Introduction

Tuberculosis (TB) is the second leading infectious cause of morbidity and mortality worldwide, trailing after SARS-CoV-2, with an estimated 9.9 million new TB cases and 1.5 million deaths in 2020 [1]. In 2014, the World Health Organization’s End TB Strategy called for a 90% reduction in TB incidence and 95% reduction in TB mortality rates by 2035 [2]. Similarly, the Stop TB Partnership’s Global Plan to End TB, launched in 2019, calls for UN member states to successfully treat 40 million people with TB and provide TB preventive therapy to at least 30 million people by 2022 [3]. But despite these ambitious goals, TB incidence and mortality are falling at no more than 2–4% per year—far below the reduction needed (>10%) to achieve global targets [1].

Currently, it is estimated that about 30% of people who develop active TB every year will not be notified to public health authorities–largely reflecting underdiagnosis and undertreatment [4]. As people with TB who are missed can perpetuate transmission and suffer the adverse consequences of untreated disease (including death), it is imperative to identify these individuals early and ensure the rapid uptake of TB treatment, particularly among at-risk populations. Public health interventions such as intensified case finding (ICF), active case finding (ACF) and other approaches to improve gaps in the TB care cascade are therefore critical components of a comprehensive strategy to reduce the burden of TB worldwide [5, 6].

Since 2010, the TB REACH initiative of the Stop TB partnership (UNOPS), supported by Global Affairs Canada, USAID, and the Bill and Melinda Gates Foundation, has funded 8 waves consisting of 313 projects in 54 countries that focus on adopting innovative (both technological and process) approaches to improve TB case detection and treatment. In particular, Wave 5 –the focus of this analysis–focused on innovative approaches to case finding. These projects have made important contributions in innovating and promoting TB case finding activities in many high-burden TB countries [7, 8]. However, TB case finding and treatment projects are resource intensive, and their cost-effectiveness–both in absolute and comparative terms–remains uncertain [9, 10].

Given the substantial investment made in these projects (over 155 million USD since the inception of the initiative), it is critical to understand the relative value generated by different types of case finding interventions. In this study, we used the project database of the TB REACH Wave 5 funded projects to comprehensively and systematically evaluate costs and cost-effectiveness across the wide range of case finding and treatment support projects supported by this initiative. In doing so, we compared cost-effectiveness ratios (CERs) [6] assessed for common outcome units that are most direct and readily calculated to measure ACF performance based on each project’s scope of operations (1. TB case finding; 2. Linkage to care; and 3. TB treatment support) to rank and compare various factors influencing CERs.

Methods

During the wave 5 funding cycle, TB REACH funded 32 case finding projects across 20 different countries with total support of 16 million USD. An additional six small grants were provided to develop tools or products. While the main focus of the funding cycle was TB case detection, the overall scope of projects funded was broad; some examples of these innovations included novel approaches to case finding (e.g. recruiting civilians as TB finders in the community and strengthening public-private mix (PPM) partnerships), scaling-up previously proven concepts (e.g., engaging community health workers in rural areas in active case finding), improving treatment referral and adherence among individuals diagnosed with TB through active case finding, and increasing awareness regarding TB infection in the community (e.g. involvement of mass media, implementation of educational programs, and community engagement).

Screening and data extraction

We first obtained a complete set of administrative documentation–including project applications, reviews of project activity and financial reports–for each project funded during TB REACH wave 5. We then created a standardized data extraction spreadsheet, the components of which were based on a complete review of 32 selected case finding projects prior to data extraction.

Wave 5 projects that did not report relevant cost or program yield/performance data to TB REACH were excluded from the study. After consultation with TB REACH technical officers, three additional projects were excluded from the analysis. The first project (NTRL, EPHI) was a lab-based assessment of a novel transport and decontaminating reagent for TB testing, called OMNIgene® SPUTUM. The second project (Ifakara Health Institute) was intended to assess the use of Xpert Omni/Xpert MTB Ultra cartridges. However, these cartridges were unavailable during the wave 5 funding cycle; therefore, the project was not able to begin activities. The third project (AIGHD) aimed to establish TB screening in an HIV community testing project that was similarly postponed.

Two authors (IG and CD) independently performed the data extraction by reviewing all relevant documentation and data for each project. All disagreements between the two authors were resolved by discussion. If a consensus could not be reached, the two senior investigators (DD and HS) were consulted. During these meetings, the four authors re-evaluated the financial report in question and/or sought additional information by the TB REACH technical team (PV, AK, and JC), who provided further detail and clarification on data discrepancies and any project-specific interpretations (e.g. successes and/or challenges reported by each project impacting interpretation of data parameters).

For each program, we assigned a unique code (Table 1) and extracted data from the finalized financial statement, determined based on the last update date for each project’s financial statement (Dec 31^st, 2018).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Project characteristics and description.

https://doi.org/10.1371/journal.pone.0270816.t001

Variables directly collected from financial statements included the characteristics of each program, the country in which the project was executed, a brief description of the program’s primary activities (i.e. community-based screening, scale-up of previous concepts, testing of new sample transport or drug delivery systems, etc.); location of project operations (facility-based, door-to-door screening by community health workers, etc.); detail regarding the program’s target population (i.e. general population, gender-based or geographical subpopulations, etc.); screening, diagnosis and treatment services; and available technology (e.g. mobile-health tools for screening, mobile chest X-rays, Xpert MTB/RIF assay), as well as reported expenditures. For each project, all financial items reported, including total budget, income received, and cumulative expenditure, were extracted separately and reported in 2017 US dollars.

Upon full review of all of the projects included in our study, we defined five major categories for further subgroup analyses. These subgroupings included: 1. Technology Innovation, 2. Public-Private Mixed (PPM) Partnerships (or Private Sector Involvement), 3. Hard-to-reach Populations (e.g. villages, camps, geographically isolated regions), 4. Pregnant Women or Pediatric Cases, and 5. Door-to-Door Screening. We also noted the projects that supplied or linked patients to preventive therapy and projects implementing ICF activities (Table 2); however, specific data or information to apportion costs or assess programmatic yields for preventive therapy were not explicitly reported. Therefore, cost-effectiveness ratios could not be estimated for provision of preventive therapy.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Projects organized by subgroup.

https://doi.org/10.1371/journal.pone.0270816.t002

Additionally, we extracted data on project service outputs, including number of people diagnosed with any type of TB, number of people started on TB treatment (notifications), and number of patients successfully treated (Table 3, S1 and S3 Tables in S1 File).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Cost-effectiveness of TB REACH Wave 5 projects by project type.

https://doi.org/10.1371/journal.pone.0270816.t003

Results

Of the 29 projects included in our analysis, 11 solely focused on case finding while 18 had additional programmatic aims (nine included linkage-to-care, and nine also included patient treatment support). Most projects were implemented in the African region (n = 12, 41%) or South-East Asia region (n = 9, 31%). Ten projects (34%) were implemented in rural areas, and the other 19 focused on urban settings. In addition, 11 projects (38%) specifically targeted vulnerable populations (e.g. internally displaced persons, children, miners, female sex workers, people living with HIV, pregnant women, etc.) 8 projects (28%) focused on scalability (Table 1).

The weighted average CER was $184 (Range: $30 –$10,497; n = 29) per TB case detected across all projects (Table 3). For case-finding only projects, the average cost per case detected was $132 (Range: $30 –$10,497). For projects that included both case finding and treatment initiation, the weighted CER was $342 (Range: $103 –$9,913). Projects with additional programmatic efforts toward treatment adherence and patient support had a weighted mean cost per case detected of $254 (Range: $141 –$3,233).

For those projects that included linkage to care efforts after TB diagnosis, the average cost per patient referred for treatment was $30 (Range: $8 –$695) (S2 Table in S1 File). For projects that included treatment adherence and patient support programs, the average cost per TB patient completing the treatment was $40 (Range: $8 –$160) (S3 Table in S1 File).

Six projects were identified as projects with CERs above a $1,000-per-case-detected threshold. Two of these projects–NAANK and ASOCI–included treatment support efforts and were implemented in upper-middle income countries. Thus, only three case-finding only projects (CHEAS, IRDSA, and FUNDA) and one project with linkage-to-care (GLOHI) had an estimated cost per case diagnosed higher than the corresponding country’s per-capita GDP (Tables 1 and 3). Programmatic setbacks were the potential reasons for these inflated costs and will be discussed further in the Discussion section.

CERs also varied with characteristics of the underlying setting. Projects in urban settings had lower CERs than those in rural contexts (e.g., $169 vs. $242 per case detected). Moreover, the CERs of projects in the African region were generally higher than of projects performed in Southeast Asia ($554 vs. $95 per case detected) (Fig 1). This finding reflected two data trends. First, a small number of African projects had very high CERs, and these projects had greater influence on average CER values. Second, on average, projects from Southeast Asian countries diagnosed more people with TB, thereby lowering the estimated cost per TB case diagnosed compared to projects from the African region (Table 3). Cost per case diagnosed increased with the corresponding country’s per-capita GDP ($543 per $1000 increase, 95% confidence interval: -$53, $1138).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1.

Cost-effectiveness (cost per case diagnosed) of TB REACH Wave 5 projects focused on a) case-finding only and b) case-finding and treatment support.

https://doi.org/10.1371/journal.pone.0270816.g001

This plot illustrates the cost-effectiveness ratio (2017 US dollars per case of tuberculosis diagnosed) associated with each project, according to the gross domestic product (GDP) per capita in each corresponding country. Letters represent the geographic region in which the projects were performed, and numbers order projects from largest (1) to smallest within each region. In each panel, there was one project that was not shown because its associated cost-effectiveness ratio was exceptionally high (Panel A, project A10/IRDSA, cost per case diagnosed $10,497, GDP per capita: $6,133; Panel B, project A12/GLOHI, cost per case diagnosed $9,913, GDP per capita: $515).

Subgroup analyses (S5 Table in S1 File) suggested that the average cost per case detected was highest for the projects that involved door-door screening ($361, Range: $160 –$589; n = 5), followed by projects targeting pregnant women and children ($192, Range: $46 –$10497; n = 13), projects serving hard-to-reach populations ($187, Range: $103 –$9913; n = 7), and those involving technology innovation ($169, Range: $30 –$10497; n = 10). Projects that aimed to engage the private sector through PPM partnership had an average cost of $68 per case detected (Range: $30-$739, n = 5).

In sensitivity analysis, variation in effectiveness estimates tended to have greater influence on estimated CERs than variation in costs. Varying both costs and outcomes by +/- 25% did not affect on the characterization of projects as cost-effective (based on a cost per case diagnosed below a threshold of GDP per capita), with the sole exception of the ACREO project, which fell above this threshold when total costs increased, or case detection decreased by 25% (S1 and S3 Tables in S1 File).

Discussion

This comparative assessment of 29 projects designed to strengthen the TB care cascade highlights the heterogeneity in cost and cost-effectiveness observed when implementing interventions with similar aims in diverse contexts. Specifically, while the majority of projects diagnosed people with TB at a cost of less than $1000 per case detected, quantitative estimates varied over 100-fold depending on the local setting, target population, specific technology or other intervention employed, methodology of implementation and assessment, and objective. These findings demonstrate the importance of considering local context and realities of implementation when evaluating cost-effectiveness and argue against making blanket statements about the cost-effectiveness of certain interventions (e.g., TB case detection). Furthermore, our specific results can be helpful to implementers and funders seeking to introduce interventions to strengthen the TB cascade of care in cost-effective fashion across a variety of diverse settings.

Despite this heterogeneity, important generalizable insights were discernible in these data. ACF projects included in our study had an overall weighted average cost per case detected under $300, below the midpoint of the corresponding opportunity-cost-based cost-effectiveness thresholds (CETs) for low-and-middle-income countries as estimated by Woods et al [12]. In most high-burden contexts, the long-term cost per disability-adjusted life year (DALY) averted has been estimated to be only modestly higher than the cost per case detected through active case-finding [6]. Thus, it is likely that these interventions would fall below country-specific CETs assessed in terms of GDP per capita in most settings with high TB burden. CETs have known limitations, and these data should not be used on their own to suggest that any specific TB case-finding intervention is cost-effective [13]. Nevertheless, these data can provide some guidance regarding the value for money of interventions to strengthen the cascade of TB diagnosis and care. Cost-effectiveness ratios were lower for projects implemented in low-income settings (where TB incidence is higher) and rural areas. Targeting hard-to-reach populations was generally not associated with an increase in cost per case detected. The cost of treatment support was generally lower than the cost to diagnose a TB case, suggesting that closing case-finding gaps requires considerably larger resource dedication than support of patients who have already been diagnosed and started on treatment. Cost-effectiveness is also more favorable in settings with higher TB incidence, as the costs of screening result in more people with TB detected and treated. These findings can help funders and policymakers prioritize project implementation and evaluation in the future. In particular, for global funding mechanisms such as TB REACH and Global Fund, interventions focused on the most disadvantaged populations (e.g., the rural poor in low-income settings) may offer optimal value for money.

Our methodology–including systematic extraction of data from standardized financial and annual reports–facilitates comparison across projects and may be useful to funding agencies when seeking to draw comparative insight on cost-effectiveness across a pre-defined set of projects. As an external benchmark of the validity of this approach, we compared our estimates with the results from a recent study by Jo et al which comprehensively evaluated the costs and cost-effectiveness of one of the projects included in this analysis (CIDRZ Zambia) [14]. Jo and colleagues used a much more detailed approach to estimate the costs of each programmatic component of this intervention and reported a cost of $435 per patient initiating treatment [14]. Our simplified budget/finance statement-based calculation of the CER for this project was $420 per TB case diagnosed and $486 per patient initiating treatment, thus showing good agreement. The similarity of these findings may add a degree of external validation to the simple finance-report-based calculations performed here.

Four projects were identified as having higher cost-effectiveness ratios. On deeper review, these projects each confronted major operational difficulties during implementation. CHEAS experienced a major delay in implementation owing to challenges in hiring and retaining staff and underlying political unrest. IRDSA reported challenges in making household visits due to security concerns and a lower-than-expected number of patients with undiagnosed TB seeking services in the clinics. GLOHI reported that TB incidence in their region may have been overestimated (thereby resulting in fewer patients identified) and experienced multiple technical failures in drone usage, which hindered efficiency and implementation. The FUNDA project was substantially delayed in receipt of ethical approval, leading to lower-than-expected rates of TB diagnosis. These logistical challenges speak to the unpredictable nature of scaling up health interventions in real-world settings and the resulting variation in cost-effectiveness that will likely be observed in actual implementation. These project-specific findings should not be interpreted as favoring one intervention over another–as such barriers to implementation are generally unexpected and often not related to the actual type of intervention performed.

There are several limitations to consider when interpreting our study’s findings. Firstly, cost and project activity data available (level of data) for each project were highly heterogenous in that systematic assessment of activity-based costs and declassification cost data (e.g. capital assets, fixed and variable costs) across the projects were not possible. As such, we were only able to use simplified apportionment criteria to allocate total costs for major programmatic activities only: case finding, treatment initiation (linkage to care), and treatment management. This may have resulted in over-estimation of costs as certain capital assets and fixed cost items may have value beyond project activity (e.g. laboratory equipment). Additionally, several Wave 5 projects reported that provision of (e.g., to contacts of TB patients identified during ACF campaign) TB preventive therapy (TPT). However, we were unable to categorically assess costs nor performance outcomes explicitly for activities relating to TPT from data sources available to our team, including project’s financial statements. Therefore, misclassification of TPT related costs, while it may be small, may have resulted in over-estimation of CERs for these projects. We also were not able to extract quantifiable data on human resource involvement that are were not compensated through the project budget (e.g. routine healthcare worker tasked to screen TB symptoms at antenatal clinic for ICF). In this regard, our results may be an under-estimate of real-world human resource costs of ACF interventions. These data limitations in our study shed light on the need to improve and standardize the types and depth of the data necessary to perform comparative in-depth empiric cost and cost-effectiveness assessment of ACF interventions [9, 14, 15]. In particular, the cost per person screened (or initiating treatment) is also useful to programs for budgeting purposes and should be a priority for future research.

Secondly, we did not use estimates of health utility such as QALYs and DALYs, as conversion from cases detected to these measures is inherently context-specific. Therefore, our study findings may not be comparable to projects with other health outcomes. However, the cost per TB case detected is arguably the most direct and readily calculated measure of ACF performance without the need to make additional strong assumptions (e.g., future trajectory of people with TB who are not detected by ACF, impact on transmission of delayed diagnosis). Earlier analyses that do make these assumptions suggest that, in high-burden countries where missed TB diagnosis is common, the cost per case detected is about 25% higher than the cost per DALY averted (when considered over a 10-year time horizon) [6]. Thus, our cost-effectiveness ratios in terms of cost per case detected are likely to be modestly higher than the corresponding cost per DALY averted–but still well within most country-specific cost-effectiveness thresholds [12].

Thirdly, a simplified, standardized analytic approach taken in this study can be helpful for high-level comparisons on factors influencing cost-effectiveness (or costs associated programmatic performance) of interventions with similar/same objective. However, as described in our study, ACF projects are becoming more diversified to address gaps in the complex TB care cascade that are specific to the needs in the settings where these projects are implemented. Therefore, more detailed evaluation that includes projection of future benefits of key programmatic components (e.g., TPT provision to contacts of TB patients identified through ACF interventions or treatment linkages and adherence programs) is needed for each individual projects to precisely estimate the overall value and incremental cost-effectiveness of ACF interventions in each context [6, 7, 15–18].

In conclusion, this systematic comparative analysis demonstrates that the costs and cost-effectiveness of projects designed to improve the cascade of TB care are heterogeneous and context specific. This heterogeneity reflects the diversity of programmatic approaches and designs, target populations, and local settings in which these interventions are implemented. As future interventions introduce novel technology and processes, efforts to collect more detailed data that can provide insights on mechanistic, epidemiological, and operational factors influencing costs and impact of ACF interventions on the TB care cascade should be prioritized. Such data will be useful to improve our understanding of the costs and cost-effectiveness of interventions–including those interventions that fail to achieve targets for cost-effectiveness. In the meantime, projects to strengthen the TB care cascade–if implemented in locally relevant fashion–appear to offer reasonable value for money and should continue to be prioritized as part of a comprehensive approach to ending TB in high-burden settings.

Supporting information