Protocol for an economic evaluation of perennial malaria chemoprevention delivered at scale in Benin, Cameroon, and Côte d’Ivoire

Interventions

Stakeholders from the National Malaria Control Programme (NCMP), EPI, and other government bodies, non-governmental organizations (NGO), donors, and community organizations collaboratively co-designed PMC delivery models in each country through workshops in October-December 2021 and dialogue before and after. The funder was not present in the co-design but required a target population of children under the age of 2 years, use of SP, and ≤8 doses. Benin and Cameroon opted for an 8-dose schedule and Côte d’Ivoire a 5-dose schedule, co-delivered within the existing EPI delivery schedule (vaccines, Vitamin A supplementation, or growth checks) ( Table 1); the latter decision by Côte d’Ivoire was based on the assessment that coverage at other contact points was insufficient for effective PMC co-delivery.

Country stakeholders also selected the PMC pilot implementation districts ( Table 2). Areas receiving seasonal malaria chemoprevention (SMC) were excluded. The remaining districts were scored using the following criteria: duration of malaria transmission >6 months, high malaria prevalence, high vaccination coverage, and acceptable SP resistance (<50% prevalence of 540E). In Benin and Côte d’Ivoire, stakeholders selected one health district in each of the three regions (north, center, and south) to account for variation in epidemiological profiles. In Cameroon, stakeholders selected the Centre Region because it has the country’s highest malaria burden and chose six health districts from the 32 within the region based on high malaria incidence and the majority of rural populations ( Figure 1). Across the three countries, the ownership mix of facilities delivering PMC varied ( Table 2). Cameroon’s national policy supports a 5-dose PMC schedule in areas with perennial transmission, while Benin and Côte d’Ivoire do not currently implement PMC beyond the Plus Project-supported districts.

Figure 1. Maps of Plus Project-supported districts in a) Benin; b) Cameroon (inset shows the Centre region); and c) Côte d’Ivoire.

Dark blue indicates phase 1 districts; light blue indicates phase 2 districts.

The Ministry of Health (MoH) is delivering PMC with support from Population Services International (PSI), an international NGO, in the creation of training and routine monitoring materials, procurement of SP, and co-delivery of supportive supervision. In all three countries, health workers deliver PMC through health facility-based EPI clinics and outreach clinics, and community health workers (CHWs) support community mobilization activities. In Cameroon, CHWs have also been trained to administer PMC to children aged >6 months who received their first dose at a health facility, with priority given to areas with gaps in outreach coverage.

PMC implementation was planned in two phases: Phase 1 districts began delivering PMC in November-December 2022, with Phase 2 districts starting in March-June 2023 ( Table 2). Plus Project support for implementation in phase 1 and 2 districts will continue until March 2025 in Benin and Côte d’Ivoire, and until June 2025 in Cameroon.

Settings

The 12 selected implementation districts across the three countries have a combined target population of 226,000 children under 2 years. The Plus Project is supporting PMC delivery at 435 health facilities. Malaria transmission is perennial and intense, with malaria prevalence among children aged 6-59 months of around 50% (range across districts: 34-51%) ( Table 2). The administrative coverage of DTP1, a marker of EPI coverage, was ≥68% in all districts (range across districts: 69-156%) in 2023.

In Benin, the project will support PMC delivery in all 99 public health facilities (and their outreach services) currently providing EPI in the three selected health districts, whose combined target population is approximately 90,000 children under two years; preventive services (EPI and PMC) are not routinely provided in private health facilities in Benin (although they represent an important source of care for curative services, including malaria treatment¹³) ( Table 2). In Cameroon, the project will support PMC delivery in 224 health facilities (115 public and 109 private), approximately two-thirds of the total facilities providing EPI in the six selected districts. Around half of the facilities are public and half are private, including both faith-based not-for-profit and for-profit facilities, reflecting the distribution of healthcare providers in the intervention area. The total target population is approximately 46,000 children under two years. In Côte d’Ivoire, the project will support PMC delivery in all 112 public health facilities and in the four faith-based not-for-profit health facilities currently providing EPI in the three selected districts. The total target population is approximately 90,000 children under two years.

Study design

The study period for the economic evaluation was from January 2022 to July 2025 ( Figure 2). The decision problem motivating the economic evaluation is whether decision-makers in Benin, Cameroon, and Côte d’Ivoire should add PMC to their current package of malaria interventions, and if so, where and using which delivery model(s), reflecting the opportunity costs of this investment. Key decision-makers include managers within the NMCP and EPI of the Ministry of Health, as well as international donors supporting malaria control. We will take a disaggregated societal perspective, meaning that we will consider the costs to the provider (including the donor and national health service) and the household, both separately and together. We assess incremental financial costs, which reflect additional expenditure, and incremental economic costs, which reflect the opportunity cost of all resources used. The latter includes, for example, the time of health service staff, whose contributions may not result in direct financial expenditure but still represent an important cost to the health service. We will include the cost savings from reductions in malaria and anemia cases. Impact evaluations were used to assess the effectiveness of PMC in Cameroon and Côte d’Ivoire in reducing malaria and anemia incidence.¹⁶ In Benin, effectiveness will be modelled using efficacy estimates from genotype-specific modelling of the duration of protection of SP¹⁷ (further detail provided in the “Identification, measurement, and valuation of effects” section below).

Figure 2. Timeline of perennial malaria chemoprevention implementation and economic evaluation activities in Benin, Cameroon and Côte d’Ivoire.

Cost-effectiveness analyses will compare the incremental costs and incremental effects in the intervention areas of the co-designed PMC delivery models relative to a hypothetical counterfactual comprising the absence of PMC or, in Cameroon, the national 5-dose delivery model if they had been delivered in the same area. We present the incremental cost per malaria case averted and disability-adjusted life-year (DALY) averted. DALYs combine length and quality of life into a single measure of effect, and are a standard metric used in economic evaluation to allow comparison of results with other studies. We will model the costs and effects of scaling up PMC to all regions of each focus country where PMC may be suitable using information on malaria epidemiology and health system structure.

As detailed below, we will collect primary data, combine it with secondary data, and use various forms of modelling to meet our objectives.

Identification, measurement, and valuation of costs

Resource use data will be collected from multiple sources using various methods ( Table 3). We prospectively collected detailed financial data from PSI on project expenditures by activity. The economic costs of MoH personnel time involved in co-design, training, etc. will be captured through detailed record keeping of workshops and meetings by PSI. Key informant interviews (KIIs) will be conducted with managers of malaria and EPI services at national, regional and district levels to capture resources provided by the public health service in the implementation of PMC. Drug procurement costs are obtained from national and international price lists.

To capture resources required for PMC delivery and malaria and anaemia treatment at health facilities, we had a target sample size of 12-18 facilities per country; this allows for a purposive sample that is sufficiently large to include health facilities with varying characteristics (ownership, size, and rurality), while remaining logistically feasible. The final number selected in each country differed owing to the geographic spread and the available budget. We purposively sampled 45 health facilities (18 in Benin, 15 in Cameroon, and 12 in Côte d’Ivoire) from the 435 designated to provide PMC to obtain the widest possible variation in PMC costs, thereby allowing us to understand the key drivers of the variation in costs. Therefore, we sampled health facilities from three districts in each country with varying ownership, size, and rurality. This type of purposive (rather than random) sampling is standard in economic evaluation and allows maximum possible insights to be gleaned from a relatively small number of health facilities. Therefore, we did not perform power-based sample size calculations. Two data collection rounds (February-June 2023 and March-July 2024) will allow comparison of Phase 1, when only approximately one-third of the facilities had initiated early stage implementation, and Phase 2, when PMC implementation was established. The same facilities were visited in both the rounds. Small field teams of data collectors and their supervisors spent approximately 10 days in each health facility carrying out the following activities:

Other costs of accessing PMC (and vaccinations and Vitamin A) and costs of malaria treatment to households will be collected through household surveys within the process evaluation activities described elsewhere.

Identification, measurement, and valuation of effects

To calculate the number of DALYs averted by PMC, we will estimate the total number of malaria-related DALYs in children <2.5 years in a standardized population of 10,000 children <2.5 years (i) for each comparator and (ii) for the PMC intervention model(s) implemented. Although the intervention targets children aged <2 years, we will include children <2.5 years because some doses scheduled for children aged ~24 months may be delivered slightly later and may confer benefits at the start of the third year of life.

For each PMC model explored in each country, the number of incident malaria cases per 10,000 children <2.5 years will be estimated by multiplying estimates of malaria case incidence in the comparator scenario by the rate ratio (RR) of malaria cases with implementation of PMC. Effectiveness will be estimated in two ways: 1) directly, based on impact evaluations in focus countries where these are conducted (Cameroon and Côte d’Ivoire), and 2) indirectly, by using modelling to combine estimates of coverage and efficacy of SP doses.

Efficacy estimates will be obtained from genotype-specific modelling of the duration of protection of SP,¹⁷ using data from parasite clearance and prevention of infection (PCPI) studies being conducted in Cameroon and Zambia,¹⁸ and parasite resistance genotype mapping activities. Detailed estimates and an understanding of coverage will be obtained from the Plus Project process evaluations. The direct estimates from the impact evaluations will be used to validate and calibrate the approach to indirectly modelling effectiveness. The (indirect) modelling will then be used to estimate effectiveness in all areas of the focus countries where PMC is considered potentially suitable based on SP efficacy, the absence of SMC, and other factors. The incidence of mild, moderate, and severe anemia with PMC will be estimated in similar ways.

Data management

For each health facility, we entered expenditure records directly into an Excel-based spreadsheet developed for the project. At least two supervisors reviewed the data for quality control purposes. We will use Visual Basic for Applications (VBA) macros to extract data from these individual health facility spreadsheets into a single Excel workbook in a long format for analysis, alongside data from Mozambique. We entered register data directly into ODK forms on encrypted tablets and use paper forms for healthcare worker observations and caregiver time logs before entry into ODK forms. ODK forms had built-in data validation controls and checks with additional regular reviews by the data manager. Further checks were conducted by the data manager.

We recorded KIIs using the same encrypted tablets and transcribed them in French, ensuring the use of unique identification codes without names to preserve the anonymity of the participants. The transcripts were checked against the original audio for quality control.

All electronic data were captured using encrypted devices and stored in password-protected devices and database systems. After verification, and as soon as the Internet was available, data was be uploaded to a central secure server. Data – whether on central computers and servers, remote computers, or handheld devices – was backed up daily to a central secure data server. We will delete the original audio recordings from all devices once the transcription is complete. A data manager in each country has full access to data through the server.

Detailed financial data from PSI will be provided regularly by the PSI program management team as an Excel-based transaction list from their financial reporting system. Transactions will be coded against activities and expenditure categories using standard fields and reporting codes within the reports and verified through discussion with the PSI team.

Data analysis

Costs of implementing PMC (objective 1)

We will code and analyze KIIs thematically in French using NVivo 14¹⁹; we will seek to understand how EPI and PMC services are organized and identify information on resource use.²⁰ We will analyze quantitative data captured via ODK in Stata 18²¹ using descriptive statistics to explore variations within and between health facilities and over time with respect to health worker and caregiver time on EPI clinic days, which will then be incorporated into cost analyses in Excel.

We will analyze resource use and cost data collected from health facility expenditure records and PSI records in Excel. We estimate the provider costs incurred within health facilities of delivering PMC using a combination of step-down costing (to estimate the cost per EPI visit) and micro-costing (to estimate the cost of SP and associated supplies).^8,22 Financial data from PSI will be analyzed to estimate the fixed and variable provider costs of supporting PMC delivery, including SP procurement.

We will estimate the cost per PMC dose delivered and the total costs for PMC implementation across the eligible areas of each country. We will disaggregate costs in various ways, including by cost center (personnel, materials, SP drugs and storage, meetings and training, equipment, transport, and monitoring), perspective (i.e., donor, national health service, EPI, NMCP, community level, household, etc.), financial and economic costs, and start-up and recurrent costs.

Cost-effectiveness of PMC (objective 2)

Qualitative analysis will be used to refine our understanding of decision makers and decision-making processes around PMC, which will inform modelling choices. Our main outcome metric is the incremental cost-effectiveness ratio (ICER).

The costs of the intervention and comparator strategies reflect the combined costs of implementing PMC and treating malaria and anemia cases:

In this way, the ICER will reflect the additional costs of implementing PMC in the intervention strategy, minus the cost savings from having fewer malaria and anemia cases to treat than in the comparator strategy.

The cost to the public health provider per malaria case and per anemia case diagnosed and treated will be estimated separately for outpatients and inpatients from our health facility data following a similar approach as for PMC costs. Step-down costing will be used to estimate the cost per outpatient consultation and per inpatient day for children <2.5 years at each health facility. For each case sampled in the detailed register review, we will generate a cost by applying these unit costs per outpatient or inpatient day and unit costs of diagnostics and medicines obtained from national and international sources to patient-specific resource use data. Analyses of these individual patient data reflect clustering at health facilities and purposive sampling.

A decision tree model will be used to combine primary data on costs and effects with additional primary and secondary data. We express all estimates of costs and effects for a standardized population of 10,000 children under the age of 2.5 years. The implementation costs and cost savings will be presented separately and in combination.

A base case analysis for each country will explore the cost-effectiveness of PMC using the most plausible set of assumptions. We will use deterministic analyses to explore heterogeneity; that is, how our results may vary across settings. We use deterministic and probabilistic analyses to explore the sensitivity of our findings to uncertainty in methodological choices, including input parameters.

The costs of implementing PMC will be estimated over a period of five to ten years to be decided in consultation with relevant stakeholders. The costs and effects of incident malaria cases and anemia cases occurring during this implementation period will be modelled over a lifetime. A 3% annual discount rate will be used for both costs and effects in the base case.²³ DALYs will be modelled without age weighting and using standard DALY weights²⁴ and life tables²⁵ as well as other secondary data (e.g., duration of an episode of malaria or anemia).

To understand whether PMC is likely to represent an efficient use of scarce resources, we will compare the ICERs we estimate with appropriate cost-effectiveness thresholds, such as those recommended by Ochalek et al.²⁶ Cost-effectiveness acceptability curves will be generated, and probabilities calculated that adopting PMC in a given context and for a given cost-effectiveness threshold will represent an efficient policy choice. ICERs for PMC will also be compared to ICERs from alternative interventions targeting children aged <2 years, based on a review of evidence at the time of analysis.

Budget impact (objective 3)

A qualitative analysis of interviews with national- and regional-level stakeholders will inform our understanding of budgets and budgeting processes. A budget impact analysis will be performed to explore whether introducing PMC may be affordable within the relevant provider budgets in each country. We will compare the net incremental financial costs of PMC (including both implementation costs and cost savings from reduced malaria incidence) to specific budgets where possible. The time horizon will be limited to a period relevant to the budget holder, usually 5-10 years.

Equity (objective 4)

We will assess the equity of implementing PMC by exploring the variation in costs and the effects of PMC models across sub-populations. The main equity analyses will use wealth quintiles as a measure of socioeconomic status. Wealth quintiles will be created using principal component analysis²⁷ based on standard asset ownership questions included in the impact evaluation cohort census survey and process evaluation household surveys.

Ethical considerations

Ethical review committees approved this study in Benin (Comité national d’éthique pour la recherche en santé) (103/MS/DC/SGM/CNERS/SA), Cameroon (Comité national d’éthique de la recherche pour la santé humaine) (2023/01/1512/CE/CNERSH/SP), Côte d’Ivoire (Comité national d’éthique des sciences de la vie et de la santé) (001-23/MSHPCMU/CNESVS-km), WHO (ERC.0003756), and LSHTM (27531). Following sensitization to the study and presentation of ethics approvals, administrative authorization was obtained from the relevant national, regional, and/or district authorities.

Before commencing data collection from health facilities, written informed consent was obtained from the head of the health facility. We also obtained written informed consent from all participants in the primary data collection (i.e., key informant interviews, health worker observations, and caregiver time capture). For participants who were unable to read the information sheet, a witness verified that the information had been explained correctly. We used assent forms for participants younger than the age of the majority (<21 years in Cameroon; <18 years elsewhere).

We ensured that the locations of the interviews provided privacy. Respondents may participate but choose not to quote, even anonymously; specific consent was sought for the use of anonymous quotes. We will present the results in ways that prevent the identification of individuals or health facilities.

Dissemination of findings

Research team members attend quarterly chemoprevention Technical Working Groups (TWGs) in each country to present updates on evaluation activities and are in regular communication with PSI country teams for two-way communication on implementation and evaluation updates. After each round of activities, we will disseminate preliminary results through the national as well as international chemoprevention TWGs attended by a range of relevant stakeholders, including the MoH, particularly the NMCP and EPI, and their partner agencies (NGOs, technical agencies and implementing partners, donor organizations, WHO, UNICEF). The results will also be widely disseminated in each country through a series of meetings that will include regional and district health authorities, as well as the national level, and the health facilities and communities where we are working. The results will be shared more widely through scientific publications in peer-reviewed journals, at national and international conferences, and other fora, as appropriate.