Association of Rapid Molecular Testing With Duration of Respiratory Isolation for Patients With Possible Tuberculosis in a US Hospital

Key Points

Question  What is the feasibility, safety, and clinical impact of implementing molecular testing strategies to guide discontinuation of respiratory isolation among hospitalized patients undergoing evaluation for active tuberculosis (TB)?

Findings  In this cohort study with a pragmatic, before-and-after implementation design, a molecular testing strategy and the assay was associated with a significant reduction in median time to isolation discontinuation (2.9 vs 2.5 days), and hospital discharge (6.0 vs 4.9 days), and saved approximately $13 347 per isolated non-TB patient.

Meaning  Molecular testing appears to be associated with facilitating faster, more patient-centered care for individuals placed in respiratory isolation while undergoing evaluation for active TB in US hospitals.

Importance  New guidelines recommend that molecular testing replace sputum-smear microscopy to guide discontinuation of respiratory isolation in patients undergoing evaluation for active tuberculosis (TB) in health care settings.

Objective  To evaluate the implementation and impact of a molecular testing strategy to guide discontinuation of isolation.

Design, Setting, and Participants  Prospective cohort study with a pragmatic, before-and-after-implementation design of 621 consecutive patients hospitalized at Zuckerberg San Francisco General Hospital and Trauma Center who were undergoing sputum examination for evaluation for active pulmonary TB from January 2014 to January 2016.

Interventions  Implementation of a sputum molecular testing algorithm using GeneXpert MTB/RIF (Xpert; Cepheid) to guide discontinuation of isolation.

Main Outcomes and Measures  We measured the proportion of patients with molecular testing ordered and completed; the accuracy of the molecular testing algorithm in reference to mycobacterial culture; the duration of each component of the testing and isolation processes; length of stay; mean days in isolation and in hospital; and mean cost. We extracted data from hospital records and compared measures before and after implementation.

Results  Clinicians ordered sputum testing for TB for 621 patients at ZSFG during the 2-year study period. Of 301 patients in the preimplementation period with at least 1 sputum microscopy and culture ordered, clinicians completed the rapid TB testing evaluation process for 233 (77%).Among 320 patients evaluated in the postimplementation period, clinicians ordered molecular testing for 234 (73%) patients and received results for 295 of 302 (98%) tests ordered. Median age was 54 years (interquartile range, 44-63 years), and 161 (26%) were women. The molecular testing algorithm accurately diagnosed all 7 patients with culture-confirmed TB and excluded TB in all 251 patients with Mycobacterium tuberculosis (MTB) culture-negative results. Compared with the preimplementation period, there were significant decreases in median times to final rapid test result (39.1 vs 22.4 hours, P < .001), discontinuation of isolation (2.9 vs 2.5 days, P = .001), and hospital discharge (6.0 vs 4.9 days, P = .003), on average saving $13 347 per isolated TB-negative patient.

Conclusions and Relevance  A sputum molecular testing algorithm to guide discontinuation of respiratory isolation for patients undergoing evaluation for active TB was safe, feasible, widely and sustainably adopted, and provided substantial clinical and economic benefits. Molecular testing may facilitate more efficient, patient-centered evaluation for possible TB in US hospitals.

Nosocomial transmission of tuberculosis (TB) is one of the most feared public health consequences of a delayed TB diagnosis. Following several hospital outbreaks in the 1980s,^1-3 the US Centers for Disease Control and Prevention (CDC) issued guidelines on risk-stratification and infection-control measures to prevent such events.⁴ Last updated in 2005, these guidelines recommend use of administrative screening measures, personal-protective equipment including high-efficiency particulate respirators, and environmental controls including airborne infection isolation until highly infectious TB can be excluded.⁵ These procedures are resource-intensive, requiring private rooms with negative-pressure ventilation systems. Although these policies have helped reduce nosocomial TB transmission,^6,7 prolonged stays in isolation rooms are common because conventional rapid diagnostic testing for TB requires serial sputum collection for microscopic examination over 2 or more days.

A novel approach employs nucleic-acid amplification testing to guide discontinuation of respiratory isolation.⁸ Following introduction of a semiautomated, cartridge-based molecular testing assay (GeneXpert [Xpert] MTB/RIF; Cepheid)⁹ that provides testing results in less than 2 hours, we and others have identified the potential to substantially decrease the duration of isolation^10-12 and hospital costs^13,14 required to evaluate inpatients for active TB. Based on high-quality diagnostic accuracy and modeling studies,^10,11,13-15 regulatory authorities¹⁶ and professional societies¹⁷ have endorsed molecular testing strategies employing 1 or 2 sputum molecular testing assays, but little is known about their impact in routine practice. Therefore, we introduced a molecular testing assay strategy to guide discontinuation of respiratory isolation for patients undergoing evaluation for active pulmonary TB at a public hospital. We evaluated implementation outcomes, including adoption, feasibility, safety,¹⁸ and impact on time to completion of TB evaluation, time in isolation and in hospital, and hospital costs.

About 300 patients per year initiate and 250 patients complete rapid TB testing and respiratory isolation during evaluation for active TB at Zuckerberg San Francisco General Hospital and Trauma Center (ZSFG), a public teaching hospital serving the City and County of San Francisco, California. Prior to introducing molecular testing, ZSFG TB infection-control policies required all possible TB patients to stay in isolation for collection of 2 or more expectorated or induced sputa over 2 separate days for concentrated acid-fast bacilli smear microscopy and mycobacterial culture. Sputum concentration, smear preparation, and slide examination were carried out in a single batch once daily. Patients with a high clinical probability of TB were placed in airborne infection isolation; patients considered to have a low clinical probability of TB could be placed in respiratory isolation in conventional private rooms without negative-pressure ventilation systems if no airborne infection isolation rooms were available. Isolation could be discontinued for TB-negative patients when 2 or more sputa tested negative. Hospital discharges of possible TB patients required 3 or more negative and no positive smear results; 3 or more pending mycobacterial cultures; and authorization from the San Francisco TB Control Program.

In 2015, leaders from multiple departments at ZSFG and from the San Francisco TB Control Program developed a revised algorithm for discontinuing respiratory isolation incorporating sputum molecular testing. In constructing the new algorithm, stakeholders placed the highest priority on avoiding false-negative results and the next highest priority on shortening the time to final test results and the duration of respiratory isolation. The final algorithm recommended clinical assessments to guide how many sputum assay tests should be ordered and required that individuals be isolated for collection of 2 or more sputa for mycobacterial culture on 2 separate days. The algorithm allowed discontinuation of isolation after negative smear and/or molecular assay examination results of 2 sputa for patients with low-probability clinical presentations, or of 3 negative sputa results for patients with high-probability clinical presentations, as determined by the bedside clinicians (Figure 1). Finally, the algorithm recommended that clinicians assess the public health risk of TB transmission to determine whether 2 (if low risk) or 3 (if high risk) negative sputum examination results would be required before hospital discharge. We disseminated this algorithm to clinicians and bedside nurses via information sessions, handouts, wall posters in clinician work areas, a website,¹⁹ and prompts in the electronic order-entry system. Laboratory staff completed training on Xpert MTB/RIF assay procedures; there were no other laboratory interventions. Two physicians with expertise in TB (A.L. and J.L.D.) and an emergency medicine physician (D.D.) worked with stakeholders and with facilitators from the UCSF Caring Wisely Initiative (M.A.H., D.S., and L.E.G.) to plan implementation.

From January 28, 2014, to January 27, 2016, we performed a prospective, pragmatic, before-and-after implementation study to evaluate the molecular testing strategy introduced on January 28, 2015. We also assessed program sustainability from January 1, 2017 to December 31, 2017. We evaluated consecutive adults (≥18 years) undergoing sputum examination for Mycobacterium tuberculosis in the ZSFG Emergency Department or on the Inpatient Medicine or Family Medicine Services. We excluded patients with positive rapid TB test results (ie, by smear microscopy or Xpert) from our analyses of clinical efficiency and impact because they were not the target population for our intervention; discontinuing isolation for patients with active TB follows a longer process not reducible by assay testing. We included all medical inpatients admitted January 28, 2014, to January 27, 2016, in assessments for underlying temporal trends in study outcomes. All ordering, testing, and decision making were carried out by routine clinical and laboratory staff. All data were collected through routine hospital-information systems.

The University of California San Francisco Committee on Human Research approved the study protocol as quality improvement research and waived the requirement for informed consent. The Yale University Human Investigation Committee approved the study for analysis only. Cepheid, the manufacturer of the molecular testing assay, was not involved in study design or analysis.

In the postimplementation period, a clinical laboratory scientist performed Xpert MTB/RIF assay testing on unprocessed sputum according to manufacturer instructions using a GeneXpert XVI (Cepheid) instrument already in routine use for a variety of microbiologic assays. Previously developed laboratory protocols required 2 separate sputum samples of at least 1.0 mL each for molecular and conventional microbiologic testing by concentrated acid-fast bacilli smear microscopy and mycobacterial culture.¹⁰ If the number or volume of samples was insufficient, staff prioritized available specimens for molecular testing. Laboratory operating procedures stated that molecular assay testing would be performed and reported in the electronic medical record as soon as specimens were received in the laboratory on weekdays during daytime working hours. After hours and on weekends, molecular testing would be completed by the on-duty clinical laboratory scientist as soon as possible pending other requests for rapid microbiologic testing. As in the preimplementation period, smear microscopy results were entered into the electronic medical record once-daily as soon as they became available.

We calculated the proportion of patients with 1 or more molecular assays ordered in the postimplementation period as a measure of adoption of molecular testing by clinicians. We recorded the proportion of samples with adequate volume for analysis to determine the feasibility of simultaneously collecting 2 separate sputa for molecular and conventional microbiologic testing. We defined the final smear result as positive if there was any positive result among the first 3 sputa collected and negative if there were 2 or more negative results by smear examination and no positive smear results. We defined the final molecular assay result as positive if there were any positive result among the sputa examined and negative if all sputa examined were negative. We excluded patients who had fewer than 2 sputa examined by microscopy, if negative or missing, or no sputum molecular assay result for having an incomplete TB examination. We determined the accuracy of the microscopy and molecular testing strategies in reference to a gold standard of serial sputum mycobacterial culture, excluding those with fewer than 2 culture results unless culture-positive. We compared frequencies of false-negative results to assess the relative safety of each strategy.

To measure clinical efficiency and clinical impact, we calculated time intervals from the hospital admission order to several important time points in the TB evaluation process: (1) sputum collection, (2) sputum receipt in the laboratory, (3) reporting of first and final test results, and (4) hospital discharge. In addition, we calculated time spent in isolation from the order for its initiation until the order for its discontinuation. To measure impact on bed utilization, we calculated the mean number of (1) days in isolation and (2) days in hospital per rapid TB test-negative patient. We estimated mean costs per day for all participants using the US Centers for Medicare and Medicaid Services’ Principles of Reasonable Cost Reimbursement.²⁰ Finally, using these mean values, we projected annual hospital savings in isolation days, hospital days, and total costs, assuming 250 patients complete TB evaluation each year.

We defined time to first test result using the reporting time for the first smear result in the preimplementation period and the reporting time for the first Xpert result (if molecular analysis was performed) or the first smear result (if molecular analysis was not performed) in the postimplementation period (eTable 1 in the Supplement). We defined the time to final result in the preimplementation period using the reporting time for the second negative smear result. In the postimplementation period, we defined the time to final result using the reporting time for the second smear result if only microscopy was performed, for the second Xpert result if 2 or more molecular assays were performed, or for the second rapid test result (Xpert or smear) if only 1 molecular assay was performed.

We compared clinical and demographic characteristics; median time intervals for each component of the sputum testing process, respiratory isolation, and hospitalization; and measures of clinical efficiency and impact between the preimplementation and postimplementation periods. We evaluated statistical significance using χ² tests, Wilcoxon rank-sum tests, or t tests, as appropriate. We performed linear regression to assess trends in time in isolation and hospital length of stay in the preimplementation period. All analyses were performed using Stata statistical software (version 13, Stata).

Clinicians ordered sputum testing for TB for 621 patients at ZSFG during the 2-year study period (Figure 2). Of 301 patients in the preimplementation period with at least 1 sputum microscopy and culture ordered, clinicians completed the rapid TB testing evaluation process for 233 (77%). A similar proportion (259 of 320 [81%]) had TB evaluation terminated prior to completion during the postimplementation period (difference, −4%; 95% CI, −10% to 3%; P = .28). After introduction of molecular testing, clinicians ordered assays for 234 (73%) patients and smear microscopy without assay testing for 86 (27%) patients. Of those with assay testing ordered, 172 (74%) had 1, 56 (24%) had 2, and 6 (3%) had 3, for a total of 302 tests ordered. Results were reported for 295 (98%) tests; 6 (2%) samples had insufficient sputum for testing, and 1 sample provided indeterminate results. Overall, 228 (71%) patients received molecular assay results.

Median age was similar in the 2 periods (54 years vs 53 years, P = .76), as were the proportions of women (27% vs 22%, P = .20), homeless patients (20% vs 25%, P = .19), and persons living with HIV (34% vs 33%, P = .45) (Table 1). Ten (4.3%) patients before implementation and 9 (2.7%) after had positve rapid TB test results, including 6 (2.3%) assay-positive and 8 (3.7%) smear-positive after implementation. Eight (3.4%) patients evaluated before implementation and 7 (2.7%) evaluated after had positive M tuberculosis culture results. Forty-three (18%) patients evaluated before implementation and 58 (22%) evaluated after had sputum that grew nontuberculous mycobacteria. One patient with positive M tuberculosis culture results with a high clinical probability of TB initially had negative assay results and positive smear results, but subsequent sputa had positive assay results (eResults in the Supplement). Among 168 patients who completed both smear and assay, 1 patient with a positive assay result and negative smear result tested M tuberculosis culture-positive and 1 patient with a negative assay result tested positive for Mycobacterium kansasii by smear microscopy (eTable 2 in the Supplement).

Median time from hospital admission until initial sputum collection was 19.1 hours (interquartile range [IQR], 10.3-40.3) before implementation and 18.0 hours (IQR, 9.2-41.8) after (P = .62) (Table 2). Median time to first test result after sputum collection decreased from 18.4 hours (IQR, 15.5-23.6) before implementation to 4.6 hours (IQR, 3.4-6.9) after (P < .001). Median time to final test result after sputum collection decreased from 39.1 hours (IQR, 35.6-42.9) before implementation to 22.4 hours (IQR, 13.7-30.6) after (P < .001). Median time to hospital discharge after final test results were reported was 66.5 hours (IQR, 26.6-160.3) before implementation and 49.6 hours (IQR, 21.5-139.8) after (P = .08). Median hospital length of stay decreased from 6.0 days (IQR, 3.8-10.9) before implementation to 4.9 days (IQR, 2.9-8.9) after (P = .003). There were no significant temporal trends in hospital length of stay during the preimplementation period for patients who had negative rapid TB test results (P = .17). Moreover, median length of stay for all medical inpatients did not change from the preimplementation (3 days; IQR, 2-4; n = 11 287) to the postimplementation period (3 days; IQR, 2-4; n = 10 950).

Respiratory isolation data were available for 207 (93%) patients with negative rapid TB testing results before implementation and 226 (90%) after (P = .34). Median time from hospital admission to initiation of respiratory isolation was 2.4 hours (IQR, 1.2-15.7) before implementation and 1.8 hours (IQR, 1.0-9.0) after (P = .06) (Table 3). Median time between initiation of isolation and sputum collection was 12.9 hours (IQR, 6.6-19.3) before implementation and 13.5 hours (IQR, 5.1-29.1) after (P = .50). Median time from initial sputum collection to reporting of a final negative rapid TB test result decreased from 39.3 hours (IQR, 36.3-43.4) before implementation to 21.9 hours (IQR, 13.4-30.0) after (P < .001). Median time from a final negative rapid TB test result until discharge from isolation was 13.9 hours (IQR, 1.7-32.3) before implementation and 15.9 hours (IQR, 2.3-34.4) after (P = .52). Median duration of respiratory isolation decreased from 2.9 days (IQR, 2.0-3.7) before implementation to 2.5 days (IQR, 1.7-3.4) after (P = .001). There were no significant trends in length of stay in isolation in the preimplementation period (P = .29).

Among patients with negative rapid TB test results, mean time in isolation decreased 29%, from 3.9 days per patient before implementation to 2.8 days after (P = .03), and mean hospital length of stay decreased 27%, from 10.4 days before implementation to 7.5 days after (P = .01). Mean hospital costs per patient with negative rapid TB test results decreased from $46 921 before implementation to $33 574 after, providing average savings of $13 347 per patient. Estimating utilization and costs for approximately 250 inpatients completing TB evaluation each year, we project total annual savings to the hospital of 278 inpatient days in isolation, 705 inpatient days in hospital, and $3.3 million.

From January 2017 to December 2017, 293 patients had sputum examination for active TB ordered, including 205 (70%) with molecular assay testing. Compared with the postimplementation period, the proportion with molecular assay ordered was unchanged (−3.2%; 95% CI, −10.0% to 4.0%; P = .39).

Respiratory isolation is effective for reducing nosocomial TB transmission, but delays care and places a considerable burden on patients, clinicians, and hospitals. Molecular testing is simpler, faster, and more accurate than conventional microbiologic testing and has been deemed a public health priority, although it has not been widely adopted.^21-23 In this implementation study, we demonstrated that using a molecular assay testing algorithm to guide discontinuation of isolation for patients undergoing evaluation for active TB was safe and associated with meaningful reductions in time in respiratory isolation and in length of hospital stay compared with the conventional, microscopy-based testing strategy.

We documented favorable implementation outcomes and changes in several important process measures that emphasize the key role molecular assay testing had in increasing clinical efficiency and clinical impact. First, a large proportion of clinicians adopted the molecular testing strategy and usage was sustained 2 years after implementation. Second, we found that implementing a molecular assay to reduce turnaround time for testing, isolation, and hospital length of stay was highly feasible and did not affect the ability to complete culture-based evaluation. Finally, the molecular testing algorithm was cost saving compared with the conventional microscopy-based testing strategy. Together, these measures of impact place rapid molecular testing for TB among a select group of interventions that have been shown to advance the “quadruple aim”: improved population health, a better patient experience, a better clinician experience, and lower costs.²⁴

We previously predicted in a hypothetical study in the same setting that use of molecular assay testing could reduce time in respiratory isolation by approximately 2 days.¹⁰ During this real-world implementation study, however, we observed more modest reductions (median, 0.4 days; mean, 1.1 days). There are several possible explanations for these differences. First, clinicians did not order assay testing in about one-quarter of admissions, for reasons we did not evaluate. Second, our algorithm for discontinuing isolation required not 1 negative test by molecular assay as in the prior modeling study, but 2 negative tests by molecular assay and/or smear on 2 separate days, as well as completion of sputum collection in isolation. These more stringent requirements were intended to provide a margin of safety because rare false-negative Xpert results have been reported.^11,15 For similar reasons, current guidelines require 2 negative Xpert results.¹⁷

Among 168 patients who completed both smear and assay testing, we observed only 1 patient with a false-negative assay result, and risk-stratification in the molecular testing algorithm allowed this individual to be safely diagnosed on an additional sputum sent for molecular assay testing. The algorithm also detected 1 patient with TB who had a negative smear evaluation, who would have otherwise gone undetected. There were no false-positive molecular assay results. These results support the labeling of Xpert as safe and accurate for guiding discontinuation of isolation.¹⁶ Furthermore, they support the findings of multiple prior diagnostic accuracy studies^10-12,14,15 showing that 1 assay is likely sufficient in almost all patients, especially those with a low clinical probability of active TB. Given the low yield and substantial delays we observed when 2 assays were performed instead of 1, the recommendation from professional societies that all patients undergo 2 assay tests prior to discontinuation of isolation may be overly conservative.¹⁷ Our data, along with additional high-quality implementation studies to identify molecular testing algorithms that are not only safe but also patient centered, should inform revision of TB infection control guidelines from the CDC. Revision is urgently needed, because these guidelines have not been updated since the introduction of semiautomated testing by molecular assay.⁵ Because these guidelines determine the policies enforced by hospital accreditation agencies, updating them would likely help advance the CDC’s longstanding goal of increasing the proportion of patients with possible TB undergoing molecular testing.^22,25 In the interim, collecting sputum samples 8 hours apart as recommended by professional societies and the CDC may reduce the time to a final rapid TB test result.^17,22 Clinical efficiency and clinical impact of molecular testing algorithms may be further enhanced by increasing clinician adoption of molecular testing, and by decreasing time from hospital admission to sputum collection and from final results reporting to discontinuation of isolation, each of which delayed completion of TB evaluation by two-thirds of a day.

Our study had several limitations. First, it was conducted at a single academic center where clinicians have substantial experience with TB evaluation, potentially limiting generalizability. Nevertheless, our interdisciplinary approach of involving clinicians, laboratory leaders, and public health leaders from the TB and hospital infection control programs provides a model for implementation in different contexts. Second, before-and-after implementation designs are susceptible to false inferences if underlying temporal trends are driving changes attributed to the intervention. To reduce this risk, we compared 2 twelve-month periods before and after implementation to minimize the effects of seasonal variations in hospital census or experience among physicians in training. Furthermore, we identified no significant underlying temporal trends before implementation. Finally, we may have misestimated local cost savings because the reasonable costs methodology accounts for average rather than individual costs for services. Thus, we were unable to provide line-item comparisons of costs. However, we have previously used empirical costing to show that a shorter length of stay leads to cost savings (−$2483) for the molecular strategy ($15 285) compared with the microscopy strategy ($17 768), and that these savings outweigh the higher testing costs ($203 difference) for the molecular strategy ($218) compared with the microscopy strategy ($15, all costs in 2009 USD).^22,25 Moreover, the reasonable costs methodology may provide more relevant estimates of cost savings for hospital administrators than empirical costing because it is the approach recommended by the Centers for Medicare and Medicaid Services for determining cost-based reimbursement.²⁰

Our study had numerous strengths. First, we provide what we believe are the first published data on actual impact and implementation outcomes¹⁸ of molecular testing to guide discontinuation of isolation in a US hospital. Second, we employed a pragmatic, real-world study design that included consecutive, unselected patients referred by usual clinicians.^26,27 Clinicians were free to decide whether to use the molecular assay to guide discontinuation of isolation or not, and we extracted data on process measures, implementation outcomes, and service outcomes from routine hospital records. These design features enhance generalizability. Finally, we assessed outcomes important to both patients and hospital leaders, including clinical impact, safety, clinical efficiency, and costs.

Introducing molecular assay testing to guide discontinuation of respiratory isolation for patients undergoing evaluation for active TB appears to be effective for reducing time spent in isolation for patients in a US hospital where the frequency of active TB is low. Routine use of molecular assay testing should be strongly considered to provide faster, more patient-centered care to hospitalized patients undergoing evaluation for TB in the United States and other low TB-burden settings.

Corresponding Author: J. Lucian Davis, MD, Department of Epidemiology of Microbial Diseases, Yale School of Public Health, 60 College St, New Haven, CT 06510 (lucian.davis@yale.edu).

Accepted for Publication: June 9, 2018.

Published Online: August 27, 2018. doi:10.1001/jamainternmed.2018.3638

Author Contributions: Ms Chaisson and Dr Davis had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.

Study concept and design: Cattamanchi, Roemer, Handley, Goldman, Higashi, Winston, Haller, Luetkemeyer, Davis.

Acquisition, analysis, or interpretation of data: Chaisson, Roemer, Handley, Schillinger, Sur, Pham, Lin, Goldman, Quan, Perez, Healy, Higashi, Winston, Davis.

Drafting of the manuscript: Chaisson, Cattamanchi, Healy, Davis.

Critical revision of the manuscript for important intellectual content: Chaisson, Cattamanchi, Roemer, Handley, Schillinger, Sur, Pham, Lin, Goldman, Quan, Perez, Higashi, Winston, Haller, Luetkemeyer, Davis.

Statistical analysis: Chaisson, Sur, Pham, Davis.

Obtained funding: Handley, Schillinger, Davis.

Administrative, technical, or material support: Chaisson, Roemer, Handley, Schillinger, Sur, Pham, Lin, Quan, Perez, Healy, Higashi, Winston, Haller, Luetkemeyer, Davis.

Study supervision: Handley, Haller, Davis.

Conflict of Interest Disclosures: None reported.

Funding/Support: We acknowledge the support of grants from the University of California San Francisco/San Francisco Department of Public Health Caring Wisely Program (Drs Luetkemeyer and Davis); the San Francisco Hearts Foundation (Drs Haller and Davis); and the Johns Hopkins HIV Epidemiology and Prevention Sciences Training Program (Ms Chaisson, T32AI102623). We also acknowledge Cepheid for donating GeneXpert MTB/RIF test kits.

Role of the Funder/Sponsor: The Johns Hopkins HIV Epidemiology and Prevention Sciences Training Program and Cepheid had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication. The University of California San Francisco/San Francisco Department of Public Health Caring Wisely Program were advisors and coauthors on the project (M.A.H., L.E.G., D.S.).

Additional Contributions: We thank the patients, physicians, and staff of Zuckerberg San Francisco General Hospital and Trauma Center, especially the many individuals who contributed to the design and implementation of the molecular testing strategy for evaluating patients for possible TB.