Tuberculosis screening costs and cost-effectiveness in high-risk groups: a systematic review

Background Systematic screening for active tuberculosis (TB) is a strategy which requires the health system to seek out individuals, rather than waiting for individuals to self-present with symptoms (i.e., passive case finding). Our review aimed to summarize the current economic evidence and understand the costs and cost-effectiveness of systematic screening approaches among high-risk groups and settings. Methods We conducted a systematic review on economic evaluations of screening for TB disease targeting persons with clinical and/or structural risk factors, such as persons living with HIV (PLHIV) or persons experiencing homelessness. We searched three databases for studies published between January 1, 2010 and February 1, 2020. Studies were included if they reported cost and a key outcome measure. Owing to considerable heterogeneity in settings and type of screening strategy, we synthesized data descriptively. Results A total of 27 articles were included in our review; 19/27 (70%) took place in high TB burden countries. Seventeen studies took place among persons with clinical risk factors, including 14 among PLHIV, while 13 studies were among persons with structural risk factors. Nine studies reported incremental cost-effectiveness ratios (ICERs) ranging from US$51 to $1980 per disability-adjusted life year (DALY) averted. Screening was most cost-effective among PLHIV. Among persons with clinical and structural risk factors there was limited evidence, but screening was generally not shown to be cost-effective. Conclusions Studies showed that screening is most likely to be cost-effective in a high TB prevalence population. Our review highlights that to reach the “missing millions” TB programmes should focus on simple, cheaper initial screening tools (i.e., symptom screen and CXR) followed by molecular diagnostic tools (i.e., Xpert®) among the highest risk groups in the local setting (i.e., PLHIV, urban slums). Programmatic costs greatly impact cost-effectiveness thus future research should provide both fixed and variable costs of screening interventions to improve comparability. Supplementary Information The online version contains supplementary material available at 10.1186/s12879-021-06633-3.


Introduction
The 2020 Global Tuberculosis (TB) Report noted 3 million undiagnosed TB cases annually worldwide [1]. There is an urgent need to strengthen TB programme efforts to improve TB case detection of these missing millions. These efforts involve scaling up systematic screening programmes targeting household contacts and other Open Access *Correspondence: azwerlin@uottawa.ca 1 School of Epidemiology and Public Health, University of Ottawa, 600 Peter Morand Cresent, Ottawa, Canada Full list of author information is available at the end of the article high-risk groups, including people living with HIV (PLHIV), thereby linking the "missing millions" with diagnosis and treatment. The recent COVID-19 pandemic has dramatically decreased access to TB services and treatment globally, and thus there is an urgent need to get TB programmes back on track to avoid missing even more people who need access to care and treatment.
Passive case finding (PCF) has been the standard for TB diagnosis, and relies on symptomatic individuals selfpresenting to healthcare facilities for diagnosis and treatment [2]. Systematic screening is defined by the WHO as the "systematic identification of people at risk for TB disease, in a predetermined target group, by assessing symptoms and using test, examinations or other procedures that can be applied rapidly" [3]. Active case finding (ACF) is a term for systematic screening of people who do not seek out healthcare services [4] and intensive case finding (ICF) is a term for the regular screening for signs and symptoms of TB among people living with HIV [5]. Systematic screening is a broader term which encompasses all TB screening interventions that actively seek to identify people at risk of TB disease. Screening individuals earlier in their disease course decreases the time during which transmission can occur, with the hopes of reducing future TB incidence [6].
Individual programmatic components of systematic screening vary by setting, but may include: door-todoor symptom screening, targeted testing of asymptomatic household contacts of persons with TB disease, or screening campaigns among high-risk subpopulations [7]. The algorithms used for systematic screening can involve different screening tools, such as symptom screening or chest X-ray (CXR), and a variety of diagnostic tools including computer-automated detection (CAD) software for X-ray films, GeneXpert MTB/RIF assay (Xpert ® ), detection of mycobacterial lipoarabinomannan (LAM) antigen in urine, sputum smear microscopy and inflammatory blood work. The success of screening is dependent on both programme and population factors, such as the underlying prevalence of tuberculosis within a population (i.e., pretest probability of detecting TB). The generalizability of one screening strategy to another community cannot be assumed [8].
PLHIV are a key high-risk group for developing TB disease, accounting for approximately 8% of all TB cases globally [1]. Clinical risk factors, like diabetes mellitus, [9] and structural risk factors, such as people residing in prisons, [5] also greatly increase the risk of developing TB and have been suggested as high-risk groups to target for screening programmes. Screening programmes typically incur large costs [10] since high-risk groups for TB are often marginalized or living in difficult to access regions [11].
Economic evaluations may help in ensuring that limited resources are used wisely. The current published literature includes systematic reviews of screening interventions and their economic impact but focused on specific populations (i.e., persons experiencing homelessness or incarceration, PLHIV, immigrants or other high-risk groups) [12][13][14][15][16][17]. These reviews highlight that although systematic screening and active case finding is recommended in high-risk groups, there is a need for clearer guidance on which specific tools and screening algorithms or strategies are cost-effective, essentially highlighting the gap in knowledge that still exists despite WHO's endorsement of systematic screening. Two systematic reviews of ICF among PLHIV showed significant variability across countries and target groups of patients, but highlighted that ICF was cost-saving compared to PCF in high TB/HIV burden countries, though authors noted the lack of standardized methods for cost data collection [5,17]. The other reviews of the cost of TB screening were focused on a specific population and setting (i.e., immigrants in low-TB burden settings or contacts in Eastern Europe) [13][14][15] and recommended systematic screening for high-risk groups, but noted that there was limited data which was heterogeneous and of low quality. Therefore, our objective was to comprehensively synthesize economic evaluations of systematic screening for TB disease to inform a guideline development meeting leading to the updated guidance on TB screening, "WHO consolidated guidelines on tuberculosis. Module 2: screening-systematic screening for tuberculosis disease" [3] (see full list of PICO questions in the Additional file 1). Our study aimed to provide an up-to-date and comprehensive review of the economic evidence for all systematic screening interventions that was not limited to one sub-population, high-risk group, or setting.

Protocol
We performed a systematic review of the published literature on economic evaluations for TB screening with a focus on high-risk populations of persons with clinical risk factors such as PLHIV, diabetes or other respiratory diseases. We also focused on persons with structural risk factors, defined by the WHO as, "the circumstances in which people are born, grow up, live, work and age, " [18] including persons experiencing homelessness or residing in prisons, miners, elderly or indigenous persons. We sought to understand costs, cost-effectiveness, and affordability of screening approaches in key high-risk populations from the health system perspective. We performed this review according to the PRISMA guidelines (see Additional file 1) [19,20].

Information sources
We searched three online databases: Ovid, EMBASE and Scopus for new studies published within the past ten years (i.e., January 1, 2010 through February 1, 2020). This review was intended to inform an update to the current WHO guidelines. Technologies used for screening and diagnosis of TB have significantly improved in recent years, such as the development of rapid molecular tests for TB including Xpert ® MTB/ RIF (Xpert) and other diagnostics that were not available before 2010. A search strategy was developed to identify cost and cost-effectiveness studies of systematic screening in high-risk groups. We reviewed citations of all eligible articles, guidelines, and reviews for additional studies (see Additional file 1 for search terms).

Study selection
Studies were included if they evaluated any type of systematic screening activities among persons with clinical or structural risk factors and included costs. Our search terms were designed to broadly capture any economic evaluations or studies including costs and an outcome, such as disability-adjusted life years (DALYs) or qualityadjusted life years (QALYs) and was not limited to costeffectiveness or cost-utility analyses. Studies including utilities, such as DALYs, without costs were not included. Relevant studies were identified through electronic searches of the online databases, and duplicates were removed. Articles were excluded if they did not evaluate screening activities, or were reviews, letters, or opinion pieces.

Data collection
Articles were excluded if they only screened for latent TB infection, did not report costs per person screened or diagnosed or did not report the costs for the screening intervention separately from standard care (i.e., PCF). Studies were excluded if they evaluated screening activities in the general population, among contacts or children, since these groups were included in a separate manuscript (accepted, in press). Full text review was done independently by two reviewers (HA and BE) on remaining articles that met predetermined inclusion criteria, with all disagreements resolved by discussion with a third reviewer (AZ). No language filter was applied. Assessment of the quality of each economic evaluation and study quality was guided by the Consensus Health Economic Criteria (CHEC) [21,22] and CHEERS checklist, respectively (see Additional file 1: Tables S1 and S2) [21].

Data extraction
The study data elements extracted from each study included: primary research question, country and setting, year of study, patient population, clinical setting, type of intervention, comparison diagnostic scenarios, economic analysis perspective, analytic time horizon, type of economic evaluation, source of costing, primary and secondary outcome measures, type of model, sensitivity and uncertainty analyses performed and willingness-topay (WTP) threshold. The WHO's World Health Report guidelines on Choosing Interventions that are Cost-Effective (CHOICE) are the most commonly referenced WTP threshold among cost-effectiveness studies, particularly in low-and middle-income countries (LMICs) and are typically based on 1-3 times the country gross domestic product (GDP) per capita [23]. However, the use of these GDP-based thresholds has been challenged by many experts, the thresholds are considered overly simplistic and too easily attained when an intervention is effective [24]. Another key criticism of GDP-based WTP thresholds is the lack of their value and usefulness in assessing the trade-offs that decision-makers face in allocating limited healthcare resources [24]. Although the WHO no longer endorses GDP-based WTP thresholds [25], the challenge remains for clinicians, programme managers and researchers to determine the best metric for assessing value and reporting outcomes for cost-effectiveness and affordability of healthcare interventions. For the purposes of this review and to enable comparisons across currently published studies, we have included the GDPbased WTP thresholds used by study authors. However, we support the need for better decision-making tools for resource allocation in the local context, bearing in mind opportunity costs and the burden of disease.
Model parameters were extracted including epidemiologic, treatment and outcome parameters. Key outcomes included: cost per patient diagnosed, and incremental cost-effectiveness ratio (ICER) per utility measure (e.g., DALY averted). Costs are presented in United States Dollars (USD) adjusted to 2019 [27,28]. Given the heterogeneity of study setting, year and type of screening strategies employed, there was no plan to calculate global estimates or pool data.

Study selection
We identified a total of 3481 articles through database searching (Fig. 1). After duplicate removal, we screened 2318 citations by title and abstract for inclusion. Of these, we assessed 145 full-text publications against our inclusion criteria and excluded 118 publications. Exclusions were mainly due to the wrong intervention or no economic evaluation. A total of 27 studies were identified for inclusion in the review [11,.

Study findings
We present the study findings stratified by high-risk subgroup, including persons with clinical and structural risk factors and PLHIV, as well as by screening tool used below.

Screening PLHIV with molecular rapid diagnostics
All seven studies that assessed screening in PLHIV using Xpert ® MTB/RIF as an initial test were conducted in Sub-Saharan Africa, with 6/7 concluding screening was cost-effective. Two studies among PLHIV in South Africa conducted initial WHO-recommended four symptom screen (W4SS), including screening for cough of any duration, weight loss, fever or night sweats [55], followed by screening with Xpert ® and found screening to be costeffective  [40].

Screening among PLHIV using SSM
Kranzer et al. performed an intervention of adding TB symptom screening and SSM to an existing mobile HIV testing clinic in South Africa, [37] with an average cost of US$762 per TB case diagnosed [37].      In Sub-Saharan Africa, the average cost per person screened using rapid molecular diagnostic tests ranged from $3 to $116. Screening programs were found to be cost-effective in 6/7 (86%) of studies, with the following ICERs reported: $324 per additional TB case diagnosed; $37-$79 per DALY averted; $596-$4096 per YLS; and $517 per YLG. However, one study in Malawi found that screening with Xpert was not cost-effective at low test volumes, with an ICER of $1,980 per DALY averted. Cost-effectiveness depended on the volume of tests conducted annually and prevalence of TB b Data was available for among outpatient PLHIV using CXR as a screening tool from six studies of good quality overall. Three studies, from South Africa, Russia and Vietnam, included health system costs as well as unit tests costs, and reported an average cost per person screened using CXR ranging from $4 to $29. Three studies in Sub-Saharan Africa reported the average cost per person screened using CXR ranged from $6 to $52. The average cost per TB case diagnosed ranged from $106-$570. Screening with CXR was shown to be cost-effective among PLHIV with ICERs of $6245 per QALY and $536 per YLS c Limited data was available for screening among outpatient PLHIV using CRPs from two studies both of which were of good quality. The average cost for test costs alone of outpatient PLHIV in Uganda screened using CRP with Xpert, TB-LAM and culture for diagnosis ranged from $13 to $34 per person screened and was shown to be cost-effective with an ICER of $517 per YLS or $69-$92 per additional TB case diagnosed in Uganda d Limited data was available from one study conducted at an HIV testing clinic in South Africa for screening using SSM alone. This study reported an average cost of $762 per TB case diagnosed at a mobile testing clinic

Persons who live in urban slums
Three studies examined door-to-door screening interventions in urban slums. James [43].

Members of indigenous populations
Sohn et al. reported a cost of US$3-$5 per person detected with TB for a programme of home visits by community health workers (CHWs) to indigenous persons in rural India. This study included the costs for CHWs time to travel to homes, conduct screening and diagnostic visits, provide directly observed therapy (DOT) and care, as well as the laboratory and administrative services, [54] but did not account for routine TB care and medications provided by the Indian government.

Persons residing in prisons
Four studies provided direct evidence for the costs of screening among persons residing in prisons [38,47,48,51]. Zishiri [50]. Among elderly persons in China, additional risk factors (i.e., male, tobacco use or close TB contact) were associated with higher average costs per patient diagnosed [50].

People living in remote areas
In Papua New Guinea, Karki

Discussion
Our review of the published literature identified 27 studies of systematic screening among high-risk groups for TB, such as PLHIV, miners, persons residing in prisons, and the elderly. Our review found that systematic screening approaches are most likely to be cost-effective in settings and or populations with high prevalence of TB, such as PLHIV, persons residing in prisons and urban slum dwellers. Studies demonstrated that initial screening with more costly diagnostic tests was not cost-effective in high-risk groups, except among PLHIV. Simple, inexpensive initial screening methods (i.e., W4SS or CXR) followed by molecular diagnostic tools (i.e., Xpert ® ) among the highest risk groups in the local setting are the most cost-effective approaches to systematic screening. In high TB prevalence settings, door-to-door symptom screening in densely populated areas (i.e., slums), was generally shown to be cost-effective. However, mobile CXR units were not cost-effective due to high programmatic costs, particularly when interventions were targeting hard to reach populations (i.e., persons experiencing homelessness or IDUs). There was limited evidence identified for each high-risk group included in this review, thus caution should be used when extrapolating from a small number of studies. Our review found the most evidence for cost-effectiveness of screening programmes among PLHIV. Despite varying screening strategies (i.e., W4SS, CRP, CXR, Xpert ® MTB/RIF) and patient settings (i.e., in-patient or outpatient clinic), the majority 9/10 (90%) of PLHIV studies that calculated an ICER found screening interventions to be cost-effective among PLHIV using an author determined WTP threshold. Key drivers of costs among PLHIV included: annual test volume and diagnostic test costs, [52] underlying prevalence of TB and HIV, [42,52] clinic setting, [32,44,45] and programmatic costs included (i.e., transportation, mobile van units, or staffing costs) [31,32,55]. In sensitivity analyses of low-TB prevalence settings among PLHIV, screening strategies with simple tools, such as W4SS and SSM, were cost-saving compared to more expensive tools such as CXR or Xpert ® , [39] highlighting the importance of the local setting.
Screening interventions in high TB prevalence settings, such as urban slums or among persons residing in prisons, increased identification and diagnosis of people with TB and were shown to be cost-effective [46,48]. Equally important was ensuring proper follow-up to avoid treatment failure or loss post-screening [44]. However, among groups with other medical conditions, high programmatic screening costs, coupled with low TB prevalence meant screening interventions were not costeffective, but was limited to two studies [32,38].
Systematic screening is an expensive undertaking, particularly compared to PCF, when it involves mobilizing staff to go into the community. Evidence from this review suggests that community-based interventions [43] had higher costs compared to systematic screening programmes targeting persons presenting to healthcare facilities [35]. The use of artificial intelligence (AI) to inform screening tools through the use of CAD software is an exciting prospect for improving the efficiency and affordability of screening from standard CXR. However, our review did not find any studies which used CAD software in the context of a systematic TB screening programme. Indirect evidence, not presented in this review, has shown that the unit costs for each CXR read with CAD software are likely to be small but require significant investment in equipment and maintenance costs as well as the purchase of CAD software. These new technologies are still being evaluated in many programmes and highlight the need for costing and cost-effectiveness studies to inform their use in the programmatic setting.
The WTP thresholds, which are noted a priori by study authors and based on WHO recommendations, determine whether a given ICER is considered cost-effective. Among the papers reviewed in this analysis, many used either a country's GDP, or twice the GDP, as the WTP threshold, however there was significant heterogeneity due to the range of country settings. This variability makes comparisons challenging, particularly since setting a higher threshold increases the likelihood of an intervention being considered cost-effective. Careful attention should be paid to the WTP threshold employed by study authors when interpreting cost-effectiveness. Furthermore, there are concerns about using GDP per capita as the basis of determining cost-effectiveness, particularly in LMICs where there are more stringent resources constraints [23]. A key concern is that using a threshold that is too low (i.e., GDP per capita) may result in health systems choosing not to adopt an intervention that would generate net health benefits because the threshold does not take into account health opportunity costs [56]. Current efforts to develop country-specific estimates that account for opportunity costs, as well as updated data on population and economic growth, are underway and aim to provide better options for informing decision-making and resource allocation for health interventions [24,25,56].

Limitations
The heterogeneity around reporting of costs and costing components made comparisons across studies challenging. For instance, some costing analyses accounted for all operational and personnel-related costs, and thus reported higher total costs, while other analyses report only direct costs related to diagnostics testing [11,38]. There were no standardized screening algorithms, even across similar high-risk groups, and employed different standards of care (i.e., PCF or symptom screen alone or with CXR) which limits comparability of studies and generalizability to other settings. Not all studies describe cost components in the same manner [11] and comparisons across studies is further impeded by a range of primary outcomes from cost per case diagnosed, $/ DALYs averted, $/QALY, or $/YLS. Thus, even among the studies that do calculate an ICER, direct comparison is not necessarily appropriate. No included studies assessed the impacts of screening on earlier case detection and proper TB treatment, but this is an area that merits additional evidence to better understand the impact on cost and cost-effectiveness of preventing additional disease transmission.
Our study was restricted to the published literature and thus is likely impacted by publication bias towards those interventions that were shown to be cost-effective; programmes that were not deemed cost-effective may not have been published. A recent Task Force Report from the Professional Society for Health Economics and Outcomes Research (ISPOR) suggests benchmarking approaches, such as reviewing trial protocols, to better explore the potential for publication bias but also notes the need to develop new approaches to assess publication bias [57]. Further, more recent economic evaluations of screening interventions, particularly for novel diagnostic tests such as Xpert ® , may have been more likely to demonstrate cost effectiveness than earlier studies due to consistently decreasing test costs.
Our review highlights key gaps in the existing economic evidence, namely the need for more studies on the costs and cost-effectiveness of systematic TB screening programmes from Latin America and Asia, since the majority of included studies took place in SSA. Access to various screening and diagnostic tools was not consistent across study settings. Increased efforts should be made to ensure availability of newer diagnostic technologies to TB programmes globally. In addition, standardization of systematic screening interventions, along with fixed and variable costs included in economic analyses of programmes, is needed for better evidence generation and comparability across studies.

Conclusions
The COVID-19 pandemic has dramatically impacted TB services globally. Modelling has shown that COVID-19 related restrictions and interruptions to TB programmes may result in an increase in TB incidence up to 6.3 million, and mortality to 1.5 million, by 2025 [58,59]. Our review is the first to summarize the economic evidence for systematic screening for TB disease among high-risk groups. Our review highlights that to reach the "missing millions", and address the setbacks to TB services due to the COVID-19 pandemic, TB programmes should focus on simple, cheaper initial screening tools (i.e., symptom screen and CXR) followed by molecular diagnostic tools (i.e., Xpert ® ) among the highest risk groups in the local setting (i.e., PLHIV, urban slums). Programmatic costs greatly impact cost-effectiveness thus future research should provide both fixed and variable costs of screening interventions to improve comparability. COVID-19 has dramatically increased the number of digital applications for contact screening, as well as other video or telehealth options for service delivery. Expanded use of such digital technologies can be leveraged for TB screening to improve identification and treatment options for patients globally.