overall rates of active TB among HCWs have fallen to that of the general population, with the major risk factor being foreign birth [12]. The large and frequent nosocomial epidemics of the 1980s and 1990s no longer occur. Most TB now arising among North American HCWs likely reflects non-occupational exposure, most often in other countries before immigration. This reinforces the importance of baseline worker screening, but casts further doubt on the need for routine serial testing.

Using a decision analysis model, we compared the cost-effectiveness of current serial screening practices for LTBI among North American HCWs who work in settings where they may encounter TB patients, with that of more targeted approaches, where only HCWs involved in high-risk activities undergo routine annual testing, or where all workers are retested only after identified TB exposures. We did not evaluate the yield or cost-effectiveness of worker screening at hiring – this remains important to detect previously acquired latent infection, notably among foreign-born workers and also as a baseline in the event that repeat testing is considered after identified exposure.

Our analysis suggests that, in North American settings where patients with TB are likely encountered, routine annual screening for HCWs who perform typical patient care activities provides very limited benefit at high cost, compared to more targeted approaches, provided there is reasonable recognition of exposures after they occur. To prevent less than one additional TB case per 5000 workers over 20 years, the cost of annual screening for intermediate-risk workers is nearly triple that of a targeted strategy where only the highest risk workers are routinely screened. Less intensive screening strategies also appear to be associated with slightly better quality-adjusted survival. These findings reflect the fact that most North American HCWs now face a very small annual risk of TB infection. On a group level, unnecessary LTBI treatment produces small decrements in quality-adjusted survival that outweigh small gains from active TB cases averted in a much smaller number of workers.

In our base case, we assumed that workers were 100% adherent with annual testing, i.e. that it was mandatory for renewal of employment. This conservative assumption may not be applicable to all settings, in which case annual screening would be even less cost-effective compared to the targeted strategies. In addition, steps to ensure adherence with screening after recognised exposures will likely offset any minor decrements in TB prevention if annual screening is stopped. Our results also suggest that the use of QFT is systematically more expensive than the TST, regardless of the screening frequency, and most often does not provide any gain in TB prevention.

In 2015, the World Health Organization published guidelines for screening and management of LTBI in higher-income, lower-incidence countries [32]. The guidelines suggest that HCWs represent one of the groups where testing may be considered; this was a “conditional” recommendation, based on “low to very low quality evidence” [32]. The guidelines further indicate that screening policy decisions should reflect local epidemiology and context. Along these lines, our findings suggest that, with current epidemiologic conditions in the United States, Canada, and other similar countries, most HCWs derive limited benefit from routine serial screening. On the other hand, annual worker screening will be appropriate, and potentially much more cost-effective, in higher-incidence settings where there is substantial ongoing exposure and infection [33].

With any testing method, false positive results on serial testing [6, 8, 9] lead to a greater number of workers receiving unnecessary and potentially toxic treatment. In addition, they lead to missed opportunities for prevention after significant exposures, potentially contributing to slightly higher numbers of expected cases. Indeed, a recent retrospective cohort study of California first responders estimated a cumulative frequency of positive QFT tests exceeding 25% over 7 years of routine annual testing. The vast majority of these were likely false positives, meaning that individuals who were deemed no longer eligible for subsequent screens could then be missed after true exposure and infection [10]. A recently published Markov modelling analysis, based on extensive review of published data for the QFT test, reinforces these concerns about the possible extent of false positive results with serial testing, and about the potential to miss true positives accordingly [34]. This is distinct from potential advantages of the QFT for single screens.

The use of a second, confirmatory QFT in serial testing leads to fewer false-positive results, and thus fewer individuals placed on unnecessary treatment for LTBI. Our results likewise suggest that it leads to fewer individuals inappropriately excluded from future testing, such that there are ultimately more infections correctly detected among exposed workers. The use of a confirmatory QFT in the annual and targeted screening strategies was associated with higher testing but lower treatment costs, resulting in net savings when compared to treatment after a single positive QFT. There is some possibility that the T-SPOT.TB test may also lead to improved specificity and hence less unnecessary treatment [7], although this was not confirmed in another study [6]. It is also possible that improvements in IGRA manufacturing, and consistent procurement and processing, will lead to better specificity [7].

In recent years, both cohort studies and decision analyses have addressed TB screening of HCWs, but these have focused largely on choice of screening tools, i.e. the use of IGRAs versus the TST [6, 7, 20, 35‐38]; they have not directly examined the need for screening as such. As a clinical trial comparing active TB after alternative screening strategies is impossible, a decision analysis is well suited to address the need for screening.

In addition, sensitivity analyses identified key drivers of cost and clinical effectiveness for serial screening. The most important was exposure risk, followed by recognition of exposure, and adherence to post-exposure and annual screening protocols. However, even when we considered high levels of adherence to screening, and high test sensitivity and specificity, cost-effectiveness of annual screening was limited, largely because of low ongoing exposure risk.

del Campo et al. [38] conducted a cross-sectional study of HCWs in Madrid, in a hospital with substantially more admissions for microbiologically confirmed TB (47–66/year) than in most North American hospitals. Participating workers underwent both TST and QFT testing. The authors then used a decision analysis model to project costs and TB cases associated with several potential strategies for testing, based on their study results. They suggested that the most cost-effective strategy for LTBI screening and treatment employed the TST (5 mm cut-off) followed by a confirmatory QFT, with an estimated cost of €14,650 per active TB case prevented (compared to no screening) [38]. However, many differences explain this lower cost estimate. Firstly, the Spanish analysis did not distinguish baseline from subsequent screens, it assumed a high true prevalence of LTBI among workers (30%) and it assumed a constant high risk of progression to active TB among infected workers (2.5%/year). In addition, the strongest predictor of positive test results was employment as a janitor, suggesting that non-occupational factors likely accounted for most of the observed infections, as also suggested by the US registry study of TB in HCWs [12].

As with any decision analysis, our results have several limitations. The most important relates to the quality and variability of published data used for model parameters (probabilities and costs). For this reason, we performed extensive sensitivity analyses, which suggested that our major findings were robust. In addition, we made several simplifying assumptions. For example, while active TB cases were extremely rare among HCWs, we did not model the impact of potential transmission to patients or other workers. This is because, in recent years, the extent and attendant cost of transmission from workers in low-incidence settings has been poorly defined, with a small number of publications detailing extensive contact investigations in extreme circumstances, e.g. a HCW with multi-drug resistant TB. Even in such instances, the true extent of transmission may be difficult to capture [39]. Nonetheless, our analysis suggests that, even if each case of active TB in a worker cost an additional $200,000 for very large-scale contact investigation and treatment, the incremental cost of annual screening would still exceed $1 million per case averted.

We assumed that all cases of active TB in workers would ultimately be diagnosed on the basis of symptoms or other findings. We made this assumption because the goal of screening with the TST or the QFT is to identify and treat latent infection, rather than to diagnose active TB.

We also did not address the cost-effectiveness of baseline screening at hiring, which is needed to identify workers who would benefit from any subsequent tests, and to interpret such tests, regardless of the specific test and frequency adopted. Indeed, the continued use of baseline tests is also essential in identifying the much larger group of HCWs who begin employment with latent TB infection, and who may be suitable candidates for treatment accordingly. Most active TB in US HCWs now appears to reflect infection acquired outside the workplace (e.g. before hire). Foreign-born HCWs have a ten-fold higher TB incidence than US-born HCWs, and both groups now have TB incidence rates similar to their non-HCW counterparts in the general population [12]. Baseline testing is even more important in areas where a large proportion of workers are foreign-born, such as Western European countries. Extended residence and/or previous healthcare work in countries with higher TB prevalence are also highly relevant to baseline TB risk. The use of newer LTBI treatment regimens, such as isoniazid-rifapentine [40] or rifampin [41], will enhance the impact of baseline testing, to the extent that treatment adherence improves and serious adverse events become less frequent.

US data suggest that only about 14.5% of US HCWs are foreign-born [12], while in an older nationwide Canadian study, 18% of workers had positive TSTs at hiring [13]. More recent data from US cohorts report 95% or more to have negative baseline tests [6, 7, 11]. This implies that most North American HCWs are expected to have negative baseline tests, and hence, in many settings, the number potentially eligible for repeat annual testing is large. Reducing the number of workers retested could provide substantial savings to occupational health and TB prevention programmes.

For most North American HCWs, annual screening for latent TB infection appears expensive with limited health gains. IGRAs improve the specificity of baseline testing, but are unlikely to improve the cost-effectiveness of subsequent screens due to low annual risks of infection. Annual worker screening may no longer be appropriate in most settings, and reconsideration of this longstanding recommendation may be warranted. The resources currently allocated to routine TB testing for HCWs may be more productively used for other TB prevention activities.