INTRODUCTION
Malaria vector control interventions have contributed to decreases in malaria transmission since the early 2000s.1 Although global malaria case incidence decreased by 27% from 81 to 59 cases per 1,000 population at risk from 2000 to 2015, progress has stalled in recent years, with malaria incidence holding at 59 cases per 1,000 population at risk in 2020.2 Some potential reasons for this lack of progress include reduced funding for malaria control, low political will or engagement, plateaus in vector control coverage, remote areas not being reached by vector control interventions, and mosquito populations becoming increasingly resistant to the insecticides used in existing vector control tools.3–5 However, new vector control tools are available to address these challenges. Since 2017, the WHO has prequalified two new active ingredients for indoor residual spraying (IRS) products—pirimiphos-methyl and clothianidin—and three new active ingredients that are combined with pyrethroids in insecticide-treated net (ITN) products—piperonyl butoxide (PBO), chlorfenapyr, and pyriproxyfen.6 But these new tools often come at a greater financial cost.7–9
With recent changes in WHO policy to accelerate product evaluation and rapid scaling of new interventions, not all new vector control products are required to undergo cluster randomized controlled trials (CRCTs) with epidemiological outcomes to receive WHO prequalification.10 The goal of CRCTs is to assess the efficacy of the vector control product under ideal conditions, which may not be representative of routine implementation. Cluster randomized controlled trials, conducted primarily before the introduction and scale-up of a new tool, provide the highest quality of evidence of effectiveness. However, primarily because of cost, randomized controlled trials (RCTs) are often completed in only a few countries.11–14 When the deployment of new vector control tools is considered, national malaria control programs (NMCPs) need to be able to make data-driven, cost-effective decisions to make more efficient use of limited resources and maximize impact. Ministries of health (MOHs) and NMCPs are increasingly supporting subnational stratification and targeted implementation of selected vector control interventions, with data sources and strategies varying across countries.15–18 However, there is often limited evidence regarding the impact of these new tools, or combinations of tools, within specific country contexts to support product choice and targeting decisions. Quasi-experimental designs are often used to evaluate the impact of vector control tools during program scale-up by assessing any observed changes in trends and testing association with predictor variables.19,20
The increased availability and quality of routine malaria case data from health management information systems (HMIS) has greatly improved, with health facility and in some cases community- or individual-level data available within national databases, such as District Health Information Software 2 (DHIS2), a web-based open-source software that is used as the HMIS in 73 countries.21–24 There have also been recent improvements in data collection and management systems for vector control, with increased availability of entomological and vector control program data from NMCPs, research institutions, and implementing partners. Although further work is needed to strengthen vector control data systems in many countries, increasingly MOHs and NMCPs are integrating entomology and vector control data into their DHIS2 systems. This is often implemented through (1) the addition of custom-built programs or (2) the use and adaptation of entomology and vector control DHIS2 modules developed by WHO or VectorLink Collect DHIS2 modules developed by the U.S. President’s Malaria Initiative (PMI) VectorLink Project.25 These efforts have improved the availability of routine data to help address important malaria control questions.
Through well-designed, quasi-experimental evaluations using routine epidemiological, entomological, and programmatic (i.e., campaign data) data sources, NMCPs can assess the impact of vector control interventions in their countries and inform national policy, strategy, and implementation. Leveraging these routine data sources can allow NMCPs to conduct impact evaluations at a lower cost, at a more granular level, and with greater breadth across the country than through RCTs or other primary data collection methods. Although general guidance exists for using routine data to evaluate malaria interventions, there is insufficient practical and specific guidance on the design and implementation of vector control evaluations using routine data.22,26 Informed by a narrative review of published literature and project experience in conducting vector control evaluations, this article provides considerations for (1) selecting study objectives and designs, including follow-up periods; (2) defining outcome and exposure variables; (3) identifying covariates and contextual factors, their calculations, and data sources to strengthen the analysis; (4) addressing common data quality challenges; and (5) using evaluation findings to inform national policy, strategy, and program implementation. This practical guidance can increase the frequency, rigor, and utilization of routine data to support national and subnational vector control decision-making as well as help standardize analysis approaches to allow for cross-country comparisons.
MATERIALS AND METHODS
From 2016 to 2022, 22 retrospective longitudinal studies assessing the impact of vector control interventions on malaria case incidence using routine data across 14 countries in sub-Saharan Africa have been conducted or are still underway by the Next Generation IRS (NgenIRS) project, the New Nets Project, and the PMI VectorLink Project, in partnership with NMCPs and local research institutions (see Figure 1 and Table 1).27–35 In this analysis, the lessons learned and methodologies from these evaluations were collated and reviewed through discussions with eight program staff members, including details on how various epidemiological, entomological, and vector control data were used to assess the impact of interventions on malaria disease burden. To supplement information from our experience conducting evaluations, we searched PubMed for retrospective longitudinal studies assessing the impact of IRS and ITNs, focusing on evaluations that use outcome indicators from routinely collected data, including malaria cases, malaria case incidence, malaria hospitalization or death, test positivity rate, or low birth weight. Through this process, 11 vector control impact evaluations that used routine health systems data were selected, as they described key characteristics of the evaluation for each of the areas of consideration in this manuscript, including the evaluation questions, time period of analysis, evaluation design and analysis, unit of analysis, outcome, exposure, covariates included in the model, contextual variables, and data quality checks conducted (see Supplemental Table 1).
Map of countries with vector control evaluations under NgenIRS, the New Nets Project, and PMI VectorLink. NgenIRS = Next Generation IRS; PMI = U.S. President’s Malaria Initiative.
Citation: The American Journal of Tropical Medicine and Hygiene 112, 1_Suppl; 10.4269/ajtmh.22-0757
Vector control evaluations using routine HMIS data conducted or planned by NgenIRS, the New Nets Project, and PMI VL*
Country | Project | Evaluation Question(s) | Time Period | Design and Analysis | Unit of Analysis | Outcome | Exposure | Covariates | Contextual Variables | Data Quality Checks |
---|---|---|---|---|---|---|---|---|---|---|
Burkina Faso | PMI VL |
|
July 2016–Apr 2021; 25 months pre-2018 IRS, 11 months post-2018 pre-2020, and 10 months post-2020 IRS | Interrupted time series analysis using a log-linear regression model; control group was two districts that did not receive IRS | Health facility; 65 health facilities in four districts | Laboratory-confirmed malaria cases per 100 non- OPD visits | IRS status (binary) | Rainfall (lagged 2 months), EVI (lagged 1 month), regional high-transmission season | IRS spray coverage (structures sprayed/found), IRS population coverage (population protected/total population) | Data from June 2019 to May 2020 were removed owing to HMIS reporting issues; checked completeness and for outliers |
Cote d’Ivoire* | PMI VL | What is the impact of the 2020 and 2021 IRS campaigns? | Sept 2018–Apr 2022; 24 months pre-2020 IRS, 11 months between 2020 and 2021 IRS, and 8 months post-2021 IRS | Interrupted time series analysis using a log-linear regression model; control group was two districts that did not receive IRS | Health facility; 88 health facilities in four districts | Confirmed malaria cases per 1,000 population | IRS status (binary) | RDT stockouts, rainfall (lagged 2 months), temperature (lagged 2 months) EVI (lagged 2 months), high transmission season, proportion of cases reported by CHWs, non-malaria outpatient visits. | IRS spray coverage (structures sprayed/found), insecticide resistance, IRS residual efficacy | Data were collected from registries owing to District Health Information Software 2 quality concerns; data from September 2019 to August 2020 removed owing to significant RDT stockouts |
Ethiopia | PMI VL | What is the impact of the 2015–2019 IRS campaigns? | May 2015–Apr 2019; 48 months post-IRS | Dose-response approach with a negative binomial distribution and a log population offset | District; 34 districts in two regions | Confirmed malaria cases per 1,000 population | IRS population coverage (% of population protected by IRS) | Rainfall (lagged 2 months), NDVI (lagged 1 month), terms for month and spray year | Human biting rate, IRS residual efficacy, ITN distribution | Excluded district-months if IRS population coverage was low (< 50%), were missing malaria case data, or were not sprayed that year |
Ethiopia* | PMI VL | Are PBO ITNs as effective as IRS plus standard pyrethroid ITNs? | July 2019–June 2024; 24 months pre- and 36 months post-intervention | Cluster-randomized noninferiority trial | Health facility; 57 health posts in five districts | Confirmed malaria cases per 100 non-OPD visits† | Intervention status (binary) | ITN distribution, ITN usage, IRS coverage, rainfall, EVI, elevation | Human biting rate, species composition, sporozoite rate, insecticide residual efficacy, insecticide resistance | Data quality checks to determine the percentage of facilities with available data |
Liberia* | PMI VL | What is the impact of IG2 ITNs? | Aug 2019–June 2022; 24 months pre- and 22 months post-intervention | Interrupted time series analysis using a negative binomial model with no control group | Separate analyses at the health facility and district levels; 500+ facilities in 15 districts | Confirmed malaria cases per 1,000 population | ITN status (binary) | ITN distribution, rainfall, NDVI, RDT stockouts, ACT stockouts | Vector density, species composition, EIR, insecticide susceptibility | Data quality checks to determine the percentage of facilities with available data |
Madagascar | PMI VL |
|
July 2016–June 2021; 60 months, with pre/post period varying by location | Counterfactual scenarios of no IRS vs. IRS generated using multilevel, negative binomial, generalized linear mixed-effects model | Health facility; 2,524 facilities across all districts | Confirmed malaria cases per 1,000 population | IRS status—set of binary variables indicating months post-campaign (1–6 or 7–12 months) and number of consecutive IRS years; IRS coverage (sprayed out of found structures) | ITN survivorship, mass drug administration (binary), rainfall (lags varied by part of country), EVI | None | Facilities missing geocoordinates, population, and private health facilities and hospitals were excluded; health facility months missing RDT-confirmed case data were excluded from initial effect estimates; cases were imputed using the model, and these facility months were included in the final model |
Malawi* | PMI VL |
|
Jan 2018–May 2024; ∼46–53 months pre- and 24–30 months post-interventions; varies by month of implementation | Interrupted time series analysis with comparison group using a multilevel, negative binomial, generalized linear mixed-effects model | Health facility; 998 facilities in all 29 districts | Confirmed malaria cases per 1,000 population | Intervention status (IRS, Royal Guard ITN, IG2 ITN, PBO ITN, mixed ITNs) | Age (< 5, 5+) years, rainfall (lagged), EVI (lagged), elevation, IRS spray coverage, ITN population coverage, non-malaria outpatient attendance, % treated at community level | Species composition, indoor resting density, human biting rate, EIR, insecticide resistance, parity rate | Analysis restricted to facilities with 100% complete reporting; imputation methods will be explored if <60% of facilities do not meet this criterion |
Mali | PMI VL | What is the impact of the 2017–2019 IRS campaigns? | June 2016–May 2019; 12 months pre- and 24 months post-campaign; separate analysis for annual transmission and high transmission season only | Difference-in-difference using a negative binomial model; control group was those facilities that did not receive IRS | Health facility; 47 facilities (annual transmission) and 63 facilities (high-transmission season only) | Confirmed malaria cases per 1,000 population | IRS status (binary) | Rainfall (lagged 1 month), NDVI (lagged 2 months) | Human biting rate, IRS residual efficacy, insecticide susceptibility, < 5 malaria prevalence, health-seeking behavior, ITN ownership | Analysis restricted to facilities with 100% complete reporting during the study period |
Mali* | PMI VL | What is the impact of IG2 ITNs? | Aug 2018–July 2022; 24 months pre- and 24 months post-campaign | Interrupted time series analysis with comparison group using a multilevel, negative binomial, generalized linear mixed-effects model | Health facility; 240 facilities in 10 districts | Confirmed malaria cases per 1,000 population | ITN status (binary) | ITN distribution, rainfall, NDVI, SMC coverage, intermittent preventive therapy in pregnancy coverage, RDT and ACT stockouts | ITN ownership and use, vector composition, vector density, vector behavior, insecticide susceptibility | Data completeness checks will be done for each facility; imputation may be done for facilities missing 1 month of data during study period if completeness falls below 60% |
Nigeria* | PMI VL | What is the impact of PBO ITNs? | Dec 2017–Nov 2021; 24 months pre- and 24 months post-campaign | Interrupted time series analysis using a negative binomial model with control group | Ward; 216 wards in one state | Confirmed malaria cases per 1,000 population | ITN status (binary) | Rainfall (lagged), EVI (lagged), RDT stockouts, ITN use | Vector density, vector behavior, vector composition, parity rates, EIR, insecticide susceptibility | Data completeness checks done for each ward and trends by completeness threshold reported |
Nigeria* | PMI VL | What are the relative impacts of the IG2 and PBO ITNs? | Nov 2018–Oct 2023; 36 months pre- and 24 months post-campaign | Interrupted time series analysis using a negative binomial model with comparison group | Facility; 1,709 facilities in two states | Confirmed malaria cases per 1,000 population | ITN status (binary); ITN type (binary) | Rainfall (lagged), EVI (lagged), RDT stockouts, ITN use | Vector density, vector behavior, vector composition, parity rates, EIR, insecticide susceptibility | Data completeness checks done for each ward and trends by completeness threshold reported |
Sierra Leone* | PMI VL | Is IRS plus PBO ITNs more effective than PBO ITNs only? | May 2018–Apr 2023; 36 months pre- and 24 months post-campaign | Interrupted time series analysis using a negative binomial model with comparison group | Health facilities in four districts | Confirmed malaria cases per 1,000 population† | Intervention status (PBO ITN vs. IRS + PBO ITN) | Non-malaria outpatient visits, ITN coverage, ANC and EPI ITN coverage, rainfall, EVI, RDT stockouts | Entomological indicators (i.e., vector density, vector behavior, vector composition, insecticide resistance) | Data completeness checks will be done for each facility |
Ghana27 | NgenIRS |
|
Jan 2015–Dec 2017; 6-month follow-up after each round of IRS | Difference-in-difference approach used to assess implementation and withdrawal of IRS | District; 35 districts in three regions | RDT-confirmed malaria cases per 10,000 person-months | IRS status (binary) | None reported | Temperature | Data completeness was reported for each region |
Mali28 | NgenIRS |
|
Jan 2016–Mar 2018; 6-month follow-up after each round of IRS | Difference-in-difference approach used to assess both implementation and withdrawal of IRS | Health facility; 314 facilities in two regions | RDT-confirmed malaria cases (< 5 years) per 10,000 child-months | IRS status (binary) | None reported | SMC campaign data, % of suspected cases receiving diagnosis, % of confirmed cases treated with ACT, RDT stockouts | Data completeness was reported for each district |
Mali29 | NgenIRS | What is the impact of IRS and SMC, either individually or in combination, on malaria case incidence? | June 2014–Mar 2015; 6-months post-IRS and first round of SMC | Post-only comparison using a negative binomial regression model to compare intervention districts to districts receiving neither IRS nor SMC | District; six districts in one region | RDT-confirmed malaria cases per 10,000 person-months, by all ages and < 5 years | Intervention status (IRS, SMC, IRS + SMC, none) | None reported | None reported | Censored facility-months with no data from the analysis |
Mali30 | NgenIRS |
|
Jan 2012–Aug 2015; 6 months post-IRS each year | Post-only comparison between IRS and non-IRS districts | District; seven districts in one region | RDT-confirmed malaria cases per 10,000 person-months | IRS status (binary) | None reported | ITN access and use, SMC coverage, RDT stockout, entomology surveillance, rainfall | Removed years when a district received SMC to independently evaluate IRS; district-months with no data excluded |
Mozambique31,32 | NgenIRS |
|
Nov 2016–Oct 2018; 10 months post-first IRS campaign, 13 months post-second IRS, 17 months post-ITNs | Negative binomial regression model using generalized estimating equations to compare IRS and non-IRS clusters | Health facility | Confirmed malaria cases per 10,000 person-months† | IRS status (binary) | Net usage, net integrity and washes, and ecological, sociological, and economic variables | Malaria prevalence, vector density, human biting rate, sporozoite rate, insecticide susceptibility, EIR | None reported |
Zambia33 | NgenIRS |
|
Dec 2013–June 2017; 7 months after each IRS campaign over 3 years | Random effects Poisson model used to assess the relationship between IRS exposure and incidence; comparison group was facility catchment areas not receiving IRS | Facility | Confirmed malaria case incidence per 1,000 population | IRS and mSpray status (no IRS, IRS only, IRS w/ mSpray) | EVI (lagged 1 month), precipitation (lagged 2 months), province and year fixed effects, total number of tests performed | IRS coverage | Facilities were dropped if they could not be linked for geographical and environmental data; facilities only included if data were available for at least 3 of the 7 months |
Burkina Faso, Rwanda, Mozambique, Nigeria34 | New Nets |
|
Varies by country 2018–2022; 12 months pre- and 24 months post-campaign | Pre/post comparison using a difference-in-difference approach to compare impact of dual-AI ITNs to standard ITNs | Health facility; district; LGA (varies by country) | Confirmed malaria case incidence per 10,000 person-months | ITN status by type | Appropriate covariates will be identified for each country | ITN survivorship, attrition, and physical integrity; human behavior | Data quality assessments being conducted in some country to assess quality and completeness of malaria surveillance data |
ACT = artemisinin combination therapy; ANC = antenatal coverage; EIR = entomological inoculation rate; EPI = Expanded Programme on Immunization; EVI = enhanced vegetation index; HMIS = health management information system; IG2 = interceptor G2; IRS = indoor residual spraying; ITN = insecticide-treated net; LGA = local government area; NDVI = normalized difference vegetation index; NgenIRS = Next Generation IRS; OPD = outpatient department; PBO = piperonyl butoxide; PMI = U.S. President’s Malaria Initiative; RDT = rapid diagnostic test; SMC = seasonal malaria chemoprevention; VL = VectorLink.
Evaluations are ongoing, and details are subject to change.
In addition to the listed epidemiological outcomes from routine HMIS data, these studies also include entomological outcomes that are prospectively collected from sentinel sites.
RESULTS AND DISCUSSION
Selecting study objectives and designs.
Objectives of vector control evaluations.
Vector control impact evaluations typically assess the impact of vector control tools on clinical outcomes such as malaria cases or malaria case incidence.27–34,36–58 Beyond assessing the impact of vector control interventions, vector control and NMCP leadership should identify the most relevant policy questions, including any cost-benefit analyses, that can be answered with the available data. The evaluations should be done as often as needed to effectively inform national policy and decision-making. In addition to using routine malaria case data collected from health facilities to assess overall impact, these data could be used to address secondary questions such as (1) heterogeneity in impact between different regions,59 (2) withdrawal of interventions,27,30,51,55 (3) level of coverage of the intervention,38 (4) duration of impact after a single campaign,55 (5) effects of multiple campaigns,55,56,59 (6) superiority, noninferiority, or equivalency trials of new and existing tools,34,57 (7) incorporating new systems for targeting and increasing coverage,33,43,57 (8) combinations or layering of interventions,29 and (9) cost-effectiveness comparisons between interventions.9,57 For many of these questions, there is currently a lack of evidence within the existing literature to inform NMCP policy. For example, the few published studies that have examined the impact of IRS withdrawal found increases in malaria case incidence, but the extent of the increase compared with the pre-intervention period or measures to mitigate the impact of withdrawal is less well characterized.27,30,51,55 Similarly, there is limited evidence regarding the relationship between the level of IRS coverage and the reduction in malaria burden, with existing studies demonstrating that coverage as low as 50% reduced the risk of malaria,60 whereas the WHO has recommended coverage of 80% or above.61
Vector control evaluation study design considerations.
Effective use of high-quality routine data for observational analyses requires evaluators to think creatively about the natural experiments produced during intervention implementation to answer key questions as defined by policymakers. Observational studies leveraging routine data sources are typically conducted as retrospective longitudinal analyses. However, these studies can also benefit from prospective planning, particularly when considering the selection of areas that will receive each type of intervention or combination interventions and their appropriate comparison areas. Several resources already describe the various study designs available when evaluations are conducted using routine data, including before-and-after (with or without a comparison group), descriptive, interrupted time series, stepped-wedge, and dose-response studies.22,26,62 In some cases, evaluators may combine a CRCT design with routine data sources to provide additional information beyond what is available through the CRCT.31,32 The selection of the study design may vary based on the study objectives and the context in which the study is being implemented. Beyond the study design, several other factors should be considered when planning a vector control impact evaluation.
Evaluation period.
To account for seasonal changes in malaria case incidence, it is best to include at least 2 to 3 years of a baseline period before the intervention occurs. However, this is challenging in areas that have routinely received ITNs or IRS campaigns or where surveillance systems have experienced rapid improvements or changes in data systems and structures. If an appropriate baseline period is not available, use of a dose-response analysis can be considered if there is sufficient variation in intervention coverage. For example, a dose-response approach was used in an ITN evaluation conducted in Zambia in which no baseline period was available, as full-coverage ITN campaigns had been implemented in the country since 2005.38
When determining the follow-up period after the intervention, options are to use the intended period of protection, the functional period of protection, or a model that combines both. The intended period of protection can be defined based on the frequency of implementation. Insecticide-treated net mass campaigns are recommended to be repeated every 3 years.63 Indoor residual spraying campaigns are typically conducted once a year, and in some cases twice a year.55 Although using the intended period of protection represents the full period between interventions, evaluations are designed to assess the real-world effectiveness of the vector control intervention. Based on studies conducted in multiple countries, there is evidence that IRS and ITN products may last for a longer or shorter period of time than the intended period of protection. Depending on the ITN type, net use environment, and net care, one study has shown that ITN survival ranged from 1.6 to 5.3 years across four countries, with another study estimating median survival at 1.64 years across 40 countries.64,65 Similarly, IRS residual efficacy can range from 1 month to more than 12 months depending on the environment, wall type, and insecticide, with some newer active ingredients, such as clothianidin, lasting longer than 12 months.66–68 For this reason, when evaluations are conducted, it is useful to evaluate a product for the functional period of protection in addition to the intended period. Locally collected data on ITN durability and IRS residual efficacy can provide important information on the expected duration of the product and are often available through public sources, including ITN durability monitoring reports and PMI-funded entomological monitoring reports.69,70 Inclusion of both the intended and functional periods of protection in the analysis can help policymakers answer the more nuanced question of whether the intervention provides adequate protection until the next intervention is implemented or whether the intervention is effective but over a shorter time period and may need to be carried out more frequently. For example, an IRS impact evaluation conducted in Madagascar included an exposure variable to distinguish between the first 6 months after IRS and the 7 to 12 months since the most recent IRS campaign, reflecting the documented IRS residual efficacy.59 These intended and functional periods can also be used to assess the impact of discontinuing vector control activities. Two studies in Ghana and Mali compared the 6 months after IRS campaigns with the same 6 months the following year to assess the impact of IRS withdrawal.27,30 Follow-up periods can also be extended to assess longer-term impacts of vector control interventions when interventions are discontinued. In Uganda, evaluators assessed the impact of IRS for 31 months among health facility catchments after IRS withdrawal.55 The policy questions should guide the selection of the follow-up period, with evaluators clearly explaining in the resulting documentation why the follow-up period was chosen.
Unit of analysis (time, spatial).
Routine health systems data are often available at the health facility–month level within a national HMIS. In some instances, data may be available at the level of the community health worker (CHW) or reported weekly, such as in the community health program in Zambia or the Malaria Epidemic Early Detection System (MEEDS) system in Zanzibar or even at the individual level in areas where case investigations are conducted.71,72 Wherever feasible, the spatial unit of analysis should match the level at which the interventions were conducted. Many recent evaluations have used either district-month or facility-month as the unit of analysis for the study (see Supplemental Table 1). Facility-month is often preferred, as it provides a greater number of units for analysis than district-month. This can be particularly useful for IRS evaluations, when IRS is not implemented nationwide but instead in select districts. The facility-month analysis is also preferred when the exposure variable of interest (i.e., ITN or IRS coverage) is available for each health facility. The main barrier to analyzing the facility level is often the lack of reliable data on the health facility catchment population.26,27 As long as a suitable population data source is available, which is increasingly available with geospatial population estimates (see Defining outcome and exposure variables), the more granular data at the health facility–month or lower level provide a larger set of observations for data analysis and reduce the risk of bias or misclassification from incomplete data at lower levels.
Defining outcome and exposure variables.
Outcome: Calculating malaria case incidence.
Although some vector control evaluations using routine data use other indicators of malaria burden, such as malaria hospitalization,42,45,73–75 death,36,39,42,48,49,74,75 test positivity rate,50 or low birth weight,76 the majority use malaria cases36,39,40,54,58,74,75,77–80 or malaria case incidence.27–34,37,38,41,43,45–53 With the increase in access to rapid diagnostic testing (RDT), microscopy or RDT-confirmed malaria case incidence is widely available and is typically preferred over the use of clinically diagnosed or total malaria cases (see Table 2).
Outcome and exposure variables
Variable type | Indicator | Calculation | Data Source(s) | Notes/Limitations |
---|---|---|---|---|
Outcome variable | Malaria case incidence per 1,000 population | Numerator: number of confirmed malaria cases | HMIS |
|
Denominator: population at risk | Government census, local head counts, geospatial data sources matched to health facility coordinates | Geospatial sources are usually the most complete at the health facility catchment level; population data sources should be triangulated for verification. | ||
Exposure variables | Intervention status | Binary variable (yes/no) | Campaign or program documentation | Variable should be collected and calculated at the same unit of analysis as the outcome indicator (i.e., district or health facility level). |
Intervention coverage: proportion of the population at risk potentially protected by the intervention |
|
Campaign or program documentation | Coverage rates are useful in combination with binary intervention status variables to assess the impact of higher coverage rates and for dose-response analyses when there is no suitable comparison group. | |
Denominator: population at risk | Government census, local head counts, geospatial data sources matched to health facility coordinates | Population data should be from the same source used to calculate the outcome variable. |
HMIS = health management information system; IRS = indoor residual spraying; ITN = insecticide-treated net; RDT = rapid diagnostic testing.
Malaria cases.
Confirmed malaria case incidence, normally defined as cases per 1,000 population, is the most common outcome indicator used in vector control evaluations that use routine data. There are often subjective decisions to be made about which variables within a national HMIS should be used to determine this indicator. In some cases, there is a confirmed malaria cases data element, as well as data elements for the number of positive RDT results and microscopy tests. Data may also be stratified by age group, and evaluators must determine whether to include individual age groups or combine indicators into an all-ages analysis. A national HMIS data dictionary or other tools that support data systems exploration can be useful to understand the data available.72,81 Evaluators who are not NMCP staff members or highly familiar with the data systems should identify a contact person who is involved with the ongoing management of the system who can support the interpretation of key indicators, particularly considering any changes in variable definitions or data collection processes over time.
Many HMIS systems store zero values as missing data as a space-saving measure for national servers. This practice makes it difficult to distinguish between true zero values and missing data for malaria cases and other indicators. In high burden areas, null values are more likely to be missing data. However, in low burden areas, zeros are more likely to be true values and could constitute a larger portion of the dataset. When data completeness is low (see Table 3), sensitivity analyses could be done, treating the values as missing and then as true zeros to test the robustness of the findings.
Covariates and contextual factors
Indicator | Definition | Data Source | Purpose |
---|---|---|---|
HMIS* | |||
Testing rates among febrile patients | % of febrile patients who tested for malaria (number of patients tested/Number of febrile patients) | HMIS | To account for fluctuations in parasitological diagnosis |
Non-malaria outpatient attendance | Total outpatient attendance minus confirmed malaria cases | HMIS | To control for variations in treatment-seeking behavior |
IPTp | % of women receiving full course of IPTp per national guidelines | HMIS | To control for other malaria prevention measures |
Continuous ITN distribution | % of pregnant women and infants receiving ITNs during health facility visits | HMIS | To control for other malaria prevention measures |
RDT stockout | Varies by country but is often (1) number of days with an RDT stockout, (2) RDT stockout that lasted for > 7 days (yes/no), or (3) any RDT stockout (yes/no) | HMIS or LMIS | To control for periods where RDT stockouts can impact the number of confirmed cases |
Data completeness | % of months during the study period with data available | HMIS | To determine how missing or inconsistent data may skew the results |
Climate and ecology† | |||
Precipitation/rainfall | Monthly averages calculated in millimeters; values should be lagged 1–3 months | Climate Hazards Group InfraRed Precipitation with Station Data; local weather stations | Climate variables are often positively correlated with malaria cases and should be considered for inclusion as a covariate in the evaluation model. |
Vegetation | Index values calculated as monthly averages; values should be lagged 1–3 months | Famine Early Warning Systems Network; MODIS Vegetation Index Products (NDVI and EVI) | |
Temperature | Monthly averages calculated as °C; values should be lagged 1–3 months | MODIS Land Surface Temperature and Emissivity; local weather stations | |
Elevation | Static value calculated in meters | U.S. Geological Survey’s Terra Advanced Spaceborne Thermal Emission and Reflection Radiometer; Shutter Radar Topography Mission | |
Routine survey data‡ | |||
Treatment-seeking behavior81 | Proportion of children under 5 years old with fever in the last 2 weeks for whom advice or treatment was sought | DHS, MIS | To adjust for differences in treatment-seeking behavior by region |
Private sector treatment3 | Proportion of children under 5 years old with fever in the last 2 weeks for whom advice or treatment was sought from the private health sector | DHS, MIS | To adjust for private treatment seeking, which is often not accounted for within HMIS data |
Population ITN access82 | Total number of individuals who could sleep under an ITN if each ITN in the household is used by two people/Total number of individuals who spent the previous night in surveyed households | DHS, MIS; ITN durability monitoring surveys; malaria behavioral surveys | To adjust for population access to ITNs per national surveys; can be used in collaboration with or instead of ITN program distribution data |
ITN use81 | Number of individuals who slept under an ITN the previous night/Total number of individuals who spent the previous night in surveyed households | DHS, MIS; ITN durability monitoring surveys; malaria behavioral surveys | To explain why ITNs may or may not have been effective by determining how often they are used and where |
ITN survivorship83 | Proportion of distributed nets still available for use as intended in the households to which they were given after a defined period | ITN durability monitoring surveys | To explain why ITNs may or may not have been effective by determining whether ITNs distributed during the previous campaign are still available |
Entomological surveillance data§ | |||
Human biting rate84 | Number of adult female vectors collected divided by the total number of collector nights (disaggregated by indoor/outdoor collection) | Entomological surveillance | To assess the impact of vector control interventions |
Vector density84 | Total number of vectors collected resting inside per house per day | To assess the impact of vector control interventions | |
Indoor resting density84 | Total number of vectors collected resting inside per house per day | To assess the impact of vector control interventions | |
Sporozoite rate84 | Number of adult female vectors positive for CSP divided by the total number of adult female vectors tested for CSP | To assess the impact of vector control interventions | |
Entomological inoculation rate84 | Number of bites per person per unit time multiplied by sporozoite rate | To assess the impact of vector control interventions | |
IRS residual efficacy |
|
To explain the duration of IRS efficacy | |
Resistance frequency84 |
|
To explain why interventions may or may not have been effective by determining whether local mosquito populations are susceptible to the insecticide used | |
Resistance status84 | Classification of adult female vector populations as confirmed resistant (< 90%), possibly resistant (90–97%), or susceptible (≥ 98%) | To explain why interventions may or may not have been effective by determining whether local mosquito populations are susceptible to the insecticide used | |
Resistance intensity84 | Classification based on proportion of mosquitoes dead or incapacitated after exposure to 5× and 10× intensity concentrations of an insecticide in a standard bioassay, whereby < 98% after 10× exposure = high-intensity resistance; ≥ 98% after 10× exposure but < 98% after 5× exposure = moderate intensity resistance; ≥ 98% after 10× and 5× exposure but < 98% after 1× exposure = low-intensity resistance | To explain why interventions may or may not have been effective by determining whether local mosquito populations are susceptible to the insecticide used | |
Biting time84 | Number of adult female vectors that attempt to feed or are freshly blood-fed, per person per unit time, usually expressed per 2-hour increments | To explain why ITNs may or may not have been effective by determining whether individuals are likely to be asleep and under ITNs; newer studies compare biting time with human behavior studies to determine when individuals are indoors and asleep | |
Biting location84 | Proportion of attempted bites or successful blood-feeds by adult female vectors indoors and outdoors, per unit time | To explain why interventions may or may not have been effective by determining how often mosquitoes bite indoors, where IRS and ITNs are effective | |
Population ITN access82 | Total number of individuals who could sleep under an ITN if each ITN in the household is used by two people/Total number of individuals who spent the previous night in surveyed households | To adjust for population access to ITNs per national surveys; can be used in collaboration with or instead of ITN program distribution data | |
ITN use81 | Number of individuals who slept under an ITN the previous night/Total number of individuals who spent the previous night in surveyed households | To explain why ITNs may or may not have been effective by determining how often they are used and where |
CSP = circumsporozoite protein; DHS = demographic and health survey; EVI = enhanced vegetation index; HMIS = health management information system; IRS = indoor residual spraying; ITN = insecticide-treated net; IPTp = intermittent preventive therapy in pregnancy; LMIS = logistics management information system; MIS = Malaria Indicator Survey; NDVI = normalized difference vegetation index; RDT = rapid diagnostic testing.
A designated point of contact can help to determine the correct indicator and source for RDT stockout data. Data completeness should be used as a covariate in the model when there is a wide variation across the unit of analysis.
Data should be collected at the same unit of analysis as the outcome and exposure variables.
Data on treatment-seeking behavior are often available only at the regional or provincial level and will often be used contextually. An identified point of contact can help to determine if data from the private sector are included in the HMIS or not.
All of the entomological indicators below can be considered for both ITN and IRS evaluations, with the exception of residual efficacy, which is for IRS only.
Another common challenge in calculating malaria cases for an evaluation occurs when CHWs contribute case data to the HMIS. Community-based malaria case management typically expands gradually over time, and with that expansion, malaria case counts often increase. If CHW data are stored within the HMIS and disaggregated from health facility reports, the percentage of cases seen by CHWs could be used as a covariate to control for the potential increase in diagnoses where CHWs are more active.73,85 In countries or subnational areas where CHWs are active but case data are not stored or disaggregated, evaluators may have difficulty controlling for the effects of CHW activity on case counts, which could lead to bias in the evaluation results.
Population.
Many HMIS systems include population data through extrapolations from census data, local head counts, or both. Census data are the official data source for population statistics in most countries and are usually available for use by national programs through national statistics offices. However, longer time periods between the original census data collection and the current evaluation year increase the potential risk for data inaccuracy, including geographical boundaries that no longer match current administrative structures and population estimates that are often lower than those observed through other sources. For example, vector control interventions using census projections as the areal denominator to estimate the proportion of the population protected often find coverage rates over 100%, as do other mass campaigns such as seasonal malaria chemoprevention and immunization programs.86
For local head counts, the MOH or NMCP often asks local governments to provide a count of the populations in their geographic area either on an annual basis or when countries are preparing for mass campaigns, such as ITN or IRS campaigns. Local governments may use multiple methods to obtain these estimates, including receiving counts from village or area chiefs, using results from a recent enumeration for another intervention (e.g., a recent public health intervention or research project that required household registration), or conducting a local census or ad hoc registration process for the campaigns (which may include primary data collection, a projection based upon national census, or other estimate methods). The methods used to produce these estimates are not often documented, may not cover all the evaluation areas, and may vary by local governments, making it difficult to compare estimates from one area with those from another. This method of estimation is also vulnerable to over- and underestimation. For example, in a study of mass drug administration in an urban area of Liberia, a Médecins Sans Frontières census was used in collaboration with estimates from community leaders to initially project a population of 300,000. The population was later verified as 551,971 by distribution teams during the first round and corrected to 558,483 in the second round—a 186% increase from the initial projection.87 Local head counts are usually a trusted source of information, particularly when the census is outdated and may not reflect recent population movements or changes.
Many HMIS systems have census extrapolations or local head counts available at the district level. Although health facility catchment populations within an HMIS are increasingly available, they are still not available or complete in many countries. Where they are available, these sources may be preferred by NMCPs as the official population sources. However, the process for deriving facility-level head count estimates may differ between localities and is often not well documented. In addition, if facility-level head counts are not systematically and regularly updated, it may be difficult or impossible to determine how the opening or closing of facilities may impact population estimates of neighboring facilities.
Where health facility catchment populations are not available or complete within an HMIS, populations can be estimated using a variety of well-documented geospatial techniques.26,86,88,89 These techniques use publicly available gridded population estimates to assign a population to a health facility catchment based on its geospatial coordinates. Maintenance of complete facility geo-registries is necessary for geospatial-based methods. However, unlike head counts, facility locations are static and do not depend on variable estimation protocols. Geospatially derived population estimates can be readily updated and designed to consider only active facilities. Triangulation of population data sources and review by national stakeholders can help to determine the best population option to use and can build confidence in the evaluation results. Where civil registration and vital statistics systems are actively maintained, they can also be used to support the calculation and revision of population estimates. Although these systems are currently weak in many malaria-endemic countries, this could be a useful source in future evaluations as these systems improve.90
Exposure: Calculating ITN and IRS coverage.
Common methods of calculating ITN and IRS coverage for evaluations include binary variables that indicate whether an area received an intervention, a continuous variable that measures the proportion of the population that is expected to be protected by the intervention, or a more complex model that attempts to capture both initial population coverage and the estimated effective duration of the intervention over time (see Table 2).
Binary intervention implementation variables are the simplest measures to collect and report, and they are less subject to biases in coverage estimations. As noted under the unit of analysis, ideally, the data on whether a given area received an intervention will match the lowest level of data available for the outcome variable. For example, if the intervention was implemented and data are available at the village level, then it would be ideal to have village-level case data available. Similarly, if health facility catchment is the lowest level of case data, then it is preferable to have data on whether each health facility received the intervention. A main challenge in using the health facility level or lower as the unit of analysis is the ability to match the spatial units between the intervention and case data. As intervention data are often stored outside of an HMIS, the spatial unit names or hierarchies may not be the same as those used in the HMIS. If vector control implementation data are not available at the same level as the outcome variable (i.e., only at the district level and not at the heath facility level) or coverage of the intervention is not uniform, with areas receiving higher or lower coverage of the intervention levels, the impact of the intervention may be misspecified.
Population coverage can be defined as the proportion of the population at risk that is protected by the intervention.86 A commonly used indicator for assessing IRS coverage is the proportion of households or structures sprayed out of the total targeted or found.86 Although such indicators are useful for program operations and monitoring, they cannot easily be extrapolated to the population for use in impact evaluations.86 Ideally, IRS population coverage would be calculated as the number of individuals residing in sprayed households divided by the estimated population at risk or the population in the spatial area of interest.77,91 However, this is feasible only if the population protected is recorded during IRS interventions and population estimates are available. The ITN population distribution coverage can be calculated as the number of ITNs distributed multiplied by 2 and divided by the estimated population, based on the WHO recommendation that one ITN should be distributed for every two people (i.e., universal coverage).64 It may also be multiplied by 1.8 to account for households with an uneven number of members.64,66 The accuracy of both IRS and ITN coverage rates relies on the quality of the program intervention data collected and, as with malaria case incidence, the selection of the population data source. Population coverage estimates are useful for dose-response analyses when there is not a suitable comparison area and there is adequate coverage variation to detect a difference.38 Similar to binary estimates, coverage that is available at the unit of analysis (i.e., health facility or district) is preferred. In many cases, evaluators will use the binary variable as the exposure variable within the statistical model and use continuous coverage variables to provide additional context for interpretation.30,33,43,47,52,74
Newer statistical models of estimating coverage are incorporating multiple factors to better estimate the duration and use of the interventions. One study in Madagascar, which assessed the impact of IRS on malaria case incidence, incorporated the IRS status at the health facility level (binary), time since the IRS intervention (0–6 months versus 7–12 months based on insecticide residual efficacy data), number of past consecutive years of IRS, and number of structures sprayed/number of structures found.59 The study also controlled for the impact of ITNs using ITN survivorship from an ITN durability monitoring study to project ITN survivorship across the study areas.
In addition, modelers have produced global maps of ITN and IRS coverage for sub-Saharan Africa by blending multiple datasets.65,91 These studies integrated program data, demographic health survey data, published articles, WHO World Malaria Reports, and data from ITN manufacturers to produce granular datasets to calculate estimates of intervention coverage from the national level to the 5 by 5–km pixel.
Finally, some areas that receive IRS and ITN interventions will be closer to the border of the intervention than others. Health facilities or CHWs in border areas are more likely to include patients from areas that did not receive the vector control intervention being studied. In these situations, sensitivity analyses can be conducted including and excluding prespecified border facilities to test the robustness of the findings. For example, in an ongoing analysis in Sierra Leone, sensitivity analyses will also be conducted excluding health facilities where <80% of the catchment population, as determined through geospatial modeling, is located within the intervention or comparison district. If individual-level data are collected, with the village of origin for each resident recorded and matched to the intervention areas, the primary analysis can be restricted to only those patients who originate in areas that received the intervention. However, this level of detail is often not available without collecting data directly from health facility registers that contain village level information.32
Identifying covariates and contextual factors.
It is important to be able to account for potential biases in the exposure, outcome, or impact estimates. Ashton et al.26 provide a detailed list of such factors to consider, including access to health care and treatment seeking for malaria, malaria diagnostic practices, and climate data. For vector control evaluations, there are a variety of routine data sources and indicators that can be useful in the analysis and interpretation of results (see Table 3). These factors can be broadly divided into two groups, covariates that are included directly within the statistical analysis and contextual factors, which are descriptive summaries of key data used to describe the evaluation context and support the explanation of key findings. Whether to include a variable as a covariate or contextual factor can be determined by the availability of the data (are they available at the unit of analysis?) and generalizability (if they are not available at the unit of analysis, can they feasibly be extrapolated and applied?). Given the many covariates and contextual factors that should be considered in vector control evaluations, the triangulation and interpretation of these factors can be complex. However, it is important to review these data sources and consider, in collaboration with key stakeholders, their potential impact on the intended outcome.
Other health service delivery data.
Beyond malaria cases and population, health service delivery data often contain additional data elements that are useful either as covariates in the model or as contextual variables. Total outpatient visits can be used to calculate non-malaria outpatient visits to adjust for differences in treatment-seeking between health facilities and over time. Malaria testing rates (proportion of malaria suspects receiving a diagnostic test) or RDT stock data, available within the HMIS or a separate logistics management information system, should also be incorporated to adjust for differences in access to diagnostic testing, which could potentially bias the number of confirmed malaria cases. The HMIS also often contains data on other malaria prevention interventions, including intermittent preventive therapy in pregnancy and continuous distribution of ITNs through antenatal care and the expanded program on immunization.
Entomological surveillance data.
Entomological surveillance data, such as that routinely collected by national programs or by research or implementing partners, can be useful to explain trends in malaria case incidence.78 Where data are available for the entire evaluation period and across both intervention and comparison areas (if any), vector density, human biting rate, indoor resting density (for IRS), sporozoite rate, and entomological inoculation rate (EIR) can be used to assess the impact of vector control interventions on vector populations.92 Although the sporozoite rate and EIR are malaria transmission measures (before post-inoculation individual factors such as immunity, age, nutrition, and genetics and systems factors such as access to treatment), they require more complex analyses, can be difficult to measure in low-transmission settings, and have limited generalizability beyond the sites in which they were collected.82,92
Other entomological indicators are as follows: Resistance frequency and intensity can be useful in determining whether an insecticide is still effective against mosquitoes in the areas of interest. Biting time and location can be useful in determining where and when humans are likely to be protected by ITN and IRS interventions, namely indoors and at night, as well as potential gaps in coverage.83,84 Entomological surveys may also contain information on ITN availability and use among the few households visited for entomological collections.83 This varies from collecting the type and number of ITNs and the proportion of household members who slept under an ITN to collecting hourly information on human behavior to quantify when individuals are indoors and under ITNs.34,83 Although surveys of these factors may not necessarily capture the full spectrum and extent of human behaviors in relation to vector control interventions owing to their limited sampling, they may still be useful in partially explaining HMIS data trends.
Routine surveys.
Nationally representative surveys or vector control monitoring and evaluation surveys can be another useful source of contextual information. National household surveys, such as demographic and health surveys, are often ill-suited for evaluations of specific campaigns, as they are usually collected every 3 to 5 years during time periods that may not align with the evaluation and data are often available only at the provincial or regional level.86 As IRS is often implemented at the subregional level, the most recent guidance for household surveys no longer recommends including questions related to IRS.93 However, these surveys can be useful in calculating coverage and use indicators for ITNs and understanding care-seeking behavior, and they can provide useful contextual information on how these factors may differ by region.
Insecticide-treated net durability monitoring surveys are typically conducted on an annual basis up to 3 years after an ITN campaign, with the primary purpose of evaluating the bioefficacy and physical integrity of ITNs.64,69 If conducted within designated areas for vector control evaluation, these small-scale surveys can also be a useful source of information on ITN survivorship, attrition, and use. Sampling for ITN durability monitoring is based on a cohort of ITNs and is not representative of the human population. Malaria behavioral surveys aim to understand drivers of behavior and provide key information on ITN use, care, and repair; care seeking for malaria testing and treatment; and IRS acceptance.94 These surveys collect data on human behavior but are restricted to select areas within a given country. Despite the subnational sampling methods, ITN durability monitoring and malaria behavioral surveys can still be useful in supporting the interpretation of key findings.
Climate and ecology.
Key climate and ecological variables include precipitation (rainfall), vegetation indices, temperature, and elevation. Climate variables are available from modeled geospatial data, which are publicly available for use from sources such as the Climate Hazards Group InfraRed Precipitation with Station Data for rainfall88; enhanced vegetation index for vegetation95; land surface temperature and emissivity for temperature96; and the Global Multi-resolution Terrain Elevation Data for elevation.97 These variables can also be accessed from national meteorological departments and local weather station data.51 Climate and ecology variables should be calculated at the same unit of analysis as the outcome and exposure variables, especially when climate and ecology variables are used as covariates in the model. Similar to geospatial population data, health facility coordinates are used to match climate and ecology data to health facility catchment areas. Climate and ecology variables are calculated as total,39 mean values,58,74 or anomalies, or they use a mix of approaches.33,37 It is also considered best practice to incorporate climate variables as lagged values, given that changes in climate may not impact changes in malaria cases until 1 to 3 months later.26
Other programmatic data.
In the analysis, NMCPs and evaluators should include other programs or interventions relevant to the evaluation areas. This could include other malaria interventions such as seasonal malaria chemoprevention29 or mass drug administration, mass screening, or other health programs. Data for these can often be obtained from the relevant national health programs. For other events that may impact the intervention or data sources—such as health worker strikes, political instability and insecurity, or population migration—detailed data may not be available. Public news sources can often provide broad information and context for the analysis.
Potential data quality challenges and how to address them.
Although the availability and quality of routine health systems data continue to improve, data quality is still a noted limitation that could lead to biased results for vector control evaluations. When conducting evaluations using routine data, evaluators should characterize the quality of the data.
Increasingly, NMCPs and implementing partners are conducting routine data quality assessments in which timeliness, completeness, accuracy, consistency, and validity are assessed through review and verification of data within health facilities and the HMIS. Where recent data quality assessments or audits are available, these results can provide important contextual information to inform the interpretation and use of findings. Before or as part of an evaluation, evaluators may also conduct data quality assessments within evaluation areas. Where tools are not already available from national sources, global tools can be adapted to support data quality assessment efforts.99,100 In some cases where initial data quality assessments reveal gaps in data from HMIS sources, evaluators may recommend collecting routine data directly from health facility registers. Finally, when on-the-ground data quality assessments are not possible, remote data quality checks in which HMIS data are reviewed for key data quality indicators can still be useful in assessing data quality before an evaluation is conducted. Data quality checks on all outcome, exposure, covariate, and contextual variables should be incorporated into any evaluation and steps taken to mitigate their impact. Plotting raw data can help to identify erroneous values and outliers, and clear rules can be documented in a data cleaning protocol to flag data points that fall outside of the pre-identified criteria, which can be reviewed for potential exclusion. Data completeness can also be assessed by calculating the proportion of observations with missing data. Health facilities or other observational units with missing data can be removed from the analysis if they do not meet a data completeness threshold.33,52,73 For example, an evaluation of standard and PBO ITNs on pregnancy outcomes in Uganda removed facilities that did not have data for at least 25 months of the 38-month evaluation period or when covariates were systematically missing.76 Alternatively, imputation methods can be applied to the missing data.43,51 For example, an evaluation of scaling up anti-malaria interventions in Ghana found that data for 10% of the months were missing. The researchers then imputed data for missing months by taking the average of the same month from the other years of the evaluation period.74 Finally, data from all health facilities can be included in the evaluation, and reporting completeness rates can be integrated as a covariate in the evaluation.37
Using evaluation findings to inform national policy, strategy, and program implementation.
Evaluations using routine data are intended to inform national malaria policies, strategies, and implementation. Without this crucial step, the evaluation becomes an academic exercise, with little practical use. There are several steps that evaluation teams can take to increase the potential for evaluation findings that can inform national decision-making. First, evaluators should define and engage relevant stakeholders and decision-makers early in the evaluation process. As noted under study objectives above, vector control and NMCP leadership should identify the specific policy decisions that the evaluation will inform. Evaluators can facilitate a discussion of evaluation use by asking stakeholders three questions: (1) How do you anticipate the results of the evaluation will be used? (2) Who is interested in the outcome(s) of this evaluation? (3) What do you want to do differently as a result of engaging with the evaluation findings? In some cases, it may also be helpful to facilitate a thought experiment with decision-makers exploring potential outcomes of the evaluation and how they might use the results in each case. For example, when the impact of standard pyrethroid and new types of ITNs is being compared, what decision might the NMCP make if the results show that the new ITNs performed better or if the evaluation results show no difference in impact?
Evaluators should also consult decision-makers on what type of evaluation product—a report, presentation, evidence briefs, peer reviewed publication, or other product—they would prefer to inform their decision and how and where these products should be presented. Previous research has demonstrated that short, solution-oriented take-home messages, such as evidence briefs, tend to be more effective than research reports.98 Finally, evaluators should map out key decision points, such as malaria strategic planning processes, midterm strategy reviews, and donor cycles such as Global Fund applications or PMI malaria operational plans. Evaluators should set expectations with decision-makers regarding when interim or final results will be available to inform these key strategy points and determine what specific content decision-makers would like to include.
During the evaluation process, evaluators can engage key stakeholders and decision-makers in reviewing initial descriptive findings and any interim results. This early exploration can help to refine variable definitions and suggest practical solutions to data quality challenges, ensuring that the final product meets NMCP needs. Finally, when the evaluation is complete, the results should be presented in a way that is easily interpretable and usable by program staff. For example, rather than presenting risk ratios, results can be presented in terms of cases averted per population or a percent change in malaria case incidence, comparing intervention and comparison areas where feasible. It may also be helpful to describe whether the intervention resulted in changes in stratification level (i.e., from high to medium burden) according to national or WHO stratification guidelines.
CONCLUSION
Although CRCTs provide the highest quality of evidence for impact, they are costly. Well-designed observational, quasi-experimental evaluations using routine epidemiological, entomological, and programmatic data sources can allow NMCPs to gain important insights into the impact of vector control interventions at a lower cost, at a more granular level, and with greater breadth across the country than commonly available through RCTs or other primary data collection methods.
This article provides methodological recommendations to support MOHs and NMCPs in conducting strong vector control evaluations using routine data that lead to evidence-based policy development and program implementation. Evaluations should aim to address the most relevant policy questions, going beyond the impact of vector control on malaria case burden to address questions such as optimal timing, duration, dosage, combinations of interventions, and the intended and functional periods of protection.
Key recommendations for designing vector control evaluations and selecting variables include (1) using the health facility catchment population month or lower as the unit of analysis, when that level of granular data is available, (2) using binary measures of intervention coverage and variables that incorporate the duration and use of the intervention as exposure variables, and (3) incorporating both covariates and contextual factors, including entomological data, to describe the evaluation context and support the explanation of key findings.
Key recommendations for the process of implementing vector control evaluations include (1) ensuring an in-depth understanding of the data systems being used, including changes in systems and data or variable structure over time, (2) assessing data quality through either facility-based data quality audits or remote data quality checks and taking steps to reduce bias in the data, such as removal of incomplete data for health facilities or imputation, and (3) triangulating population data sources to determine the best population option to use and build stakeholder confidence in the evaluation results.
Finally, key recommendations for ensuring that evaluation results are used include engaging stakeholders to define specific (1) evaluation objectives and use cases for evaluation findings, (2) evaluation products that would be useful for decision-makers, and (3) key decision points and their timing to ensure that evaluation results are incorporated into the national malaria policy, strategy, and program implementation.
Continued strengthening of malaria surveillance and vector control data systems, data integration, including publicly available sources of climate and population data, and the increasing focus on data-driven vector control decision-making will strengthen the design and implementation of vector control evaluations and ensure their use in guiding NMCP policy and strategy. The increased availability of routine data also reduces the cost of conducting these observational impact evaluations, making them more within reach for NMCPs seeking an evidence-based policy to support national vector control decisions. Future evaluations should continue to build upon this guidance, further identifying methods and considerations that lead to strong vector control evaluations and data use.
Supplemental Materials
ACKNOWLEDGMENTS
We acknowledge the support of the National Malaria Program staff involved in the NgenIRS, New Nets Project, and PMI VectorLink evaluations, including Baltazar Candrinho (Mozambique), Mutinta Mudenda Chilufya (Zambia), Austin Gumbo (Malawi), Busiku Hamainza (Zambia), Michael Kayange (Malawi), Keziah L. Malm (Ghana), Vincent Sanogo (Mali), and Chrispin Williams (Liberia). We also acknowledge the support of the staff at the U.S. President’s Malaria Initiative who provided critical review and feedback in the development of this manuscript, including Daniel Impoinvil, Allison Belemvire, Jenny Carlson, Lilia Gerberg, Kristen George, John Gimnig, Audrey Lenhart, Rebecca Levine, Paula Marcet, Melissa Yoshimizu, and Sarah Zohdy.
REFERENCES
- 1.↑
Bhatt S et al., 2015. The effect of malaria control on Plasmodium falciparum in Africa between 2000 and 2015. Nature 526: 207–211.
- 2.↑
World Health Organization, 2021. World Malaria Report 2021. Available at: https://apps.who.int/iris/handle/10665/350147. Accessed January 19, 2022.
- 3.↑
Churcher TS, Lissenden N, Griffin JT, Worrall E, Ranson H, 2016. The impact of pyrethroid resistance on the efficacy and effectiveness of bednets for malaria control in Africa. eLife 5: e16090.
- 4.↑
Ranson H, 2017. Current and future prospects for preventing malaria transmission via the use of insecticides. Cold Spring Harb Perspect Med 7: a026823.
- 5.↑
van den Berg H, da Silva Bezerra HS, Al-Eryani S, Chanda E, Nagpal BN, Knox TB, Velayudhan R, Yadav RS, 2021. Recent trends in global insecticide use for disease vector control and potential implications for resistance management. Sci Rep 11: 23867.
- 6.↑
World Health Organization, 2022. Prequalified Vector Control Products. Available at: https://extranet.who.int/pqweb/vector-control-products/prequalified-product-list. Accessed February 28, 2022.
- 7.↑
Oxborough RM, 2016. Trends in US President’s Malaria Initiative-funded indoor residual spray coverage and insecticide choice in sub-Saharan Africa (2008–2015): urgent need for affordable, long-lasting insecticides. Malar J 15: 146.
- 8.↑
Mnzava AP, Knox TB, Temu EA, Trett A, Fornadel C, Hemingway J, Renshaw M, 2015. Implementation of the global plan for insecticide resistance management in malaria vectors: progress, challenges, and the way forward. Malar J 14: 173.
- 9.↑
Shepard DS, Odumah JU, Awolola ST, 2021. Cost-effectiveness of PBO versus conventional long-lasting insecticidal bed nets in preventing symptomatic malaria in Nigeria: results of a pragmatic randomized trial. Am J Trop Med Hyg 104: 979–986.
- 10.↑
World Health Organization, 2017. The Evaluation Process for Vector Control Products. Available at: http://apps.who.int/iris/bitstream/handle/10665/255644/WHO-HTM-GMP-2017.13-eng.pdf. Accessed March 3, 2022.
- 11.↑
Gleave K, Lissenden N, Chaplin M, Choi L, Ranson H, 2021. Piperonyl butoxide (PBO) combined with pyrethroids in insecticide‐treated nets to prevent malaria in Africa. Cochrane Database Syst Rev 5: CD012776.
- 12.↑
Accrombessi M et al., 2021. Assessing the efficacy of two dual-active ingredients long-lasting insecticidal nets for the control of malaria transmitted by pyrethroid-resistant vectors in Benin: study protocol for a three-arm, single-blinded, parallel, cluster-randomized controlled trial. BMC Infect Dis 21: 194.
- 13.↑
Mosha JF et al., 2021. Protocol for a four parallel-arm, single-blind, cluster-randomised trial to assess the effectiveness of three types of dual active ingredient treated nets compared to pyrethroid-only long-lasting insecticidal nets to prevent malaria transmitted by pyrethroid insecticide-resistant vector mosquitoes in Tanzania. BMJ Open 11: e046664.
- 14.↑
Choi L, Pryce J, Garner P, 2022. Indoor residual spraying for preventing malaria in communities using insecticide-treated nets. Cochrane Database Syst Rev 1: CD012688.
- 15.↑
Thawer SG et al., 2020. Sub-national stratification of malaria risk in mainland Tanzania: a simplified assembly of survey and routine data. Malar J 19: 177.
- 16.↑
Cox J, Sovannaroth S, Dy Soley L, Ngor P, Mellor S, Roca-Feltrer A, 2014. Novel approaches to risk stratification to support malaria elimination: an example from Cambodia. Malar J 13: 371.
- 17.↑
Kifle MM, Teklemariam TT, Teweldeberhan AM, Tesfamariam EH, Andegiorgish AK, Azaria Kidane E, 2019. Malaria risk stratification and modeling the effect of rainfall on malaria incidence in Eritrea. J Environ Public Health 2019: 7314129.
- 18.↑
Alegana VA et al., 2021. Plasmodium falciparum parasite prevalence in East Africa: updating data for malaria stratification. PLOS Glob Public Health 1: e0000014.
- 19.↑
World Health Organization, 2017. Design of Epidemiological Trials for Vector Control Products: Report of a WHO Expert Advisory Group, Château de Penthes, Geneva, 24–25 April 2017. Available at: https://apps.who.int/iris/handle/10665/255854. Accessed April 27, 2022.
- 20.↑
World Health Organization, 2017. How to Design Vector Control Efficacy Trials: Guidance on Phase III Vector Control Field Trial Design Provided by the Vector Control Advisory Group. Available at: https://apps.who.int/iris/handle/10665/259688. Accessed April 27, 2022.
- 21.↑
Kigozi SP et al., 2020. Practical implications of a relationship between health management information system and community cohort–based malaria incidence rates. Am J Trop Med Hyg 103: 404–414.
- 22.↑
Ashton RA, Prosnitz D, Andrada A, Herrera S, Yé Y, 2020. Evaluating malaria programmes in moderate- and low-transmission settings: practical ways to generate robust evidence. Malar J 19: 75.
- 23.↑
PATH, 2021. PMI Digital Community Health Initiative Cross-Country Landscape Report: Understanding the Use of Digital Technologies in Community Health Programs. Available at: https://static1.squarespace.com/static/59bc3457ccc5c5890fe7cacd/t/620be09ed784a362a9e421ce/1644945569802/PMI_DCHI_Global_Report.pdf. Accessed February 28, 2022.
- 24.↑
University of Oslo, 2022. About DHIS2. Available at: https://dhis2.org/about. Accessed April 27, 2022.
- 25.↑
World Health Organization, DHIS2-Based Entomology and Vector Control Data Collection and Collation Tools. Available at: https://www.who.int/teams/global-malaria-programme/prevention/vector-control/dhis-data-collection-and-collation-tools. Accessed February 28, 2022.
- 26.↑
Ashton RA, Bennett A, Yukich J, Bhattarai A, Keating J, Eisele TP, 2017. Methodological considerations for use of routine health information system data to evaluate malaria program impact in an era of declining malaria transmission. Am J Trop Med Hyg 97 (Suppl ):46–57.
- 27.↑
Gogue C et al., 2020. An observational analysis of the impact of indoor residual spraying in Northern, Upper East, and Upper West Regions of Ghana: 2014 through 2017. Malar J 19: 242.
- 28.↑
Wagman J et al., 2020. Rapid reduction of malaria transmission following the introduction of indoor residual spraying in previously unsprayed districts: an observational analysis of Mopti Region, Mali, in 2017. Malar J 19: 340.
- 29.↑
Wagman J et al., 2020. Combining next-generation indoor residual spraying and drug-based malaria control strategies: observational evidence of a combined effect in Mali. Malar J 19: 293.
- 30.↑
Wagman J et al., 2018. An observational analysis of the impact of indoor residual spraying with non-pyrethroid insecticides on the incidence of malaria in Ségou Region, Mali: 2012–2015. Malar J 17: 19.
- 31.↑
Alonso S, Chaccour CJ, Wagman J, Candrinho B, Muthoni R, Saifodine A, Saute F, Robertson M, Zulliger R, 2021. Cost and cost-effectiveness of indoor residual spraying with pirimiphos-methyl in a high malaria transmission district of Mozambique with high access to standard insecticide-treated nets. Malar J 20: 143.
- 32.↑
Chaccour CJ et al., 2018. Combination of indoor residual spraying with long-lasting insecticide-treated nets for malaria control in Zambezia, Mozambique: a cluster randomised trial and cost-effectiveness study protocol. BMJ Glob Health 3: e000610.
- 33.↑
Keating J, Yukich JO, Miller JM, Scates S, Hamainza B, Eisele TP, Bennett A, 2021. Retrospective evaluation of the effectiveness of indoor residual spray with pirimiphos‐methyl (Actellic) on malaria transmission in Zambia. Malar J 20: 173.
- 34.↑
Gansané A et al., 2022. Design and methods for a quasi-experimental pilot study to evaluate the impact of dual active ingredient insecticide-treated nets on malaria burden in five regions in sub-Saharan Africa. Malar J 21: 19.
- 35.↑
The PMI VectorLink Project, 2022. Integrated Data Analytics and Visualization. Available at: https://pmivectorlink.org/technical-areas/integrated-data-analytics-and-visualization. Accessed February 28, 2022.
- 36.↑
Aregawi MW et al., 2011. Reductions in malaria and anaemia case and death burden at hospitals following scale-up of malaria control in Zanzibar, 1999–2008. Malar J 10: 46.
- 37.↑
Ashton RA et al., 2019. Use of routine health information system data to evaluate impact of malaria control interventions in Zanzibar, Tanzania from 2000 to 2015. EclinicalMedicine 12: 11–19.
- 38.↑
Bennett A et al., 2014. A methodological framework for the improved use of routine health system data to evaluate national malaria control programs: evidence from Zambia. Popul Health Metr 12: 30.
- 39.↑
Chanda E, Coleman M, Kleinschmidt I, Hemingway J, Hamainza B, Masaninga F, Chanda-Kapata P, Baboo KS, Dürrheim DN, Coleman M, 2012. Impact assessment of malaria vector control using routine surveillance data in Zambia: implications for monitoring and evaluation. Malar J 11: 437.
- 40.↑
Kamuliwo M et al., 2013. The changing burden of malaria and association with vector control interventions in Zambia using district-level surveillance data, 2006–2011. Malar J 12: 437.
- 41.↑
Kanyangarara M et al., 2016. Reduction in malaria incidence following indoor residual spraying with Actellic 300 CS in a setting with pyrethroid resistance: Mutasa District, Zimbabwe. PLoS One 11: e0151971.
- 42.↑
Karema C et al., 2012. Trends in malaria cases, hospital admissions and deaths following scale-up of anti-malarial interventions, 2000–2010, Rwanda. Malar J 11: 236.
- 43.↑
Larsen DA, Martin A, Pollard D, Nielsen CF, Hamainza B, Burns M, Stevenson J, Winters A, 2020. Leveraging risk maps of malaria vector abundance to guide control efforts reduces malaria incidence in Eastern Province, Zambia. Sci Rep 10: 10307.
- 44.↑
Louis VR et al., 2015. An insecticide-treated bed-net campaign and childhood malaria in Burkina Faso. Bull World Health Organ 93: 750–758.
- 45.↑
Maes P, Harries AD, Van den Bergh R, Noor A, Snow RW, Tayler-Smith K, Hinderaker SG, Zachariah R, Allan R, 2014. Can timely vector control interventions triggered by atypical environmental conditions prevent malaria epidemics? A case-study from Wajir County, Kenya. PLoS One 9: e92386.
- 46.↑
Mufunda J et al., 2007. Roll back malaria – an African success story in Eritrea. S Afr Med J 97: 46–50.
- 47.↑
Mumbengegwi DR et al., 2018. Is there a correlation between malaria incidence and IRS coverage in western Zambezi region, Namibia? Public Health Action 8 (Suppl 1 ):S44–S49.
- 48.↑
Ngomane L, de Jager C, 2012. Changes in malaria morbidity and mortality in Mpumalanga Province, South Africa (2001–2009): a retrospective study. Malar J 11: 19.
- 49.↑
Nyarango PM et al., 2006. A steep decline of malaria morbidity and mortality trends in Eritrea between 2000 and 2004: the effect of combination of control methods. Malar J 5: 33.
- 50.↑
Oguttu DW, Matovu JKB, Okumu DC, Ario AR, Okullo AE, Opigo J, Nankabirwa V, 2017. Rapid reduction of malaria following introduction of vector control interventions in Tororo District, Uganda: a descriptive study. Malar J 16: 227.
- 51.↑
Okullo AE, Matovu JKB, Ario AR, Opigo J, Wanzira H, Oguttu DW, Kalyango JN, 2017. Malaria incidence among children less than 5 years during and after cessation of indoor residual spraying in Northern Uganda. Malar J 16: 319.
- 52.↑
Topazian HM et al., 2021. Effectiveness of a national mass distribution campaign of long-lasting insecticide-treated nets and indoor residual spraying on clinical malaria in Malawi, 2018–2020. BMJ Glob Health 6: e005447.
- 53.↑
Yapabandara MAMG, Sarmento R, de Fatima Mota MdR, don Bosco J, Martins N, Wickremasinghe AR, 2015. Evidence-based malaria control in Timor Leste from 2006 to 2012. Malar J 14: 109.
- 54.↑
Graves PM et al., 2008. Effectiveness of malaria control during changing climate conditions in Eritrea, 1998–2003. Trop Med Int Health 13: 218–228.
- 55.↑
Namuganga JF et al., 2021. The impact of stopping and starting indoor residual spraying on malaria burden in Uganda. Nat Commun 12: 2635.
- 56.↑
Epstein A et al., 2022. Resurgence of malaria in Uganda despite sustained indoor residual spraying and repeated long lasting insecticidal net distributions. PLOS Glob Public Health 2: e0000676.
- 57.↑
Bath D et al., 2021. Effectiveness and cost-effectiveness of reactive, targeted indoor residual spraying for malaria control in low-transmission settings: a cluster-randomised, non-inferiority trial in South Africa. Lancet 397: 816–827.
- 58.↑
Machini B, Waqo E, Kizito W, Edwards JK, Owiti PO, Takarinda KC, 2016. Trends in outpatient malaria cases, following Mass Long Lasting Insecticidal Nets (LLIN) distribution in epidemic prone and endemic areas of Kenya. East Afr Med J 93: 10–15.
- 59.↑
Hilton ER et al., 2023. Using routine health data to evaluate the impact of indoor residual spraying on malaria transmission in Madagascar. BMJ Glob Health 8: e010818.
- 60.↑
Rehman AM et al., 2011. How much does malaria vector control quality matter: the epidemiological impact of holed nets and inadequate indoor residual spraying. PLoS One 6: e19205.
- 61.↑
World Health Organization, 2015. Indoor Residual Spraying: An Operational Manual for Indoor Residual Spraying (IRS) for Malaria Transmission Control and Elimination. Available at: https://apps.who.int/iris/handle/10665/177242. Accessed January 19, 2022.
- 62.↑
Gertler PJ, Martinez S, Premand P, Rawlings LB, Vermeersch CMJ. 2016. Impact Evaluation in Practice, Second Edition. Washington, DC: Inter-American Development Bank and World Bank. Available at: http://hdl.handle.net/10986/25030 License: CC BY 3.0 IGO. Accessed February 28, 2022.
- 63.↑
World Health Organization, 2017. Achieving and Maintaining Universal Coverage with Long-Lasting Insecticidal Nets for Malaria Control. Available at: https://apps.who.int/iris/handle/10665/259478. Accessed March 3, 2022.
- 64.↑
Kilian A, Obi E, Mansiangi P, Abílio AP, Haji KA, Balufuss S, Olapeju B, Babalola S, Koenker H, 2021. Variation of physical durability between LLIN products and net use environments: summary of findings from four African countries. Malar J 20: 26.
- 65.↑
Bertozzi-Villa A et al., 2021. Maps and metrics of insecticide-treated net access, use, and nets-per-capita in Africa from 2000–2020. Nat Commun 12: 3589.
- 66.↑
Dengela D, Seyoum A, Lucas B, Johns B, George K, Belemvire A, Caranci A, Norris LC, Fornadel CM, 2018. Multi-country assessment of residual bio-efficacy of insecticides used for indoor residual spraying in malaria control on different surface types: results from program monitoring in 17 PMI/USAID-supported IRS countries. Parasit Vectors 11: 71.
- 67.↑
Marti-Soler H et al., 2021. Effect of wall type, delayed mortality and mosquito age on the residual efficacy of a clothianidin-based indoor residual spray formulation (SumiShieldTM 50WG) in southern Mozambique. PLoS One 16: e0248604.
- 68.↑
Lees RS, Praulins G, Lissenden N, South A, Carson J, Brown F, Lucas J, Malone D, 2022. The residual efficacy of SumiShieldTM 50WG and K-Othrine® WG250 IRS formulations applied to different building materials against Anopheles and Aedes mosquitoes. Insects 13: 112.
- 69.↑
U.S. President’s Malaria Initiative, LLIN Durability Monitoring: Toolkit and Data Repository for PMI-funded LLIN Durability Monitoring Studies. 2022. Available at: https://www.durabilitymonitoring.org. Accessed March 1, 2022.
- 70.↑
U.S. President’s Malaria Initiative, Resources. Available at: https://www.pmi.gov/resources. Accessed March 1, 2022.
- 71.↑
Biemba G, Chiluba B, Yeboah-Antwi K, Silavwe V, Lunze K, Mwale RK, Russpatrick S, Hamer DH, 2017. A mobile-based community health management information system for community health workers and their supervisors in 2 districts of Zambia. Glob Health Sci Pract 5: 486–494.
- 72.↑
Cressman GM, McKay MV, Al-Mafazy A-W, Ramsan MM, Ali AS, Garimo IA, Mkali H, Ngondi JJ, 2015. Using mobile technology to help eliminate malaria in Zanzibar. Online J Public Health Inform 7: e15.
- 73.↑
Kamau A, Nyaga V, Bauni E, Tsofa B, Noor AM, Bejon P, Scott JAG, Hammitt LL, 2017. Trends in bednet ownership and usage, and the effect of bednets on malaria hospitalization in the Kilifi Health and Demographic Surveillance System (KHDSS): 2008–2015. BMC Infect Dis 17: 720.
- 74.↑
Aregawi M et al., 2017. Effect of anti-malarial interventions on trends of malaria cases, hospital admissions and deaths, 2005–2015, Ghana. Malar J 16: 177.
- 75.↑
Aregawi M et al., 2014. Time series analysis of trends in malaria cases and deaths at hospitals and the effect of antimalarial interventions, 2001–2011, Ethiopia. PLoS One 9: e106359.
- 76.↑
Roh ME, Oundo B, Dorsey G, Shiboski S, Gosling R, Glymour MM, Staedke SG, Bennett A, Sturrock H, Mpimbaza A, 2022. A quasi-experimental study estimating the impact of long-lasting insecticidal nets with and without piperonyl butoxide on pregnancy outcomes. Malar J 21: 5.
- 77.↑
Sande S, Zimba M, Mberikunashe J, Tangwena A, Chimusoro A, 2017. Progress towards malaria elimination in Zimbabwe with special reference to the period 2003–2015. Malar J 16: 295.
- 78.↑
Abong’o B et al., 2020. Impact of indoor residual spraying with pirimiphos-methyl (Actellic 300CS) on entomological indicators of transmission and malaria case burden in Migori County, western Kenya. Sci Rep 10: 4518.
- 79.↑
Masaninga F et al., 2013. Review of the malaria epidemiology and trends in Zambia. Asian Pac J Trop Biomed 3: 89–94.
- 80.↑
Maude RJ et al., 2014. Spatial and temporal epidemiology of clinical malaria in Cambodia 2004–2013. Malar J 13: 385.
- 81.↑
DHIS2 Magic GlassesAvailable at https://github.com/jpainter/MagicGlasses2. Accessed March 2, 2022.
- 82.↑
Killeen GF, McKenzie FE, Foy BD, Schiefflin C, Billingsley PF, Beier JC, 2000. A simplified model for predicting malaria entomologic inoculation rates based on entomologic and parasitologic parameters relevant to control. Am J Trop Med Hyg 62: 535–544.
- 83.↑
The Malaria Elimination Initiative, 2020. Entomological Surveillance Planning Tool (ESPT). Available at: http://www.shrinkingthemalariamap.org/sites/default/files/tools/espt-eng-feb21-final.pdf. Accessed March 4, 2022.
- 84.↑
World Health Organization, 2018. Malaria Surveillance, Monitoring and Evaluation: A Reference Manual. Available at: https://apps.who.int/iris/bitstream/handle/10665/272284/9789241565578-eng.pdf. Accessed March 3, 2022.
- 85.↑
Fraser M et al., 2020. Evaluating the impact of programmatic mass drug administration for malaria in Zambia using routine incidence data. J Infect Dis 225: 1415–1423.
- 86.↑
Nice J, Nahusenay H, Eckert E, Eisele TP, Ashton RA, 2020. Estimating malaria chemoprevention and vector control coverage using program and campaign data: a scoping review of current practices and opportunities. J Glob Health 10: 020413.
- 87.↑
Kuehne A, Tiffany A, Lasry E, Janssens M, Besse C, Okonta C, Larbi K, Pah AC, Danis K, Porten K, 2016. Impact and lessons learned from mass drug administrations of malaria chemoprevention during the Ebola outbreak in Monrovia, Liberia, 2014. PLoS One 11: e0161311.
- 88.↑
Wardrop NA, Jochem WC, Bird TJ, Chamberlain HR, Clarke D, Kerr D, Bengtsson L, Juran S, Seaman V, Tatem AJ, 2018. Spatially disaggregated population estimates in the absence of national population and housing census data. Proc Natl Acad Sci USA 115: 3529–3537.
- 89.↑
Thematic Research Network on Data and Statistics (TReNDS), 2020. Leaving No One Off The Map: A Guide For Gridded Population Data For Sustainable Development. Available at: https://www.unsdsn.org/leaving-no-one-off-the-map-a-guide-for-gridded-population-data-for-sustainable-development. Accessed March 3, 2022.
- 90.↑
AbouZahr C, de Savigny D, Mikkelsen L, Setel PW, Lozano R, Nichols E, Notzon F, Lopez AD, 2015. Civil registration and vital statistics: progress in the data revolution for counting and accountability. Lancet 386: 1373–1385.
- 91.↑
Tangena JAA et al., 2020. Indoor residual spraying for malaria control in sub-Saharan Africa 1997 to 2017: an adjusted retrospective analysis. Malar J 19: 150.
- 92.↑
Amoah B et al., 2021. Identifying Plasmodium falciparum transmission patterns through parasite prevalence and entomological inoculation rate. eLife 10: e65682.
- 93.↑
Measure Evaluation, 2018. Household Survey Indicators for Malaria Control. Available at: https://www.malariasurveys.org/documents/Household%20Survey%20Indicators%20for%20Malaria%20Control_FINAL.pdf. Accessed March 3, 2022.
- 94.↑
Harper S, 2015. Malaria Behavior Survey. Available at: https://malariabehaviorsurvey.org. Accessed March 4, 2022.
- 95.↑
U.S. Geological Survey, 2022. Landsat Enhanced Vegetation Index. Available at: https://www.usgs.gov/landsat-missions/landsat-enhanced-vegetation-index. Accessed March 13, 2022.
- 96.↑
Wan Z, Hook S, Hulley G, 2015. MOD11A1 MODIS/Terra Land Surface Temperature/Emissivity Daily L3 Global 1 km SIN Grid V006. Available at: https://lpdaac.usgs.gov/products/mod11a1v006. Accessed March 3, 2022.
- 97.↑
U.S. Geological Survey, U.S. National Geospatial-Intelligence Agency, 2010. GMTED2010. Available at: https://www.usgs.gov/coastal-changes-and-impacts/gmted2010?qt-science_support_page_related_con=0#qt-science_support_page_related_con. Accessed March 14, 2022.
- 98.↑
Ssempiira J, Kissa J, Nambuusi B, Kyozira C, Rutazaana D, Mukooyo E, Opigo J, Makumbi F, Kasasa S, Vounatsou P, 2018. The effect of case management and vector-control interventions on space–time patterns of malaria incidence in Uganda. Malar J 17: 162.
- 99.↑
MEASURE Evaluation, 2020. Malaria Routine Data Quality Assessment Tool: User Manual. Available at: https://www.measuremalaria.org/wp-content/uploads/2020/04/ms-20-190.pdf. Accessed March 13, 2022.
- 100.↑
World Health Organization, 2022. Malaria Surveillance Assessment Toolkit. Available at: https://www.who.int/publications-detail-redirect/9789240055278. Accessed September 29, 2022.
- 101.↑
Lavis JN, Lomas J, Hamid M, Sewankambo NK, 2006. Assessing country-level efforts to link research to action. Bull World Health Organ 84: 620–628. Erratum in: Bull World Health Organ (2006) 84:840.