Tuberculosis active case finding in Cambodia: a pragmatic, cost-effectiveness comparison of three implementation models

Tuberculosis (TB) control remains a critical global challenge, becoming the leading cause of death from an infectious disease in 2014. TB principally affects adults in lower income countries, with 10.4 million new cases estimated in 2015, of which more than half occurred in people living in South-East Asia and Western Pacific regions [1]. The Directly Observed Therapy Short-course (DOTS) strategy to control TB was launched by the World Health Organisation (WHO) in the early 1990’s and was based on the premise that early detection and treatment of new cases would prevent spread and reduce the burden of disease. Global targets for the proportion of new TB cases detected and treated successfully under DOTS were set at 70 and 85% respectively. Despite decades of investment towards these targets and 100% DOTS coverage by 2004 [2], early case finding has been a key challenge for TB program implementers. Currently, an estimated 37% of TB patients globally remain undiagnosed or are not reported [1], and prolonged delays to diagnosis are reported among some patients who do eventually access DOTS services. In recent years, the WHO has raised concerns that the impact of current interventions to improve early detection of TB may have been saturated [3].

It is indeed likely that the current DOTS passive case finding (PCF) strategy, which relies on patients presenting to the health facility for screening of their own accord, has limitations in contexts where primary health services are weak and awareness of TB is low; evidence from a multi-country analysis indicated that DOTS programs may have no impact on improving case detection [4]. PCF approaches may be resulting in both missed diagnoses and delayed diagnosis, leading to transmission of infection, which would be consistent with the small (1–2% per annum) observed annual decline in incidence of TB [5, 6]. At the current rate of decline in incidence, it will take more than 150 years to meet the WHO targets of reducing TB deaths by 95% and incidence by 90%, compared to 2015. In light of this, there is a push for new models of case finding – collectively referred to as Active Case Finding (ACF) approaches- which involve specific out-reach activities by the health system and its staff, in order to identify and test patients who are either symptomatic or considered to be at higher risk of having TB [7, 8]. The advantages of ACF include not only that it potentially identifies TB cases which would not have been diagnosed without the intervention, but may also identify cases at an earlier stage than if relying on PCF alone, thereby reducing transmission [9].

Although referred to using the same terms, numerous different ACF implementation models exist, ranging from population-wide screening to targeted case-finding in high-risk groups [10]. Compared to PCF, ACF requires higher investment and more human resources from National TB Programs, which is a major concern in resource-constrained settings [9]. Reviews of studies have shown that the mode of implementation of TB control activities has a large impact on cost-effectiveness, and researchers have highlighted the need for comparisons of different health service delivery models to inform planning and expansion of programmatic approaches to TB control. This is particularly true for ACF interventions. While globally there is some research comparing ACF implementation models and yields [11, 12], with a study in Zimbabwe finding that 77% of symptomatic, smear-positive study participants had their first investigation for TB through community-based ACF [12], there is very limited information about relative cost-effectiveness of different ACF strategies. A review of over one hundred ACF publications found that no studies conducted a cost effectiveness analysis using established guidelines [13].

Cambodia presents a unique opportunity for addressing this gap in evidence on relative cost-effectiveness of alternative ACF implementation models. The country has one of the highest TB incidence rates in the region, at 380 per 100,000 population and approximately 40% of smear-positive patients remain undiagnosed [2]. Many different ACF interventions have been piloted in Cambodia since 2010 [14‐17], and studies thus far have only looked at the cost effectiveness of a single ACF strategies [15, 18]. The National Center for Tuberculosis and Leprosy Control in Cambodia (CENAT), which is the government body responsible for resource allocation decisions on TB control activities, highlighted the urgent need to compare ACF strategies in order to assess which models identify the largest number of new TB cases per unit of investment. In close collaboration with, and in order to inform, CENAT policy makers, we conducted a pragmatic comparison of the cost effectiveness of three large-scale ACF programs from a health system perspective. These programs have been piloted in Cambodia and decisions on scale up need to be made.

We compared three ACF programs using retrospective data on case-finding during the implementation period of the pilot program, and primary data collected on costs of implementation, as described below.

Our comparison of three distinct ACF models in Cambodia revealed that the cost per case detected was 249 USD, 308 USD and 316 USD, for the HOPE, CENAT and CATA ACF models, respectively (Table 3). HOPE’s program, working in poor urban areas of Phnom Penh, was the most cost-effective of the three, by a small margin, followed by the model of testing contacts of TB patients (CENAT) and the CATA model focusing on older rural populations.

CENAT’s ACF activities involved providing CXRs to 33,029 people and diagnosed 807 B+ patients. HOPE tested 10,301 symptomatic individuals using sputum microscopy and 1894 with GeneXpert, leading to 737 confirmed B+ cases. Finally, CATA focused on older rural subjects, providing 11,650 CXRs after verbal symptom screening with 2518 going onto have GeneXpert testing, leading to 397 confirmed B+ cases.

The costing analysis, summarized in Table 4, showed that the contact tracing ACF model implemented by CENAT had a total cost of USD 248,222, while HOPE and CATA’s models cost 183,180 and 125,297 respectively. Distribution of costs varied between ACF models, with HOPE spending a higher proportion on salaries than the other organizations, while CATA and CENAT spent more on consumables. HOPE used some of its budget on equipment not routinely supplied by CENAT, which included mobile phones for their TB workers and cool boxes for sputum transportation.

The results show that despite differences in implementation models and populations, cost-effectiveness of the three interventions was relatively similar, with HOPE being the most cost-effective at less than 250 USD per new case diagnosed. The two rural programs of CENAT and CATA, which applied more costly diagnostic protocols involving chest x-rays, had very similar cost-effectiveness results, at USD 308 and 316 per case detected respectively.

Published ACF costs, per case detected, show that a wide range exists globally, from USD 108 in Cambodia (2010) [18] to USD 444–1493 in South Africa (2015) [23] and USD 1581–4000 in Russia (1995) [24]. TB REACH guidance suggests that the cost per additional case (those TB cases detected by the intervention above the local historical trend) detected and treated, for interventions they fund should be up to USD 350, however, this is a different measure to the costing in our study [1, 22].

Increased case detection and high treatment success rates are the cornerstones of global TB control policy [1] and as one of the top 22 high burden countries, Cambodia has implemented multiple strategies to improve its case detection rate. This is the first cost-effective analysis directly comparing alternative models of ACF in Cambodia, and, to our knowledge, no similar comparison has been conducted in other high TB burden countries.

Our study provides timely information in light of calls for cost-effectiveness data from ‘routine conditions’ to be included when programmatic recommendations for TB diagnosis are made [25]. Such cost-effectiveness information is essential to inform resource allocation decisions of national policy makers. Our method of analyzing the three ACF strategies was designed to provide data that is directly relevant to policy makers, taking a pragmatic approach towards costing the interventions from Cambodia’s national TB program’s (CENAT’s) perspective and using case finding data from program activities, rather than experimental settings. It is important to consider that our analysis, by design, compares heterogeneous ACF models with different population characteristics, the most important being a likely variable TB prevalence [22]. Different screening algorithms were used and the urban poor screening by HOPE used cheaper diagnostic techniques compared to the other two; thus the logistics and costs of each project varied. HOPE also focused exclusively on urban areas where the higher population density meant that larger numbers of people could be screened within a smaller geographical area.

A potential limitation relates to the estimation of costs by CENAT and an assumption that scale up of CATA and HOPE strategies by CENAT will be possible and provide the same efficiency, without substantial extra investment in the current services and health centres. However, there were also advantages to the method employed for cost calculations, with the unit costs being standardized according to national program estimates. We acknowledge that reductions in patient costs owing to outreach activities are not accounted for, and that the cost-effectiveness comparison does not consider equity and accessing of hard to reach populations, which may justify a higher cost per case detected. Finally, we recognize that wider questions regarding ACF itself still remain and highlight that ACF strategies will only be effective if cases are identified early such that onward transmission is prevented. Thus, before launching into large ACF programs in resource-limited settings, pragmatic comparisons of different ACF models – looking not only at cost per case detected as we have done but also at forecasts of long term impact on TB incidence- are required to inform resource allocation decisions.

Alternative ACF implementation models for TB are being introduced in numerous resource-constraint settings, but evidence on relative cost-effectiveness, which is essential to inform resource allocation decisions of national policy makers, is lacking. We applied a novel, pragmatic methodological approach to compare the cost-effectiveness of three ACF implementation models that had been implemented between 2012 and 2013 in Cambodia, allowing the models to be ranked in terms of potential to identify the largest number of new bacteriologically-positive TB cases per unit of investment. We found that implementation of ACF activities in urban areas was slightly more cost-effective than in rural areas, and that the pragmatic cost-effectiveness assessment approach provided information that was called for by local policy makers.