Stratication At The Health District Level And Targeting of Malaria Control Interventions In Mali

From 2017 to 2019, the median incidence across 75 health districts of Mali was 129.34 cases per 1,000 36 person-year (IQR=86.48). The results showed different periods of high malaria transmission in health 37 districts level and durations varying from 2 to 6 months, showing a double peak for some health 38 districts, which were located in the flooded areas. environmental variables such as rainfall, vegetation 39 index (NDVI), maximum temperature and relative humidity were significantly associated at malaria 40 incidence with a lag of around one month. A strata defines a geographical area with similar 41 epidemiological, environmental and socio-economic factors. Stratification resulted in 12 health 42 districts of very low transmission, 19 low transmission, 20 moderate transmission and 24 in high 43 transmission areas. The number of rounds of season malaria chemoprevention will be based on the 44 number of months in the high transmission period. 45 This first stratification in Mali will allow targeting malaria control strategies. This approach will be 47 dynamic and revised yearly in order to integrate information from the national epidemiological 48 surveillance. 49


Introduction 23
Malaria has been the leading cause of morbidity and mortality for several decades in Mali, with an 24 increase from 2017 to 2020 (2,884,837 confirmed cases and 1,454 deaths). On the recommendation of 25 the World Health Organization (WHO) and in the interests of efficient use of resources, Mali has 26 begun a process of stratifying the health districts to target malaria control strategies. 27

Method 28
Malaria, entomological and environmental data were collected through the local health information 29 system (LHIS), the Demographic and Health Survey (DHS 2018), research institutions and MALI-30 METEO services. The WHO has recommended stratification at the district level consisted of assigning 31 each district to one of 4 classes according to criteria based on incidence adjusted for attendance rate. 32 Variables associated with monthly malaria incidence at the district level were identified using a 33 general additive non-linear regression model. 34 Background 51 Malaria remains the leading cause of morbidity and mortality in Mali. In 2020, the surveillance system 52 reported 2,884,827 confirmed cases out of 4,252,213 people tested, of which 871,265 were severe 53 cases. The number of deaths was 1,454 notified by the health facilities in 2019, still being the primary 54 reason for health care visits with 36% of consultation due to malaria [1]. The main parasites 55 responsible for malaria in Mali are: Plasmodium falciparum (more than 85%), Plasmodium malariae 56 (10-15%) and Plasmodium ovale (1%) [2]. However, cases of Plasmodium vivax have been observed 57 and documented in Mali [3,4]. Malaria  week of pregnancy. 70 Currently the period of SMC has been defined on a national basis, from July to October. However, the 71 high transmission period may be different from place to place, notably the onset and the duration, 72 according to different environmental and population characteristics. This may hinder the effectiveness 73 of SMC in the situation of asynchronism between intervention and high transmission period. 74 WHO recommends that malaria-endemic countries adapt interventions as transmission evolves [7]. 75 The intensity of malaria transmission is strongly dependent on environmental, socio-demographic 76 characteristics and the malaria interventions implemented [8][9][10][11][12]. It is generally measured by the manifestations. This stratification allows better targeting of malaria control strategies in Mali. 87 From the environmental perspective, Mali is divided into 5 zones corresponding to the initial 88 epidemiological facies described in 1989: i) a Sudano-Guinean zone with a long seasonal transmission 89 of 4 to 6 months; ii) a zone of short seasonal transmission of 3 to 4 months; iii) a zone of sporadic or 90 epidemic transmission in the northern regions and some localities in the regions of Koulikoro, Segou, 91 Mopti and Kayes; iv) zones of bi or multimodal transmission including the inner delta of the Niger 92 River and the dam zones; v) zones that are less propitious to malaria transmission, particularly in 93 urban areas as Bamako  There are several dams and flooded areas used for agriculture. The flooded areas extend along the 108 Niger River: the inner delta of the Niger River extends from the Segou region in the center, through 109 Mopti to the Timbuktu region in the north (Figure 1), and covers a maximum area of 41,000 km 2 , 110 including numerous lakes, ponds and swamps [15,16]. Annual rainfall data measured vary from less 111 than 200 mm in the Saharan desertic zone to more than 1100 mm in the pre-Guinean zone (Figure 1): 112 desertic zone (less than 200 mm), Sahelian zone (200 to 600 mm), Sudanian zone (600 to 800 mm), 113 Sudano-Guinean zone (800 to 1100 mm) and the pre-Guinean zone (more than 1100 mm) [17]. Mali is composed of 11 regions including the District of Bamako. The health system is divided into 117 75 health districts. At the institutional plan, the health system is structured on three levels: 118 • the operational: the health district is the operational unit responsible for planning, budgeting 119 and managing health development; 120 • the regional: is the technical support level at the first level; 121 • the national: is the strategic level that defines strategic orientations and determines 122 investments and operations. wind speed in m/s (MERRA-2, resolution 0.5 x 0.625°) [18]. 149 The land use data, i.e. percentage of cultivated area, percentage of urbanization, percentage of area 150 occupied by trees and surface water, were obtained from the 2017 Copernicus Africa classification [19]. To determine the periods of transmission at the national level, a breakdown of the malaria incidence 166 curve in Mali was made using the change point analysis of the mean, which looks for points of change 167 on the monthly incidence curve. The algorithm PELT was used [22], allowing the identification of 168 periods of low and high transmission. 169 The determination of the periods of high transmission at the health district level was carried out to 170 identify the periods of transmission and the endemic health districts. We presented time series data 171 from 2017 to 2019 as a graph in a box plot by month. The duration of the high transmission period 172 was defined for each health district to determine the number of SMC rounds (one per month). 173 As the impacts of environmental data on malaria incidence are well documented in Mali [9,23,24], we 174 have presented malaria incidence and environmental data to observe a possible relationship between 175 increasing incidence and lags. It was investigated with which lagged environmental factors best 176 predicted the incidence time series. 177 To assess the relationship between environmental factors and the malaria case series, we provided 178 synthetic meteorological indices (SMI) by using principal component analysis (PCA). This approach 179 was necessary to take into account the curse of dimensionality and collinearities. The number SMI will 180 be chosen according to the elbow criterion [25]. 181 We analyzed the relationships between synthetic meteorological indices and malaria incidences by 182 using Generalized Additive Model (GAM). This approach allowed to use Spline smoothing functions 183 modeling non-linear relationships, Negative-binomial distribution for count data considering 184 overdispersion, and log (Population) as offset to provide standardized incidence ratios (SIR). The time 185 series of monthly cases was modeled as a function of meteorological factors grouped into synthetic 186 meteorological indices by the PCA method with a lag of 1 to 4 months. The criterion for selecting the 187 best model was the high explained deviance and the low model generalized cross-validation score 188 • The low completeness of the data by location. 201 Adjusted incidence was calculated per 1,000 person-years by adjusting for the health district usage rate. 202 To account for this, the adjusted incidence was calculated. 203 = ( health facility usage rate ) × 100 (Eq3) 204 Entomological data were presented in the form of maps to report the distribution of species and 206 insecticide resistance. 207 Stratification and targeting of interventions 208 Adjusted incidence data from the health districts were used to determine the strata with meteorological 209 and land use data. The classes were determined by the statistical method of Classification and 210  Uni and multivariate regression analysis 249 The univariate regression analysis by using general additive modeling (GAM). Showed a month lag 250 between SEI 1 (rainfall, relative humidity, vegetation index and soil moisture) and malaria and a zero-251 month lag between SEI 2 (temperature) and malaria incidence. 252 Considering these lags, the multivariate analysis showed a linear relationship between malaria 253 incidence and the SEI 1 (rainfall, NDVI, relative humidity and soil moisture) (p-value < 0.001). 254 Malaria incidence first decreased at low temperatures with SEI 2, and then rapidly decreased with high 255 temperature values (SEI 2) (p-value < 0.01). High resistance to pyrethroids and organochlorines was observed in several health districts. Resistance 325 to carbamates (bendiocarb and propoxur) was also noted, although it appears to be less extensive than 326 resistance to pyrethroids. The bioassays showed that the major malaria vector was sensitive to  (Table S1-2) with a monthly incidence of between 20 and 60 cases per 1,000 person-387 months. Health districts in the center, which were in the majority, had the standard high transmission 388 period from July to December with an incidence of between 10 and 30 cases per 1,000 person-months. 389 It represents the majority of health districts in Mali. For the northern health districts, the high 390 transmission period was late, starting in August or even September and ending in January with an 391 incidence drop for some in October. The incidence was often low, less than 5 cases per 1,000 person-392 months (Figure 3). These are generally health districts in the north and those located in flooded areas 393 (Dire, Niafunké, Youwarou, Bourem, Gao, Timbuktu) [23]. 394 In a randomized trial carried out in the Segou health district between 2014-2016, incidences were 395 similar to these routine data [32]. 396 The duration of the high transmission period varied from 2 to 6 months depending on the health 397 district. The number of rounds for the SMC should be based on these results to be effective ( Figure  398 15). In a context where financial resources are low, health districts needing more than four rounds may 399 conduct a pilot cost/benefit study before scaling up [33,34]. 400 This heterogeneity of transmission must be considered when implementing preventive interventions as 401 seasonal malaria chemoprevention in children (SMC) and the number of rounds for an effective and 402 efficient use of malaria control resources. Some health districts are not seasonal, which makes them 403 ineligible for certain interventions targeting periods of high transmission as defined by WHO. This 404 was the case in some health districts in Niger that have a low incidence of malaria [35]. These health 405 districts must benefit from continuous weekly epidemiological surveillance to move towards case-by-406 case surveillance because of the low incidence level but also the risk of epidemics due to the lack of 407 malaria premunition in the population. 408 To find a link between the incidence of malaria and environmental factors, the principal component 409 analysis allowed us to select two axes that explained more than 80% of inertia (Figure 4). SEI 1 410 consisted of rainfall, soil moisture, relative humidity and NDVI. SEI 2 consisted of the maximum 411 temperature [8,23,36,37]. 412 The adjusted incidence was calculated considering the frequentation rate to correct for the bias of 413 under-attendance for several reasons. 414 Mapping of the adjusted incidence showed that more than a third of the health districts are the zone of 415 high transmission in Mali with more than 450 cases per 1,000 person-years. These health districts are 416 located in areas of high rainfall, along rivers and flooded areas. Health districts with moderate 417 transmission were generally located in the Sahel with an incidence from 250 to 450 cases per 1,000 418 person-years. Health districts in areas of low and very low transmission account for 41%. They were 419 located in areas of low rainfall with a very low population density of less than one habitat per km² 420 (Figure 7). They are located in northern Mali (Figure 6 and 13). These low transmission areas could 421 be potential candidates for the use of the mass treatment administration strategy to move towards 422 malaria pre-elimination in . A Study carried out in Dire, in the Sahelian zone of Mali 423 showed between incidence from 20 to 120 cases per 1,000 person-years [23].  (Figure 7). It is 40 436 to 60% in the health districts of the center of the country and more than 80% in the northern health 437 districts (Sahara zone). 438 Alternatives to improve accessibility outside of 5km, such as the introduction of community health 439 agents, must be strengthened to compensate for this lack of access to healthcare for children and increase surveillance on insecticides to continue to benefit from their efficacy (Figure 10-11). 458 Stratification has been done using data from Epidemiological surveillance or prevalence surveys [58]. is located south of the Sahara and has a high malaria risk than Class 1 and 2 because of the presence of 469 the river. Class 5 is similar to class 3 in terms of incidence but there is the absence of the river which 470 is replaced by the presence of low vegetation. This is a Saharan area with epidemic potential [62]. 471 The health districts of the moderate transmission zone are located in the Sahelian zone of Mali, and the 472 zone of high malaria transmission is located in the south, the Sudanian zone (Figure 1, 13-14). with an environmental risk as in class 4, but transmission is low due to urbanization and better access 478 to care. This is the capital city of Bamako and the health districts of Kayes, and Yélimané. The two 479 latter have many health care's structures through the investment of expatriates (Figure 8). The number of rounds per health district will depend on the duration of the period of high transmission 504 (Figure 15). Health districts with a long transmission period will be able to make 4-5 cycles and 505 health districts with a short transmission period will be able to make a maximum of 3 cycles. The 506 number of SMC rounds can also take into account the bimodal transmission of malaria, especially in 507 flooded areas [23]. 508 The main limitation of this study was the choice of the health district scale. Malaria transmission is 509 heterogeneous within a health district (several health areas) and interventions can be inappropriate in 510 the whole district. However, this stratification is dynamic, a short revision mechanism must be put in 511 place to adjust for insufficiencies. 512 The limitations of this study are the period of data collection of confirmed cases and the time scale 513 which was monthly. 514 Another limitation is the unavailability of meteorological observation data in some health districts. 515 However, we used remote sensing data for these health districts. It is a dynamic document that must be revised periodically to consider observations from 525 epidemiological surveillance and research activities. It provides the starting point for a phase towards 526 pre-elimination in targeting control strategies to a specific stratum which is, therefore, more optimized 527 in terms of efficiency and resource-saving. 528 Availability of data and materials 529 The data and background maps are available at the level of the authors and the NMCP of Mali. 530 Anyone wishing to consult or use these data are requested to contact us. The grey band shows periods of high malaria transmission from 2017 to 2019. Rainfall in blue histogram on the left axis, and Malaria incidence in a red curve, the average monthly temperature in a black curve on the right axis.    Changes in malaria incidence as a function of meteorological factors; GAM multivariate analysis.

Figure 6
Annual malaria incidence at the health district level, adjusted for health facility access, Mali, 2019.   Malaria prevalence map for children aged 6-59 months according to DHS VI, Mali 2018.

Figure 10
Map of mortality of children less than 5 years old from DHS 2018 survey in Mali. The regions of Kidal and Gao and the District of Bamako showed the lowest mortality rates of 20%, 78% and 55% respectively.

Figure 11
Malaria vector distribution map based on data available from 2010 to 2019.

Figure 12
Map distribution of resistance of Anopheles gambiae s.l., a major malaria vector to different insecticides from 2010 to 2019 in Mali.

Figure 13
Malaria transmission zones in Mali by adjusted incidence and prevalence

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download.