Prioritizing the scale-up of interventions for malaria control and elimination

The World Health Organization (WHO) Global Technical Strategy for Malaria 2016–2030 (WHO-GTS) sets the goal of universal access to malaria prevention, treatment and diagnosis [1]. This includes a core package of recommended interventions for reducing malaria-related morbidity and mortality: diagnosis and treatment of clinical and severe malaria, vector control with long-lasting insecticide-treated bed nets (LLINs) or indoor residual spraying (IRS) and chemoprevention for high-risk groups (infants, children in areas of seasonal transmission, pregnant women). The costs of providing these core interventions to all individuals residing in areas of malaria risk were estimated at US$2.7 billion per year in 2016, which will rise to $6.4 billion (95% UI US$4.5–9.0 billion) by 2020 if intermediate scale-up targets on the road to universal coverage are reached [1‐3]. This remains significantly higher than the current estimated malaria spend of US$2.5–3 billion per year since 2010 [3]. Given this funding gap, both country-level and international decisions regarding which interventions to fund and where are shaped by the shortfalls in required funds.

Ensuring universal access to prompt and effective diagnosis and treatment of malaria is both a health system priority as well as an ethical goal [1]. In the majority of malaria-endemic countries, artemisinin-based combination therapy (ACT) is used for effective first-line treatment of uncomplicated cases of falciparum malaria and intravenous artesunate for severe malaria. Vector control is implemented in all malaria-endemic countries, with the specific intervention mix between LLINs, IRS and other methods influenced by the distribution, relative abundance and behaviour of different vector species, as well as other context-specific factors such as the historical use of a particular form of vector control. The WHO recommends universal coverage of the population at risk with either LLINs or IRS [1] and whilst a combination of the two is deployed in some areas (primarily for insecticide resistance management), WHO recommends attaining high coverage of a single method before a second form is deployed.

In areas of moderate to high transmission, chemoprevention of high-risk groups is recommended. This includes intermittent preventive treatment of infants (IPTi), intermittent preventive treatment of pregnant women (IPTp) and seasonal malaria chemoprevention (SMC) for children under 5 years in highly seasonal areas of the Sahel region [1]. Although not currently recommended by WHO, the RTS,S vaccine is a potential additional tool that is in the process of large-scale post phase III pilot implementation in Ghana, Kenya and Malawi [4, 5].

Whilst coverage and access for these core recommended interventions have increased over time, significant gaps still exist. In 2015, it was estimated that only 19.7% (95% CI 15.6–24.8%) of febrile children under 5 years with Plasmodium falciparum received ACT [6]. In 2016 in sub-Saharan Africa 54% (95% CI 50–58%) of the population at risk slept under an insecticide-treated bed net (ITN) [3]. The proportion of the population at risk protected by IRS has declined from 5.8% in 2010 to 2.9% in 2016 [3]. Of the recommended chemoprevention measures, SMC has been implemented at scale and at high coverage in a sub-set of eligible areas [3, 7], IPTi has yet to be implemented programmatically at scale [3] and the coverage of three or more doses of IPTp remains less than 20% [3].

These interventions constitute a core cost-effective toolkit for malaria control [8]. LLINs remain a highly cost-effective intervention and their mass production and distribution has led to falling costs since their introduction [8, 9]. Cost-effective prevention is coupled with cost-effective front-line treatment in the form of ACT [8]. The chemo-preventative interventions have been shown to be highly cost-effective in a number of settings, a trend largely driven by the low cost of drugs recommended for SMC [10], iPTi [11] and IPTp [12]. The RTS,S vaccine has also been estimated to be an additional cost-effective tool (base on an assumed cost-per-dose of $5) in the near future [13].

Despite the availability of cost-effective tools, ambitious coverage targets combined with limited budgets mean that national malaria programmes and other key stakeholders face difficult decisions on where to focus available funds. Being able to effectively prioritize which interventions to fund and on what scale becomes increasingly important in these circumstances. It is therefore helpful for such decisions to be quantitatively assessed to ensure that the maximum marginal impact is gained for the least marginal cost. The best prioritization of interventions that will ensure effective allocation of resources for malaria control and elimination has been considered. By combining data on the efficacy and cost of delivering individual interventions at varying levels of coverage within a mathematical model, estimates were made of the most cost-effective packages of interventions for a given budget in order to maximize the reduction in malaria transmission, case incidence and mortality. The study also highlights the complex interactions between vector control, treatment and the measure of burden.

Figure 1 shows the most cost-effective pathways for scaling up interventions for three different initial transmission settings: low (panel A), medium (panel B) and high (panel C) initial parasite prevalence, respectively. As the budget is increased (shown on the x-axis), the mean coverage of two core interventions (LLINs and treatment) across all uncertainty runs is shown in the blue and red bars, respectively.

Across all settings, LLINs are estimated to be the most cost-effective intervention and are therefore scaled first in the majority of simulation runs (Fig. 1, Additional file 2: Figure S1). In both medium and high transmission settings, achieving high levels of LLIN coverage prior to scaling-up treatment is estimated to be the most efficient. This is because LLIN coverage works synergistically with treatment coverage scale-up by reducing disease burden and therefore the number of treatments that are required (Fig. 2). This effect is most pronounced at high transmission levels where increasing LLIN coverage always leads to substantial reductions in transmission (and therefore cases that need to be treated). In the low transmission setting, treatment is introduced at a lower level of LLIN coverage. In this setting the marginal benefit of increasing LLIN coverage, quantified in terms of reduced treatment costs due to fewer cases, diminished with increasing LLIN coverage (Fig. 2a), at which point introducing or scaling treatment coverage becomes relatively more cost effective. Examples of posterior draws for the non-linear costs are shown in Additional file 9: Figure S8.

Figure 3 shows the same prioritization curves with the addition of SMC (in seasonal settings) or IPTi (in non-seasonal settings). In both medium and high transmission settings in which these interventions are recommended, chemoprevention is scaled up after scale-up of LLINs to high coverage levels. Introduction (especially of SMC) is selected earlier in the high transmission settings (Fig. 3b, d), driven by the concentration of cases in younger age groups at higher transmission levels. In the medium transmission setting (Fig. 3a, c) treatment is selected alongside SMC as cases are less concentrated in the younger age group.

Figure 4 shows how the introduction of the RTS,S vaccine (at an assumed cost of $5 per dose) compares as an alternative disease prevention method to SMC or IPTi. In all settings the vaccine is selected late in the cost-effective pathway (Fig. 4). Whilst the vaccine remains a cost-effective measure [13], it is relatively less cost effective than LLINs or treatment and hence enters the prioritization pathway after these two interventions.

Scale-up trends are broadly similar for the alternative outcome measure of clinical incidence only (Additional file 3: Figure S2, Additional file 4: Figure S3, Additional file 5: Figure S4) or a weighted combination of incidence and mortality rates only (Additional file 6: Figure S5, Additional file 7: Figure S6, Additional file 8: Figure S7, Additional file 9: Fig. S8). When maximizing deaths averted, the drug-based strategies (treatment and chemoprevention) are introduced earlier in the pathway as they impact the rate of progression from clinical disease to severe disease and death. Increasing treatment coverage has a positive impact on mortality rate, regardless of the transmission setting (Fig. 5). Treatment has a relatively larger impact on case incidence in the lower transmission setting, where infections are more likely to be symptomatic and lead to treatment seeking. In high transmission settings, the small impact on transmission that clearing cases (and the subsequent prophylactic period) has is outweighed by the high force of infection.

A mathematical model of malaria transmission, combined with costing data was used to estimate the most cost-effective prioritization of combination of malaria interventions including treatment, LLINs, SMC, iPTI, and the RTS,S vaccine.

At the lowest budget levels, the modelling indicates that implementation and scaling prevention, in the form of vector control (LLINs) is the most cost-effective option. With increasing transmission, scaling prevention also has an important role of reducing the burden of malaria upon the routine health system. As budgets increase to moderate levels, treatment coverage is scaled up. As the nominal budget is increased further, both LLINs and treatment are scaled up together. The cost of both LLINs and treatment is modelled to rise non-linearly at higher coverage. This reflects inefficiencies and challenges associated with attaining very high access and coverage of LLINs and treatment. As the budget is increased, the point at which an intervention is scaled is therefore sensitive to the relative inflection points in the cost curves. The switch from LLINs to treatment is, in part, driven by the increasing marginal costs of pushing LLIN access to mid to high levels. Likewise, attaining the maximum coverage of treatment is assumed to be increasingly expensive (as health system capacity, scope and scale must be increased) and therefore the final steps where LLIN access and treatment coverage are both very high, are increasingly expensive. The impact and cost-effectiveness of scaling treatment, as well as being influenced by preventative interventions, is also a complex function of the immune dynamics of a population which impact the likelihood an infection is symptomatic and would instigate treatment-seeking. In this instance, it was assumed that all populations are equally likely to access either intervention. Spatial heterogeneities in the distributions of both malaria and available interventions could negatively impact outcomes (by decreasing impact, increasing costs or both). By assessing the impact of intervention packages at equilibrium in the model, temporal factors were omitted that may differ between interventions. These could include the time-scale for scale-up and associated impact. As such, policy decisions would also need to consider the feasibility and logistics of scaling any specific intervention to a target coverage. It is likely that the ability to execute an intervention package in a timely manner will also vary with the target coverage and where there is little to distinguish between two or more packages of interventions this may be a critical factor in decision making. As such the impact of timing and dynamic shifts in transmission is an important open research question in this field.

Whilst prevention proved more cost-effective using this approach, there are ethical considerations surrounding providing treatment for all that fit with the broader picture of universal health coverage more generally [30]. The goal of strengthening the health system, and therefore being better able to achieve targets for universal health coverage (UHC) is likely to work in synergy with reducing the health system burden through prevention of malaria in affected countries.

IPTi and SMC are further cost-effective tools at the disposal of malaria control programmes in regions where their use is appropriate. Both chemoprevention options were introduced and scaled up after LLINs had reached mid to high levels of coverage, being best layered in addition to core vector control as opposed to replacing it. The setting was influential here also, as with increased intensity of transmission, cases become more concentrated within younger age groups [31]. Therefore, the relative cost-effectiveness of these child- and infant-focussed strategies (particularly SMC, but also IPTi) becomes greater at higher levels of transmission.

The potential for the RTS,S vaccine to be included as an additional new tool in medium and high transmission areas was also examined. Whilst still cost-effective, and in line with previous work [32], the vaccine was introduced later than for other interventions. This is driven by the limited age groups the vaccine provides protection to combined with the relatively high costs (at an assumed cost per dose of $5). It is, therefore, apparent that focussing resources on achieving good levels of vector control and treatment coverage would be the priority before investing in vaccination.

Surveillance, monitoring and evaluation activities were not included in this analysis despite the importance of these components to a malaria control programme. Other interventions, for example, behaviour change communication [33], could also modify the cost-curve gradient. Data on the cost of these components at varying levels of ‘coverage’ is scarce and their impact is highly dependent on the mix of interventions and level of heterogeneity in transmission. There are, therefore, a range of benefits, economic and otherwise, that a well-functioning surveillance, monitoring and evaluation programme can bring about that are not captured here.

The outputs presented here demonstrate averaged approaches for generic transmission settings, aimed at informing decision-making in the general case. In practice there will be layers of context-specific information that will influence and shape the decision-making process. Malaria control programmes are faced with optimizing the impact of a single budget across an area with heterogeneous transmission, leading to a more complex optimization problem than presented here, and other more sophisticated approaches including, for example, simulated annealing [34] have been proposed for these scenarios. Some costs, for example, inpatient and outpatient costs, vary a great deal between countries, impacting country-specific treatment coverage. As discussed previously, the local ecology, vector species and associated bionomics will impact the choice of vector control. Insecticide resistance, where known, will strongly influence vector control decisions. This decision is further influenced by other factors: the logistics associated with large-scale programmatic deployment of IRS, historical precedents and acceptance. Drug resistance plays a key role in determining the application of chemoprevention strategies. Whilst the uncertainty propagated through this analysis reflects this between county variation, the interpretation must still be made on a country-by-country basis. Key drivers of much uncertainty are the non-commodity costs (e.g., delivery costs), which are likely to vary between countries substantially. Better knowledge of these costs would improve cost-effective decision-making.

The ICER was used as the decision-step making tool in this analysis. It should be stated that this step-wise approach could potentially result in a non-optimal solution for any given budget. This is a result of taking a step-wise approach in a highly non-linear environment (both with respect to impact and cost) leading to restriction of the solution to local optima. The ICER does however benefit from simple interpretation and implementation.

This study reiterates the importance of a combination of prevention, via vector control (here LLINs), and treatment as the cornerstone of malaria control in SSA. Additional currently recommended interventions (chemoprevention) and tools in development (RTS,S) will be valuable in those many areas where vector control and treatment are themselves not enough to bring about a transition to elimination. When resources are limited the prioritization of the use of available tools can help to maximize the impact of available finances, reducing malaria morbidity and mortality in a cost-effective manner.