Using antenatal care as a platform for malaria surveillance data collection: study protocol

Background: While many malaria-endemic countries have health management information systems that can measure and report malaria trends in a timely manner, these routine systems have limitations. Periodic community cross-sectional household surveys are used to estimate malaria prevalence and intervention coverage but lack geographic granularity and are resource intensive. Incorporating malaria testing for all women at their first antenatal care (ANC) visit (i.e., ANC1) could provide a more timely and granular source of data for monitoring trends in malaria burden and intervention coverage. This article describes a protocol designed to assess if ANC-based surveillance could be a pragmatic tool to monitor malaria. Methods: This is an observational, cross-sectional study conducted in Benin, Burkina Faso, Mozambique, Nigeria, Tanzania, and Zambia. Pregnant women attending ANC1 in selected health facilities will be tested for malaria infection by rapid diagnostic test and administered a brief questionnaire to capture key indicators of malaria control intervention coverage and care-seeking behaviour. In each location, contemporaneous cross-sectional household surveys will be leveraged to assess correlations between estimates obtained using each method, and the use of ANC data as a tool to track trends in malaria burden and intervention coverage will be validated. Results: This study will assess malaria prevalence at ANC1 aggregated at health facility and district levels, and by gravidity relative to current pregnancy (i.e., gravida 1, gravida 2, and gravida 3 +). ANC1 malaria prevalence will be presented as monthly trends. Additionally, correlation between ANC1 and household survey–derived estimates of malaria prevalence, bed net ownership and use, and care-seeking will be assessed. Conclusion: ANC1-based surveillance has the potential to provide a cost-effective, localized measure of malaria prevalence that is representative of the general population and useful for tracking monthly changes in parasite prevalence, as well as providing population-representative estimates of intervention coverage and care-seeking behavior. This study will evaluate the representativeness of these measures and collect information on operational feasibility, usefulness for programmatic decision-making, and potential for scale-up of malaria ANC1 surveillance. This is a multi-country, observational, serial cross-sectional study with data from pregnant women attending ANC1 aggregated monthly. These data will be compared to similar survey questions (adapted from standard demographic health survey and malaria indicator survey questions) administered to households as part of cross-sectional surveys being conducted for evaluation of a number of different interventions (ClinicalTrials.gov Identifier: {“type”:”clinical-trial”,”attrs”:{“text”:”NCT04716387″,”term_id”:”NCT04716387″}}NCT04716387, {“type”:”clinical-trial”,”attrs”:{“text”:”NCT04157894″,”term_id”:”NCT04157894″}}NCT04157894, {“type”:”clinical-trial”,”attrs”:{“text”:”NCT04148690″,”term_id”:”NCT04148690″}}NCT04148690). Malaria prevalence by RDT among children under 5 years of age collected during the household surveys will also be compared to the prevalence among pregnant women at ANC1 in the same or a similar time period. All pregnant women of legal age and emancipated minors attending ANC1 at selected health facilities in Benin, Burkina Faso, Mozambique, Nigeria, Tanzania, and Zambia are approached for enrollment (Table ​(Table1).1). In Mozambique and Nigeria only, pregnant women below the legal age are considered eligible with the consent of a legal guardian. Pregnant women with symptoms of severe disease as determined by the clinician at the ANC clinic are excluded and referred for urgent case management. Children under 5 years of age whose parents respond to the cross-sectional household survey and provide consent for malaria testing of their children are also included. Study site, participant, and survey details This study is leveraging several planned intervention studies, including cross-sectional household surveys (modified malaria indicator surveys) at the district or village level that collect parasite prevalence data among children under 5 years (and other age groups as well, depending on country-specific epidemiology and national malaria control programme priorities). In Burkina Faso, Mozambique, and Nigeria, study activities are aligned with surveys from the New Nets Project, an observational quasi-experimental study that is evaluating the cost-effectiveness of various types of insecticide-treated nets [16]. During the New Nets Project, annual cross-sectional surveys are conducted from 2019 to 2022, including assessments of malaria prevalence, and corresponding ANC1-based surveillance occurs in a subset of health facilities in corresponding districts. Additionally, studies assessing the impact of group ANC are planned in the catchment areas of 40 health facilities each in Geita Region, Tanzania, and Atlantique Department, Benin, with corresponding cross-sectional surveys occurring as the primary method of evaluating the impact of the group ANC intervention. Due to the COVID-19 pandemic, the group intervention was stopped early in Tanzania, but the ANC1-based surveillance assessment and cross-sectional surveys continue. Cross-sectional surveys are conducted at study baseline and endline (November 2019 and June 2021 in Tanzania, February/March 2021 and September/October 2022 in Benin). In Chadiza, Zambia, this ANC1-based surveillance pilot leverages a study assessing the impact of proactive community case management in 21 health facility catchments, with surveys conducted in May 2021 and May 2023. Health facilities and their catchment areas are selected for participation in the parent studies according to the needs of each study; of these, health facilities with a minimum of 20 ANC1 per month are selected for co-inclusion in this ANC surveillance pilot. In some studies, all eligible health facilities are included, while in others a subset of eligible facilities are randomly selected for inclusion. Two activities are integrated into routine ANC1 consultations: malaria testing of eligible participants regardless of symptoms using an RDT, and administration of a questionnaire to collect data on participant demographics, gravidity, insecticide-treated net ownership and use, and care-seeking behaviour. For each facility, the corresponding catchment villages were identified in collaboration with the health facility workers. From among all the villages in the health facility catchment area, 1–2 villages or enumeration areas were selected for household sampling for the cross-sectional survey proportional to population size. The selected areas were mapped and respondents randomly selected as indicated by the parent study conducting the cross sectional survey. All pregnant women attending ANC1 are tested for malaria infection. During group counselling sessions at initial ANC1 intake, women are informed of this pilot surveillance activity. Women are consented for this pilot surveillance activity individually prior to testing. The number of women refusing malaria testing is noted on a screening form. For consenting women, a malaria RDT is administered concurrently with other routine ANC testing, using blood from the same finger prick. While awaiting the results from these tests, the short ANC surveillance questionnaire (Additional file 1) is administered by a trained study team or staff member, and the RDT result recorded when available. Women have the option of responding to the questionnaire even if they declined the RDT. During a regular ANC visit, the nurse asks questions to determine a woman’s age, gravidity, and gestational age. These data are recorded in the ANC registers and copied into the corresponding study questionnaires to avoid asking women the same questions twice. In Burkina Faso, Mozambique, and Zambia, a paid trained field worker assists with recruitment and administering the questionnaire while in the other countries, this task is completed by clinic staff. As stated above, finger-prick capillary blood samples are collected for malaria diagnosis using malaria RDTs. The RDT used for both ANC1 surveillance and the cross-sectional household surveys is determined by national procurement protocols. For ANC1 surveillance, the blood used for malaria testing is taken from the same finger prick that is used to collect blood for standard ANC testing. If positive, treatment is given to the women according to the national guidelines for malaria treatment. A sample size of 88 people from a population of 1,000 people produces a two-sided 95% confidence interval with a precision (half-width) of 0.10 when the actual proportion is near 0.50. The sample size needed for a given confidence interval around a proportion is greatest for a proportion of 0.50. This is used to provide the most conservative assumption of sample size. The health facility sampling scheme is implemented across the different countries to provide a minimum of 88 women enrolled per month per sampling unit (i.e., health facility, district, or country), for a yearly total of 1,056 women per sampling unit. In cases where this sample size is not achieved in a single month at a given health facility, the data will be aggregated across facilities to provide a district/country estimate or across months to provide quarterly estimates to have sufficient sample size to provide an accurate estimate. To calculate the correlation between the facility-level data and community-level data, an estimated minimum of 20 facilities is needed to provide an accurate estimate of the correlation coefficient (R), using a two-tailed alpha test (alpha 0.05) and with 80% power, assuming an expected correlation coefficient of 0.6 or greater (Fig. 1). This is not feasible across all individual countries, thus using the combined data from multiple countries aimed to ensure sufficient sample sizes to assess the correlation. Sample size required for given correlation coefficient This study will summarize ANC1 participant demographic characteristics using descriptive statistics for age and gravidity by health facility and district. Key indicators will include: Prevalence is calculated as the number of women with positive RDT results divided by the total number of women tested [11]. Within each health facility and district, prevalence will be disaggregated by gravidity relative to the current pregnancy (i.e., first pregnancy [gravida 1], second pregnancy [gravida 2], and third or more pregnancy [gravida 3 +]). The monthly malaria prevalence data will be used to produce time series graphs, quarterly and/or annually, for correlation with cross-sectional survey estimates. In addition, where available, the proportion of RDT-positive women who are asymptomatic and symptomatic will be calculated. To compare ANC1 malaria prevalence to prevalence from cross-sectional surveys within the same district, ANC1 malaria prevalence are obtained for the three months surrounding the community cross-sectional surveys (i.e., the month before the cross-sectional survey, the month of the survey, and the month directly following the cross-sectional survey). The relationship between malaria prevalence at ANC1 and in the community will be quantified using mixed-effects logistic regression, accounting for likely modifiers such as age, gravidity, transmission intensity, and transmission season, and adjusting for sampling effort in time and space. To quantify the incremental value of the ANC1 data, the ability of the ANC1 data to predict population prevalence using these models will be tested using leave-one-out cross-validation for each district and compared to alternative predictive models in the absence of any ANC1 data (e.g., extrapolating district-level trends from other districts sampled within the region). With support of national malaria control programmes and operational partners, district-level data on the temporal distribution of interventions within each region will be leveraged to assess the utility of ANC1 data to capture plausible trends in intervention impact. Similar analyses will be conducted to quantify the relationship between the cross-sectional surveys, the ANC1 data, reported case incidence, and the malaria test-positivity rate (obtained by dividing the number of positive malaria tests by the number of tests performed). In addition, the use of more heuristic measures such as the Kendall correlation coefficient, which provides a summary of the extent to which prevalence ranking of each district remains constant over time [12], will be used to assess the extent to which the choice of malaria metric influences operational decision-making. Estimates of care-seeking behaviors and indicators for bed net ownership, use, and access will be determined for ANC1 data and compared to the corresponding indicators from the cross-sectional surveys. Net use and ownership (proportion of households that own at least one bed net and proportion of households with one net for every two people) will be aggregated monthly for each sampling unit/district and correlated with monthly ANC1 malaria prevalence to compare trends. Once an optimal predictive model of community prevalence using ANC1 data has been obtained, this relationship will be incorporated within an existing, well-established mechanistic model of malaria transmission [17–19] in an attempt to provide a framework that can convert estimates of ANC1 malaria prevalence into continuous measures of community malaria transmission and burden with which to explore future intervention scenarios (analogous to a current framework that involves calibrating to cross-sectional data [20]). Data will be obtained from separately generated electronic or paper registers in each country. At each health facility, teams will be trained to conduct data quality assessments to monitor and improve data quality and completeness. Health facility–level ANC1 data from the DHIS2 system will include the monthly number of ANC attendees, the number tested for malaria at ANC1, and the number positive for malaria. Additional data will be collected on maternal age, gravidity, and insecticide-treated net ownership and utilization. Data on malaria prevalence among children will be obtained from representative, cross-sectional, household surveys as specified in the parent studies. Participants will be assigned a unique identifier number by the study, and will be identified only by this number in the dataset. No participant names or other information that would make the participant identifiable will be included. Individual written consent is obtained from women of legal age prior to testing and questionnaire administration during ANC1. In Mozambique and Nigeria, women under legal age are able to participate with the consent of their parent or legal guardian. Participants are told the general purpose, possible risks, and benefits of the ANC1 pilot surveillance activity in the local language. Participation is voluntary. Ethical clearance for the study was sought and obtained from the following institutional review boards: In addition, this activity was reviewed by CDC and was conducted consistent with applicable federal law and CDC policy.