Pregnant women and infants as sentinel populations to monitor prevalence of malaria: results of pilot study in Lake Zone of Tanzania

Populations living in sub-Saharan Africa are at greatest risk for malaria, where approximately 90 % of deaths are estimated to occur, with 70 % of deaths occurring in children under 5 years of age [1]. Tanzania has a substantial burden of malaria. Among children under-five, recent population-based, cross-sectional surveys in 2011/12 documented an overall national prevalence of malaria parasitaemia of 9.2 %, ranging from 0 to 31.8 % across the administrative regions [2]. Malaria parasitaemia is similarly prevalent among pregnant women, ranging from 5.5 to 11.5 % [3, 4].

Over the last decade, Tanzania has gradually scaled-up malaria control interventions. Insecticide-treated nets (ITNs) have been distributed through various campaigns, including: the Tanzania National Voucher Scheme introduced in 2004 targeting vulnerable groups, such as pregnant women and children aged under five [5]; the children under five catch-up campaign (U5CC) in 2009 [6]; and, the universal coverage campaign (UCC) in 2011 [7]. Artemisinin-based combination therapy (ACT) was introduced as first-line treatment in all public facilities in 2006 and malaria rapid diagnostic tests (RDTs) were introduced in 2009 [8]. Indoor residual spraying (IRS) was introduced in two districts of Kagera Region in 2007 and expanded to 18 districts in the Lake Zone by 2011 [8], where malaria prevalence measured by RDT in children aged 6–59 months was higher (30–41 %) than the rest of mainland Tanzania (18 %) in 2007–2008 [9]. Following the effective scale-up of multiple malaria control interventions, there has been a 45 % reduction in all-cause under-five mortality from 1999 to 2010 [8]. In addition, the Tanzania HIV and Malaria Indicator Survey (THMIS) 2011–12 found that malaria prevalence among children aged 6–59 months had declined to 9 % nationally, and 8, 19 and 26 % in Kagera, Mwanza and Mara Regions, respectively [2]. As malaria control interventions are scaled-up, rational approaches are needed for monitoring impact over time. The 2008–13 National Malaria Control Programme (NMCP) Monitoring and Evaluation Plan proposed testing of pregnant women and children under 5 years of age attending reproductive and child health (RCH) clinics as a suitable sentinel population for monitoring longitudinal malaria morbidity trends [10]. This study piloted an active screening approach to monitor malaria parasitaemia prevalence in two sentinel populations: (1) pregnant women attending their first ANC visit; and, (2) infants eligible for measles vaccination, at 9–12 months of age attending selected RCH clinics in Kagera, Mara and Mwanza Regions, as well as to assess the costs of implementing this strategy.

In an area of Tanzania with high malaria transmission where intensive vector control interventions (IRS and ITNs) were scaled up and complemented by universal access to ACT and RDT, nearly one in eight pregnant women and one in ten infants attending RCH clinics were positive for malaria infection. In general, the trends over time and space were similar for the two populations. Mapping the prevalence by location revealed the presence of localized high transmission hotspots for both the pregnant women and infants. While malaria test positivity rates from the OPD provide a source of data, which could be used to identify malaria hotspots, these data are influenced by other causes of fever. This study piloted a new approach for assessing population-based prevalence of malaria through targeted screening of readily accessible, asymptomatic populations (pregnant women and infants) that have high attendance at RCH clinics (95.9 % of pregnant women attend at least one ANC visit and 74.6 % of infants attend for measles vaccination) [19]. Compared to test positivity rate data obtained from the OPD, testing of a sentinel population (including both symptomatic and asymptomatic individuals) provides a more stable estimate over time, which is unaffected by other circulating illnesses. This data thus provides a better measure for monitoring disease trends over time. Cross-sectional household surveys are typically used to provide this data, however, these are costly and are performed at widely spaced intervals, which does not allow for fine scale tuning of intervention delivery in the interim. Ongoing, routine screening of readily accessible populations may offer a practical strategy for continuous monitoring to identify malaria hotspots and track the progress of malaria control over time. These data complement the data from the OPD, thus combining these sources could provide a more comprehensive understanding of the full epidemiologic picture.

In Tanzania, sentinel population testing of pregnant women and children under 5 years of age attending RCH clinics was initially proposed in the 2008–13 Malaria Monitoring and Evaluation Plan [10]. Following the pilot study, the 2014–2020 National Malaria Strategic Plan set a strategic intervention to establish countrywide longitudinal vigilance of malaria parasitaemia in sentinel populations: pregnant women and infants at RCH clinics, and school-age children [11]. While data obtained from these sentinel populations would not obviate the need for periodic cross-sectional population-based surveys to measure coverage of interventions, the available data suggest that the prevalence of anaemia and parasitaemia among children presenting to the health facility for routine immunization is correlated with that among children aged 6–30 months detected by household surveys [11]. Similarly, data from a recent meta-analysis by van Eijk found that the prevalence of malaria among pregnant women was strongly correlated with that from household surveys of children [20], suggesting that either of these populations could be used to provide ongoing information about community level prevalence. From a biological standpoint, monitoring of infants may be preferable to monitoring adult women. The prevalence among infants nine to 12 months of age should closely mirror that of all children 6–59 months. Since infants have not yet developed significant immunity, these infections are likely to have been recently acquired. Furthermore, infants are less likely to have travelled than pregnant women. Thus, the prevalence of infection among infants may be more likely to represent recent and local transmission dynamics than that in adult women. However, pregnant women are more likely to be infected with malaria than non-pregnant women, and are at greater risk for severe disease than non-pregnant women [21]. This is particularly true for primigravid women, who are at increased risk of severe disease, with higher risk of severe anaemia and maternal death, higher rates of miscarriage, intrauterine demise, premature delivery, low-birth-weight neonates, and neonatal death compared to multigravid women [22]. Identifying and treating infected pregnant women with asymptomatic infections who may otherwise have gone untreated may have the added benefit of improving birth outcomes [23]. In addition, this may have a benefit on transmission, as it has been suggested that pregnant women may be a reservoir of transmission [24]. Finally, during the first ante-natal care visit, a panel of blood tests is already performed, whereas children presenting for measles immunization would not routinely have a blood draw. For these reasons, in Tanzania, testing of pregnant women was felt to be a better choice for routine monitoring than infants, and in 2015, Tanzania started implementation of routine testing of pregnant women for malaria at first ANC as part of the antenatal profile [25]. In addition, recording and reporting of ANC malaria screening results has been integrated into the routine health management information system (HMIS).

The study estimated the direct costs associated with the sentinel population screening pilot. The costs included training, staff time required to perform and record the test, test kits, and travel expenses for monitoring and supervision. These estimates suggest that nearly three-quarters of direct costs were attributable to purchase and delivery costs for RDTs. Nonetheless, reporting of ANC malaria testing through routine HMIS, and integration of RDT quality assurance monitoring and supervision with other routine supervisory activities, that are undertaken on a quarterly basis by the district teams, would reduce these costs.

This study has a number of potential limitations. The low rate of testing in Kagera and Mwanza is a potential source of bias and may have underestimated the results. The low testing rates were largely attributed to stock-outs of RDTs during the study period. Several reasons were reported by RCH clinics to explain frequent RDT stock-outs, including: weak quantification and forecasting by health facilities, delayed delivery of RDTs to health facilities by MSD, and stock out at MSD central stores. The analysis suggests that malaria prevalence was not correlated with the proportion of participants tested; thus the low testing rate is not expected to have biased the results. Lack of reporting was partly related to testing activity because RCH clinics did not submit any reports where testing was absent due to RDT stock-outs. It is likely that the low reporting rates, especially among infants in Mwanza Region, may have biased the results. In this study, participating health facilities were purposively selected to provide a geographically spread sample, but were not probabilistically selected to support parametric statistics and generalizable estimates. While this may be a potential source of bias, sampling is not likely to have influenced reporting rates and RDT stock-outs. Finally, history of recent symptoms for the persons tested was not recorded, as it was not part of the routine ANC register, making it impossible to know if those who tested positive had fever or other malaria symptoms in recent days (but were asymptomatic at time of the RCH visit). Finally, there is evidence to suggest that the HRP-2/pLDH RDTs are not sufficiently sensitive to diagnose very low density malaria infections in asymptomatic pregnant women [26], thus some infections may be missed. Nonetheless, if this screening activity also worked to detect very early cases of uncomplicated malaria, then this approach could benefit low transmission areas where additional efforts are needed to reinforce passive case detection.

Routine screening of pregnant women at first ANC and infants presenting for measles vaccination may offer a practical approach for ongoing monitoring to track the progress of malaria control over time. This strategy provides benefit to the tested individual by identifying and treating those with asymptomatic but detectable parasitaemia. In addition, the continuous data generated down to the sub-district level allows for targeting malaria control interventions according to the level of transmission. Of the two groups, testing of pregnant women is likely to be less disruptive to clinic flow, as pregnant women already have blood drawn for routine ANC screening, but infants may provide a more precise estimate of transmission variability across sites and over time. Malaria screening of pregnant women at first ANC should be considered as a practical method for continuous routine surveillance of malaria prevalence at the sub-district level.