Malaria trends in districts that were targeted and not-targeted for seasonal malaria chemoprevention in children under 5 years of age in Guinea, 2014–2021

Background Seasonal malaria chemoprevention (SMC) is a main intervention to prevent and reduce childhood malaria. Since 2015, Guinea has implemented SMC targeting children aged 3–59 months (CU5) in districts with high and seasonal malaria transmission. Objective We assessed the programmatic impact of SMC in Guinea’s context of scaled up malaria intervention programming by comparing malaria-related outcomes in 14 districts that had or had not been targeted for SMC. Methods Using routine health management information system data, we compared the district-level monthly test positivity rate (TPR) and monthly uncomplicated and severe malaria incidence for the whole population and disaggregated age groups (<5 years and ≥5 years of age). Changes in malaria indicators through time were analysed by calculating the district-level compound annual growth rate (CAGR) from 2014 to 2021; we used statistical analyses to describe trends in tested clinical cases, TPR, uncomplicated malaria incidence and severe malaria incidence. Results The CAGR of TPR of all age groups was statistically lower in SMC (median=−7.8%) compared with non-SMC (median=−3.0%) districts. Similarly, the CAGR in uncomplicated malaria incidence was significantly lower in SMC (median=1.8%) compared with non-SMC (median=11.5%) districts. For both TPR and uncomplicated malaria incidence, the observed difference was also significant when age disaggregated. The CAGR of severe malaria incidence showed that all age groups experienced a decline in severe malaria in both SMC and non-SMC districts. However, this decline was significantly higher in SMC (median=−22.3%) than in non-SMC (median=−5.1%) districts for the entire population, as well as both CU5 and people over 5 years of age. Conclusion Even in an operational programming context, adding SMC to the malaria intervention package yields a positive epidemiological impact and results in a greater reduction in TPR, as well as the incidence of uncomplicated and severe malaria in CU5.


INTRODUCTION
Malaria is the leading cause of morbidity and mortality in Guinea, with 2 422 445 confirmed cases and 1029 deaths reported in 2021. 1 WHAT IS ALREADY KNOWN ON THIS TOPIC ⇒ When effectively deployed, seasonal malaria chemoprevention (SMC) has shown to reduce childhood malaria by up to 75% in areas of high, seasonal malaria transmission.

WHAT THIS STUDY ADDS
⇒ Using routinely collected data from Guinea's health management information system, we compared the compound annual growth rate (CAGR) of malariarelated outcomes in 14 districts that had or had not been targeted for SMC between 2015 and 2021.⇒ Compared with non-SMC districts, a statistically lower CAGR in all-age number of clinical malaria cases and clinical malaria test positivity rate (TPR), as well as the incidence of uncomplicated and severe malaria incidence was observed in SMC districts.But for the number of clinical malaria cases, this difference in CAGR between SMC and non-SMC district was observed for both children under 5 years of age and people above 5 years of age.

HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY
⇒ Our analyses also show that routine health management information system data and the CAGR approach can be used to continuously monitor malaria intervention effectiveness against standard malariometric indicators.⇒ Our results provide evidence to support that-even in an operational programming context-adding SMC to the comprehensive package of malaria interventions yields a positive epidemiological impact and results in a greater reduction in TPR, as well as the incidence of uncomplicated and severe malaria in CU5.

BMJ Global Health
Over the past decade, Guinea's National Malaria Control Programme (NMCP)-in collaboration with bilateral, multilateral and non-governmental partners-has scaled up malaria prevention and control efforts, including rapid diagnostic tests (RDTs), artemisinin-based combination therapies, intermittent preventive treatment for pregnant women with sulfadoxine-pyrimethamine (SP) and insecticide-treated nets. 2 Additionally, in 2015, the NMCP piloted and then rolled out seasonal malaria chemoprevention (SMC) in children aged 3-59 months (CU5).Recommended by the WHO since 2012 for countries with seasonal malaria transmission, 3 SMC is the monthly administration of a single dose of SP and three daily doses of amodiaquine (AQ) (SP-AQ) to CU5 during the peak malaria transmission season.SMC can reduce the incidence of clinical malaria in the 28 days following administration by up to 75% when effectively deployed [4][5][6][7] ; depending on the length of the rainy season, 3-5 cycles of SMC are conducted. 8As of 2021, 13 countries had adopted SMC (ie, Benin, Burkina Faso, Cameroon, Chad, Gambia, Ghana, Guinea, Guinea Bissau, Mali, Niger, Nigeria, Senegal and Togo) at different scales of implementation. 85][6][7] There is also some evidence that SMC positively affects other health outcomes (eg, anaemia and malnutrition), but such effects have not always been significant across studies. 7 9-14onthly administration of SP-AQ was found to be the drug regimen with the highest efficacy, [15][16][17] and using community health workers (CHWs) to deliver SMC was shown to be more cost-effective than using facility-based nurses, immunisation outreach clinics or outreach trekking teams. 18 19hile there is strong evidence of the effect of SMC in rigorously conducted academic research studies, evidence of the protective effect of SMC in programmatic contexts is more limited. 7 20 21This type of evaluation is challenging but very necessary-indeed, the effectiveness of interventions in the context of public health programmes often differs from the efficacy measured in academic research studies, because of operational challenges when interventions are implemented at scale. 22 23he aim of the analyses presented here was to assess the programmatic impact of SMC in Guinea's context of scaled up malaria intervention programming, specifically by comparing malaria-related outcomes in districts that had or had not been targeted for SMC between 2015 and 2021.Additionally, to do so, we wanted to use a methodological (ie, using data routinely collected by Guinea's health management information system (HMIS)) and an analytical approach (ie, using the compound annual growth rate, CAGR) that is simple and would allow the NMCP to readily monitor SMC effectiveness against standard malariometric indicators as it continues to expand the intervention across the country.

Study setting
The Republic of Guinea is located in West Africa: it covers a total land area of 245 860 km 2 and has an estimated population of 13.5 million people.The country is divided into 8 major administrative regions, which are further divided into 38 préfectures (districts), 5 of which comprise the urban areas around the capital city, Conakry.Districts are the main administrative unit where many of the public services are planned, managed and implemented, including for health and malaria.Guinea's climate is tropical and humid with a wet (June-November) and a dry season (December-May); the rainy season is followed by a peak malaria transmission season, with the highest malaria case count typically observed between July and November/December.Approximately 95% of malaria cases in Guinea are caused by Plasmodium falciparum, the principal vectors being Anopheles gambiae, Anopheles funestus and Anopheles arabiensis. 1 24 25 SMC in Guinea Guinea introduced SMC initially in 6 districts in 2015, scaling up to 17 districts by 2021, with 4 monthly community-based SMC campaigns (cycles) conducted during the peak malaria transmission season (generally from July to October).The NMCP implements SMC in partnership with nongovernmental and international organisations that support the successful execution of Guinea's National Malaria Strategic Plan and its activities.As per WHO guidance, 8 the recommended drug regimen for SMC is SP-AQ.CHWs (existing ones as well as ones mobilised specifically for the SMC campaign) administer SP and the first dose of AQ to eligible children and give the remaining two daily doses of AQ to the caregiver.
Cycles of SMC are conducted each year-once every 4 weeks during peak malaria transmission season.SMC usually starts in late July, but the exact starting date fluctuates every year depending on several factors, including logistical considerations and rainfall.The target population for SMC comprises all CU5, excluding those with known allergies to SP or AQ, those under cotrimoxazole treatment, and those severely ill or experiencing a presumptive malaria episode.Children 3-11 months of age receive one 25 mg dose of sulfadoxine, one 12.5 mg dose of pyrimethamine and three doses of 75 mg AQ given over the course of three consecutive days; children 12-59 months of age receive double doses of SP and AQ.The children's vaccination booklets are consulted to help determine their age.
In teams of two, CHWs go door-to-door to every household; sometimes they organise distribution sessions at gathering venues such as markets, churches, dwellings, mosques and fields.After explaining the SMC strategy (objective, rationale, risks and benefits, treatment BMJ Global Health instructions) to the caregiver, they administer treatments, with any children with malaria or danger signs referred to the closest health centre.CHWs fill in forms where they indicate the number of treatments administered in every household, and they keep updated stock management sheets.Drugs are supplied through the Ministry of Health (MOH) to the health centres, who then allocate them to the CHWs.Nurses in health centres are responsible for coordinating CHWs' work and for collecting SMC forms.Supervisory visits are conducted by nurses and district health authorities.

Malaria data
In the aftermath of the 2014-2016 Ebola outbreak, the MOH established a strategic plan to strengthen its routine surveillance system, which led to the adoption of the District Health Information Software (DHIS) 2, an open source, online software application as the national HMIS platform for monthly aggregate data from health facilities.DHIS2 is managed by Guinea's Système National d'Information Sanitaire (SNIS) and has been rolled out nationally 26 ; it is complemented by an infectious disease response and surveillance platform that reports on aggregate disease surveillance and individual case surveillance of epidemic-prone diseases.The NMCP uses DHIS2 to store the monthly aggregate epidemiological surveillance data collected, including for all outpatients and inpatients seen, suspected clinical malaria cases seen, suspected clinical malaria cases tested, malaria cases confirmed and malaria cases treated. 26Monthly numbers of tested clinical malaria cases, confirmed positive clinical cases and severe malaria cases are disaggregated by age group (<5 years of age (<5 years), ≥5 years of ages (≥5 years)).Data for CU5 and people ≥5 years of age (PO5) were downloaded for January 2014 to November 2021 from the SNIS portal; we also downloaded the estimated catchment population of each health facility for each year.

Study area, study population and data analysis
Analyses included eight (Labé, Koubia, Tougue, Mali, Lelouma, Gaoual, Koundara and Dinguiraye) and six (Boffa, Boké, Coyah, Dubreka, Forecariyah and Fria) districts where SMC campaigns were or were not performed from January 2014 to November 2021, respectively (figure 1).We adjusted analyses to account for different year of enrolment of districts in the SMC campaign (ie, Labé and Lelouma started SMC in 2017).The 14 study districts were among the districts supported by StopPalu and StopPalu+, projects funded by the US President's Malaria Initiative and implemented by RTI International.The remaining districts covered by these two projects were the communes of the capital Conakry (ie, Matam, Dixinn, Ratoma, Matoto and Kaloum), which were not eligible for SMC due to their low malaria prevalence; these districts were, therefore, excluded from the analyses.
Among all the study districts' facilities reporting data to SNIS, we selected those that had a high level of malaria data reporting completeness.Given the long study period and possible difficulties in consistent data reporting, an arbitrary threshold of 10 missing months within the 2014-2021 study period and no more than 3 continuously missing months per year was used to identify facilities with adequate data reporting completeness.We only included public health facilities (ie, hospitals, all health centre types and health posts) supported by StopPalu and StopPalu+.This selection method identified 131 facilities of the 149 in the dataset with adequate reporting completeness.The resulting catchment population based on the selected facilities was 4 451 792 people in 2014, increasing to 5 244 844 people in 2021.
Using the extracted indicator data from SNIS, we calculated the monthly test positivity rate (TPR) and monthly uncomplicated and severe malaria incidence at the district level for the whole population, as well as disaggregated by age groups (<5 years and ≥5 years of age).We used statistical analyses to describe the time trend of the number of tested fevers, TPR, uncomplicated malaria incidence and severe malaria incidence.The changes of these malaria indicators through time were analysed by calculating the CAGR from 2014 to 2021 at the district level.The CAGR was calculated by dividing a time series end value by its beginning value and raising the resulting figure to the inverse number of the time series years subtracting it by one.Thus, for our SMC analyses, the CAGR for each malaria indicator was calculated as follows: where n is the year of SMC implementation.Testing for differences in malaria indicator values and their CAGR between SMC and non-SMC districts were performed by using the Wilcoxon's signed rank test. 27A full description of the analysis steps is provided in online supplemental file 1.

BMJ Global Health
The reflexivity statement to promote equitable authorship in the publication of research from international partnerships is appended as online supplemental file 2.

SMC coverage
In the districts included in the analyses, 8.1 million treatment doses of SP+AQ were administered to eligible CU5 between 2015 and 2021, resulting in an average annual programmatic SMC coverage of 89% (range between years: 86%-93%).

Number of clinical malaria cases tested
From January 2014 to November 2021, 5 002 551 clinical malaria cases were tested in the health facilities included in the study, of which 1 658 637 were CU5 (33.2%) (table 1).Among all tested clinical cases, 2 484 794 (49.7%) and 2 517 757 (50.3%) were tested in SMC and non-SMC districts, respectively; in both district groups the proportion of clinical malaria in CU5 was approximately 30% of all tested clinical cases (SMC districts=29.8%vs non-SMC districts=36.5%)(table 1).From 2014 to 2021, the trend of tested clinical cases in CU5 and PO5 showed consistent increase in all districts (online supplemental figure S1).The 2014-2021 CAGR of tested clinical cases for SMC districts (median=12.2%,range=4.1%,20.6%) was significantly lower compared with that of non-SMC districts (median=19.3%;range=12.9%,22.2 %) (Wilcoxon's test, p<0.05) (figure 2A).When disaggregated by age, the CAGR of tested clinical cases was lower in SMC districts compared with non-SMC districts for both CU5 and PO5, but it was only statistically significant for CU5 (Wilcoxon's test, p<0.05) (figure 2A).

Uncomplicated malaria incidence
The median malaria incidence in SMC districts was 18.5 cases per 1000 people (IQR=14.5  S4).Thus, both SMC and non-SMC districts experienced an increase in incidence from 2014 to 2021.Comparing the incidence trend among the age groups and districts, only the incidence of CU5 in SMC

BMJ Global Health
districts showed a declining trend from 2014 to 2021.

DISCUSSION
5][6][7] Several studies have analysed the effectiveness of SMC when implemented at large operational scales, showing SMC protective effectiveness against clinical malaria ranging from 73-98% in CU5 at 28 and 42-days post-SMC administration, respectively. 28 29However, these studies relied on large samples of at-risk populations (ie, CU5), either following case/ control study designs, 28 29 or using nationally representative surveys such as Demographic and Health Surveys or Malaria Indicator Surveys. 30 31While such approaches are certainly robust, they can be onerous, time-consuming, and costly, and may not allow for continuous monitoring of the intervention across all SMC implementation areas.

BMJ Global Health
][34][35][36] Using district-level data from Guinea's HMIS collected between 2014 and 2021, we show the programmatic impact of SMC on malaria trends in targeted districts, confirming the effectiveness of SMC shown in various controlled research studies.Our analyses show that-in a context of strengthened case management and medium to high coverage of LLINs-SMC significantly reduced the TPR, uncomplicated malaria incidence, and the incidence of severe malaria in CU5 in Guinea between 2014 and 2021.These findings also corroborate other recent analyses estimating SMC effectiveness through use of routine health information system data.Thus, in Chad, using generalised additive modelling, Richardson et al 37 estimated that SMC reduced the number of suspected and confirmed malaria in CU5 by 18% (95% CI 6% to 28%) and 19% (95% CI 7% to 29%) at primary health facilities in 23 health districts during the months of SMC implementation, respectively.Similarly, an evaluation of the ACCESS-SMC programme using a difference-indifferences approach of DHIS2 data observed a 45.0% and 55.2% reduction in malaria outpatients after 2 years of SMC implementation in Burkina Faso and The Gambia, respectively; the study also reported substantial reductions in the number of malaria inpatients and malaria hospital deaths following SMC. 29nterestingly, we show that the effect of SMC is particularly pronounced in the dry season rather than the rainy season.We hypothesise that SMC results in CU5 either being fully protected from infection or, if infected, only having infections with low parasitaemia (which may be due to incomplete adherence to the 3-day regimen or drug resistance). 38Combined with the lower abundance and density of mosquitoes during the dry season, this

Global Health
then results in lower transmission rates, which leads to a lower incidence.Similar observations have been made in studies evaluating vector control interventions. 39hese observations do show, however, that in Guinea SMC should possibly start earlier than in July (ie, to have a greater effect on infection protection and transmission intensity during peak transmission season), and that studies monitoring SMC impact should potentially monitor effectiveness beyond the currently recommended 28-42 days post-SMC administration. 40oreover, unlike prior controlled academic research and other studies, our analyses also demonstrate an effect of SMC administered to CU5, although lower, on PO5, specifically for TPR, uncomplicated malaria incidence, and severe malaria incidence.Such effect seems plausible, given that-even in a context of decreasing malaria prevalence such as Guinea 1 -CU5 still represent a large proportion of all cases (ie, 55.2% for the 2014-2021 study period; table 1) and, thus, a substantial proportion of the human reservoir infecting anopheline mosquitoes.Any significant intervention effect reducing the CU5 Plasmodium reservoir will lower transmission intensity and, thus, spill over and reduce malariometric indicators in PO5.
As support continues for malaria elimination and additional means for malaria control are introduced (eg, vaccines), alternative methods for measuring impact must be explored outside trial settings as controlled and observational research studies to investigate effectiveness at scale are very costly.Multiple impact evaluations and approaches that address gaps or biases in data and triangulating between data sources strengthens the plausibility of programme impact.We used the CGAR as an analytical method.The CAGR is a commonly used methodology in the field of economics to describe changes in (economic) growth from the beginning to the end of a time series.The advantage of using CAGR compared with other analytical approaches-such as average annualised rates-is that the resulting estimate is not affected by fluctuations within the time series that could produce misleading results.2][43] An additional advantage of the approach is that, because it is algebra based, it does not require advanced data analytics or modelling expertise, and thus can be routinely applied by programme personnel with a basic understanding of calculus (eg, in MS Excel worksheets or even DHIS2).
Because SMC is consistently proving to be a high coverage and effective intervention, several countries with seasonal malaria transmission outside of the Sahel are now piloting SMC to be included in their malaria intervention package (eg, Uganda 44 and Mozambique 45 ).WHO recently amended the SMC guidelines to give countries flexibility to change the SMC regimen (eg, by adding additional SMC cycles to cover longer peak transmission seasons, or by including older age groups). 87][48] Moreover, there is increased interest to use SMC campaigns to integrate or leverage other health interventions, [49][50][51] the rationale being that (1) other health interventions that may not have as high population acceptance or coverage as SMC would benefit from the high acceptance and coverage of SMC and (2) cost-efficiencies would be obtained, since combining intervention planning, implementation and monitoring would result in various economies of scale (eg, training of health workers, transport costs, limiting health workers' time to implement campaigns rather than providing health services).

Limitations
Several potential caveats of our analyses should be highlighted.First, our analyses did not adjust for intervention coverage, either for SMC or other malaria interventions such as LLINs.Thus, we assume that programmatic coverage, use of and adherence to these interventions is relatively homogenous and that the only difference in overall malaria intervention coverage between the two sets of districts included in our analyses is the administration of SMC to CU5.Post-SMC round assessments consistently showed programmatic coverage of 82%-93% across districts included in our analyses.Moreover, as per 2021 Malaria Indicator Survey data, 52 LLIN use in CU5 in households with at least 1 LLINs was 60.4% in survey enumeration areas in SMC districts, compared with 65.6% for non-SMC districts.
Second, using routine HMIS data for impact evaluations can be problematic due to internal validity, completeness and potential bias in estimates of effect, and caution must be exercised in interpreting analytical findings. 32Additionally, challenges with HMIS data are that they are dependent on the proportion of infected individual cases seeking treatment for malaria at public sector health facilities; the proportion of patients seeking care who are parasitologically diagnosed; and facilities reporting all suspected and confirmed malaria cases consistently over time.Thus, for example, even though SMC was implemented in Guinea in 2015, an increase in the number of fevers was observed in CU5, suggesting a lack of impact of SMC.However, during this period, changes in access to health services occurred, such as improvement in malaria case detection and diagnosis at facility and community levels, removal of user access fees for CU5 and pregnant women, and improvement in the availability of malaria diagnostic and case management commodities.We controlled for such bias by excluding health facilities with incomplete reporting (ie, facilities with 10 months of missing data throughout the study period, and no more than three continuously missing months per year), and noted that these facilities were evenly spread across SMC and non-SMC districts.We also note that our analyses of the SMC effect in CU5 was BMJ Global consistently observed across all malariometric indicators included in our analysed.
Third, malaria testing to identify cases in Guinea is largely RDT based.Although used tests have a high sensitivity and specificity, it is possible that low parasitaemic and asymptomatic infections were missed.It is unlikely, however, that such infections would have clustered in a specific spatiotemporal pattern across SMC compared with non-SMC districts, and, thus, would change our findings and conclusions.
Fourth, we did not adjust for climate confounders that are known to possibly affect malaria transmission, intervention implementation or health service access, such as temperature and rainfall.Given the moderate distances between SMC and non-SMC districts, we assumed that changes in temperature and rainfall pattern would likely affect both sets of districts similarly.Outlined CGAR approach certainly represents a simple approach comparing malaria trends between districts experiencing differing variability in climate variables, Fifth, we delineated our analyses to district boundaries, rather than a defined area size (eg, 25 km 2 ).This was done because routinely collected HMIS data gets aggregated at that level and districts are the lowest administrative unit in Guinea that plan, implement and monitor malaria programming; any response to an increase in malaria cases would occur at district level.
Finally, spatial analyses performed using arbitrary spatial divisions such as district administrative boundaries can be affected by the modifiable areal unit problem. 53 54or example, cases that were reported at a health facility in a given district could stem from households in a different district, and therefore, bias district-level malaria HMIS data.Thus, the possibility remains that cases got infected outside of their district (eg, during travel).

CONCLUSION
Our study provides evidence to support that-even in an operational programming context-adding SMC to the comprehensive package of malaria interventions yields a positive epidemiological impact and results in a greater reduction in TPR, malaria incidence and hospitalisations in CU5 than when implementing an intervention package without SMC.We also show that the use of routine HMIS data represents a viable method for assessing the epidemiological impact of public health interventions in the absence of trial studies, household surveys and extensive covariate data.

Figure 1
Figure 1 Districts covered by the StopPalu and StopPalu+ in which SMC was performed.Image creator/owner: DB.SMC, seasonal malaria chemoprevention.

Figure 2
Figure 2 CAGR of tested fevers (A), TPR (B), uncomplicated malaria incidence (C), and severe malaria incidence (D) in SMC districts and non-SMC districts from 2014 to 2021.The '*' symbol represents a significant difference (p<0.05,Wilcoxon's test) between CAGR values of SMC district and non-SMC districts.Image creator/owner: DB.CAGR, compound annual growth rate; SMC, seasonal malaria chemoprevention; TPR, test positivity rate.

Figure 3
Figure 3 Boxplot comparing TPR of SMC (left column) and non-SMC (right column) districts by age group of 2014 to November 2021 for <5 years and ≥5 years age group.The '*' symbol represents a significant difference (p<0.05,Wilcoxon's test) between SMC and non-SMC districts.Image creator/owner: DB.SMC, seasonal malaria chemoprevention; TPR, test positivity rate.

Figure 4
Figure 4 Boxplot comparing uncomplicated malaria incidence of SMC (left column) and non-SMC (right column) districts per age group of 2014 to April 2019 for <5 years age group and ≥5 years age group.The '*' symbol represents a significant difference (p<0.05,Wilcoxon's test) between SMC and non-SMC districts.Image creator/owner: DB.SMC, seasonal malaria chemoprevention.

Figure 5
Figure 5 Boxplot comparing severe malaria cases per 1000 people in SMC (left column) and non-SMC (right column) districts per age group of 2014 to April 2019 <5 yrs age group and ≥5 years age group.The '*' symbol represents a significant difference (p<0.05,Wilcoxon's test) between SMC district and non-SMC districts.Image creator/owner: DB.SMC, seasonal malaria chemoprevention.

Table 1
Number of tested fevers, positive tests and positive rate divided by age group and SMC coverage during the study period (January 2014-November 2021)