Asymptomatic Plasmodium falciparum malaria prevalence among adolescents and adults in Malawi, 2015–2016

Malaria remains a significant cause of morbidity and mortality in Malawi, with an estimated 18–19% prevalence of Plasmodium falciparum in children 2–10 years in 2015–2016. While children report the highest rates of clinical disease, adults are thought to be an important reservoir to sustained transmission due to persistent asymptomatic infection. The 2015–2016 Malawi Demographic and Health Survey was a nationally representative household survey which collected dried blood spots from 15,125 asymptomatic individuals ages 15–54 between October 2015 and February 2016. We performed quantitative polymerase chain reaction on 7,393 samples, detecting an overall P. falciparum prevalence of 31.1% (SE = 1.1). Most infections (55.6%) had parasitemias ≤ 10 parasites/µL. While 66.2% of individuals lived in a household that owned a bed net, only 36.6% reported sleeping under a long-lasting insecticide-treated net (LLIN) the previous night. Protective factors included urbanicity, greater wealth, higher education, and lower environmental temperatures. Living in a household with a bed net (prevalence difference 0.02, 95% CI − 0.02 to 0.05) and sleeping under an LLIN (0.01; − 0.02 to 0.04) were not protective against infection. Our findings demonstrate a higher parasite prevalence in adults than published estimates among children. Understanding the prevalence and distribution of asymptomatic infection is essential for targeted interventions.

Spatial and ecological variables. De-identified 2015-2016 MDHS survey and geospatial data were linked to each sample's PCR results through random sample barcode. As part of the DHS methodology, GPS coordinates are collected in the field, marking the center of each cluster of households. The DHS program maintains participant confidentiality by displacing the GPS coordinates for all survey clusters: urban clusters are displaced up to a maximum of 2 km and rural clusters up to 5 km, with an additional 1% subset of rural clusters displaced up to 10 km 21 . PCR P. falciparum prevalence was mapped onto a smoothed surface using a constructed semivariogram and simple kriging to predict regional variation in malaria prevalence. Simple kriging assumes that observed malaria prevalence is spatially autocorrelated and that there is a known mean trend which is stationary across our study area.
Individual and cluster level risk factors were selected based on directed acyclic graphs and known associations from relevant literature [7][8][9]22,23 . Individual level factors included sex, age group, wealth quintile, education level, owning livestock, source of drinking water, living in a household with a bed net, sleeping under a long-lasting insecticide-treated net (LLIN), LLIN insecticide type, living in a household with at least 1 net per 1.8 household members, and anemia (women only). Cluster level covariates included region, urban/rural place of residence, elevation, month of data collection, landcover, the proportion of a cluster with bed nets, and the proportion of a cluster that slept under an LLIN. As transmission intensity is seasonal in Malawi, with peak transmission between January and May due to greater rainfall 11,24 , we also examined environmental variables at the cluster level including current month's average daily maximum temperature, and the prior month's precipitation. Modeled monthly precipitation and monthly average daily maximum temperature raster files, created from both in-situ weather station data and satellite imagery, were acquired from the Climate Hazards Center at the University of California, Santa Barbara 16,17 . Clusters were assigned precipitation and temperature values by averaging raster cells which fell within the 2 km and 10 km buffers surrounding urban and rural clusters, respectively, similar to DHS methodology for ecological variables 25 . Land cover estimates were obtained from satellite imagery classified by the Regional Center for Mapping of Resources for Development and SERVIR-Eastern and Southern Africa www.nature.com/scientificreports/ population. Community-level household bed net coverage and the community-level proportion who slept under an LLIN were not protective against P. falciparum infection in the general population. Sensitivity analyses defining malaria positivity as PCR amplification which crossed the threshold line below C T values of 37 and 38 found similar relationships between covariates and P. falciparum prevalence in bivariate and multivariate analyses (Supplementary Tables 5-8).       Table 2. Bivariate associations between demographic and environmental risk factors and P. falciparum prevalence using weighted survey data. LLIN long-lasting insecticide treated net.  www.nature.com/scientificreports/ Demographic and environmental risk factors associated with P. falciparum infection among adolescent and adults in our study resemble those previously found among asymptomatic children and individuals of all ages presenting with clinical symptoms; these risk factors in Malawi include low elevation, higher temperatures, younger age, rurality, and region [7][8][9] . Similar to the cross-sectional nationally representative Malawi Malaria Indicator Surveys (MIS) among children in 2014 and 2017 27,28 , the highest malaria prevalence was found in the Central region, followed by the Southern and Northern regions, although results are likely influenced by regional data collection during different months of the year (Supplementary Table 9). The national prevalence estimate of 31% in our study is comparable to results using similar methodology among adults from the Democratic Republic of the Congo in 2007 and 2014, which also found that younger age, male sex, and lower wealth indicators were risk factors for increased infection 29,30 . Our prevalence estimate is much higher than modeling predictions of P. falciparum annual parasite rates of 18-19% in children 2-10 years in 2015-2016, however these predictions were generated from multiple community-based survey measurements in the literature and from other unpublished sources which used RDT and microscopy as opposed to PCR 6 . While prevalence estimates were high, parasitemia values were low among our study population; however this finding is consistent with research showing that malaria infection is more likely to be submicroscopic among older children and adults versus younger children, hypothesized to stem from acquired immunity 31 . Risk factor analysis is important for identifying key populations to target for malaria prevention and control, and our results suggest that malaria prevention measures might be best focused towards men, younger age groups, poor communities, and rural areas.
Surprisingly, household ownership of bed nets, individual use of LLINs for sleeping, owning 1 net per 1.8 household members, and community bed net coverage were not associated with asymptomatic P. falciparum prevalence. While two-thirds of individuals resided in households which owned bed nets, only 36.6% reported sleeping under an LLIN the previous night, indicating low LLIN usage. Prevalence was higher among men as compared to women; females have been shown to clear asymptomatic P. falciparum infections faster than males, leading to the appearance of lower infectivity, and highlighting the importance of biological differences 32 . Other reasons for prevalence differences between males and females could relate to gender norms and behavioral factors around sleeping arrangements, gender-differentiated access to education about malaria or treatment and screening services, and gendered division of outdoor labor 33 .
Bed net ownership and use indicators were not associated with P. falciparum prevalence among our overall study population but insecticide type was found to have a near protective association among pregnant women. Pregnant women had similar patterns of sleeping under an LLIN the previous night as compared to the overall study population (38.7% vs. 36.6% respectively), however of those who slept under LLINs, 43.6% of pregnant women used nets treated with alpha-cypermethrin or deltamethrin, compared with 35.2% in the general population. Additionally, 49.0% of the nets used by pregnant women for sleeping were less than one year old, compared with 37.2% in the general population. While the general population receives bed nets every three years as part of regular mass distribution campaigns, Malawi's National Malaria Control Programme has given out free LLINs to pregnant women through antenatal care clinics since 2006 11 with 79-87% of pregnant women who attended antenatal clinics receiving LLINs in 2015-16 11 . Although WHO recommendations allow for up to three years between mass distribution campaigns 34 , field research in Benin, Malawi, Rwanda, Senegal, and Tanzania suggests that ITNs have a limited lifespan of two years before protection is compromised by holes, insecticide-resistance, and reduced concentrations of insecticide [35][36][37][38][39] . In Burkina Faso, a third of people had stopped using LLINs within a year of distribution 40 . Pregnant women in our study appear to have been the recipients of relatively newer nets which might have greater effectiveness, while households without pregnant women would have primarily received nets through Malawi's first national mass distribution campaign in 2012, or through limited follow-up mop-up campaigns in six districts in 2014 41 . In areas where nets are used at night, there is also evidence to suggest that long term use of insecticides inside the home can shift Anopheles spp. mosquitos from indoors to outdoors host seeking behavior, increasing the likelihood of an individual becoming infected, and perhaps contributing to reduced effectiveness of LLINs against malaria 42,43 . Bed net durability and mosquito biting behavior were not measured as part of the 2015-2016 MDHS, but could have contributed to the lack of association found between net ownership and use, and malaria prevalence among our study population.
The major strength of this study is the efficient use of a large number of nationally collected samples from adolescents and adults, a population which is understudied in malaria transmission research. Using molecular and epidemiologic methods, we better characterized the reservoir of asymptomatic P. falciparum, a pool of infection which is likely contributing to sustained transmission in Malawi. The results presented can serve as baseline assessment; repeated use of DHS samples over time can create a picture of shifting trends across the entire country while using resources cost-effectively to supplement intermittent MIS iterations estimating malaria prevalence among children. Characterizing the prevalence of P. falciparum among adults and identifying key target groups will be informative as the Malawi Ministry of Health designs future mass distribution campaigns and other interventions against malaria.
The primary limitation of this analysis is the cross-sectional nature of available samples. The 2015-2016 MDHS captured DBS from participants at a single time point, allowing for estimation of marginal associations between risk factors and infection to identify groups at high risk, but limiting evaluation of causal relationships. Additionally, the samples used in the study were collected during the 2015-2016 MDHS survey period from October to February, limiting inference to the remainder of the year and hindering comparison with published prevalence estimates among children from different time periods; however, as the study period occurred during the transition from dry to rainy season, we anticipate that results are somewhat representative of a yearly average. We used inverse probability of selection weights to make our results generalizable to the broader DHS cohort, but additional bias could still result due to unmeasured confounding. Our analysis was also constrained by the aggregated geographical classification of individuals. As part of DHS methodology, individuals are geolocated at the center of their study cluster, which is then displaced up to 5 km for 99% of rural clusters and 2 km for urban Scientific Reports | (2020) 10:18740 | https://doi.org/10.1038/s41598-020-75261-9 www.nature.com/scientificreports/ clusters. There is an inherent lack of precision in land cover, temperature, and precipitation data, nonetheless, our methods attempt to account for geographic displacement by using average values falling within a cluster's potential buffer area. Although our study measures presence of P. falciparum, we were not able to ascertain the presence of gametocytes within each infection, limiting the extent to which we can predict how these infections continue to sustain transmission. Results from elsewhere in Africa show that infection with P. falciparum gametocytes is associated with low asexual parasite densities and asymptomatic disease 3 . Comparisons between microscopy and PCR have found that microscopy can miss over 90% of gametocyte carriers due to limited sensitivity for low-density infections 44 , further highlighting the importance of molecular surveillance tools in understanding infection transmission dynamics in this population. Despite existing limitations, this analysis provides valuable input into an understudied yet critical group to consider in efforts to interrupt ongoing malaria transmission. Malawi spent an estimated $82 million on malaria control in 2016 45 and malaria accounts for 30% of all outpatient visits and 34% of inpatient hospital admissions 11 . Households often amass high direct and indirect costs due to clinical malarial disease, despite free diagnosis and treatment 46 . One of the primary goals of the Malawi Malaria Strategic Plan 2017-2022 is to achieve universal LLIN coverage for all households 11 . Research from Madagascar shows that while LLIN mass distribution campaigns may only provide community protection for one year, protection can be sustained when campaigns are followed by continuous LLIN distribution to eligible households, including recently married couples, immigrants, children of vaccination age, and homes with uncovered sleeping areas 47 . Malawi could benefit from education on consistent and correct use of bed nets, and through expansion of continuous LLIN distribution services to additional populations beyond pregnant women, targeting younger individuals living in rural areas with high prevalence of infection for more frequent net replacement. Treating malaria at the population level through mass drug administration can clear parasite presence and prevent transmission of gametocytes from asymptomatic infections. However, mass drug administration is only recommended in settings considering malaria elimination, and requires low malaria prevalence, effective vector control, access to treatment, and extensive community participation as prerequisites for implementation 48,49 .
This study presents unique insight into the national prevalence of asymptomatic P. falciparum infection among adolescent and adults in Malawi. Use of molecular and epidemiological surveillance methods in tandem demonstrates that demographic and environmental risk factors for infection parallel those found in children and among individuals with symptomatic disease. Within the current framework of mass distribution frequency and community education, presence of bed nets in the household and use of LLINs by the individual did not appear to provide protective benefits, regardless of insecticide type, most likely due to bed net age and low frequency of use. Results from this study provide valuable guidance to decision makers in Malawi as the National Malaria Control Programme designs bed net distribution programs following mid-term review of the 2017-2022 National Malaria Control Strategy. Future work to replicate this analysis following the 2021-2022 MDHS will enable assessment of changes in asymptomatic P. falciparum prevalence and other genetic markers in adolescents and adults across time.

Data availability
All relevant data are within the manuscript and supplement. Data that support study findings are available for download from the DHS MEASURE website, conditional on approval from DHS. Laboratory testing data are available from the authors upon reasonable request.