Implementation of GeneXpert for TB Testing in Low- and Middle-Income Countries: A Systematic Review

This review highlights a commonality of implementation barriers across geographically dispersed GeneXpert interventions for TB testing. This indicates the importance of using implementation frameworks to report findings that can improve public health outcomes across low- and middle-income countries.


INTRODUCTION
A n estimated 10 million people were affected by TB globally in 2019, with the total number of deaths reaching 1.2 million people, 1 down from 1.5 million people in 2018. 2 Although TB is prevalent in all countries, the distribution shows a significant burden in lowand middle-income countries (LMICs). It is estimated that up to two-thirds of global incident cases are found a School of Public Health, Curtin University, Perth, Western Australia, Australia.
Global Health: Science and Practice 2021 | Volume 9 | Number 3 in only 8 countries, 5 of which are classified as LMICs. 1 In 2019, the highest proportion of new cases (44%) occurred in Southeast Asia. 1,3 In addition, approximately half a million cases of rifampicin-resistant TB were diagnosed in 2019. 1,4 Of these cases, 78% had multidrug-resistant TB (MDR-TB). 1 It has also been estimated that up to one-third of global TB cases and more than threequarters of MDR-TB cases are undetected, 3 equating to upward of 3.3 million people globally living with active TB who are unaware of their status and remain undiagnosed. 5 The global public health response to TB is guided by the World Health Organization (WHO) End TB Strategy 2,6 and the Sustainable Development Goals. 2 Both of these strategic documents prioritize TB as a health issue, aiming to achieve a 95% reduction in TB deaths and 90% reduction in the incidence rate by 2035. 6 Achieving these goals will require a high level of collaboration between regional, national, and international stakeholders working in partnership across a range of interventions targeting TB risk factors and priority populations. 7 These often include interventions that aim to reduce the time to diagnosis, provide more effective contact tracing, improve treatment adherence and outcome, enhance collaborative partnerships with HIV-specific programs, and prevent TB transmission. 3 Case detection interventions are an ongoing challenge, 5 often due to restrictive testing algorithms that reduce access to testing among people with presumptive TB who fall outside the scope of the algorithm. However, in resource-limited settings, removing these restrictions comes with significant cost implications. 8 Therefore, no single intervention will result in achieving the global TB elimination targets. 7 Modelling undertaken to predict the impact of a variety of interventions indicates that improvements in diagnostic testing will make a substantial contribution. 7 Earlier detection of TB and MDR-TB allows cases to be promptly triaged into appropriate treatment and care, leading to improved patient outcomes. 3,9 When comparing point-of-care testing approaches with more conventional laboratory-based testing procedures, the longer timeframe for delivery of laboratory results is often cited as a reason for disengagement from the treatment pathway, 10 with recent estimates suggesting only 56% of diagnosed MDR-TB cases worldwide are treated successfully. 2 In 2010, the WHO recommended the use of the Xpert MTB/RIF (Cepheid, Inc., Sunnyvale, CA), a test that simultaneously detects Mycobacterium tuberculosis (MTB) and rifampicin-resistant TB (RIF) strains, as the initial diagnostic for people presumed to have MDR-TB or HIV-associated TB in highincidence countries. 3,11 These recommendations have since expanded to all people with presumed TB while acknowledging resource implications within resource-limited settings. 12 The Xpert advances a health system's ability to diagnose and respond to TB as it has improved sensitivity in comparison to sputum smear microscopy. 5 Studies indicate that Xpert can detect TB in a large number of people that routine testing services cannot detect. 13 Therefore, new models of care are possible due to this rapid diagnostic technology that returns results within 2 hours.
Since the recommendation of Xpert, the number of programs using this new diagnostic technology to improve access to testing services has increased significantly. 14 By 2017, 23 million testing cartridges had been procured for use across 6,659 Xpert machines located in 130 countries. 11,15 However, requirements for specialist staff, temperature control, and a continuous power supply, as well as the high cost of purchasing the machine and ongoing testing cartridges, restrict the installation of Xpert in many locations. 11 The challenges associated with implementation differ significantly between high-income and low-income countries. 5,16 The impact of new diagnostic testing interventions is often dependent on the functioning of the overall system in which the intervention is being introduced, as well as the operational implementation of the Xpert program. 7,17,18 In LMICs, implementation challenges often impact the effectiveness of Xpert, with many failing to achieve the expected outcomes demonstrated by Xpert evaluations in upper-middleincome countries. 5 This includes indicators such as a reduced timeframe to deliver test results to patients or successfully transitioning newly diagnosed cases into quality treatment and care. 14,17,19 As countries begin to scale up Xpert-based interventions, understanding how contextual factors influence program impact is critical. 20 Implementation research is the scientific study of methods to promote the systematic uptake of research findings and other evidence-based practices into routine practice, and hence, to improve the quality and effectiveness of health services and care. 21 Furthermore, it explores the impact that implementation has on outcomes and identifies sustainable improvements. 22 This approach recognizes that challenges often experienced in real-world health settings impede meaningful outcomes for interventions that were previously proven effective Case detection interventions are an ongoing challenge, often due to restrictive testing algorithms that reduce access to testing among people with presumptive TB who fall outside the algorithm's scope.
As countries begin to scale up Xpertbased interventions, understanding how contextual factors influence program impact is critical.
in research studies. 23 For this reason, sharing knowledge from across Xpert sites is needed to improve real-world implementation and program outcomes. 13 With a higher burden of TB in LMICs, implementation research in this setting can advance understanding of how the challenges of implementation impact effectiveness.
This review aimed to assess the use of implementation science frameworks when reporting the enablers and barriers for the implementation of Xpert for the diagnosis of TB in LMICs. The specific objectives were to (1) identify approaches to implementing Xpert in LMICs, (2) determine the barriers and enablers across the identified Xpert testing programs, and (3) assess the use of implementation science frameworks in Xpert programs in LMICs.

METHODS
This qualitative systematic review was developed in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement, which ensures transparency in the formulation of findings. 24 The data analysis consisted of de-identified, publicly available data and therefore is exempt from ethics approval.

Eligibility Criteria
The PICOS (population, interventions, comparisons, outcomes, and study design) was used to frame the research question. The population was limited to LMICs as defined by the World Bank classification of low-income economies or lower-middle-income economies. 25 This was in recognition of the unique challenges associated with implementing Xpert within this setting. 5,16 Within this population setting, interventions that used the Xpert technology for TB testing in public or private sector facilities were included. In the data analysis process, a comparison of implementation approaches used across the included studies was undertaken. Therefore, a within-study comparison group was not required to meet eligibility. The outcome variables of interest were the identified enablers and barriers to implementation, as well as the implementation approach.
As the Xpert was recommended for use by the WHO in 2010, articles were limited to those published from January 1, 2011, to March 31, 2020. Included studies were peer-reviewed and published in English where the full-text article was available. Articles were excluded if they were not single country reviews and if they did not specify the implementation approach as well as identifying barriers and/or enablers to the implementation approach. Further to this, articles focused on the implementation of Xpert for TB in closed population settings, such as prisons or pediatric clinics, multicountry reviews, and interventions occurring under a research trial were also excluded.

Search Strategy and Study Selection
To identify relevant literature, PubMed, Medline, and Scopus databases were searched over 3 weeks beginning April 4, 2020. Search terms included the keywords "implementation," "GeneXpert OR Xpert" and "Tuberculosis OR TB." The preliminary search was performed in PubMed.
In total 2,296 articles were initially identified. Articles were extracted to EndNote X9.3.2 and 116 duplicates were removed. All articles identified underwent a review to assess relevance against the eligibility criteria. This review process occurred in stages, which initially involved screening the article titles (excluding 2,029 records). Then, the abstracts of the remaining 151 articles were assessed to identify those suitable for full-text review. In completing this process, a further 133 articles were excluded. This resulted in 18 articles undergoing an eligibility assessment involving a full-text review. A total of 7 articles were excluded as they did not address the barriers and/or enablers in the implementation of the intervention. A second fulltext review process was then completed involving all 3 authors, who assessed the remaining articles against the eligibility criteria and mutually agreed on the final selection of studies for the data extraction process. A total of 11 articles were included in the review (Figure).

Data Extraction and Analysis
Data were independently extracted from selected articles and transferred into an Excel spreadsheet to maintain consistency. Before the data extraction process, variables for extraction were agreed upon among all authors. The final variables included author, year, country of implementation, implementation approach (i.e., the characteristics of how an intervention is implemented, for example, a hub-and-spoke model or a point-of-care model), health setting, testing algorithm, public health impact, as well as the identified barriers and enablers to implementation and use of an implementation framework. The hub-and spokemodel was characterized by a centrally located laboratory performing the TB testing using the Xpert machine and returning the results to the community health clinic that initially collected

Study Characteristics
The final 11 articles were published between 2015 and 2020, with Xpert implementation occurring in Asia (n=5) and Africa (n=6). More details on the characteristics of the 11 studies are summarized in Table 1.

Implementation Approaches
Among the studies included in this review, the huband-spoke implementation approach for Xpert was used more often (n=6) 15,[26][27][28][29][30] than the point-of-care model (n=5). [31][32][33][34][35] As part of the hub-and-spoke model, varying approaches to the transportation of samples between the laboratory and clinic were identified, which included transportation by a health care worker, 30 employed drivers, 26,28 and informal couriers. 27 The return of results to the clinic that collected the patient sputum sample most commonly occurred via the same transport network, while a study in Uganda examined the use of SMS to return results directly to the clinic and patient. 15 Of the remaining interventions implementing the point-ofcare model, one had 2 distinct arms involving differing approaches in active case finding of symptomatic people. 35 This included health workers actively seeking sputum samples from hospital patients or community-based screeners referring symptomatic people attending private health clinics to a TB testing center with an Xpert machine installed.

Health Setting
For the hub-and-spoke models (n=6), the Xpert machine was installed at a combination of high level facilities such as regional (n=2), district (n=3) and urban (n=1) public hospitals (n=4), 15,26,28,29 or dedicated national TB diagnostic laboratories (n=2). 27,30 Information relating to the possible colocation of TB laboratories within hospitals was not provided. The majority of the spokes in these  implemented Xpert as a secondary diagnostic after smear-negative microscopy or a combination of both. 26,29,30 Across all studies, the broadest testing algorithm was defined as any person with presumptive TB (n=2). 15,35 The remaining studies provided Xpert testing to a combination of people deemed to be within an at-risk category, such as people living with HIV, 28,33,34 children aged 15 years or younger, 33 people with presumptive MDR-TB, 32,33 a contact of a known DR/MDR-TB case, 28,32,34 TB cases at risk of resistance, 32,34 health workers, 28 pregnant women or breastfeeding mothers, 28 prisoners, 28 refugees, 28 and diabetics. 28 One study limited the testing algorithm to cases of presumptive MDR-TB, as the intervention was focused on identifying MDR-TB cases. 32 Two studies did not indicate the testing algorithm used, both of which primarily focused on the technical use of the Xpert machine and identifying the cause of error in results. 27,31 Public Health Impact Seven of the 11 studies outlined some form of public health impact achieved by the implementation of Xpert. 15,26,29,30,32,33,35 Three studies assessed the Xpert within a continuum of care and highlighted treatment-related public health impacts within the cohort of TB cases detected by Xpert. 15,26,32 For example, the primary outcome identified by Hoang et al. 32 Table 1).

Implementation Barriers
The barriers were allocated into the following 6 categories: patient-level factors, human resources, material resources, service implementation, service coordination, and technical operations (Table  2). Of the total (n=28) barriers identified, 43% (n=13) were found to have occurred in multiple studies. The greatest number of barriers were categorized as service implementation factors (n=7). However, the most cited barrier was a service coordination factor (n=8) involving the lack of communication/referral pathways between staff in laboratories and health centers. [27][28][29][30][32][33][34][35] The second most common barrier related to inadequate and/or inconsistent staff training, which was identified relatively equally by studies in both Africa (n=3) 26,28,34 and Asia (n=4). 29,30,32,35 Overall, the hub-and-spoke model generated the most commonly occurring barriers. Of the studies that implemented this approach, Cowan et al. 26 was the only study that did not identify communication barriers between the testing laboratory and the clinic collecting patient samples. Further to this, delays in patient notification of results were also a common barrier associated with the hub-and-spoke model (n=5). 15,26,[28][29][30]

Implementation Enablers
Five studies 15,30,32,33,35 identified enabling factors in the implementation of Xpert for TB testing, including strategies such as taking an active casefinding approach, 35 expanding diagnostic algorithms, 35 and the daily transport of samples 15 ( Table 3). Of the studies that identified enablers, only 2 highlighted more than 2 factors. 30,35 Of the enablers identified, the single factor highlighted by more than 1 study was the addition of human resources to support the implementation of a new program. 30,35 For the study conducted in Mongolia, this corresponds with the identified barrier that relates to the workload capacity of laboratory staff. 30 Further enablers have a correlation with an identified barrier; for example, Uganda stated the daily transport of samples as an enabling factor, while also identifying inconsistent and/or delayed transportation of samples as a barrier to successful implementation. 15

Implementation Science
Of the studies included in this review, only 3 reported being guided by a structured reporting framework, 27,30,33 and only 1 was specifically for implementation studies. 33 These included the Strengthening the Reporting of Observational studies in Epidemiology (STROBE), 27 the Consolidated Criteria for Reporting Qualitative Research, 30 and the Standards for Reporting Implementation Studies (StaRI). 33

DISCUSSION
This review aimed to assess the use of implementation science frameworks when reporting approaches to implementing Xpert in LMICs and determining the barriers and enablers across the identified Xpert testing programs. The review found 7 of the 11 identified studies outlined some form of public health impact achieved by Xpert, which included treatment success (n=3) 15,26,32 and an increase in identification of active cases (n=3). 15,30,35 Mostly consistent barriers to implementing Xpert were reported across all the identified studies. This highlights a commonality of implementation barriers across geographically dispersed Xpert interventions in LMICs. In contrast, less than half of the studies articulated enabling factors in the implementation of Xpert for TB testing in LMICs. More consistent and transparent reporting using implementation science frameworks will The greatest number of barriers were categorized as service implementation factors; however, the most cited barrier was a service coordination factor.
Of the enablers identified, the single factor highlighted by more than 1 study was the addition of human resources to support the implementation of a new program. improve access to information that supports improved public health outcomes for real-world Xpert interventions in LMICs. 13

Implementation Approach
In this review, we found an integrated and coordinated approach was required when implementing Xpert models of care into a health setting. 5,28 This was particularly apparent in the implementation of the commonly used hub-and-spoke model. In high-income countries, this model of care is considered ideal for maximizing efficiencies and effectiveness for services that require advanced medical equipment, such as the Xpert. 36 However, in LMICs the hub-and-spoke model was found to be associated with the greatest number of barriers, highlighting a lack of integration and service coordination. Therefore, when deciding on a particular implementation approach, the existing context of a health setting and the expertise and needs of key stakeholders (e.g., clinicians, laboratories, and government) should be considered. 37 Recommendations to enhance integration and coordination included supporting continuous quality improvement of systems, as well as procuring and maintaining appropriate equipment, strengthening supply chains, having reliable specimen referral networks, suitable laboratory information systems, and proper laboratory training for staff. 37 This is reinforced by Rendell et al.,30 indicating that the mere introduction of Xpert does not automatically "guarantee a natural fit into the [existing] program environment." 30 There is a need for more consistent, transparent, and collaborative information sharing regarding the suitability of strategies to support the implementation and integration of Xpert programs at the local level. 37

Barriers to Greater Public Health Impact
Across all the identified studies, 2 main barriers to the identification of active TB cases were identified, the underutilization of Xpert and the inadequate identification of eligible patients. These barriers frequently resulted from a lack of communication/referral pathways between health centers and laboratories (n=8) [27][28][29][30][32][33][34][35] and inadequate or inconsistent training to support staff awareness and knowledge of testing and/or testing processes (n=7). 26,[28][29][30]32,34,35 Restricted testing algorithms were also likely to have affected the identification of eligible patients, as this adds a level of complexity in triage that requires specific staff training, which was lacking; therefore, staff were not adequately identifying eligible patients. Furthermore, the underutilization of Xpert was exacerbated when stocks of material resources such as sputum cups for sample collection were not maintained; 7 studies had periods of downtime due to inadequate supplies. 15,26,30,[32][33][34][35] These barriers are also reflected in multicountry reviews that analyzed quarterly reports and machine data. 13 As the Xpert allows for the timely turnaround of test results, a secondary public health impact of Xpert is the immediate initiation of treatment to prevent onward transmission. 33 However, in this review, 8 studies identified a delay in either samples being delivered for testing (n=4) 15,32,34,35 or a delay in the notification of test results (n=6). 15,26,[28][29][30]34 Accordingly, a delay in the notification of results reduces the opportunity to improve treatment initiation. 16,26 In examining the difference in time to treatment between patients diagnosed by Xpert and those diagnosed under the previous smear microscopy system, several studies (n=2) found that an Xpert diagnosis resulted in a longer time to treatment or loss to follow-up. 13,26,28 These outcomes are often noted as the result of operational challenges associated with program implementation. 17 With consistent reporting of barriers that inhibit the integration of Xpert into the existing health system, the overall public health impact of Xpert implementation in LMICs can be improved. 16

Implementation Enablers
In this review, only 5 studies specifically identified enablers for the implementation of Xpert for TB testing in LMICs. 15,30,32,33,35 Identifying improvements to implementation was a theme across 3 studies. 15,31,33 Gidado et al. 31 suggested that strategies for improved functionality of Xpert machines should be prioritized over the installation of new machines. Secondly, Cattamanchi et al. 15 reported on the practical benefit of highlighting enabling factors to improve implementation, reporting quality improvement Two main barriers to the identification of active TB cases were identified, the underutilization of Xpert and the inadequate identification of eligible patients.
initiatives that positively impacted outcomes along the TB treatment and care cascade including SMS communication for the delivery of Xpert results, and implementing a process to gather monthly performance feedback from health center staff. 15 Accordingly, sharing of knowledge via standardized reporting provides valuable information that supports the ongoing integration of Xpert for TB testing in LMICs. 13 In the longer term, consistent reporting and communication of the enabling factors to improve implementation will contribute to systemic improvement in the public health impact of Xpert as programs are scaled up across LMICs. 20,38,39 Implementation Science Implementation frameworks such as the Standards for Reporting Implementation Studies (StARI) and the Consolidated Framework for Implementation Research (CFIR) were developed with the knowledge that effective interventions are often found to be ineffective when implemented in a real-world context. 40,41 This is often due to the central role that context plays in understanding how factors such as the social, cultural, economic, political, legal, and physical environment may affect the intervention. 22 For programs to improve public health impact, there is a need for the continued development of tools and strategies that support successful implementation. 42 The similar or recurring barriers being experienced across LMIC Xpert interventions indicate the need for more consistent and transparent reporting methods to facilitate knowledge-sharing. Consistent reporting through implementation frameworks will increase understanding of the most effective implementation approaches and contextual influences and enable the scaling up of Xpert interventions. 20,22 Therefore, an increase in the use of implementation frameworks for planning and evaluation of Xpert programs has the potential to improve outcomes achieved by Xpert programs and accelerate the translation of research into policy and practice. 42

Limitations and Strengths
This review had several limitations. We searched 3 databases and restricted our search to English language publications. Further, of the studies included in this review, only 3 reported being guided by a structured reporting framework 27,30,33 and only 1 was specifically for implementation studies. 33 Also, not using the word "program" or "programmatic" may have reduced the number of search results. However, this review had several strengths including being guided by PRISMA, as well as the articles for inclusion being appraised by the lead author and 2 co-authors.

CONCLUSION
With a higher burden of TB in LMICs, implementation research can advance understanding of implementation barriers and enablers. This study demonstrates the commonality of these barriers across geographically dispersed Xpert interventions in LMICs. With greater transparency of these barriers and enablers, program planners can promote a more collaborative approach and adapt interventions to reduce the impact of implementation barriers. To build the evidence base and in turn improve the implementation and effectiveness of Xpert, it is recommended that programs use implementation science frameworks when conducting research and disseminating findings. Wider use of these frameworks will provide valuable insight and support the ongoing improvement of TB programs in LMICs.