Background
In 2015, a target and indicator for human immunodeficiency virus/acquired immunodeficiency syndrome (HIV/AIDS) was set in the Sustainable Development Goals (SDGs) by the United Nations (UN), government and other partners, focusing on ending the HIV epidemic by 2030 [
1,
2]. To achieve this goal, UNAIDS established triple 90 (90–90–90) targets as the explicit global objective for HIV/AIDS [
3]. In resource-limited settings such as sub-Saharan Africa, the decentralization of HIV/AIDS testing, treatment and care services is a cornerstone of providing universal testing, treatment and care. The geographic expansion of services from secondary and tertiary central health facilities (hospital level) to peripheral primary health facilities in rural areas (rural health centre [RHC] level) is thus an urgent need [
4‐
6].
Lesotho was among the first countries in sub-Saharan Africa to decentralize services from hospitals to health centres via a national policy despite a severe human resource shortage in 2007 [
7‐
9]. The development of national guidelines enabling nurses to provide ART services in health centres facilitated the successful decentralization of services [
10‐
12]. Furthermore, the integration of HIV diagnosis, care, and treatment into primary health care services through task shifting has also been adopted as national policy in South Africa [
13], Malawi [
14] and Uganda [
15]. Task-shifting approaches to overcome the shortage of human resources were also included in national programmes and have been shown to improve patients’ treatment outcomes and save costs in two countries [
16‐
18]. A systematic paper also suggested that task shifting from doctors to nurses or from health care professionals to lay health workers could potentially reduce the costs of ART provision [
19]. In addition, these integrative approaches were found to provide more effective outcomes including earlier enrolment in treatment [
20], better adherence [
15,
21,
22], better retention in treatment [
15,
23‐
25], and higher acceptance of ART [
26]. Based on these findings, a few studies have suggested that scaling up ART services to the primary care level could be a cost-effective strategy [
15,
17,
27].
In Zambia, the HIV prevalence among 15- to 49-year-old adults is the seventh highest in the world (14.3%) [
28]. The number of people living with HIV (PLHIV) is 980,000, 68% of whom receive ART. As Zambia is a large country, PLHIV are highly scattered, even in rural areas. The urban/rural split is approximately 60/40 [
29]. Decentralization of ART services is essential to expanding service coverage, despite the severely limited resources. The Ministry of Health (MoH) in Zambia has provided HIV counselling and testing services at all health facilities nationwide since 2006 [
30]. Free ART services were introduced at the hospital level in 2005 and were further expanded to some specific RHCs as a trial in 2007 through the national ‘Mobile ART Services’ programme. Under this programme, a mobile ART team comprising medical professionals such as medical doctors, nurses and pharmacists from a district hospital conducted biweekly visits to select RHCs that had been designated as ART sites; the team assisted with ART services in terms of providing human resources and building RHC staff member capacity [
31]. Although the medical professionals on the team temporarily performed the drug prescription and dispensation and the laboratory services such as ordering exams and drawing blood samples at the start of the programme, these responsibilities were gradually assumed by the RHC staff under the supervision of the professionals. This programme contributed to ART service decentralization at the primary health care level and aimed to maximize the efficient use of extremely limited resources. The MoH approved this programme nationally and published the national guidelines for Mobile ART services in 2010 [
32,
33]. There have been several cost-effectiveness studies on facility-based ART services in Zambia [
34‐
39]. However, we are the first to analyse the cost effectiveness of the National Mobile ART Services Programme in Zambia to inform policy deliberations.
Discussion
The ICER suggested that the intervention programme could be a cost-effective option for decentralization of HIV services in Zambia in addition to the original programme.
To our knowledge, this is the first study to examine the cost effectiveness of the national ART decentralization programme in Zambia. Although a similar study conducted in Uganda also examined the cost effectiveness of different modes of ART provision, it analysed a pilot programme rather than a national programme [
15]. In addition, as the HIV prevalence in Uganda was half of that in Zambia (7.2 vs. 14.3%, respectively) [
9], our study might have a stronger impact on the development of policies for expanding ART services in countries with a high HIV/AIDS burden, particularly those in sub-Saharan Africa. Additionally, we believe that other countries could modify and use our simple model to analyse their policies and programmes to identify an optimal approach to decentralizing ART services from an economic perspective.
We found that the intervention programme could be cost effective based on the ICER results and probabilistic sensitivity analysis. However, it was not very cost effective, since the ICER was over the GDP per capita. In developing countries including Zambia, using a stricter threshold (1X GDP per capita) could provide a more realistic scenario. The Tornado analysis showed that the ICER was largely influenced by the cost of the intervention programme; the cost factor could be addressed to make the intervention programme more cost effective than the original. According to the cost per effectiveness, our study showed that the hospital-based service cost less than the other option, and this finding was consistent with another study [
15]. The intervention programme had a twofold higher annual per-person cost than the original (1259.16 vs. 2601.02 USD), although the programmes were similarly effective (6.81 vs. 7.27 QALYs). The additional costs of the intervention programme relative to the original programme included the costs of the Mobile ART services. Although the costs of buildings, furniture, facility staff, drugs, and laboratory examinations accounted for a small proportion of the total Mobile ART costs, the costs of the vehicles, maintenance and salaries of the mobile team staff might be important for controlling the total operational costs. Since these costs depend on the market price and government regulations, it could be difficult for local government officers to control the costs associated with a better programme. As Mobile ART services face greater influence from uncontrollable external factors, the long-term sustainability of these services should be further examined. However, our intervention programme had the advantages of increased access and retention despite having higher costs than the original programme, and these benefits might make the intervention programme a cost-effective option.
Although the ICER was not small enough to suggest that the intervention programme was very cost effective, to attain the UNAIDS 90-90-90 [
3] objectives and universal health coverage (UHC) [
51], the country still needs to decentralize HIV services to improve the service coverage and promote equal access to services.
Our findings regarding the cost effectiveness of the intervention programme were also largely influenced by the utility of the PLHIV retained in ART and the mortality of the PLHIV not retained in ART. A one-way sensitivity analysis revealed that the intervention programme was more cost effective than the original programme at a utility value over 0.78 and an annual mortality rate below 39.3%.
Healthcare workers should support patient adherence to ART and identify signs of adverse ART-related events and opportunistic infections as early as possible to avoid negative outcomes, as both these factors are normally associated with the quality of ART services; a reduction in negative outcomes could also improve the cost effectiveness of the intervention programme.
In our model, the reference case age was set to 30 years, based on the average age of our rural Zambian cohort. A one-way sensitivity analysis of age revealed that the intervention programme was expected to produce more value than the original programme for PLHIV aged less than 32 years at the time of ART initiation. This result suggests that the earlier patients are enrolled into ART, the more cost effective the intervention programme will be. This finding is consistent with the national policy in Zambia that promotes earlier HIV diagnosis, which could have a synergistic effect on the outcomes of the intervention programme.
Two patterns in the rate of retention in the Mobile ART service were examined in the sensitivity analysis. The first pattern was the best scenario, with a 10% reduction/year in the original programme and a 3% reduction/year in the intervention programme. The second pattern was the worst scenario, with a 3% reduction/year in the original programme and a 10% reduction/year in the intervention programme. Regardless of the changes in retention rate based on the best and worst scenarios, the expected value produced by the intervention did not change. This finding implied that the DHO could choose any Mobile ART service pattern to expand ART services in their district.
Our study had some limitations. The model included some assumptions; specifically, we assumed parameters such as transition probabilities, retention rates, and utilities. These assumed data might differ from the actual situation and could thus affect the results of the analysis. However, in order to develop an accurate model, we strived to obtain these data from appropriate sources such as other studies conducted in similar settings, national reports, and findings from our cohort analysis. In addition, a sensitivity analysis involving ranges of the assumed data was performed, and it did not affect the results in terms of the cost effectiveness of the programmes.
We simplified the analysis model to the greatest extent possible. Our simple model might therefore lack accuracy and details that would reflect real situations. Indicators of the severity of PLHIV, such as WHO staging and CD4 cell counts [
52], were not included in the model. Adherence to ART was also not considered among the PLHIV retained in ART. Only one ART regimen, the first-line ART recommended in the latest WHO guidelines [
53], was adopted in the model. Using natural history data from patients who have never been on ART to approximate the mortality of people no longer on ART could also be a limitation, as the latter might still benefit from the treatment they once received. In addition, only government health facilities such as district hospitals and RHCs were included in the model; no private facilities were included. As our target population comprised ART-eligible patients with WHO stages of 3 or 4 and CD4 cell counts <350 cells/mm
3, and as it was based on the WHO guidelines, we decided that these factors might not affect the model analysis. It is difficult to identify people with resistant HIV who require second-line ART regimens and to identify private health facilities in our study setting of rural Zambia; therefore, these factors were not included in our model. We also wanted to develop a simplified model rather than a complicated model to allow other countries to use and modify the model to analyse their specific situations.
Regarding costs, only operational costs were calculated. The costs of training and supervision were not included because the DHO conducted these activities without separating the hospital and RHCs. The incremental annual costs also did not include the treatment costs for adverse ART-related events, opportunistic infections, second-line ART regimens for people with resistant HIV, or hospitalizations, as these events and patients were rare in our study setting. Although the approaches to ART service costs have varied in other papers [
37,
54], adding these costs were expected to further improve the cost effectiveness of the intervention programme because these high costs commonly occur in the hospital.
Authors’ contributions
SM and GS conceived and designed the study. SM and KK collected and assembled the data. SM analysed and interpreted the data with GS, KK, HE and TS. All authors discussed the results and commented on the manuscript. All authors read and approved the final manuscript.