Malaria Burden Stratification in Malawi- A report of a consultative workshop to inform the 2023-2030 Malawi Malaria Strategic Plan

Background: Malawi's National Malaria Control Programme (NMCP) is developing a new strategic plan for 2023-2030 to combat malaria and recognizes that a blanket approach to malaria interventions is no longer feasible. To inform this new strategy, the NMCP set up a task force comprising 18 members from various sectors, which convened a meeting to stratify the malaria burden in Malawi and recommend interventions for each stratum. Methods: The burden stratification workshop took place from November 29 to December 2, 2022, in Blantyre, Malawi, and collated essential data on malaria burden indicators, such as incidence, prevalence, and mortality. Workshop participants reviewed the malaria burden and intervention coverage data to describe the current status and identified the districts as a appropriate administrative level for stratification and action. Two scenarios were developed for the stratification, based on composites of three variables. Scenario 1 included incidence, prevalence, and under-five all-cause mortality, while Scenario 2 included total malaria cases, prevalence, and under-five all-cause mortality counts. The task force developed four burden strata (highest, high, moderate, and low) for each scenario, resulting in a final list of districts assigned to each stratum. Results: The task force concluded with 10 districts in the highest-burden stratum (Nkhotakota, Salima, Mchinji, Dowa, Ntchisi, Mwanza, Likoma, Lilongwe, Kasungu and Mangochi) 11 districts in the high burden stratum (Chitipa, Rumphi, Nkhata Bay, Dedza, Ntcheu, Neno, Thyolo, Nsanje, Zomba, Mzimba and Mulanje) and seven districts in the moderate burden stratum (Karonga, Chikwawa, Balaka, Machinga, Phalombe, Blantyre, and Chiradzulu). There were no districts in the low-burden stratum. Conclusion: The next steps for the NMCP are to review context-specific issues driving malaria transmission and recommend interventions for each stratum. Overall, this burden stratification workshop provides a critical foundation for developing a successful malaria strategic plan for Malawi.


Introduction
Malawi aims to eliminate malaria 1 . Since the year 2000, there have been four strategic plans that have been implemented. The most recent one started in 2017 and is ending in 2022 and aims to reduce the incidence of malaria from 386 per 1000 population in 2015 to 193 per 1000 by 2022 and to reduce malaria deaths from 23 per 100,000 population in 2015 to 12 per 100,000 by 2022 1 .
Over the last two decades, malaria control efforts in Malawi have been scaled up substantially through multiple control measures that include the use of insecticide treated bed nets, Artemisinin Combination Therapies (ACTs), Intermittent Preventive Treatment in pregnancy (IPTP), and malaria rapid diagnostic Tests (mRDTs) 2,3 , indoor residual spraying (IRS) and more recently the RTS,S malaria vaccine 4 . Based on Malaria Indicator Survey (MIS) reports, substantial reductions in malaria transmission have been reported. For example, the national malaria prevalence in 2010 was 44% and subsequently reduced to 10.5% in 2021, representing a 52.2% reduction over 11 years [5][6][7] . Similarly, malaria incidence and mortality have reduced from 407 per 1,000 to 361 per 1,000 population in 2016 and 65.2% decrease in mortality from 23 per 100,000 to 8 per 100000 population in 2021 respectively 8 .
Variations in intervention coverage and user uptake, weather patterns, vector, host, and parasite factors are changing the malaria landscape leading to pockets of high transmission and low transmission within the same geographical units such as districts, a status referred to as 'heterogeneous transmission'. In a heterogeneous transmission setting, application of malaria interventions is without geographic targeting at subnational level is less cost-effective. It does not have the impact of reducing malaria impact indicators such as incidence, prevalence, and mortality 9 . Moreover, in resource limited settings such as Malawi, the scarce resources need to be appropriately allocated considering the observable flattening and reduction in donor as well as government support for malaria control and elimination. The Malawi National Malaria Control Program (NMCP) is currently performing an end-of-strategy review in preparation for developing the 2023 to 2030 malaria strategic plan. In this process, the program is poised to move to a targeted approach in implementing malaria control interventions. Some of the critical reasons for this shift are that there is an observable flattening and reducing donor support for malaria control and elimination. Additionally, the government's allocation of resources for malaria control is also decreasing over time. Recent experiences from the rollout of the RTS,S malaria vaccine serve as an example of the need to shift to a targeted approach. The programme is determined that through better analysis and strategic use of quality data, the country can pinpoint where to deploy the most effective malaria control tools for maximum impact. In 2016, the World Health Organisation (WHO) took a similar approach to malaria control as; approximately 70% of the world's malaria burden was concentrated in just 11 countries -10 in sub-Saharan Africa (Burkina Faso, Cameroon, the Democratic Republic of the Congo, Ghana, Mali, Mozambique, Niger, Nigeria, Uganda and the United Republic of Tanzania) and India. The sorting of the global cases led to the development of the high burden to high impact (HBHI) approach. The HBHI strategy was introduced to fight malaria in these 11 countries. Malaria risk stratification has been key to the targeted rollout of malaria intervention in these countries 8 .
Malaria burden stratification is a process that involves classification of the geographical areas or localities according to factors that determine receptivity and vulnerability to malaria transmission 10 . Some have argued that the malaria risk stratification is the first step to planning prevention and control 10 . Malaria burden stratification is important because it allows public health officials to target their interventions to areas where they are most needed 11 . By identifying areas with the highest burden of malaria, officials can prioritize resources and interventions to those areas, which can help to reduce the overall burden of the disease. Additionally, by monitoring changes in malaria burden over time, officials can assess the effectiveness of their interventions and make adjustments as necessary. Furthermore, Malaria burden stratification can also help to identify areas where other health issues may be present, such as poverty, poor access to healthcare, or poor housing conditions, which may need to be addressed in order to effectively combat malaria 8 . There are no standard methods for malaria burden stratification. Examples from countries that have done the process involve using one or more burden indicators, such as prevalence, incidence, mortality, and intervention coverage indicators 12-14 . WHO encourages countries to define their own contexts and indicators and track their own progress 15 .
In October 2022, the Malawi NMCP for the first time set up a task force to perform a malaria burden stratification to inform the new strategic plan to be developed as well as the Global Fund grant application scheduled in January 2023.
The following were the objectives/deliverables/terms of reference for the task force.
• Review malaria burden and intervention coverage data to describe the status to date.
• Define and document the process for malaria burden stratification and all the considerations.
• Decide on the appropriate administrative level for stratification and action, with all the considerations, including health systems factors, current administrative structures, data quality, etc.
• Stratify Malawi into appropriate burden strata and define/describe the context-specific issues driving transmission in hot spot areas.
• Recommend both current and additional interventions in each of the strata.

Ethical statement
In the workshop, we utilised secondary aggregate survey data and did not require ethical approval since the routine health information systems data belongs to the Ministry of Health and was analysed at an aggregate level. For prevalence data, the authors confirmed that the relevant institutional review boards in Malawi had approved the original sampling. Furthermore, the Ministry of Health granted approval to extract the data. Participants in the workshop provided their consent by accepting an invitation sent to them via email by the Ministry of Health. The email provided information about the workshop, including its objectives and expected outcomes.

Study design
The malaria burden stratification task force set up by the NMCP comprised 18 members from the control programme comprising of members of the NMCP, NMCP partners, academia, research, district health representatives, and frontline health care workers. These members were selected in an open process during a malaria monitoring and evaluation Technical Working Group (TWG) meeting that took place on the 21 st -23 rd of October 2022. This was a regular NMCP TWG meeting. The criteria for selecting taskforce members included willingness to participate in the meetings that the taskforce would set, extensive knowledge of the malaria landscape, membership in the NMCP for some members, and a background in academia or research related to malaria. Representation of district health services and frontline health care workers. After suggestion of names fitting into the above criteria, the final composition of the taskforce was voted by the members of the TWG. This meeting was followed by a preliminary burden stratification task force meeting held on 4 th November 2022 to select the task force leadership and plan the workshop. Similar inclusion criteria for selection of leadership was used, followed by a final vote was used to select the leadership was used.
A four-day malaria burden stratification workshop took place from 29 th November to 2 nd December 2022 in Blantyre, Malawi.
In addition to the core task force members, other key malaria stakeholders who were partners with the NMCP and were known to be interested in the stratification work were invited to attend the meeting. Of note was a member from the Tanzania NMCP who was asked to share experiences from their recently conducted malaria stratification exercise. Additionally, a team from the Imperial College London shared some background information on modelling where they explained how models help in knowing where interventions can have an impact, in evaluating interventions, developing strategies and helping in-country planning. A member the PATH/MACEPA 15 in Seattle was with the team throughout the workshop to support the data analysis process.

Data assembly and their sources
All essential data for malaria burden stratification was collated before the workshop for initial checks of data quality parameters. Table 1 below shows the data elements gathered before the workshop and their sources. These sources were selected as the repositories where the NCMP hosts their data on each subject area, for example malaria case data, and were therefore the inherent choice of data sources for this study. Data was collected by downloading the relevant data element from each NCMP source.
Data quality HMIS data. For most of the data quality indicators, the NMCP is on track with its routinely collected data in the demographic health information systems (DHIS2). Data quality checks performed before the meeting showed how substantial improvement in quality parameters as of 2022 compared to the baseline parameters in 2016. Table 2 below shows the proportions of data quality indicators for the routine HMIS data used in the analysis. Several steps were taken to come up with the strata as shown in Figure 1 above. All processes were transparent, and there was consensus before moving to each step. First, indicators were selected, and then the data was prepared into a database for analysis. These indicators were agreed prior to the meeting. Workshop participants reviewed malaria burden and intervention coverage data to describe the current status. With this process, relevant cut-off thresholds for each indicator were agreed upon, some of which were data-driven using a quartile approach, and others, such as prevalence cut-offs, were based on literature 15 . After thoroughly reviewing the pros and cons of each prospective cut-off threshold, members reached a consensus. They decided on the appropriate administrative level for stratification and action as a district, with all the considerations including health systems factors, current administrative structures, data quality etc., and that the review of this process would be repeated on an annual basis.
To classify the districts into burden strata, the team reviewed two scenarios as shown in Table 3 below. Both scenarios used a composite score approach derived from three key variables: Scenario one prioritized districts based on incidence, prevalence, and all-cause mortality rate; whereas scenario two prioritized districts based on total cases, prevalence, and total all-cause mortality. Scenario two has been used by WHO before when determining the high burden high impact countries 17 .
Workshop members determined the final stratification list. It was agreed to have three cut-offs and four strata levels: highest, high, moderate, and low. The team decided to use district level and to understand contextual factors per facility for districts with high burden. In order to achieve this the members shared tasks amongst themselves to consult district teams to obtain contextual information.   The four strata are defined in Table 4 below.
The criteria above led to the formation of district-level strata. The district burden strata for both scenarios, one and two, were discussed by the team. The workshop members then determined the final stratification list. To combine the two scenarios, the members listed combined the lists in the two scenarios and removed any duplications. The team then collated the initial district-specific contents and contextual factors with the sub-district level data guidance.
This process was initially completed during the in-person stratification workshop using 2021 malaria data, as this was what was readily available at the necessary resolution from DHIS2. In the weeks following the workshop, the process was repeated for 2018-2020 in order to provide a more comprehensive view of past malarial trends in each district of Malawi. The analysis of historical data followed the same stratification process as was established during the in-person consultative workshop, including using the same definitions for burden strata.

Results
A review of the incidence and mortality data showed that incidence and mortality have reduced from their baseline levels of 2016. Incidence has declined from 407 cases per 1000 in 2016 to 208 per 1000 population in 2022. Mortality has decreased from 23 per 100,000 population to 8 per 100,000 population. Figure 2 below shows the malaria incidence and mortality trends between 2016 and 2022.
Prevalence estimates for national and regional levels traditionally come from malaria indicator surveys. Subnational prevalence estimates for the age group two to 10 (PfPR2_10) have

Strata level Definition
Highest burden strata At least two of the highest-burden cutoffs High burden strata Only one of the highest burden cut-offs Moderate burden strata At least two of the high-burden cut-offs and zero highest-burden strata Low burden strata < 2 of the high-burden cut-offs, at least one of the moderate burden cut off and zero of the highest-burden become available in the country, initially modelled by Chipeta et al., (2017) and further updated by Mategula et al., (manuscript in preparation) for this stratification exercise. Predicted prevalence estimates show that there is varied prevalence across Malawi, as shown in the maps below in Figure 3.

Review of scenario one and two
Maps for the districts in the four strata based on both scenarios one and two were plotted and shown below in Figure 4.
The task force members agreed and endorsed the final list after reviewing scenarios one and two and combined them using the process discussed in the methodology sections ensuring that there are no duplications and that districts do not appear in two strata simultaneously. The final strata are then presented in the Table 5 below and shown in the map in Figure 5 below.
The stratification outcome concluded with 10 districts in the highest-burden stratum, 11 in the high-burden stratum, and seven in the moderate burden strata. There no districts in the low burden strata. A majority of the highest-burden districts are in the central region of the country.
The analysis of historical data prior to 2021 yielded similar patterns to those seen in the original 2021 stratification, albeit with higher combined burden strata in the south-western part of Malawi as shown in Figure 6 below,

Discussion
Malawi conducted a malaria burden stratification exercise from 29th November to 2nd December 2022, using a composite of burden indicators compiled from 2021 DHIS2 data. The final stratification placed 10 out of the 28 districts in the highest-burden strata, 11 in the high-burden strata, and seven  in the moderate-burden strata. There were no districts that ended up in the low-burden strata. Most of the highest-burden districts were in the country's central region, mostly those on the eastern side, along the lake, or those sharing a border with Zambia. The final stratification was expected based on the experiences of the NMCP and the frontline teams.   The next steps following this process are fully understanding the context-specific issues within the districts that drive transmission. The strata and contextual issues will then inform the intervention allocation in each stratum. This process will inform the 2023 to 2030 national malaria strategic plan. Since the NMCP is developing a Global Fund grant application for malaria control activities during the same period of the NSP development, this stratification will also justify resource allocation to the different geographical strata.
The approach used in this stratification has several strengths.
Using multiple indicators to decide on the strata borrows the References strengths from several indicators to develop the appropriate strata. The process of stratifying was data-driven, transparent and reproducible. This is another key strength as it makes it easy for updates to be made to the strata. We used the most recent data to inform the decisions on the strata, which has better quality indicators than historical data. WHO is encouraging country-led processes informed by the local context in dealing with health problems in low-income countries such as Malawi. This process has and continues to demonstrate in-country leadership in decision-making, effective collaboration and partnership, including the role of academics in this process.
The task force members acknowledged several limitations with the process, including the challenges with denominator data based on predictions, and sometimes has a larger variance than that of headcount for the facilities. While the most recent data was of better quality than data from previous years, we acknowledge that making decisions on one year of data can have challenges even though it is recent. As an addendum to the original workshop stratification process, we analysed historical malaria burden data from 2018-2021 to present past malarial trends at a subnational level. These historical district strata will be analysed in combination with past intervention use to inform future updates of the national stratification process, to ensure that a true malaria burden baseline is considered when considering future intervention targeting.
The task force has identified several next steps, including updating facility-level strata based on the same process and incorporating past vector control interventions to determine a baseline. Additionally, the task force will include community health care data commonly known as health surveillance assistants (HSAs) capability to handle malaria community case management. The task force will also prioritise and allocate future interventions and note criteria for their use (i.e., what level of historical burden/intervention effort/insecticide resistance would you want to target each type of intervention going forward). Furthermore, there will be a pulling out of the population in each scenario high/medium/low burden districts to determine the costs of targeting interventions. There will also be a determination of what additional analyses and modelling exercise the NMCP would be interested in to allow partners to support their implementation.

Conclusion
For the first time, the Malawi National Malaria Control Programme has developed a malaria burden stratification to open up the use of a more targeted malaria strategy. Further next steps will allocate interventions in each stratum. We recommend maintaining the agreed method in further stratification exercises, including microstratification at the health facility level that could be done subsequently.

Data availability
The malaria case, mortality, and population data used in the analysis can be made available upon request, as it is under licence by a third party, the Malawi Ministry of Health, and requires permission to access from the DHIS2. Procedures to access the data can be found here. Paragraph 4: In the first two sentences, malaria burden stratification and malaria risk stratification are both mentioned. It would be helpful for the reader if malaria risk stratification is explained if it is considered to be a different process than malaria burden stratification.

6.
Paragraph 4: "no standard methods for malaria burden stratification" -we find this to be true as well, and a challenge for NMCPs, which is why this paper is useful to the field.

7.
Paragraph 5: The authors might consider moving the task force description and TORs to the methods section.

8.
At the end of the introduction: I would like to see the authors explain the rationale for writing up this process (e.g. Task force development, burden stratification) and sharing it with the malaria community. I assume that the rationale is at least partly because there are seemingly no standard methods for this stratification process and yet NMCPs need to better allocate fewer resources to achieve larger-scale goals, such as elimination.

Methods:
Study design: It would be interesting to know a little bit more about the TWG -what are its main goals? When was that formed? Other NMCPs may like to know about the TWG. 1.
"Representation of district health services…" -this is an unfinished sentence, please review. Also, the last sentence of this paragraph repeats "was used".

2.
1. Table 1: Was it possible to collate malaria confirmed case data disaggregated by age, gender, and possibly other indicators, such as occupation? It would be interesting to see any trends in cases 14 and up, for example, especially in males. Especially in lower burden areas approaching elimination, programs tend to see more adult males with a disproportionate share of the burden, and they are often harder to access.

1.
I'm also curious if and when entomological indicators were factored in, such as primary vectors, biting habits (e.g., indoor vs outdoor). These data may be brought in later as contextual data. We are just eager to see programs integrate their epidemiological and entomological data for intervention planning.

2.
I think more detail on the prevalence modeling data would be helpful to the reader. Also if not described elsewhere, what the schedule of prevalence surveys is currently in Malawi, whether national or subnational. 3.

2.
Figure 1: this figure is referred to as the "strata" in the paragraph below, when it seems to be a figure indicating the process.
3. Table 3: In the narrative it says that the team "reviewed two scenarios," however it is unclear how these scenarios were originally developed. Were they found in the literature or guidance documents? I see that Scenario 2 was used by WHO to determine HBHI countries, but it's unclear where Scenario 1 originated.

4.
I'm also interested to read more about how these different data sources were weighed to 5. come up with a stratification system -did the task force weigh certain data heavier than others? How did the task force integrate or triangulate across data sources? It would be interesting to know more about this process especially in relation to modeled prevalence.
Contextual information: I think here in the Methods section it would be good to have a full description of the type of contextual information that was sought from the district teams/health facilities. Per my above comment, I'm very interested to know if entomological data/indicators were brought into the analysis. I believe the only time contextual factors are mentioned again is in the Discussion, which means that contextual factors seemingly were not a part of this analysis, is that correct? Perhaps the role of contextual factors needs to be clearer in the methods section.
https://doi.org/10.21956/wellcomeopenres.21186.r61204 © 2023 Nawa M. This is an open access peer review report distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

Overview:
The article reports on an innovative approach by the Malawi Malaria Control Program (NMCP) to stratify districts by malaria disease burden based on a combination of key indicators such as malaria incidence, prevalence and all-cause under-five mortality in children aged two to 10 years old. The exercise aims to target appropriate interventions based on the malaria burden as guided by the World Health Organisation Global Technical Strategy for Malaria Elimination 2017 -2030. The paper describes the objectives and methodologies that were undertaken during this exercise. It adequately identified all stakeholders involved, and the Tanzanian National Malaria Control Program did another similar stratification exercise. Other stakeholders included PATH/MACEPA from Seattle, USA, the Imperial College London, UK and Liverpool School of Tropical Medicine, Liverpool, UK.
Using data from Malawi Malaria Indicator Surveys, District Health Information System (DHIS-2) and malaria modelling data by Chipeta et al. 2017, they managed to stratify the 28 districts in Malawi into four strata that can be used to target malaria interventions. Of the 28 districts, ten were placed in the highest-burden strata, 11 in the high, seven in the moderate and none in the low-burden strata. The emerging pattern was that the central parts of Malawi, areas near Lake Malawi and the areas bordering Zambia had the highest malaria burden, which was also reflected and corroborated by historical data.
Whilst the effort is innovative and commendable, there are some concerns that the authors need to address to strengthen the article further. In this regard, my overall verdict is 'approved with moderate corrections' because they may need to re-analyse some data and amend some tables and figures on top of some minor additions and edits. The main areas of concern are in the methods section, where the authors spent more time describing the stakeholders involved and the process of the workshop. They did not elaborate more on the data that was used; for example, under the sub-heading study design, they talked more about the participants of the workshop instead of surveys and DHIS2 data used, which were, to me, the true design: were these surveys adequately powered to detect prevalence and incidence respectively at their unit of analysis (district), how were the participants in the surveys enrolled. From how they gave space to participants, you would expect their results to also talk about the participants, like how many were women, average age etc. Still, they rightly talked about the data from MIS and DHIS2. Hence, the methods also need to highlight the methodologies of MIS and DHIS2 and their adequacy to represent the population and unit of analysis.
My second moderate concern is the use of point estimates without confidence intervals; there was a wholesale qualitative interpretation of quantitative data that, for example, an increase in completeness from 92% to 98% was a substantial increase. Statistical tests must be used to test whether these were statistically significant and show the evidence using P-Values. Please see give confidence to the interpretation.

Discussion:
The whole discussion seems to be a one-way narrative from the author's perspective. I suggest the authors compare and contrast with other existing literature to give a more scholarly context. For example, PMI and GFATM have funded a lot of interventions such as IRS, LLIN mass distribution and community health workers in sub-Saharan Africa; some of these activities have been published and can be cited as they are discussing their findings.

1.
Because of not comparing and contrasting highlighted in point number 9 in the discussion, the list of references is only 17; it may be good to increase the number of references to between 25 -30.

2.
Is the work clearly and accurately presented and does it cite the current literature? Partly

If applicable, is the statistical analysis and its interpretation appropriate? Partly
Are all the source data underlying the results available to ensure full reproducibility? Yes Are the conclusions drawn adequately supported by the results? Partly