The use of malaria risk maps in the NMSPs is presented and stakeholder perceptions about the utiliy and limitation of the maps are explored in further depth.
Risk maps included in the National Malaria Strategic Plans
In all three NMSPs, malaria risk maps showing the distribution of infection and, consequently, the risk of being infected were included. However, each country used a different type of malaria risk map, based on different kinds of data: a modelled PfPR map in Kenya, an incidence map based on routine data in Malawi and a prevalence map based on the Demographic Health Survey (DHS) data in the DRC (Fig. 1). Although the three maps were developed using different methods and types of data, they were used in the NMSPs for the same purpose, that of showing epidemiological strata and identifying high risk areas where interventions would be implemented.
In the Kenya NMSP 2009–2018 (29) and Kenya NMSP 2019–2023(30) maps were used to show and describe the epidemiological strata and the change of risk in each area over time. Kenya presented seven strata of the PfPR in 2 to 10 year old children: malaria free, < 1%, 1–5%, 5–10%, 10–20%, 20 to 40%, > 40%. The strategy indicated that each vector control intervention (LLINs distribution, larval source control and IRS) would be deployed in specific areas according to the stratification (30).
The Malawian NMSP 2017–2022 (31) used maps that illustrated the evolution of risk according to the Annual Parasite Incidence and the number of cases per 1,000 population by each district (divided into 5 categories: under 150, 150–250, 250–350, 350–450, over 450). According to the strategy, interventions would be allocated universally, with the exception of IRS, which would be implemented in high burden districts or areas (31).
The Congolese NMSP 2016–2020 (32) used maps that showed the stratification of the PfPR in < 5 years old children in four strata: pre-elimination in North Kivu (PfPR < 5%), control/consolidation (PfPR 6–30%), control/intensification (PfPR > 30%) and urban malaria (in Kinshasa). The strategy stated that all interventions would be allocated universally, with the exception of IRS, which would be implemented in urban areas and in the pre-elimination areas, and LLIN distribution in schools in the tropical areas (32).
Stakeholder perceptions – what drives the use of malaria risk maps
The analysis of stakeholder interviews foucsed on two main thematic areas: 1) types of use of risk maps - including strategies, prioritisation, targeting and operational planning; 2) the drivers of the use of risk maps for strategic planning - including perception of value and limitations of the maps by malaria stakeholders.
Use of risk maps
The information derived from the interviews matched and enriched that from the NMSPs. Risk maps were primarily used for strategic planning, in particular to aid in the selection of geographic areas or population sub-groups for delivery of an intervention (targeting); however, operational planning and advocacy were also important identified as important uses.
Prioritisation and Targeting: selection of interventions and geographic areas
Intervention choices were based on evidence from efficacy and effectiveness studies, as well as WHO Global Malaria Programme (GMP) guidelines. Across the three countries, treatment was widely prioritised over prevention as it was perceived as a tool for ‘saving lives’ as opposed to preventing infection. Stakeholders, including NMCPs and donors, reported that interventions were prioritised based on their perceived efficacy, which primarily meant ensuring the availability of commodities. Anti-malarials and diagnostics for case management were the first priority, followed by LLINs and IPTp for prevention. This prioritisation was also seen in the Global Fund proposals reviewed in each country, where the majority of the requested budget was for commodities: artemisinin-based combination therapy (ACTs), rapid diagnostic test (RDTs) and LLINs (33, 34). This was confirmed in conversations with the NMCP officers who oversee Global Fund grants interviewed.
“You have to make sure that lifesaving interventions are taken care of fully… there was no question... there was no debate about that. There have to be diagnostics, there have to be medicines. That's number one, because we have to save people from dying. Then the second thing was prevention, accessibility to nets.” (NMCP officer 1, Kenya)
“Of course commodities [for case management] comes first. Next is the nets.” (NMCP officer 6, Malawi)
The consideration of the geographical areas where interventions would be implemented was based on maps. Targeting was applied to all preventative interventions (LLINs, IPTp, IRS) in the case of Kenya, while limited to IRS, or LLINs in schools, or IPTp at the community level in the case of Malawi and the DRC. In Kenya, the delivery of LLINs and IPTp was only implemented in the 16 endemic counties of the Lake and Coast regions (out of 47 counties in total). In the DRC and Malawi, LLINs and IPTp were delivered in the entire country through using a universal coverage approach, however maps were utilised to identify high burden areas where additional interventions or delivery sites were appropriate, such as LLINs distributed in schools in DRC, IRS in Malawi, and the delivery of IPTp at the community level in DRC (Table 4).
Table 4
Type and use of malaria risk maps included in the most recent National Malaria Strategic Plans by country
Type of risk map
|
Source of data
|
Map resolution
|
Main use of the map in the NMSP
|
Use for targeting
|
Kenya: NMSP 2019–2023
|
|
Modelled PfPR map (geostatistical modelling)
|
Multiple surveys and studies combined with environmental data
|
Second- level administrative division (sub-counties)
|
To show the epidemiological stratification: endemic areas (lake and coast), seasonal malaria transmission areas, malaria epidemic prone areas (western highlands of Kenya) and low risk malaria areas
|
Maps were used to identify epidemic and epidemic-prone areas where LLINs were to be delivered by mass distribution and routine channels at ANC; to identify zones where to implement IRS (lake endemic areas) and IPTp (lake and coastal endemic regions); and to identify zones where installing buffer stocks of case management commodities and IRS (epidemic prone areas) was appropriate
|
Malawi: NMSP 2017–2022
|
Descriptive incidence map (cases per 1,000 population)
|
Routine HIMS 2011–2015
|
Second-level administrative division level (Districts)
|
To show variation in incidence across districts and decline in incidence from 2011 to 2015
|
Maps were used to identify highly endemic districts where to implement IRS interventions
|
DRC: NMSP 2016–2020
|
Descriptive PfPR map
|
DHS survey 2013–2014
|
First-level administrative division (by the 26 new provinces created in late 2015)
|
To show the malaria pre-elimination, control- consolidation and control intensification areas
|
Maps were used to identify areas where LLINs were to be additionally distributed through schools (areas with prevalence > 30%, also defined as tropical regions) and areas where to implement IRS (in pre-elimination and urban areas in North Kivu and Kinshasa)
|
Participants in Malawi and DRC perceived targeting to be associated with limited resources, while in Kenya participants felt the use of a targeted approach was primarily to increase efficiency and value for money in malaria control.
In DRC, with severe resource constraints, the NMCP used what they called ‘time prioritisation’, as defined by a report developed by the African Leaders Malaria Alliance (ALMA) (35). This refers to the practice of implementing malaria control interventions in high burden regions initially, whilst searching for funding to implement the interventions in additional geographical areas. As international partners explained:
“[in] DRC: what they have done is a time-bound prioritization, so they covered their 2018 and 2019 LLINs campaigns and 2020 has gaps. This it is essentially an operational programmatic prioritization of your limited resources... This is just common sense decisions a government has to make” (Partner 2, Global)
In Malawi, although LLINs and IPTp were universally implemented, targeting was perceived, by some, as a better strategy given the resource contraints and geographical variations in risk, as explained by one official:
“We should stop actually doing the blanket interventions because it is a waste of resources in some areas where they don’t need those interventions so we need more data that can guide us to plan for targeted interventions because in Malawi we have... yes we have malaria but ... all areas are not affected equally.” (NMCP officer 6, Malawi)
However, some stakeholders perceived the prioritisation and targeting of high impact interventions (LLINs and IPTp) was not suitable in a country where the risk of malaria is high everywhere. They felt that universal coverage of LLINs and IPTp was the most appropriate strategy.
“When you are a country where everywhere is highly endemic then there is not much space anyway for prioritization.” (Partner 2, global)
Also reflected in statements by participants from DRC and Malawi:
“There is no prioritization of provinces [region] in relation to prevalence…not at the moment.” (Partner 14, DRC)
“If we had universal coverage in terms of the vector control, that would be much better so that at least we reduce the incidence, after reducing the incidence then you can now try to see where can we go….[targeting].” (Partner 14, Malawi)
Planning of operational interventions, commodity quantification and advocacy
Malaria maps were used for purposes beyond strategic planning and targeting. These included project monitoring or planning, supply quantification, financial justification, and budget advocacy purposes.
Guiding and justifying commodity quantification and operational interventions
NMCPs case management division and NGOs officers described the use of risk maps to guide the quantification of commodities, such as RDTs and ACTs, according to the level of burden. Indeed, population at risk, number of malaria cases and deaths and malaria prevalence maps were utilised conjointly to quantify malaria commodities in a specific area.
“[high endemicity areas] these areas are supplied differently than the others. For example in Haut-Uele where the endemicity is very high with seasonal upsurges, even we talked about the epidemics, the attention is different, we bring in the inputs we repositioned elsewhere.” (Partner officer 19, DRC)
Maps were also used to justify national decisions, such as where LLINs needed to be allocated, to the sub-regional government. Malaria control interventions, such as universal LLIN distribution and IRS, are highly visible and often local governments are interested in implementing the intervention in their area, independent of the level of malaria risk. Having a malaria risk map helped the central government to justify the geographical allocation of interventions to the sub-national level.
“Accepting like when your program tells you really we are not giving you nets, not because we don’t like your county or your county didn’t vote for government, no. It’s because the evidence … do not have the high prevalence in malaria in your county you don’t need this. And there’re now starting to actually understand this… because I think one of the question was why are you not spraying my county, why are you not giving me nets. You are there sitting in a national meeting and hearing that nets have been distributed why not my county. So having a document or having something that you can show them and tell them it’s because of one, two, three.” (Partner officer 8, Kenya)
Monitoring interventions and trends over time
Maps were specifically used for annual reviews and during the mid-term and final reviews of the NMSP. Malaria risk maps of consecutive years were key to showing progress and readjusting the strategy over time, as Kenyan officials explained:
“For us at the national level we're using that [the map] to show progress of malaria control over time.” (NMCP officer 1, Kenya)
“We’re able to see the map actually shrinking or is it becoming darker but I can report that actually we’re heading [in] the right direction it’s becoming lighter and lighter. …. Yeah we’ve made a lot of gains since 2010. 2007 MIS, 2010 [MIS], 2015 [MIS] we are seeing progress.”(NMCP officer 14, Kenya)
Resource mobilisation and advocacy
The visual nature of the maps was seen to encourage their use as powerful tools for resource mobilisation and advocacy, and training purposes. Respondents in Kenya and Malawi gave examples where maps were used to encourage donors or other stakeholders to provide interventions to specific areas where there were gaps.
“If any donor comes, wants to come with any interventions, we direct him to say okay, according to the... the distribution of the burden of malaria I think you go to this location.” (NMCP officer 15, Malawi)
Maps were used to advocate for funds at the sub-national level, in countries with some degree of devolution, and by community-based organisations to advocate for funds or for social accountability.
“They have used that [maps]…..to advocate for the funds from the county.” (NMCP officer 1, Kenya)
“In 2013 they allocated 113 million shillings for malaria [in Nairobi]. So now we were questioning what did you do with this money? Which interventions did you do?” (Partner officer 11, Kenya)
Finally, the ease with which malaria risk could be visualised was highly appreciated and often used for training purposes to catch attention of the audience.
“When I am training them on malaria epidemiology and decision making I project for them and give them soft copy of my presentation. In this county these are the areas that you should focus your efforts more.” (Researcher 4, Kenya)
Factors driving the use of risk maps by decision-makers
The second key theme explored in the analysis were the factors driving the use of risk maps for strategic planning. The decision to use malaria risk maps was motivated by their availability (or potential to be developed), the technical characteristics of the maps, and the alignment of the maps with stakeholder expectations. Technical characteristics of the maps included: the nature and quality of the data from which the maps were developed and the granularity of the data; while the alignment of the maps with stakeholder expectations included a range of related factors such as: alignment with the expected malaria epidemiological situation in the country (based on eco-climatic zone, routine data, or indications from sentinel sites); knowledge, trust and perceived ownership of the data and of the process of developing the maps.
Availability of maps and data for their development
Malaria risk maps were developed using either multiple PfPR data points, DHS prevalence data at one point in time, or by using routine health information system data. Modelled PfPR maps were available in all three countries at the time of development of the most recent NMSP, provided by the INFORM and LINK projects. In Kenya modelled PfPR maps were developed around the time of the mid-term NMSP review and were incorporated in the revised strategy 2015–2019; in DRC the maps were developed in 2014 and the NMSP 2016–2020 was developed in 2015–2016; in Malawi modelled PfPR maps were developed in 2014 and the new strategy 2017–2022 in 2016–2017. Timely alignment of data and maps with malaria strategic planning cycles was a key element that increased their utilisation. NMSPs are developed by NMCPs and technical partners every 5–10 years and are revised as interim strategies every couples of years. The fact that the maps were developed with recent data, at the time of the NMSP revision, facilitated their utilisation.
Routine health data were available in each country and by sub-national level. These data were perceived by some stakeholders as advantageous due to their being more recent and timely compared to national survey data, which are produced every three to five years. Timeliness of the monthly routine data (the period needed to send data from health facilities to the central level) was perceived as an issue by some policy makers.
“We still have challenges with the routine data, we have challenges with the completeness, accuracy and timeliness.” (NMCP officer 6, Malawi)
Technical characteristics and quality of the data and maps
A key concern raised about the use of maps for decision-making involved the nature, representativeness and perceived quality of the data from which the maps were developed. Stakeholders in Malawi and the DRC raised concerns about a lack of transparency and clarity on the source and breadth of data in the models used to develop the maps. This meant that stakeholders felt that they were not able to judge the quality of data from which the map was developed.
“It is the data that was included in the model that was my biggest problem …You need to choose, which is the data that you need…” (NMCP officer 3, Malawi)
“I think that there are surely some biases that have entered through different studies that have been taken into account.” (NMCP officer 1, DRC).
“Because as long as we don’t have good data, the maps will not work. The maps will not work at all.” (NMCP officer 3, Malawi)
Conversely, routine health data were perceived as understandable and useful in indicating the distribution of malaria burden and guiding decisions despite the acknowledgement that the data was not always of good quality. Perceptions of the quality of routine data varied across countries and stakeholders.
“We can go with the health zone routine data… as much as we can. I think that at least the routine data allows… to have a distribution variation from one area to another and the routine data still provide satisfactory information on the fact that such area is less affected than another. Okay, we know that in terms of accuracy, it is not very reliable but in terms of distribution of burdens, it is quite satisfactory. If we have surveys with satisfactory accuracy up to the next province, we work on the improvement of the routine for the provincial deployment. This is the compromise that seems reasonable to me for a gigantic country like the DRC.” (NMCP officer 1, DRC)
Concerns were raised about the granularity of the data and some participants questioned how representative the maps were of populations at the sub-national level. In Kenya, modelled PfPR maps for the county level were available and appreciated. Modelled PfPR maps with district-level resolution were available in the DRC and Malawi. However, in Malawi some stakeholders either did not know of their existence or did not feel that the data used to develop the map was accurate enough to be representative at the district level.
“Prevalence alone I think we don’t have enough [prevalence] data to come at district levels.” (NMCP officer 15, Malawi)
DRC officers reportedly felt that the PfPR data available were not sufficient to develop a representative map and as such they decided to use the most recent DHS data (2013–2014), despite it only providing provincial-level resolution, however, they have pushed to make the 2020 DHS/MIS survey representative of the 26 provinces.
“I'll be more comfortable with surveys at 26 provinces, at least they give me an image close to the reality …..models ….give me things that deviate from the realities.” (NMCP officer 1, DRC)
The fact that DHS and MIS data were representative at national or at the higher sub-national levels (e.g. region or province) but not at district-level was perceived as a limitation of the national survey data. As one respondent explained:
“Unfortunately, MIS [Malaria Indicator Survey] only gives us by region, it’s not by district. So the smallest you can go with analysis is by region. That’s one of the challenges it will have.”(Partner officer 14, Malawi)
NMCP officers and other stakeholders in Malawi reported using incidence data from the routine health information system because they perceived this to be the only way to develop a map at district level that they felt they could trust.
Stakeholder ownership, involvement and alignment with stakeholders expectations
In Kenya, the NMCP and other stakeholders interviewed were proud to have shifted to a targeted approach for malaria interventions and appreciated the modelled PfPR maps. Respondents felt a sense of ownership because their suggestions of what data was useful were included in the maps.
Stakeholders in all three countries highlighted the importance of the sense of ownership of the maps by NMCP and of engagement of researchers and technical advisors in supporting the NMCP to develop risk maps. Kenyan NMCP officers mentioned the long term (over 20 years) and daily collaboration with KEMRI- Wellcome Trust; how the researchers have consistently contributed to the TWGs by reviewing interventions, routine data, discussing changes in strategy and policy and generating and sharing new national and global evidence. By contrast, in the DRC and Malawi, there was a keen sense from the interviews that the NMCP did not feel sufficiently involved in research generally, and specifically in the development of the PfPR maps. More importantly, the lack of involvement of the NMCP in the development of the PfPR maps had negative implications for the use of the evidence based on the maps.
As one NGO official in Malawi explained:
“Sometimes if you don’t involve the national program at the beginning …there is unwillingness to accept whatever comes out of your study.” (Partner officer 17, Malawi)
However, in addition to trust in the data and legitimacy of the process, there were also indications that the choice of map used in the NMSP and other national documents could be influenced by whether what the maps showed was in alignment with what decision-makers expected to see based on other data sources or publications. In DRC, for example, the expectations of the NMCP, based on the routine data, aligned better with a map that was produced using DHS data than with the modelled map. A local officer commented:
“The rendering of this model did not satisfy us, because it did not add to what we expected, and what we aimed at in terms of return routine field data… so finally we chose to make our stratification on the basis of the parasitic prevalence of EDS [DHS].” (NMCP officer 1, DRC)
In Malawi, participants expressed a lack of confidence in the modelled PfPR maps as they did not show the progress that was perceived to have been achieved by stakeholders based on a Roll Back Malaria (RBM) publication (36). In that publication, a multivariate analysis and the Lives Saved Tool (LiST) were used to hypothesise that malaria interventions from 2000 to 2010 had reduced mortality in children in the country, (mortality assumed to be largely caused by malaria, which contradicted the modelled PfPR maps, as explained by a researcher:
“[the PfPR map] showed that there were no changes in in malaria prevalence in the country from 2000 to 2010…in contrast with the RBM impact series [which] showed that Malawi got a decrease prevalence and actually… Malawi was also awarded by the ALMA [African Leaders Malaria Alliance] with a prize.” (Researcher 1, Malawi)
The data and the indicators utilised by the two studies were different and not comparable. However, it is logical given alternative versions of achievement that NMCP would be less accepting of the one suggesting no change in malaria prevalence after their scale-up of interventions. Furthermore, the interpretation of conflicting data could also be a challenge. Stakeholders in both countries mentioned that they preferred maps that showed what they expected to see, a decreased number of cases in Malawi and alignment with routine data in DRC.