Impact and Cost of the HIV/AIDS National Strategic Plan for Mozambique, 2015-2019—Projections with the Spectrum/Goals Model

Introduction Mozambique continues to face a severe HIV epidemic and high cost for its control, largely born by international donors. We assessed feasible targets, likely impact and costs for the 2015–2019 national strategic HIV/AIDS plan (NSP). Methods The HIV epidemic and response was modelled in the Spectrum/Goals/Resource Needs dynamical simulation model, separately for North/Center/South regions, fitted to antenatal clinic surveillance data, household and key risk group surveys, program statistics, and financial records. Intervention targets were defined in collaboration with the National AIDS Council, Ministry of Health, technical partners and implementing NGOs, considering existing commitments. Results Implementing the NSP to meet existing coverage targets would reduce annual new infections among all ages from 105,000 in 2014 to 78,000 in 2019, and reduce annual HIV/AIDS-related deaths from 80,000 to 56,000. Additional scale-up of prevention interventions targeting high-risk groups, with improved patient retention on ART, could further reduce burden to 65,000 new infections and 51,000 HIV-related deaths in 2019. Program cost would increase from US$ 273 million in 2014, to US$ 433 million in 2019 for ‘Current targets’, or US$ 495 million in 2019 for ‘Accelerated scale-up’. The ‘Accelerated scale-up’ would lower cost per infection averted, due to an enhanced focus on behavioural prevention for high-risk groups. Cost and mortality impact are driven by ART, which accounts for 53% of resource needs in 2019. Infections averted are driven by scale-up of interventions targeting sex work (North, rising epidemic) and voluntary male circumcision (Center & South, generalized epidemics). Conclusion The NSP could aim to reduce annual new HIV infections and deaths by 2019 by 30% and 40%, respectively, from 2014 levels. Achieving incidence and mortality reductions corresponding to UNAIDS’ ‘Fast track’ targets will require increased ART coverage and additional behavioural prevention targeting key risk groups.


Introduction
The HIV epidemic continues to be a significant, if not the foremost, public health and economic and challenge for Mozambique. Significant progress has been made in fighting HIV, but high prevalence persists. Since 2013, UNAIDS and the WHO recommend expanding eligibility for antiretroviral treatment (ART) to a CD4 threshold of 500/uL [1], and in 2015 new evidence emerged that immediate ART enhances both the clinical benefits to patients (morbidity and mortality from HIV/AIDS, and TB) and community prevention effects through reduced HIV infectivity [2,3]. Also voluntary male medical circumcision (VMMC) is recommended since 2007 as an important prevention strategy [4]. In the context of an expanded set of evidencebased control strategies, but a constrained HIV/AIDS budget HIV/AIDS, Mozambique's epidemic is entering a phase where strategic balancing and prioritization between interventions, target groups and areas is required to accelerate progress and keep the response financially sustainable.
Mozambique's national AIDS program started in 2001. Key service delivery areas are condom promotion, community mobilization, HIV counseling and testing (all since program start), prevention targeted at female sex workers (FSW) and clients ( [5] and the HIV acceleration plan [5]. At 2014 years-end, the National AIDS Council (NAC) and Ministry of Health, with technical partners including UN agencies, USAID and Civil Society Organizations (CSOs), developed the fourth NSP as a guide for HIV prevention, treatment and mitigation over 2015-2019. UNAIDS and Avenir Health provided technical assistance to cost the NSP and estimate its potential impact on the epidemic. Targets for intervention coverage and corresponding expected health impact were set in dialogue among partners, considering existing national health sector targets, desired and feasible targets for additional non-medical interventions,as well as expected domestic and donor resources, based on projections using the Goals [6,7,8,9,10] and Resource Needs Models [11]in the Spectrum suite of planning tools. This paper presents the results of the epidemic and response modelling, in terms of coverage and impact targets for the NSP, as well as resource needs. We review contributions of the respective interventions in three scale-up scenarios, of which the most ambitious was proposed as Mozambique's official 2015-2019 NSP targets in March 2015, and which is pending approval by the Board of the NAC. Results are discussed in the context of the evolving evidence on ART as a key prevention contributor, the UNAIDS 90-90-90 framework and Fast Track targets for ending the global epidemic by 2030, and a growing recognition of efficiencies, prioritization and increasing domestic contributions needed to sustainably finance national AIDS responses.

Epidemic model
The modelling used the Goals model, a dynamic compartmental model, built in the Spectrum suite of models, used and validated for projections of epidemic trends and the impact of interventions in several countries as well as for global assessments [6,7,8,9,10].
Goals simulates transmission of HIV and its morbidity and mortality consequences for adult populations 15-49 years, who upon reaching a median age at first sex, are allocated into one of five risk categories: stable couples (men and women reporting a single partner in the last year), multiple partners (men and women with more than one partner in the last year), FSW and clients, men who have sex with men (MSM), and injecting drug users (IDU). HIV-infected individuals move through CD4 compartments, which correspond to ART eligibility criteria and mortality patterns. Clinical progression after HIV infection is a function of CD4 count, with associated HIV-related mortality, probability of initiating ART (considering national ART eligibility criteria and coverage levels) and infectiousness.
The impact of behavioural prevention interventions is modelled according to an impact matrix that articulates the impact of each intervention on condom use, numbers of partners and age at first sex in the different risk groups, based on meta-analysis of research studies [12,13,14,15]. The model calculates new HIV infections by sex and risk group as a function of behaviors and epidemiological factors, such as prevalence among partners and stage of infection. The risk of transmission is determined by behaviors (number of partners, contacts per partners, condom use) and biomedical factors (ART use, VMMC, prevalence of other sexually transmitted infections (STIs). Interventions can change any of these factors, and thereby affect the future course of the epidemic.
Goals is linked to the Aids Impact Model (AIM) module in Spectrum, which calculates corresponding epidemic patterns for children (0-14 years) and adults above 49 years. AIM also estimates the effects of programs preventing mother-to-child transmission [6,7].

Biomedical and natural history parameters
Goals uses internationally agreed best estimates of biomedical and natural history parameters, such as the risk of HIV transmission per act, and variations in this risk by stage of HIV infection, type of sex act, condom use, etc. By preference, these parameters are kept constant across countries, although certain parameters are allowed to vary somewhat regionally, in order to produce adequate fit. For Mozambique, most parameters were kept constant across the three sub-national models, except for the default HIV transmission probability, and the transmission multiplier effects of STIs and of the primary stage of HIV infection (S1 File).
We assumed that ART reduces infectivity of HIV patients by 80%, as an average effectiveness between relevant recent studies including a 96% reduced infectivity found in a clinical trial across multiple-mainly developed, western-countries with very high adherence [16,17], a 38% reduction in a high-coverage ART program in rural South Africa [18], and 85% virological suppression observed in Swaziland [19,20].

Model fit of the Mozambique epidemic
For the NSP, Goals was fitted to official UNAIDS estimates of Mozambique's AIDS epidemic up to 2013, which had been produced in AIM by statistical fitting of HIV sero-prevalence data from antenatal clinic (ANC) surveillance and national household surveys. Similar to the June 2014 UNAIDS AIM estimates, Goals modelling was done separately for North, Central and South regions, which were aggregated to produce national result [21].
Demographic, behavioural and epidemiological parameters were quantified based on epidemiological and behavioural data from ANC surveillance [22], program statistics, 2011 DHS [23], 2009 national AIDS Indicator Survey [24], Integrated Bio-Behavioural Surveys (IBBS) of high-risk groups [25,26,27], HIV program statistics (PMTCT by regimen, adult ART, child ART and Cotrimoxazole Prophylactic Therapy), selected population-based research studies [28,29], a Mode-of-Transmission modelling conducted for the 3 regions by UNAIDS [30], and the country's Global AIDS Response Reports of 2012 and 2014 [31,32] (detailed in S2 File). For community mobilization, the (moderate) effect on reducing multiple partnerships assumed in Goals was supported by an effectiveness evaluation of a mass media and community mobilization campaign conducted in 2010 in Mozambique's four provinces with highest HIV prevalence [33].

NSP scenarios
The NSP was modelled in three scenarios, differing with respect to coverage levels of key prevention and treatment interventions: C. Accelerated scale-up: Further increased scale-up, notably for targeted behavioural prevention reaching high-risk groups and other non-biomedical prevention programs, expansion of ART eligibility to include all FSW, and improved patient retention on ART.
Some other interventions were not targeted for scale-up, and kept at a constant coverage across scenarios over 2014-2019 (Table 1).

Coverage baselines and targets for the NSP
Coverage of NSP interventions in 2014, was estimated from program records and reports [31,32], surveys in the general population and key risk groups, and selected population-based studies [28,29,30]. Relevant targets up to 2019 were set in dialogue with the NAC, Ministry of Health, technical partners and implementing NGOs, considering existing commitments of official national plans and donor grants (Table 1).
Scenario 'Current targets' foresees increasing coverage for community mobilization [33], condom promotion and provision, VMMC, adult and pediatric ART, and PMTCT. ART is scaled-up from 56% to 76% of adults with CD4<350/uL in North region, from 65% to 81% in Center, and from 57% to 85% in South; additionally eligibility includes TB/HIV-co-infected adults and pregnant women (from 2012 and 2014, respectively), in all scenarios irrespective of CD4 count.
Scenario 'Accelerated scale-up' additionally scales-up mass media, HIV testing and counselling, behavioural prevention for youth, outreach to MSM (partly through prisons) and IDU, and intensifies outreach to FSW and VMMC. ART is further scaled-up to 85% of eligible PLWH with CD4<350/uL, and all FSW irrespective of CD4 count. In addition, this scenario assumes improving patient retention on ART, from 52% at 3 years after treatment initiation as of 2013 [38] to 70% by 2019, resulting in reduced HIV-related mortality through application of Goals survival assumptions [39,40]. Assumptions and targets for coverage of community mobilization reflect weighted averages between the general population and special groups such as miners, long-distance truck drivers and other mobile workers.
Scale-up of condom distribution and promotion considered planned increases in social marketing condoms purchased, and important increases in public distribution. Public-sector distribution (about 21 million condoms in 2014) had remained limited to community interventions distributing male and female condoms to communities and hotspots from district warehouses. Renewed Global Fund support is expected to address this bottleneck, as reflected under 'Accelerated scale-up'.
Targets for behavioural prevention for high-risk populations reflect the expected feasibility of reaching FSW, MSM and IDU in each region. For outreach to FSW and their clients, the 'Accelerated scale-up' target is higher in South (80%) than in North and Center (60%), in view of higher baseline coverage (Table 1). These targets average sub-targets for prisoners in main prisons (80% by 2019), and for miners, truck drivers and other mobile workers, who form part of FSW clients.
For MSM, the overall target reflected sub-targets for MSM as a hidden group, and male prisoners. For 'hidden' MSM, the sub-target was limited to 15% for all 3 regions, up from 0.2% in 2014, in line with experiences and views of key implementing NGOs. For prisoners, building on activities implemented by the Ministry of Justice in main prisons, a scale-up from 38% in 2014 to 80% by 2019 is foreseen.
For outreach to IDUs, the 'Accelerated scale-up' foresees 50% (3 873 out of 7 746 IDUs) reached with harm reduction and HIV testing and counselling by 2019, against 5.2% (N = 374) Legend to Table 1

Unit Costs
Resource needs for the NSP (in 2014 US $) were estimated by multiplying average service unit costs ( Table 2) with annual volumes of service delivery, within the Resource Needs Model (RNM) module of Spectrum [11].
Where possible, service unit costs were extracted from cost estimates from national HIV/ AIDS programs [34,35,37,41].When information was not available, these were estimated from financial records and programmatic data of key implementers including unpublished expenditure analysis by PEPFAR, the 2010-2011 National AIDS Spending Assessment [45], the concept note application submitted to the Global Fund in October 2014 [36], and data from implementing NGOs. Final estimates were agreed with implementing partners, assembled in a working group convened by UNAIDS.
Costs of ARVs, laboratory tests and other essential commodities refer to actual procurements in Mozambique and were taken from the US Government/ PEPFAR Supply Chain Management System [43]. Price changes over 2014-2019 for first-line and second-line ARVs were estimated by extrapolation of trends in PEPFAR ARV procurements globally, giving a 16% reduction in unit price from 2015 to 2019. Laboratory cost per adult patient-year (including TB diagnosis) was assumed to increase from US$ 40.6 in 2014 to US$ 76.3 by 2019, reflecting increasing use of viral load testing (at US$ 30 per patient per year-i.e. without considering pending price reductions for viral load tests).
Costs of essential commodities are inclusive of service delivery and supply chain costs, at a fixed 17.5% [34]. For health sector-based services (ART, PMTCT and Testing & Counselling), human resource costs were integrated in the respective unit costs through a normative bottomup approach, assuming a fixed health worker time per service unit and fixed health worker capacity.
For simplicity, unit costs were kept fixed over 2014-2019 for most other interventions and budget items. This likely overestimates costs for interventions with falling commodity prices and/or which are likely to achieve economies of scale or efficiencies with program maturation. An exception was made for VMMC, for which cost per man circumcised was assumed to fall from US$ 104 in 2014 to US$ 50 by 2019, reflecting an expected reduction in investments needed in infrastructure and devices. The $ 104 was obtained from PEPFAR, Mozambique's main VMMC funder and implementing partner. Of this baseline cost, 43% covered capital investment (health unit rehabilitation and construction) and program management, which progressively reduced to 0 by 2019. Furthermore, the 'Accelerated scale-up' scenario assumed another US$ 10 unit cost reduction from transitioning to a cheaper VMMC device.
For ART, service delivery unit costs were kept constant. The additional cost of improving patient retention under 'Accelerated scale-up' was not explicitly modelled, but reflected under increasing cost of second-line ARVs (for patients who failed first-line ARVs, accumulating with years on treatment), which are more expensive than first-line ARVs.
Program support costs were determined based on actual expenditures (national program coordination), extrapolated from existing plans and identified needs (M&E, Training, Health & Community Systems Strengthening). Since these costs do not increase proportionally with intervention scale-up, an annual US$ 48 million was assumed throughout 2014-2019, in all scenarios.
Mitigation includes support to orphans and vulnerable children (OVCs), HIV-related human rights and gender programs and home-based care. OVC was costed as a lump sum,  Human Rights and Gender 0 0.34M 3.62M Implementing partners, following UNAIDS guidelines for human rights programs in the areas: stigma reduction, legal services, law review and reform, legal literacy, training of health care providers, sensitization of law enforcement agents and women's rights in the context of HIV) [41].
Legend to Table 2 allocated across regions by their numbers of PLWH. The nation-wide lump sum was based on the price of support kits provided to OVCs by the Ministry of Gender and Social Affairs, which is consistent with unit expenditures on OVC by US Government implementing partners multiplied with OVC numbers (from 2012 and 2013 program data) and a targeted 70% coverage nation-wide by 2019. For HIV-related human right and gender programs, no package of services was yet defined for Mozambique. These were costed using a normative approach for intervention activities and budgets. Activities were aligned to seven programs (stigma reduction, legal services, law review and reform, legal literacy, training of health care providers, sensitization of law enforcement agents and women's rights in the context of HIV) recommended by the UNAIDS Human Rights Reference Group [46].
For home-based care, unit cost corresponded to PEPFAR recurrent expenditures (excluding training and program management), multiplied with 34% coverage of known PLWH according to 2013 programmatic data, and assuming a gradual reduction to 22% by 2017, to reflect decreasing need for home-based patient care provided by community organizations following expansion of facility-based ART. The resulting national lump sum was allocated across regions proportional to their ART patient numbers in 2014.

Fitted sub-national epidemics, up to 2014
With the behavioural, biomedical and natural history quantifications specified (methods and S1 and S2 Files), Goals closely fitted HIV sero-prevalence trends over 1985 to 2013 for all three regions (Fig 1). To achieve this, condom usage was adjusted downward and numbers of partners adjusted upward compared to survey responses, which is justified in view of suspected social desirability bias, due to which respondents often over-report condom usage and underreport partner numbers. HIV prevalence has risen highest in Southern region (which is most developed, and neighbors South Africa, where many Mozambican workers migrate); is moderately high but with a strong reversal and ongoing decline in Center region, and lower but rising in the more remote North.

Behavioural risk reductions
As a result of scale-up of behavioural prevention interventions, Goals estimates marked increases in condom use among all risk groups, in all regions but most notably in the North (Table 1). Based on these projected condom usage rates, the NAC proposed as behavioural outcome target for the NSP to increase condom usage among people with two or more partners (including commercial partners/sex work) by 50% from the 2014 level (of 12%).
Similarly, intervention scale-up is projected to reduce numbers of partners, most notably for medium-risk groups and secondarily for high-risk groups.

HIV incidence and mortality under NSP scenarios
Under 'Constant coverage', the Goals model projects a fairly stable HIV incidence over 2015-2019, close to rates in 2014, in all 3 regions (Fig 2A-2C). In the advanced epidemics in Southern and Center regions, this stable incidence follows a progressive decline from peaks around and before 2003, respectively. In the North, incidence is stable at a level similar to that since 2002, although with a slight temporary decline around 2014-2015 which follows rapid ART scale-up over 2013-2014.
Given population growth, these incidence patterns combine into a progressive rise in annual HIV infections nation-wide, from 2015 as a temporary minimum, back to the level of 2012 by 2019 (Fig 2D).
Program scale-up under 'Current targets' reduces new infections considerably, reaching a low around 2019, two years after reaching 2017 ART targets. Slight further scale-up of prevention interventions over 2018-2019 does not entirely prevent a subsequent rise in annual infections, which approaches 2014 levels again by 2028. Only 'Accelerated scale-up' sustainably reduces annual infections to 30% below the 2014 baseline by 2019, with a fairly stable annual number also over next years.  Current targets will reduce annual new infections (among all ages) from 105,000 in 2014 to 78,000 (19% reduction) in 2019. Accelerated scale-up will reduce annual new infections to 65,000, 33% below the 2014 level.
Similar patterns are apparent for HIV-related deaths. Under 'Constant coverage', mortality is fairly stable over 2015-2019 (slightly rising in low-prevalence North; slightly decreasing in the advanced epidemics with higher ART coverage of Center and South; Fig 2E-2G). As a result, annual deaths continue a slow increase as before 2014 (Fig 2H). Temporary mortality fluctuations (e.g. dips in 2007 and 2014, and a peak in 2013) reflect strong, immediate effects of recent ART scale-up, whose impact in individual patient peaks 1-2 years following treatment initiation, followed however by mortality resurgence over subsequent years, if fewer new patients start ART while patients enrolled earlier begin failing treatment or get lost. Nationwide, 'Current targets' will reduce deaths (among all ages) from 80,000 in 2014 to 56,000 (a 33% reduction) by 2019.
'Accelerated scale-up' will reduce all-age deaths to 51,000 in 2019, 45% below the 2014 number. Over 2015-2019, this scenario averts 125,000 deaths, of which 12,000 through improved patient retention on ART.

Contribution of risk groups and interventions
Up to two-thirds of new infections occur in adults who have multiple partners and/or who engage in commercial sex. Especially in Northern region, most new infections continue to occur in these higher-risk groups (Fig 3A), and NSP scenarios avert most new infections in these higher-risk groups.
Comparing infections averted among interventions, in the North, behavioural/prevention outreach to sex workers and their clients, and condom promotion (which reduces risk among people with multiple partners) are key contributors to infections averted over 2015-2019 ( Fig  3B). In Center and especially South, with comparatively higher prevalences in the general population, VMMC (from baseline levels lower than in North), and mass media are relatively more important. ART scale-up is a key contributor to infections averted in all three regions.

Resource needs
From US$ 273 million program cost in 2014, NSP scenarios require increased resources, to US $ 319 million in 2019 under constant coverage (due to population growth, and increasing need of ART and PMTCT as PLWH progress clinically), US$ 433 million for 'Current targets', or US$ 495 million for 'Accelerated scale-up' (Fig 4A).
Cost increases from 2015 to 2019 largely reflect increasing numbers on ART (Fig 4B), including an increasing need for second-line ARVs (Fig 4C) for an estimated 0.65% of ART patients in 2014 up to 5.1% of patients in 2019.

Cost per infection or death averted
Compared to constant coverage, scale-up to current targets will avert 110,000 infections over 2015-2019, whereas Accelerated scale-up would avert 145,000 infections (discounted, across all ages). This corresponds to a discounted cost per infection averted of US$ 16,612 (Current targets) or US$ 13,380 (Accelerated scale-up).  Discounted costs per death averted (across all ages) are higher (US$ 18,966 and US$ 17,377, respectively), since over 2015-2019 the NSP will avert fewer deaths than infections (96,000 deaths under 'Current coverage', and 114,000 deaths under 'Accelerated scale-up'). Accelerated scale-up lowers costs per infection and death averted compared to 'Current targets', thanks to an enhanced focus on behavioural prevention for high-risk groups, which saves many infections and deaths at relatively low cost.
Among interventions, scale-up of prevention outreach for sex workers and IDU, mass media, and HIV testing and counselling save relatively many infections at low cost (Fig 5) in all regions. Scaling-up condom promotion and school-based prevention are also relatively costeffective. Costs per infection averted are higher for community mobilization and outreach to MSM. ART scale-up to above 2014 levels saves many additional infections but at high cost, so is relatively costly per infection averted. Nation-wide, VMMC scale-up averts most additional infections, at low incremental cost in all regions despite varying baseline circumcision prevalences.

Discussion
Mozambique's proposed NSP, supported by an evidence-based epidemiological projection, represents a sensible next phase in the country's AIDS response. If coverage targets as adopted for 'Accelerated scale-up' are reached, Mozambique could feasibly expect to reduce new infections by 30-33%, and deaths by 40-45%, in 2019 compared to 2014 levels. The NSP's modelbased target setting builds on previous work but benefits from more and better data, particularly on local intervention unit costs and epidemic dynamics, and national policy analysts used a more integrated set of tools linking interventions with cost and impact, than ever before.
A formal cost-effectiveness analysis-which should consider not only costs but also financial savings from health care averted or postponed beyond 2019 -was beyond the scope of this analysis. Presented incremental costs per infection or death prevented (over 2015-2019 alone, relative to a scenario with coverage maintained at 2014 levels) are in the range of values generally considered cost-effective for countries with incomes similar to Mozambique, with a GDP per capita around 1,100 international dollars over 2010-2014 [47,48,49]. In addition, infections averted bring future savings as people will not need treatment. ART costs per patient, combined with expected survival and a 3% annual discount, imply a net present value of the lifetime cost of treatment of about US$ 3,100. Each infection averted within 2015-2019 produces these savings, which partially off-set costs of scaling-up behavioural prevention. Other future costs averted (for example, orphan support) will offset even more of the needed investments. Despite the simplicity of our cost-related outcomes, the comparison of NSP scenarios illustrated the good value for money of program packages prioritizing prevention for key groupsleading the NAC and partners to, for the first time, recognize these as key complements to existing targets for biomedical interventions.
While NSP modelling and target setting was undertaken separately for the three regions with their varying epidemics and baseline service coverage (notably for behavioural prevention for FSW and clients), final coverage targets came out similar across regions. As a result, impact contributions from incremental scale-up of the respective interventions, when compared to their 2014 baseline levels, differed somewhat across regions. Up to 2014, male circumcision through traditional practice as was prevalent before scale-up of VMMC was already preventing many infections in Northern region (not reflected in the VMMC shares in Fig 3B), whereas behavioural prevention for sex workers and clients already prevented many infections in Southern region with its advanced epidemic and response.
Nevertheless, country-wide the proposed NSP coverage targets for 'Accelerated scale-up' seem a sensible and rational balance between behavioural and biomedical interventions, which should consolidate incidence and mortality reductions started recently, and enhance value for money and sustainability of the national AIDS response. The NSP projections also form a basis for further refined prioritization tailored to regions' evolving epidemic, response and service costs.
NSP targets for behavioural prevention in key groups (except for marginalized, difficult to reach IDU and MSM) as well as outreach to youth, VMMC, HTC and PMTCT, are in line with global Fast Track targets. The 2015-2019 NSP as proposed is an important step toward reaching the global goal of ending AIDS by 2030. Yet, targets defined for Accelerated scale-up will not permit reaching the Fast Track target of reducing new HIV infections by 80% over 2010−2020 and by 90% over 2010−2030 [50]).To reduce new infections by more than 33% by 2019 and further thereafter, requires a more aggressive program package.
Expanding ART to PLWH with CD4 between 350-500/uL and/or sero-discordant couples, according to curren WHO recommendations [1] or to all PLWH asunder UNAIDS' 90-90-90 targets (90% of PLWH diagnosed, of whom 90% enrolled and retained on ART, with 90% adherent [50]) would greatly enhance population-level impact. However, unlike higher-income countries in Southern and Eastern Africa, the Mozambican Government has not yet considered such treatment expansion, given resource and system constraints. From current prevalence trends, that implies over 1.5 million PLWH on ART by 2019 i.e. 600,000 more than in the Accelerated scale-up scenario-which at a gross US$ 230 per patient-year would require an additional US$ 138 million annually (28% above 'Accelerated Scale-up').
Reaching 'Fast track' targets could also be supported by intensifying behavioural prevention efforts among highest-risk groups, further or quicker scale-up of VMMC in Center and South (if judged feasible), and/or higher coverage and/or smarter targeting to youth-especially girlsin and out of schools.
Reaching targets for 'Accelerated scale-up' will, in 2019 alone, require US$ 130 million more than the US$ 350 million available for HIV in Mozambique in 2014 [32]. Increasing commitment of PEPFAR for fiscal year 2016, along with the Global Fund concept note grant signed in June 2015, possible additional Global Fund funding from 2017, and an increased state budget allocation for HIV in 2015 [36] could enlarge the resource envelope to US$ 400-450 million per year for 2015-2017. The remaining gap could be closed with additional government expenditure on health (8% only in 2012 [51]) and increased budget allocation from the Health Ministry for HIV, estimated at just 10% in 2013 [36], especially as absolute government revenues are expected to increase. Reaching 'Fast track' targets would require additional resources which Mozambique's economic growth may not be able to cover, and other innovative funding mechanisms need to be considered.
These findings should be interpreted in the context of certain limitations. The model was structured and quantified on available evidence from Mozambique and worldwide. However, sexual mixing between risk groups is limited in the Goals model, which may cause over-estimation of intervention impacts on new infections.
The projected mortality impact is driven by ART, modelled based on Southern Africa region-wide survival patterns, with mortality adjusted upward by 20% to account for relatively poor patient retention in Mozambique [52]. This effectiveness remains to be validated against longitudinal patient cohorts and/or population-based data from Mozambique, as in recent mortality impact assessments by neighboring countries [53,54]. Similarly, Mozambique lacks population-based data to validate assumptions on effectiveness of ART in reducing HIV incidence, nevertheless a significant contributor to infections averted over 2015-2019. Assumptions on mortality and transmission impacts of ART in Mozambique are more conservative than in earlier models of global and African country epidemics, reflecting relatively low patient retention on ART in Mozambique [38,52]. Nevertheless, the assumed 80% reduction in infectivity on ART (across scenarios) is higher than the 70% assumed by UNAIDS in its 2015 estimation of national HIV burdens and trends across low-and middle-income countries.
For preventive interventions, Goals may under-estimate impact especially in longer-term, as it does not fully capture the dynamics of how reduced risk behaviours impact HIV transmission both directly and indirectly by lowering prevalence of cofactor STIs [55]. Goals may also under-appreciate MSM interventions, because MSM are assumed to not have non-marital heterosexual relationships-which may not apply in Mozambique. Lacking an age stratification in coverage, Goals projections assume VMMC to be homogeneously distributed through the male population 15-49 years. If VMMC preferentially reaches young men-per Mozambique's program policy and targets-impact may be different: e.g. greater in the long-term, but possibly less in the short-term, when boys circumcised have not yet started high-risk behaviours.
In costing, not all unit costs or cost components corresponded to observed expenditures (e.g. for human rights programs). For health sector-based services, human resource costs assumed fixed health worker time per service unit and fixed health worker capacity, without considering possible inefficiencies in health worker performance. High unit costs for outreach to MSM and IDUs reflect low current coverage; these were applied throughout 2015-2019 without efficiencies that may be realized through economies of scale [10]. Finally, projections ignored cost variation among regions (e.g. higher logistics and transport cost in the North, where commodities are trucked from Southern harbors instead of directly shipped).
In conclusion, the NSP is an ambitious but sensible strategy for reducing health and economic burdens of HIV in Mozambique. Projections support the feasibility of reducing new infections by 40%, and HIV-related deaths by 30% by 2019 from 2014 numbers, if stated coverage targets for both treatment and behavioural prevention, notably outreach to key groups, can be achieved. The proposed NSP will not turn the epidemic into elimination yet (as aimed for in UNAIDS' 90-90-90 and Fast Track targets); that will require further scale-up of targeted behavioural prevention and expanded treatment.