In 2020 the Sysmex XN-31 automated malaria detection analyser was evaluated at the MBTS, utilising samples from 5281 blood donors, collected nationally from 24 of the 28 administrative districts in Malawi . In this current study, an in-depth analysis of the same dataset was conducted utilizing donor demographic and blood donation site location to explore the feasibility of using blood donor malaria screening as a complementary source of malaria surveillance data.
For malaria elimination to succeed, it is essential that countries generate and use high quality surveillance data to accurately define malaria incidence over time and space . Depending on availability of quality data, the WHO uses two different methods to estimate malaria burden for its annual World Malaria report . The first method estimates malaria incidence from reports of malaria cases compiled by national health ministries. These data are then statistically adjusted for reporting completeness, positivity rate among suspected cases, and the rate at which febrile patients seek public healthcare. However, in many malaria-endemic countries (about 30 Sub-Saharan African countries), precise estimation of malaria burden is hindered by data scarcity due to weak surveillance systems . As a result, the WHO has resorted to geospatial epidemiological models to estimate malaria burden in these countries . These models rely on asymptomatic parasite infection prevalence data collected largely from children under the age of 5 years, via infrequent, costly, and logistically complex surveys.
The most recent Malawi MIS took place in 2021, and prior to that in 2017 , and 2014 . The MIS tests children aged 6–59 months for anaemia and malaria parasitaemia. Malaria prevalence in this age group was 10.5% in 2021 (unpublished 2021 Malawi Malaria Indicator Survey), a significant decline from 24% in 2017. Many individuals, including children, as demonstrated in these household surveys, are parasitaemic without showing any signs of infection. Besides placing infected individuals at risk of developing anaemia and associated morbidity, asymptomatic infections play an integral role in sustaining transmission in regions where malaria is endemic as they largely outnumber symptomatic infections . Systematic reviews of studies conducted in malaria-endemic countries have shown that asymptomatic malaria is common in healthy adult blood donors, and that the prevalence of malaria-infected donors generally mirrors the overall malaria burden of the region [8, 20].
The overall donor malaria prevalence rate in this study was 11.6%, increasing progressively from December (8.6%) to April (18.3%) (Fig. 2a). This seasonal trend is in line with what has been previously reported to be driven largely by the annual rains that typically begin in November–December and last through March–April in most parts of the country . However, when viewed by weather zones (Fig. 3), this trend was only observed for the high prevalence Central (17.1%) and Lakeshore (27.8%) zones, with the latter having the highest prevalence rate in Malawi, which has been consistently reported [22, 23]. In contrast, the low prevalence Southern Highlands (6.1%) zone, showed the opposite, with a consistent downward trend with the lowest donor malaria prevalence in April, whereas the Northern (7%) and Shire Valley (12%) zones had variable patterns. These data are consistent with those of others [17, 18, 24] with the Central administrative region having the highest and the Northern region the lowest malaria prevalence rates (Fig. 1b), but the same does not hold true for the Southern region. This region had a malaria donor prevalence below that of the Northern region in contrast to that of the 2015–2016 Malawi Demographic and Health Survey (which assessed parasitaemia using PCR in adolescent and adult males and females)  and the 2017 MIS , which reported high malaria prevalence rates for the Southern region, close or equal to that of the Central region. Whilst the absolute prevalence rates of the current study and these two Malawian surveys [17, 24] are not comparable, because of temporal (year and month/s of collection) and test group differences (children versus adults) and differing sensitivities of diagnostic test used for parasitaemia assessment (PCR > XN-31 > microscopy/rapid diagnostic test), the patterns observed are informative, and reflect similar trends. As these study data were obtained 3 years after the 2017 Malawi MIS, it is proposed that these changes reflect rapidly changing malaria dynamics in Malawi , rather than unreliable data. When comparing the 2014  and 2017 Malawi MIS data , the overall prevalence of malaria in the under 5-year age group dropped from 33 to 24%, but the rate of change was highly variable amongst the regions. The Northern region had a 68% decline (29% to 11%), in contrast to more modest reductions of 28% and 21% respectively for the Central (36% to 26%) and Southern (33% to 26%) regions. It is thus conceivable that the low malaria prevalence finding in this study for the Southern region (7.9%), more in keeping with the Northern (8.6%) rather than Central region (18.5%), reflect a regional acceleration of decline here, in line with what was observed for the Northern region between 2014 and 2017 [17, 18]. This view is supported by the preliminary data from the still unpublished 2021 Malawi MIS report which indicate a significant decline in malaria parasitaemia in the 6–59 month age group, with the Southern region showing the greatest reduction from 26% to 10% (62% decline), and overall from 24% to 10.5%, in the 4 year period since the last survey. The 2021 MIS overall prevalence rate of 10.5% is not dissimilar to the 11.6% blood donor malaria parasitaemia prevalence in the current study (where 85.6% of samples were collected in 2020, and the remainder in December 2019).
These observations highlight that the episodic nature and short duration (1–2 months) of data collection utilizing epidemiological household surveys, and the delay in such results becoming available, makes accurate tracking of rapid, year-to-year changes in malaria burden a difficult task .
The current study data concur with the generally reported higher malaria risk associated with non-urban areas [26, 27]. They are also in keeping with previous Malawi survey findings [17, 24], notwithstanding that weighting was not applied and the definition of urban versus non-urban was crudely based on blood donation site location and not place of residence, and cognisant that non-urban versus urban malaria risk association in Malawi is more complex than simply considering physical location .
The study blood donor pool was 80.2% male with 95% of donors between 16 and 47 years. Female donors were slightly younger (median age 19 years) compared with males (median age 21 years). Urban donors comprised 45.9% of the pool with an even stronger male representation (90.2%), and significantly (p < 0.001) older (median age 25 years) compared with non-urban donors (median age 20 years). This urban male dominance and older age, accounted for the age difference between male and female donors overall. Parasitaemia rates were significantly (p < 0.001) and consistently higher in male donors (13.1%) compared with female donors (7.4%), irrespective of location or month of collection. This male bias of parasitologically confirmed malaria infection prevalence is in keeping with several other studies conducted in Africa [24, 27, 29, 30]. Interestingly, different temporal trends for male and female donor malaria prevalence were observed, with males mirroring the generally reported seasonal increase from November to April and female donors having a more stable malaria prevalence over time (Fig. 5). A study involving intensive follow-up of P. falciparum infections in Eastern Uganda , utilizing ultrasensitive molecular testing to distinguish new from persistent infections, demonstrated that whilst the incidence of new infections was the same in both sexes, males had a slower clearance of asymptomatic infections attributed to biological sex-based differences as no behavioural differences were recorded. Behavioural differences that lead to increased exposure to mosquito bites in men can however not be excluded in the current study. In contrast to asymptomatic parasitaemia, the incidence of clinical malaria appears to be higher in females  with pregnancy-associated loss of malarial immunity  and increased health-seeking behaviour in females  playing a contributory role. It is thus speculated that the observed increasing prevalence of parasitaemia in male donors only could be explained by the slower clearance of parasites in males compounded by the growing incidence of new infections as the seasonal changes become more conducive to promoting mosquito breeding sites and thus increased malaria transmission. Also, the reported increased incidence in symptomatic infections in females , would result in self-deferral from blood donation or rejection through health questionnaire screening at the donation sites. Furthermore, the enhanced health-seeking behaviour of females  would increase treatment-induced clearance of parasites in those diagnosed with malaria. Taken together with the biologically faster parasite clearance in females , these factors may be a plausible explanation for the relatively stable malaria parasitaemia prevalence observed over time in the female donors in this study. The peak in April, observed also in the female donors, may well be the compounded effect of a marked increase in incidence driven by rainfall and humidity. Whilst parasite rates differed amongst the sexes, the parasite burden (MI-RBC count) in infected donors did not reach significance (p = 0.7933) in keeping with what has been reported elsewhere .
For surveillance data to be translated into meaningful interventions and to monitor their impact, it needs to be representative of the community at risk of malaria. Sample size determination and sampling for a typical MIS is a highly complex process, and whilst the bigger the better, it is a balancing act between the demands of the analysis, implementation team capacity and budget constraints . The 2017 Malawi MIS  sampled 3750 households, representing 0.13% of total households, and tested 2031 children under the age of 5, representing 0.08% of children in this age group, based on the 2018 Malawi population and housing census data . In the current study 5281 blood donors (aged 15–64) were tested, representing 0.06% of this age group (using the same census data) for both sexes combined, and 0.1% and 0.02% males and females respectively. There was also a reasonable match between donor distribution by region and population density. The split of donors by the Northern, Central and Southern regions gave a proportion of 9%, 39% and 52% respectively, compared with 13%, 43% and 44% for the total population of the respective regions. The disproportionate representation of the regions is best explained by the non-random nature of sampling due to preferential inclusion of blood donor samples from low workload days, as the original study was not designed with a surveillance data collection objective in mind. This may also account for the proportional over-representation of urban donors relative to the urban population size (urban 0.13%, non-urban 0.04%). Only ~ 25% of the total number of donations received during the study period were included, as study-related sample processing was restricted to low workload days. If XN-31 blood donor screening had been integrated into MBTS in 2018 a total of 66931 [34,35,36,37] individual donor blood samples would have been screened for malaria, which represents 0.73% of the total population aged 15 to 64 years. Stratified by sex and urban/non-urban location utilizing the current study donor pool distribution, all sub-groups would have greater representativity (ranging from 2.97% for urban males to 0.23% for non-urban females) than the MIS. In this regard routine malaria screening of blood donors would serve as an excellent data source to complement traditional household surveys. Notably, the XN-31 is a high throughput automated analyser which could be easily integrated into blood transfusion services for the testing of subjectively healthy asymptomatic adolescents and adults to provide real-time information on the trends of the size, location, age, and gender distribution of the asymptomatic malaria parasite reservoir to complement existing malaria surveillance activities. Furthermore, M’baya et al.  have demonstrated that XN-31 has superior sensitivity compared to routine microscopy for malaria screening in blood donors. As screening and data collection would be a continuous activity, the analysis of trends which would have a comparable baseline, more so than absolute values, would be of most value to inform malaria control interventions. Furthermore, as blood transfusion is a core life-saving intervention globally, blood donation is a vital activity undertaken in all communities. The source of “survey” participants is thus self-selected and in continuous supply. Notably also, blood donor surveillance data would be highly representative of males, a group currently excluded from traditional household MIS, yet with their greater persistence of parasitaemia and thus higher parasite rates, are an important source of ongoing malaria transmission. Recognizing both the value and limitations of periodic household surveys, the need for complementary data, for example school-based surveys  or from antenatal clinics , has been highlighted by others. School-based surveys, whilst targeting a unique population would still be limited by their episodic nature, whereas antenatal clinic screening may provide continuous data.
Besides providing quantitative parasite counts, the XN-31 analyser also provides information on the presence of gametocytes, as a research-use-only parameter. Gametocytes were observed equally in malaria-infected male and female donors, with the parasite burden (MI-RBC count) being significantly higher in gametocyte carriers compared with malaria-infected donors where gametocytes were not detected (Fig. 6) in line with what has been previously reported [40, 41]. Interestingly, donors in whom only gametocytes were detected had significantly lower MI-RBC counts than those with ring forms and gametocytes. The gametocyte carriage, as a percentage of malaria-infected asymptomatic donors, was not statistically different between high and low prevalence regions. Monitoring trends of gametocyte carriage is an important component of surveillance as it provides insights into the pool of individuals capable of onward transmission .
This study has limitations. Firstly, the original study was designed exclusively to confirm analytical performance of the XN-31 analyser as a malaria diagnostic test, hence the focus was on inclusion of a sufficient number of blood donors over a certain time-period with no regard for proportional representativity of donors based on districts (blood collection site), urban versus non-urban location nor month of collection. Consequently, reliable parasitaemia prevalence rates could not be established for administrative districts with too few or no donor inclusions in this study nor could weighting be applied to adjust for this disproportional representation. Also, only donor samples received on low workload days (~ 25% of total donations during the study period), were included, meaning that the study sample composition may differ from the complete donor pool. Furthermore, the data collection did not span a full year hence it does not cover the full seasonal range of malaria transmission. Topazian et al. , using quantitative PCR testing, showed that 55.6% of malaria-infected individuals identified in the 2015–2016 Demographic Health Survey had parasitaemia of ≤ 10 p/µL. It is thus acknowledged that a substantial number of malaria-infected blood donors would have very low level parasitaemia falling below the limit of detection of the XN-31 (20 p/µL). PCR is however not an accessible diagnostic tool, because of cost and complexity, and thus rapid diagnostic tests and microscopy are widely utilized for malaria surveillance parasite screening. As shown in the original blood donor screening study at MBTS , XN-31 detected almost twice as many malaria-infected donors compared with microscopy. In this regard, although lacking the sensitivity of PCR, the XN-31, due to its automated nature, superior sensitivity to routine microscopy, and ease of use, is an excellent diagnostic tool for malaria screening within blood transfusion services, and thus ideal for generating complementary malaria surveillance data, where the focus is on the observation of trends. In addition, the populations surveyed in this study (blood donors) and in the MIS (children under the age of 5 years) are different, and thus not directly comparable. However, malaria screening of blood donors brings a different dimension to malaria surveillance since it targets the detection of the asymptomatic reservoir of parasites in adults, an important metric for malaria control strategies. Furthermore, blood donation is an ongoing activity, and thus implementation of XN-31 malaria screening of donors, would generate data continuously in real-time, allowing the rapid detection of changing trends which is not possible with episodic household surveys. Also, as blood donors are recruited across a wide geographic location, this facilitates a high spatial resolution which is currently only obtainable from clinical case reporting. Thus, XN-31 blood donor screening would provide an easily accessible source of complementary malaria surveillance data, since this secondary data analysis has shown that the malaria prevalence in blood donors mirrors the patterns and trends detected by other means of surveillance in Malawi.