Discussion

Instituting airborne isolation for the purposes of excluding TB is a resource-intensive task, particularly involving personal protective equipment, human resources, and bed capacity. Rapid molecular testing to guide the discontinuation of airborne isolation has been shown to be beneficial in multiple single-center studies, by reducing airborne isolation duration without compromising the risk of transmission. We found that the median difference in airborne isolation time was reduced by approximately 38 hours when compared to smear microscopy. There was insufficient evidence to analyze whether this translated to reduced LOS, although one study did find a significant difference.^{Reference Chaisson, Duong and Cattamanchi26} In keeping with prior studies, our analysis demonstrated higher sensitivity of GeneXpert compared to serial AFB smear microscopy, and the specificity of GeneXpert was 99%.^{Reference Chaisson, Roemer and Cantu20,Reference Poonawala, Leekha and Medina-Moreno21,Reference Cowan, Chandler and Kracen25}

Unsurprisingly, there was significant heterogeneity across studies. Although all studies had a similar comparator, being AFB smear microscopy, there were significant differences in the study designs. Due to the lack of implementation of GeneXpert across centers, many utilized a hypothetical design, where GeneXpert was performed on residual specimens and assumptions were made about the estimated time to discontinue isolation. Poonawala et al also compared multiple hypothetical scenarios for reporting the Xpert result, which likely contributed to the higher observed variance in the median difference in isolation compared to the other studies. Although hypothetical trials can explore the impact of GeneXpert-based de-isolation, the findings may overestimate or underestimate the true effects in practice. Hospitals and labs have varying logistics with regard to processing specimens, conducting assays, and relaying the information to clinicians for action. Study protocols differed in terms of whether sputum specimens were concentrated or not before performing GeneXpert. The current manufacturer’s instructions state that GeneXpert may be performed on either raw or concentrated sputum.²⁷ To our knowledge, there have been no studies to determine whether test accuracy changes if done on concentrated sputum. As sputum concentration often occurs before inoculation onto culture medium, both options seem feasible. The importance of specimen quality for determining isolation decisions cannot be overstated. Regardless of decision making via smears or Xpert, the volume and quality of the respiratory specimen should be adequate to avoid decisions that lead to premature de-isolation.

The number of GeneXperts required for de-isolation is somewhat contentious. The 2016 recommendations from the TB controllers in the United States recommend 2 GeneXperts.^{Reference Page, McKenzie and Bossuyt13} However, almost half of the studies utilized a single GeneXpert as the intervention, along with 2 studies examining 1- or multiple GeneXpert strategies simultaneously. A recent review evaluating diagnostic accuracy for 1 versus 2-Xpert strategies found that a 2-Xpert strategy only improved pooled sensitivity by 8% (86% vs 92%) while increasing AII duration and cost.^{Reference Gandhi, Shah, Gautam and Meena28} Institutions may need to decide between single versus 2-Xpert strategies balancing operational efficiency with diagnostic accuracy. An existing gap in knowledge is whether GeneXpert Ultra may perform well as a single specimen strategy for discontinuing isolation, as this newer version has improved sensitivity. Regardless of the method, clinicians should always practice clinical judgement, particularly in cases where there is higher suspicion for TB, such as among immunosuppressed individuals or in areas with higher prevalence of TB. In these cases, a second GeneXpert could be utilized in addition to careful IPC review before discontinuing isolation. The evidence suggests that GeneXpert would reduce the number of false negative cases compared to serial microscopy, though the sensitivity is notably lower than culture, which remains the gold standard for diagnosing TB. Fortunately, the potential transmissibility of those with negative AFB and NAAT is low, and reported cases of transmission from smear-negative cases from genotype-confirmed transmission reports are extremely rare in acute-care hospitals.^{Reference Khan, Nakasone and Ghajar29}

Although GeneXpert has been available for more than a decade, its uptake for guiding de-isolation decisions has been slow. This is likely due to the upfront increased lab costs and labor associated with performing the assay on all admitted patients with possible TB. This is shortsighted though, as there have been multiple cost-benefit analyses that have shown cost savings driven by a reduction in isolation bed use, thereby saving total hospitalization cost.^{Reference Zifodya, Kreniske and Schiller11,Reference Cowan, Chandler and Kracen25,Reference Chaisson, Duong and Cattamanchi26} Chaisson et al determined an average savings of $13,347 per patient in their real-world study, while Cowan et al estimated a savings of $7,002 per patient in their hypothetical trial. ^{Reference Cowan, Chandler and Kracen25,Reference Chaisson, Duong and Cattamanchi26} Although our results demonstrate a median reduction in isolation time of 38 hours, there is not a lot of real-world evidence that such a reduction would translate to cost savings. There is also a paucity of evidence to determine whether Xpert strategy reduces LOS. As many hospitals across North America are experiencing capacity issues, it seems prudent to expedite the evaluation of individuals with suspected TB, and an Xpert-strategy is likely to improve health-system costs and patient outcomes.

Aside from an Xpert-strategy, additional measures may further improve the process of ruling out TB in hospitals. A recent study by Dugdale et al developed and validated an innovative clinical decision support tool (TBorNotTB) to inform whether AII could be discontinued, following negative AFB smears and polymerase chain reactions.^{Reference Dugdale, Zachary and Craig30} They evaluated certain TB risk factors within a cohort of 208 matched case-controls and demonstrated that their tool had 100% sensitivity and 27% specificity to detect TB. Other strategies like reducing the number of AFB smears required for de-isolation from 3 to 2 have also been studied.^{Reference Dugdale, Zachary and McEvoy31} Simple means of further reducing the time in isolation may include faster orders for sputum collection/testing or direct communication between the laboratory and clinician, as was assessed in one study.^{Reference Poonawala, Leekha and Medina-Moreno21} Electronic order set implementation has also been shown to reduce isolation and admission durations.^{Reference Eddy, Seth, White, Sulis, Fagan and Bernardo32}

There are several limitations in this study. There was significant heterogeneity among studies for reasons discussed above. Risk of bias included residual confounding and/or missing data, which may have affected the results. Most of the studies were hypothetical in nature when determining time to de-isolation using GeneXpert; therefore, the external validity and generalizability of the results are limited, although the few real-world implementation studies available show similar benefits. Due to the nature of the data, we could not assess for reporting bias due to risk of false-positive results.^{Reference Zwetsloot, Van Der Naald and Sena33}