Study Sites
This study covered eight districts from four regions; Geita, Kigoma, Mtwara and Ruvuma regions (Fig. 1), with persistently high malaria transmission as shown by different surveys conducted in Tanzania between 2011and 2017. Such surveys include the Tanzania HIV and Malaria indicator survey (THMIS2011-12)(32), Tanzania demographic and health survey – Malaria Indicator survey (TDHS-MIS 2015–2016)(34) and the 2015/2016 school malaria prevalence surveys (SMPS)(35)( Figs. 2 and 3).
Based on the 2011-12 THMIS, the highest malaria prevalence (by malaria Rapid Diagnostic Test(mRDT)) was observed in the Southern (20.6%), Western (16.7%) and Lake (14.8%) zones(2). A similar pattern was observed 3 years later during the 2015-16 TDHS-MIS which showed that the highest prevalence was within the same zones; Western (27.7%), Lake (23.5%) and Southern (18.8%)(3). Regions with high (> 10%) prevalence during the 2011-12 survey included (in descending order); Geita (31.8%), Lindi (26.3%), Kigoma (26.0%), Mara (25.4%), Mwanza (18.6%), Mtwara (17.4%), Morogoro (13.0%) and Ruvuma (12.0%)(2). The same set of regions presented high prevalence in 2015-16 in additional to Kagera, Pwani, Tabora, Shinyanga, Mwanza, Katavi and Simiyu(3). The results of SMPS brought a similar set of zones and regions affected mostly with malaria. Regions with the highest prevalence included Geita (53.7%), Pwani (48.4%), Mwanza (40.0%), Katavi (39.9%), Mara (36.4%), Mtwara (36.2%), Shinyanga (35.1%), Kagera (31.1%), Simiyu (31.0%), Lindi (30.3%), Kigoma (30.3%), Tabora (30.0%) and Ruvuma (22.6%)(35). The findings of these surveys indicated that some regions consistently maintained high malaria transmission in the past 5 years, including Geita (Lake zone), Ruvuma and Mtwara (Southern zone) and Kigoma (Western zone) (Fig. 1).
Thus, four regions were selected for the study and two districts with high and low malaria prevalence based on the SMPS of 2015/2016 were selected from each region(Fig. 1); Buhigwe and Uvinza (Kigoma), Mtwara DC and Nanyumbu (Mtwara), Nyasa and Tunduru (Ruvuma) and Nyang’hwale and Chato (Geita). Districts in Geita region were chosen with an added criterion of taking part in IRS programme which was done shortly before this study. Two villages with high malaria burden, based on the number of cases reported by health facilities in 2016, were selected from each district, making 16 villages. Health facilities serving the respective villages were also visited for the health system survey.
Study design
This was a cross-sectional survey which evaluated entomological, parasitological, health systems, socio-economic and socio-anthropological components (Fig. 4).
Plannning and organization of the study teams
Training of the teams on the study protocol and the tools was done before initiation of the study; and was done separately according to the component to be implemented. Dry runs were conducted in Magoda village in Muheza district, Tanga region. Community members of Magoda village have experience of taking part in different surveys of malaria transmission for the past three decades (37). The demographic and socio-economic (census) team conducted first the dryrun, other teams conducted the dry runs one week later using the data generated by the census team. The exercise was followed-up by a post-test assessment to sort out any issues encountered during the dry runs before data collection was launched. The main study was conducted between July and November 2017 following a similar order. In each village, the teams conducted the survey in 3 to 4 days and the entire survey took about 65 days.
Sample size, sampling procedures, and collection of data and samples
To create a sampling frame, a demographic census was undertaken and all households (HHs) from the selected villages were enumerated, and each member of the HH was given a unique identification number. Systematic sampling was used to identify HHs to be included in other components of the study as described below.
Demographic and socio-economic survey
A demographic census was undertaken in the selected villages whereby all HHs (between 200 to 700HHs per village) were enumerated and assessed for socio-economic status (SES) and risk factors associated with malaria transmission. Demographic data including name, sex, date of birth, education status and occupation were collected for all members residing in each HH. Malaria risk mapping was undertaken, and it included collection of data on the housing quality based on construction materials, presence of eaves and open windows, location of the HH in the village and environmental features associated with malaria transmission (e.g. mosquito breeding sites). Furthermore, information on malaria prevention strategies [e.g. screened windows, availability and use of bed nets (ITNs and LLINs), IRS and intermittent presumptive treatment in pregnancy using SP (IPTp-SP)] were collected. The SES of each HH was also assessed and it focused on ownership of assetssuch as livestock, farming land, electronic equipment (e.g. mobile phones, radio, television), means of transport (e.g. bicycle, motorcycle), source of energy for cooking and lighting, source of drinking water and type of latrines.
Parasitological survey
A total of 120 HHs from each village were sampled from the census list and all members from the selected HHs were enrolled in the parasitological survey that was conducted at a central point. A comprehensive parasitological assessment was conducted to obtain baseline data on malaria transmission intensity in the study population. A parasitological questionnaire was used to collect information on demographics, malaria prevention, clinical and treatment history. Additionally, vital signs and anthropometrics measures were taken, splenomegaly was evaluated for children up to 12-years. Blood samples were collected by finger prick, thin and thick smears were prepared, malaria rapid test (mRDT) (Carestart™ Somerset, NJ USA) was done for detection of malaria parasites and haemoglobin levels were evaluated using HaemoCue® machines. Dried blood spots (DBS) on filter papers were collected for further analysis of malaria parasites and host genetic factors that might affect susceptibility/resistance to malaria parasite infection and clinical disease. Participants with mRDT positive results were treated according to the National malaria treatment guidelines(38) and participants thought to need further investigation and management were referred to the nearest health facility.
Socio-anthropological survey
A structured questionnaire was used to collect quantitative data while qualitative data was collected through focus group discussions (FGDs) and in-depth interviews (IDIs) using interview topic guides. Heads of the HHs which were involved in the parasitological survey were interviewed to capture information on knowledge, attitude, practices and beliefs (KAP&B) towards malaria and its control. Specifically, respondents provided information on key issues such as malaria transmission, signs, symptoms and treatment; malaria prevention (e.g. use of bed-nets and chemoprevention such as IPTp) and health seeking behaviour. For the qualitative study, two FGDs were conducted in each village, whereby each FGD included 8–12 individuals aged > 18 years (male and female participants separately). The information collected during FGDs included; knowledge on malaria (its transmission, signs, symptoms and treatment), community’s practices and attitudes on malaria prevention, e.g. use of bed-nets, use of antimalarial drugs and health seeking behaviour. Participants of the FGDs were randomly selected from HHs which were not involved in the parasitological survey and KAP&B interviews. Indepth interviews were also undertaken with at least two officials involved in malaria control at the district level, including the district medical officer and/or district-malaria coordinators/focal persons.Repondents provided information on availability, accessibility and utilization of malaria prevention, treatment and control services in their areas.
Entomological survey
A sub-sample of 25 HHs were selected randomly (from the census list) in each of the study villages for entomological assessments. The selected HHs had open eves, were close to the breeding sites and with medium size windows and/or with no mosquito netting material. Among the 25 HHs; 20, 10 and 5 HHs were randomly selected for collection of indoor host seeking malaria vectors using the Centres for Disease Control (CDC) light traps(John W Hock Co, Gainesville, FL, USA), indoor resting malaria vectors using pyrethrum spray catch and outdoor biting malaria vectors using tent traps, respectively. Assessment of immature mosquitoes was also done through larvae search. Details of each collection method used are provided below:
Light Trapping: After obtaining a written informed consent from the homeowner, a CDC light trap was installed at the foot of a sleeping place occupied by a family member sleeping under an insecticide-treated bed-net as described previously(39). In brief, trapping was done in 20 HHs from each village for three consecutive nights and the traps were set between 18.00 and 19.00 hours, and retrieved the following morning between 06:00 and 07:00 hours.
Pyrethrum Spray Catches
Spray catch collection was done for one day in 10 HHs from each study village. The HHs used here were a sub-sample of the same HHs used for CDC light trapping. Mosquitoes were collected by spraying a non-residual crude pyrethrum solution premixed with kerosene at the concentration of 0.5%. Food and small items were removed from the houses and white cotton sheets were spread to cover the floor and large furniture. Ten to fifteen minutes after spraying, researchers collected knocked down mosquitoes and preserved them in petri-dishes for later identification and score of their physiological status.
Furvela Tent Trapping: These were used for three nights consecutively in 5 HHs from each village. Furvela tent traps consisted of a tent and a CDC light trap with the light bulb removed. A volunteer slept in the tent with a small opening in the tent door to allow host odours to escape and attract mosquitoes as previously described (40). The traps were set each night between 18:00 and 19:00 hours and retrieved the following morning between 06:00 and 07:00 hours.
Larval Surveillance
Mosquito larval breeding habitats were mapped using global positioning system (GPS) in Open data kit (ODK) (GPS; Trimble Geoexplorer II, USA) while Larvae search was conducted using the World Health Organization (WHO) 350 ml standard mosquito dipper. Depending on the size of the aquatic habitat, five to twenty dips were taken from each larval habitat after physically inspecting the habitat for the presence or absence of mosquito larvae. Five dips were taken in small larval habitats of ≤ 1 m2, 10 dips for medium-sized habitats (2–15 m2) and20 dips for relatively large habitats (> 15 m2). The immature mosquitoes were classified into three categories as; early instars (first and second stage larvae), late instars (third and fourth stage larvae) and pupae, after which they were sorted by genus, counted and recorded.
Mosquito identification
Mosquitoes collected were transported to the field laboratory which was set up in each village (within a school or village administration office)and identified to species using morphological criteria (41). Malaria vectors were further classified according to their abdominal status as unfed, blood-fed, semi-gravid or gravid. Female mosquitoes (Anopheles gambiae complex and An. funestus group) were stored in individual Eppendorf tubes containing silica gel desiccant for sibling species identification, detection of sporozoite and molecular markers of insecticide resistance. All non-malaria mosquitoes were identified, counted and discarded.
Health system survey
The health system(HS) assessment survey targeted all health facilities serving the study villages including Accredited Drug Dispensing Outlets (ADDO), dispensaries, health centres and district hospitals where village members were getting initial consultations and/or refferal services. The facilities include those located within the study villages or nearby. The assessment was based on the WHO recommended guidelines on service provision assessment for malaria (SPAM)(42) which collects information on availability, accessibility, affordability and quality of malaria prevention and case management services. Data collection tools, adopted from the WHO guidelines, were used for this assessment. These included, Health facility Profile, Consistency check, Antenatal clinic (ANC), Out-patient department (OPD), Inpatient department (IPD), Laboratory and Pharmacy questionnaires. Data were collected from the facilities through observations and documentary review, health workers interview and exit interviews with clients. A critical assessment of different units/sections and departments at the health facilities (including outpatient and inpatient departments, laboratory and reproductive child health – RCH units) and within ADDOs was done. Health workers interviewed included the facility in-charges and heads of units while exit interviews were held with patients’/care-givers who were attended at the mentioned units on the day of the visit (interviewees were selected based on history of fever or diagnosis of malaria). In summary, observations were done for 12 patients attended at each facility including;5 malaria patients from outpatient department, 5 from RCH and 2 patients who received artesunate injection for treatment of severe/complicated malaria.
Laboratory Analyses
Detection of malaria parasites by microscopyand Polymerase chain reaction(PCR)
Blood smears and DBS sample were sent to laboratory in Tanga for further analysis. The blood slides were dried at room temperature, thin film fixed with 99% Methanol and the smears were stained with 5% Giemsa solution for 45 minutes the following day while in the field. Stained blood slides were then packed and shipped to the laboratory in Tanga where the smears were examined under high power objective (oil immersion objective) for identification and quantification of malaria parasites. Asexual and sexual parasites were counted against 200 and 500 white blood cells (WBCs), respectively. Parasite density was obtained by multiplying the counts by 40 or 16 for asexual and sexual parasites, respectively; assuming each micro litre of blood contained 8000 WBCs(43). A smear was declared negative after examining 200 high power fields.
DNA was extracted from all microscopy positive DBS samples using QIAamp DNA Mini Kits (QIAGEN, Valencia, CA, USA) as described by the manufacturer, and stored at-200C before analysis. Other samples will be extracted later using the same methods and the photo-induced electron transfer PCR (PET-PCR) assay will be used to detect the presence of sub-microscopic infections. The assay relies on self-quenching primers for the detection of Plasmodium spp using Plasmodium genus and species specific primers as previously described (44).
Analysis of malaria vectors
Preserved mosquito samples were shipped to the laboratory in Tanga for further processing and analysis. Polymerase chain reaction for identification of sibling species of the An. gambiae complex and An. funestus group was conducted on a sub-sample of adult mosquito selected randomly from each village. Genomic Deoxyribonecleic acid (gDNA) was extracted from wings or legs of a female An. gambiae complex and An. funestus group according to methods previously described (45). The DNA samples were used for identification of sibling species of An. gambiae complex and An. funestus group using species specific diagnostic primers for An. gambiae complex (An. gambiae ss., An. arabiensis, An. quadriannulatus, An. melas, An. merus) (46). For the sibling species of the An. funestus group, the samples were identified based on species-specific primers in the internal transcribed spacer 2 (ITS2) region of the recombinant DNA (rDNA) genes, using a method previously developed to identify An. funestus, An.vaneedeni, An. rivulorum, An. leesoni, and An. parensis(47). Genotyping results were analysed using MXPro software (Agilent technologies, Santa Clara, CA, USA).Further analysis of malaria vector will be performed later including i) use of real time PCR (using TaqMan assays)to detect mutations in the knock down resistance (kdr) gene associated with pyrethroid resistance and determine the presence of kdr east and west genotypes in An. gambiae complex (48); ii) single nucleotide polymorphisms (SNPs)genotyping to assess the population structure of mosquitoes and additional markers of insecticide resistance; and, iii) Detection of P. falciparum circumsporozoite protein (Pf-CSP) and source of blood meal (for blood fed vectors)on a sub-sample of mosquitoes (An. gambiae complex and An. funestus group) by enzyme-linked immunosorbent assay (ELISA) technique as previously described(49).
Analysis of parasite population structure and molecular markers of antimalarial resistance
The samples were genotyped for parasite population structure at the University of North Carolina (USA) using molecular inversion probes (MIP) protocol as described earlier(51). The data has been used to assess the population structure and other population genetics metrics in the parasite populations from the study areas (Mosers et al. Manuscript in Preparation). The assay based on MIPs was also used to detect molecular markers of drug resistance. The analysis will focus on markers of artemisinin resistance (P. falciparumkelch 13, k-13), partner drugs particularly lumefantrine and amodiaquine (P. falciparummultidrug resistance – Pfmdr1 gene) while piperaquine (Plasmepsin 2 and 3) will be assessed later. Markers of resistance to the old drugs which were previously used in Tanzania – chloroquine (chloroquine transporter gene Pfcrt) and sulphadoxine/pyrimethamine – SP (dihypteorate synthase and dihydrofolate reductase, Pfdhps and Pfdhfr genes) were also assessed. The Pfmdr1 copy number variation will be detected by real-time PCR following a previously described assay(52).
Additional analysis was done to detect the levels of P.falciparum Histidine rich protein 2 (HRP2) and the prevalence of hrp2/3 gene deletion and their impacts on the performance of HRP2based mRDTs in Tanzania. Analysis of HRP2 will be done using a bead luminex assay described earlier (Rogier et al)(53), while the gene deletion will be assessed using the molecular assay that has been optimized at CDC (Bakari et al. Manuscript in Preparation). Polymorphisms in the hrp2/3 genes was analysed by Sanger sequencing as previously described(53).
Human genetic analyses
Human DNA will be extracted from DBS and analysed for polymorphisms in different genes which are known to be associated with susceptibility/ resistance to malaria. Such analyses will target sickle cell (β-globin), a-thalasemia (α-globin), glucose 6-phosphate dehydrogenase (G6PD) genes and others polymorphisms which will be deemed relevant.
Data management
Quantitative data was collected using tablets which were connected through internet to the central server located at NIMR Tanga Centre. The ODK software was used to create the database and data collection applications. The data manager reviewed the data on a daily basis for precision and consistency. Queries generated on a daily basis were sent back to the head of each study team for resolution. Data cleaning and validation was done continuously and daily/weekly as well as periodic reports generated. Final cleaning was done after the survey and analysis is being performed according to the data analysis plan which was prepared at the beginning and finalized at the end of the study.
Qualitative data were collected using tape recorders and later the audio files were stored in the computer. Audio files were transcribed from Kiswahili and translated into English language. Thematic approach was used for data analysis.
Data analysis
The Analytical framework and modelling approaches which will be used for the study is shown in Fig. 5.
The proposed analytical framework for determining the intrinsic and extrinsic drivers of malaria resilience can be derived starting from any of the five components depending on what action need to be taken or from which point of view implementation of the action targets. The model in Fig. 5 illustrates the interlinks that exists between these components. Health system is assumed to affect all components, but with different magnitude and assumed to impacts more socio-anthropological and the parasitological factors. Availability, access and utilization of malaria preventive measures is a across-cuttingcomponent between HS, entomology and parasitology, and slightly between anthropological aspects, with more weight given to its linkage with performance of the HS and affects the entomological and parasitological components.
Prior to analysis, the following will be done: i) potential indicators/measures to be derived from each component will be identified. A list with details of data to be used in calculation of the indicators and what information is expected to contribute to the analytical framework will be created; ii) information from demographic/socio-economic census, available for all individuals in all study villages, will be used to create a community-based vulnerability index to malaria resilience in the study districts. The index will be build based on "who are we dealing with" -concept, i.e. the study population. A sensitivity analysis will be done to assess how the index change when adding information from other components. Malaria prevalence levels (at finest scale possible, e.g. village/ward/district) from the parasitological component will be used as a response to a vulnerability model. Once the preliminary step is finalized, the team will gather to refine the analytical model and to generate linkages between components. This exercise will be used as a platform to refine set of models and approaches proposed here.
Different approaches are proposed tobe considered for analysis. Multiple analyses are considered to accommodate different context on which drivers for malaria resilience can be studied. With that, response variable(s) will be defined from one or multiple components then use data from other components as determinants or modifiers to create associations. All models will consider spatial differences of the study districts which will be defined using climate and environmental information.
The first approach will consider entomological parameters such as adult mosquito density, their infectivity (transmission) rates, resistance and information from breeding sites (number, spread and larval density) as a response. Once mapped, these will be linked to socio-demographic, socio-anthropological, parasitological and lastly HS parameters. Levels of entomological metrics are expected to be driven by rate of utilization of preventive measures, malaria infection levels, governance and institutional actions (taken by administrative team in collaboration with the HS), e.g. requesting for additional interventions such as larviciding; and lastly, human activities such as type of economic activity, choices on where to live/build a house, type of houses constructed, sources of water, knowledge and burden of disease observed in the community. This model assesses the problem from the vector aspects hence will derive actions focusing more on vector control
The second approach will take HS as the central point of deriving the understanding on the key drivers of persistence of malaria burden. Here, it is assumed that the success or failure in other components depends directly on the performance of the HS or vice versa. Taking it from here, a HS performance indicator will be built for all districts and modelled against selected/relevant indicators from all the remaining components. One of the hypothesis here could be that, if governance, institutional policies, actions and practices are not sufficiently implemented, they may result into insufficient community malaria knowledge, negative attitudes, high levels of entomological indicators, poor health seeking behaviours, poor distribution and utilization of effective interventions, and finally high burden (parasitological). This model is expected to derive perspectives that assess malaria stagnant problem from the governance actors than the beneficiaries.
The last model will approach the analysis from the perspective of availability, access and utilization of malaria preventive measures. This model will try to divide the weight of the drivers to multiple players, i.e. the HS performance, community knowledge and practices and patterns of entomological parameters for example seasonality. Since the analyses will be done using the data collected at the village level, an attempt will be made to extend the analysis to the entire district. Data from the health management information system (HMIS) which is routinely collected and reported through the Demographic Health Information system 2 (DHIS2) will be used. Such data will include test positivity rates (TPR), incidence rates, prevalence of malaria in under-fives (from MIS), school children and pregnant women using ANC data. The analyses will support development of district-wide maps of malaria burden and identify key factors to be incorporated in the ongoing efforts in the micro-stratification of malaria.