- Record: found
- Abstract: found
- Article: found
Population coverage of artemisinin-based combination treatment in children younger than 5 years with fever and Plasmodium falciparum infection in Africa, 2003–2015: a modelling study using data from national surveys
research-article
Author(s):
Adam Bennett , PhD,
Donal Bisanzio , PhD,
Joshua O Yukich , PhD,
Bonnie Mappin , MSc,
Cristin A Fergus , MPH,
Michael Lynch , MD,
Richard E Cibulskis , PhD,
Samir Bhatt , PhD,
Daniel J Weiss , PhD,
Ewan Cameron , PhD,
Peter W Gething , PhD,
Thomas P Eisele , PhD
Publication date PMC-release:
01 April 2018
Journal:
The Lancet. Global health
Read this article at
There is no author summary for this article yet. Authors can add summaries to their articles on ScienceOpen to make them more accessible to a non-specialist audience.
Summary
Background
Artemisinin-based combination therapies (ACTs) are the most effective treatment for uncomplicated Plasmodium falciparum malaria infection. A commonly used indicator for monitoring and assessing progress in coverage of malaria treatment is the proportion of children younger than 5 years with reported fever in the previous 14 days who have received an ACT. We propose an improved indicator that incorporates parasite infection status (as assessed by a rapid diagnostic test [RDT]), which is available in recent household surveys. In this study we estimated the annual proportion of children younger than 5 years with fever and a positive RDT in Africa who received an ACT in 2003–15.
Methods
Our modelling study used cross-sectional data on treatment for fever and RDT status for children younger than 5 years compiled from all nationally available representative household surveys (the Malaria Indicator Surveys, Demographic and Health Surveys, and Multiple Indicator Cluster Surveys) across sub-Saharan Africa between 2003 and 2015. Estimates for the proportion of children younger than 5 years with a fever within the previous 14 days and P falciparum infection assessed by RDT who received an ACT were incorporated in a generalised additive mixed model, including data on ACT distributions, to estimate coverage across all countries and time periods. We did random effects meta-analyses to examine individual, household, and community effects associated with ACT coverage.
Findings
We obtained data on 201 704 children younger than 5 years from 103 surveys (22 MIS, 61 DHS, and 20 MICS) across 33 countries. RDT results were available for 40 of these surveys including 40 261 (20%) children, and we predicted RDT status for the remaining 161 443 (80%) children. Our results showed that ACT coverage in children younger than 5 years with a fever and P falciparum infection increased across sub-Saharan Africa in 2003–15, but even in 2015, only 19.7% (95% CI 15.6–24.8) of children younger than 5 years with a fever and P falciparum infection received an ACT. In meta-analyses, children younger than 5 years were more likely to receive an ACT for fever and P falciparum infection if they lived in an urban area ( vs rural area; odds ratio [OR] 1.18, 95% CI 1.06–1.31), had household wealth above the national median ( vs wealth below the median; OR 1.26, 1.16–1.39), had a caregiver with any education ( vs no education; OR 1.31, 1.22–1.41), had a household insecticide-treated net (ITN; vs no ITN; OR 1.21, 1.13–1.29), were older than 2 years ( vs ≤2 years; OR 1.09, 1.01–1.17), or lived in an area with a higher mean P falciparum prevalence in children aged 2–10 years (OR 1.12, 1.02–1.23). In the subgroup of children for whom treatment was sought, those who sought treatment in the public sector were more likely to receive an ACT ( vs the private sector; OR 3.18, 2.67–3.78).
Interpretation
Despite progress during the 2003–15 malaria programme, ACT treatment for children with malaria remains unacceptably low. More work is needed at the country level to understand how health-care access, service delivery, and ACT supply might be improved to ensure appropriate treatment for all children with malaria.
Related collections
Most cited references28
- Record: found
- Abstract: found
- Article: not found
Impact of Artemisinin-Based Combination Therapy and Insecticide-Treated Nets on Malaria Burden in Zanzibar
Achuyt Bhattarai, Abdullah S. Ali, S Patrick Kachur … (2007)
Introduction The increased malaria-related morbidity and mortality, especially in children under the age of 5 y (“under five”), due to emerging resistance of Plasmodium falciparum to conventional antimalarial drugs calls for immediate actions to “Roll Back Malaria” in sub-Saharan Africa. This need has been clearly recognized in the Millennium Development Goals “to halt and begin to reverse malaria incidence” [1] as well as in the Abuja Declaration objective to halve malaria mortality in Africa by 2010 through implementation of combined control strategies [2]. In the year 2000, the overall treatment failure of chloroquine was found to be 60% in a 14-d efficacy trial; consequently the Zanzibar Ministry of Health and Social Welfare decided in November 2001 to change both first- and second-line treatment guidelines for uncomplicated malaria from chloroquine and sulfadoxine-pyrimethamine to artemisinin-based combination therapies (ACT) [3]. The ACT policy was implemented in September 2003, when Zanzibar became one of the first regions in sub-Saharan Africa to recommend routine use of ACT. This action was followed by strengthened vector control, culminating in a nation-wide distribution campaign of long-lasting insecticidal nets (LLINs) from early 2006. Both ACT and vector control measures have independently proven to be efficacious malaria control strategies. Ecological studies have credited ACT with enhancing treatment efficacy, reducing malaria transmission, and possibly forestalling drug resistance in low-endemicity areas [4,5]. Moreover, specific African trials have indicated that the use of insecticide-treated nets (ITNs) or indoor residual spraying can reduce mortality of children under five in Africa [6–9]. This is, however, to our knowledge the first study to examine the public health impact of wide-scale deployment of ACTs alone and combined with ITNs through the general health structure/channels on malaria indices and general health parameters in an endemic area in sub-Saharan Africa. Methods Study Site The study was conducted in North A District, Zanzibar, situated just off the coast of mainland Tanzania. The district is rural and has a population of about 85,000. Subsistence farming and fishing are the main occupations. Plasmodium falciparum is the predominant malaria species and Anopheles gambiae complex is considered the main vector. Malaria transmission is stable with seasonal peaks related to rainfall in March–May and October–December. Malaria transmission in the district prior to the interventions has been reported to be high, but specific entomological data are not available to allow a precise characterization of malaria transmission intensity. However, during the screening process of a major antimalarial drug trial conducted in 2002–2003, a P. falciparum prevalence exceeding 30% was observed in febrile children under five [10], suggesting that North A District had been a high transmission area prior to ACT implementation in September 2003. North A District has one Primary Health Care Centre, which includes a hospital with inpatient and laboratory services, e.g., blood transfusion and malaria microscopy services. Basic medical treatment services without laboratory support are provided in 12 Primary Health Care Units located in different shehias (the smallest political administrative unit in Zanzibar). Drugs, including conventional and artemisinin monotherapies, are also available in private shops throughout the district. Malaria Control Interventions Figure 1 illustrates time of implementation of the two malaria control interventions. Figure 1 Malaria Interventions, Cross-Sectional Surveys, Monthly Rainfall, and Reported Clinical Malaria Diagnoses in Children under 5 Years of Age in North A District, Zanzibar (A) Start of the implementation of artemisinin-based combination therapy for treatment of uncomplicated malaria in September 2003. (B) Introduction of LLINs in February 2006. Promotion of ITNs started in January 2004; the use of conventional ITNs, however, remained low, until the introduction of LLINs. Outpatient data for 2006 are up to June. First intervention—ACT. A loose combination of artesunate and amodiaquine (AS+AQ; from various suppliers with preapproval from WHO) and a fixed combination of artemether–lumefantrine (Coartem; Novartis, Basel, Switzerland), were implemented as first- and second-line treatment, respectively, for uncomplicated malaria in all public health facilities from September 2003. In a pre-implementation assessment of the new treatment policy, partly conducted in North A District 2002–2003, both AS+AQ and artemether–lumefantrine were highly efficacious with PCR-adjusted cure rates by day 28 above 90% [10]. Quinine remained the drug of choice for severe malaria and sulfadoxine-pyrimethamine for intermittent preventive treatment during pregnancy. From September 2003, chloroquine was withdrawn from all health facilities and replaced by free provision of ACT to all malaria patients. Total treatment courses of AS+AQ dispensed in North A 2004 and 2005 were 34,724 and 12,819, respectively. The supply of ACT has been uninterrupted, with no reports of AS+AQ being out of stock from any public health facility in the district during 2003–2006 (unpublished data). ACTs were purchased with support from African Development Bank and Global Fund to Fight AIDS, Tuberculosis and Malaria (GFATM). Second intervention—vector control. A policy to distribute conventional ITNs to the most vulnerable groups—children under five and pregnant women—free of charge through antenatal clinics or local shehia leaders was officially launched in 2004. However, ITN coverage and use remained low in North A District 2004 and 2005 due to limited number of ITNs distributed, 4,026 and 1,550, respectively. A mass campaign was therefore initiated early 2006, with distribution of 23,000 LLINs to the two most vulnerable groups in North A. This campaign was supported by GFATM and the US Agency for International Development. Cross-Sectional Surveys Three cross-sectional surveys with the primary objective to determine P. falciparum prevalences were conducted in North A District between 2003 and 2006. A two-stage cluster sample technique was used. First shehias and then the households were randomly selected from the sampling frame obtained from the Office of Chief Government Statistician, Zanzibar. The sampling frame was updated before each survey. The first exploratory survey, conducted in May 2003, included 625 households and provided baseline data prior to ACT and widespread ITN implementation. Sample size calculations for the follow-up surveys conducted in May 2005 and 2006 were based on the proportion of children under five with malaria parasitemia in 2003, about 9%, and an assumed relative error of 20%. The calculated number of households to be included was 490 after adjusting for a design effect of 2. Trained interviewers visited all selected households. Interviews and blood sample collection were initiated upon written consent from head of each household and proxy consent from the mother or guardian of each child. Information was recorded using a structured questionnaire on recent febrile illness, mosquito net use, and care-seeking behavior from each individual present in the household at the time of the survey. We did not replace households in which residents were not present at time of survey, could not be located, or refused to participate. Thick blood films were collected from all consenting participants, stained with 5% Giemsa for 30 min, and examined by experienced microscopists for presence and density of P. falciparum parasites. If fewer than ten parasites were detected per 200 white blood cells, examinations were extended to 500 white blood cells. Blood slides were considered negative if no asexual parasites were found in 200 high-power fields. High-density parasitemia was defined as presence of ≥ 5,000 parasites/μl [11]. Quality control was conducted for all positive slides and 10% of the negative slides [12]. Health Facility Records Malaria-related indicators, i.e., outpatient attendances, hospital admissions and blood transfusions, from all 13 public health facilities in North A District were obtained from the Health Management and Information System (HMIS) records of the Zanzibar Ministry of Health and Social Welfare. The existing HMIS records were about 90% complete for the period 2000–2004. Data were validated and missing information retrieved by retrospective review of source documents from all 13 health facilities. This confirmed the HMIS records and resolved missing or inconsistent data, which increased the completeness to nearly 100%. A database of malaria-related indicators was created on the basis of this retrospective review. Data from 2005 were abstracted on quarterly basis. Vital Statistics Records of vital events, i.e., births and deaths, for the period 1998–2005 were obtained from the District Commissioner's Office (DCO) in North A. Annual crude mortalities of children under five were estimated from these data. Demographic estimates were obtained from Tanzania National Population and Housing Census 2002. Rainfall Complete records of monthly rainfall during 1999–2005 were obtained from official registers of the Tanzania Metrological Agency of the Ministry of Communications and Transport. On Unguja island, rainfall is centrally measured in one weather station, situated 26 km (radially) from North A District. The mean annual rainfalls recorded in 2003, 2004, 2005, and 2006 were 702, 1,934, 1,231, and 1,214 mm, respectively. The corresponding mean seasonal rainfall (March–May) between 2003 and 2006 was 285, 786, 890, and 613 mm, respectively. During the post-ACT intervention period (2004–2006) the mean annual and seasonal rainfall was 8%–12% lower than the pre-ACT intervention period (2000–2002). However, the only year with a marked reduction in the mean annual and seasonal rainfalls was the year 2003 with two- to three-fold lower rainfall, as compared to both the preceding and subsequent 3 y. Data Processing and Analysis Data were entered and validated using Microsoft Access and Excel. Statistical analyses for cross-sectional surveys, health facility records, vital statistics, and rainfall data were performed using Stata version 8. Analysis for the surveys was corrected for multi-stage sampling errors using the Rao-Scott second order correction [13]. A logistic regression model with robust standard errors (robust cluster) was used to adjust for the effect of age, sex, sleeping under a mosquito-net, and asset index on asexual P. falciparum prevalence and gametocyte carriage across the study years. Households were the primary sampling units in the surveys and were defined as clusters. Wald test was used to assess the fit of the model and interactions between covariates incorporated in the model. Odds ratios were adjusted for the complex sampling design and covariates listed above. Pearson correlation coefficients were calculated to assess the linear relationships between monthly rainfall and outpatient malaria diagnosis, and malaria-attributed deaths. Ethical Approval Protocols for the household surveys were reviewed and approved by the Medical Research Coordinating Committee of the Tanzanian Commission on Science and Technology, the Zanzibar Health Research Council and the institutional review board of US Centers for Disease Control and Prevention. Results Cross-Sectional Surveys The timings of the cross-sectional surveys in relation to start of each malaria control intervention and seasonal rainfalls are presented in Figure 1. The number of households enrolled and participant characteristics in the respective surveys are shown in Table 1. Over 95% of all participants agreed to both answer questionnaires and provide blood samples in the respective surveys. Table 1 Number of Households Surveyed and Characteristics of Survey Participants The parasite prevalences and odds ratios (ORs) of asexual P. falciparum parasitemia and gametocyte carriage at the time of cross-sectional surveys are shown in Table 2. Between 2003 and 2005 the parasite prevalence was reduced by about 50% in children under five. A further 10-fold decrease in P. falciparum prevalence was observed between 2005 and 2006, following mass distribution of LLINs specifically targeting this age group. Concomitant reductions of parasite prevalence were observed in children over the age of 5 y, although only by about 3-fold, between 2005 and 2006 (OR 0.41, 95% confidence interval [CI] 0.13–1.21), p = 0.08). Table 2 Parasite Prevalence and ORs of P. falciparum Asexual Parasitemia and Gametocytemia in Children 0–14 Years of Age in North A District, Zanzibar, in May 2003, 2005, and 2006 High-density parasitemia (≥5,000/μl) was found in 14 (2.7%) and 2 (0.6%) children under five in 2003 and 2005, respectively. No child carried high-density parasitemia in 2006. Reported fever within 14 d prior to the survey was similar in 2003 and 2006 among children under five (2003, 13% [95% CI 11–17]; 2006, 12% [95% CI 9–16]), whereas care-seeking at public health facilities by recently febrile children under five increased significantly (2003 was reference year; 2005, OR 3.91 [95% CI 0.85–17.9]; 2006, OR 5.5 [95% CI 2.3–13.3]; p-value for trend < 0.001). The proportions of children under five sleeping under effective ITNs were below 10% in both 2003 and 2005 (Table 1), whereas in 2006, 90% were reported sleeping under an LLIN on the night before survey. Health Facility Surveillance All reported clinical outpatient malaria diagnoses in North A District between January 1999 and June 2006 among children under five are shown by month in Figure 1 and by year in Table 3. Between 2002 and 2005 the total number of out-patient malaria diagnoses decreased by 77%. The annual incidences of malaria diagnoses standardized per 1,000 children under five in North A District were 843, 786, and 233 in 2003, 2004, and 2005, respectively. The total number of children under five attending public health facilities for any cause during 1999 and 2005 remained relatively constant, ranging from 31,069 to 39,374 annually. Up to 2003 malaria accounted for about 50% of all outpatient diagnoses in this age group, whereas in 2005 this proportion had decreased to 13%. Table 3 Outpatient Malaria Diagnoses, Hospital Admissions, Blood Transfusions, and Malaria-Attributed Deaths in North A District, Zanzibar, between 2000 and 2005 Malaria-related hospital admissions, non-malaria admissions, and blood transfusions in children under five between 2000 and 2005 are also shown in Table 3. From 2002 to 2005, malaria-related admissions, blood transfusions, and malaria-attributed mortality decreased by 77%, 67%, and 75%, respectively. Crude Mortality Data A total of 23,200 live births and 1,032 deaths in children under five (49% females) were registered between January 1998 and December 2005. The annual mortality figures for children under five, children (1–4 y), and infants (0–1 y) are shown in Table 4. Between 2002 and 2005, crude under five, infant, and child mortality decreased by 52%, 33%, and 71%, respectively. Table 4 Mortality of Children under 5 Years of Age in North A District, Zanzibar between 1998 and 2005 Relationships between Rainfall and Malaria Diagnosis and Deaths In the pre-ACT intervention period (2000–2002), significant positive correlations were found between monthly rainfall and both outpatient malaria diagnoses (Pearson correlation coefficient [r p] = 0.59, p < 0.001) and malaria-attributed deaths (r p = 0.75, p < 0.001), when data were adjusted to allow for a 1-mo lag between rainfall and malaria diagnoses and deaths. However, in the post-ACT intervention period (2003–2005), no significant correlations were found between monthly rainfall and outpatient malaria diagnosis (r p = −0.05; p = 0.75) or malaria-attributed deaths (rp = 0.23; p = 0.20). Discussion Malaria burden in Zanzibar, as in most parts of sub-Saharan Africa, has remained high and in many areas even increased during the last 10–20 y, a major reason being rapid spread of resistance to commonly used monotherapies against malaria. This problem has necessitated urgent implementation of new and effective control strategies to “Roll Back Malaria.” Two main cornerstones in this effort are the introduction of ACTs for treatment of uncomplicated malaria and the promotion of ITN use. The targets for the implementation of these new strategies have been defined by the UN Millennium Development Goals [1] and the Abuja Declaration [2], to be achieved by the years 2015 and 2010, respectively. Deployment of ACTs The ACTs were dispensed free of charge to all patients in the study area through public health facilities from September 2003 onwards. The ACT implementation and deployment was very rapid, effective, and with high coverage. Monitoring of drug supplies confirmed that ACTs were available throughout the study period in all 13 public health care settings in North A District. This outcome also indicates that estimates were adequate of the needed and thus deployed numbers of ACT treatments in the district. This result was accomplished despite an apparent two-fold increase in care seeking among children under the age of 5 y at public health facilities as observed in the cross-sectional surveys. We believe that the observed shift in treatment-seeking behavior at public facilities may be related to availability of free, effective ACTs. A previous study in Zanzibar showed that people's attitudes towards health seeking at public health facilities (biomedical practices) are negatively influenced by the distribution of ineffective antimalarial drugs [14]. High ACT coverage was rapidly achieved in malaria patients despite availability of other drugs in the private sector. This achievement was probably influenced both by comprehensive information to the public and health care staff and by the strong commitment of the Zanzibar government to rapidly ensure free coverage of the ACTs. Also, in North A District, as well as in Zanzibar generally, the entire population has relatively easy access to public health facilities, which are located within 5 km from any community and are served by good transport links. However, the absence of co-formulation or even of co-blistering of the two compounds in the first-line treatment, artesunate and amodiaquine, may have resulted in some degree of monotherapy with either compound. Mortality Impact Our study provides the first, to our knowledge, observation of a reduction in mortality of children under five following introduction of ACTs solely in a stable malaria-endemic setting. The highly significant reduction of 52% in crude under-five mortality according to vital statistics between 2002 and 2005 also highlights the importance of malaria as a major cause of death among children in malaria-endemic areas. The 71% reduction among children aged 1–4 y indicates that the relative contribution of malaria to crude mortality is particularly important in this age group. Major reductions in crude under-five mortality has also been observed in previous randomized intervention studies with ITNs [6,7] and community-based malaria treatment [15,16], but the reduction rates (between 25% and 40%) have been less pronounced than those in our study in Zanzibar. We believe our findings are valid and represent a true picture of the effects of ACT deployment in North A District, Zanzibar. No other major political, socioeconomic, or health-care change with the potential to halve mortality in children under five occurred in Zanzibar after 2002. This includes Expanded Programme on Immunization coverage, which remained constantly above 80% in the district during 1999–2005. Furthermore, there was no significant change in rainfall that may have contributed to the observed reduction in malaria transmission. Indeed, the only year with reduced rainfall with potential influence on vector capacity occurred before the introduction of ACTs—in 2003. Increased use of ITNs may also represent a potential confounding factor in our study. However, the ITN use was below 10% during 2004 and 2005 as reported and observed during the cross-sectional surveys. A significant improvement in ITN coverage was only achieved in 2006 after the introduction of LLINs (see further below) and only affected the 2006 cross-sectional results. We chose 2002 as reference year in our analyses of health facility surveillance and under-five mortality, because 2002 represents the last complete year before ACT introduction in September 2003. Routinely collected mortality statistics may underestimate the true values. However, such data have been shown to provide valid mortality trends [17,18]. Morbidity Impact A significant reduction was found with regard to hospitalization of malaria patients and incidence of blood transfusions, which may be considered proxy indicators of severe malaria. The reduction of severe malaria showing a similar pattern thus supports the under-five mortality trends. This health impact probably represents effects of improved case management of uncomplicated malaria with ACT, thus preventing the development of severe manifestations of the disease. The decrease in malaria morbidity (and mortality) at health facilities between 2003 and 2005 confirms the therapeutic efficacy of ACT [10], but the reduction in outpatient malaria diagnoses may also reflect some transmission blocking effect of artemisinin derivatives through its gametocytocidal activity. Reduction in transmission potential has been suggested after the introduction of artemisinin derivatives (before vector control) for routine treatment in a low and seasonal malaria transmission setting in Thailand [4]. Data obtained from routine health facility records have inherent potential pitfalls and need to be interpreted cautiously. However, the fact that they all show the same downward trend after improved coverage of malaria prevention and treatment interventions, and with no change in the climatic conditions that are favorable for malaria transmission, supports the plausible conclusion that enhanced malaria control interventions contributed to the observed public health benefits. Deployment of ITNs The deployment of LLINs in early 2006 provided a high coverage, i.e., over 90% reported use in children under five in the cross-sectional survey in May 2006. Importantly, this high mosquito-net use was observed after strong government commitment and after free LLIN distribution to children under five and pregnant women. The most significant decrease in prevalence of asymptomatic parasitemia was achieved in 2006, when LLINs were widely used by the children under five, whereas the major impact on the under-five mortality was achieved earlier with ACT use only. Strengthened vector control and the use of ACT also resulted in marked and sustained malaria control in South Africa [5]. The similar public health benefits observed in North A supports the concomitant use of vector control and ACT for malaria control. However, it should be emphasized that our study captures short-term trends in malaria control in North A, which may be too short to generalize long-term trends in the burden of malaria. Sustained coverage and use of LLINs by vulnerable groups is yet to be demonstrated, especially under declining malaria endemicity and if the free LLIN distribution scheme were to be changed. Conclusions The declining under-five mortality, malaria morbidity, and malaria prevalence observed in our study is the first comprehensive evidence supporting the major public health benefits of ACT and ITNs in a stable endemic malaria transmission setting in sub-Saharan Africa. The findings suggest that ACTs with high coverage of ITN use may potentially even eliminate malaria as a public health problem in highly endemic areas of sub-Saharan Africa. High community uptake of the two interventions is probably required but indeed achievable if, as in our study, they are easily available free of charge. The UN Millennium Development Goals to alleviate malaria as a major public health problem and substantially reduce the under-five mortality in sub-Saharan Africa are thus achievable even in settings with historically intense malaria transmission. The sustainability of these efforts as well as surveillance to prevent resurgence of malaria represent key research and programmatic follow-up issues of malaria control in Africa.
- Record: found
- Abstract: found
- Article: found
Measuring Coverage in MNCH: Tracking Progress in Health for Women and Children Using DHS and MICS Household Surveys
Attila Hancioglu, Fred Arnold (2013)
This paper is part of the PLOS Medicine “Measuring Coverage in MNCH" Collection. Introduction Considerable progress has been made since the mid-1990s in reducing maternal and child mortality [1]. However, there are still many unnecessary deaths among women and children (about 287,000 maternal deaths and 7.6 million child deaths in 2010 [2]), even though effective and affordable interventions are available. Tracking coverage indicators for women and children is necessary to guide global, regional, and country efforts to improve health so that scarce resources are directed to where they are most needed and will be most effective in saving lives. Coverage measurements reflect the proportion of individuals needing an intervention, and must therefore be representative of the reference population. Two recent reviews assessed potential sources of coverage data for reproductive, maternal, newborn, and child health (RMNCH) interventions in the 75 countries that account for over 95% of maternal and child deaths [2],[3]. These reviews concluded that although routine data from health management information systems are the preferred source of coverage data because they provide information on a continuous basis at lower administrative levels such as districts, these systems are currently too weak in these countries to provide data of adequate quality for assessing and guiding health programmes. Demographic surveillance systems often produce higher quality data but for limited geographic areas that become progressively less representative of national populations over time if health intervention trials are conducted in these areas. Both reviews identified high-quality, nationally representative household surveys as the method of choice for measuring RMNCH coverage for the foreseeable future in most low- and middle-income countries. Importantly, even after health management information systems become reasonably complete and accurate, national household surveys will need to continue as a complementary data source since these surveys are representative of the general population and provide vital information on background characteristics and the determinants of population and health conditions. Household surveys are also needed for measuring inequalities in coverage [4]. A large majority of household surveys that have produced coverage estimates in low-income countries have been conducted under the USAID-supported Demographic and Health Surveys (DHS) [5] and the UNICEF-supported Multiple Indicator Cluster Surveys (MICS) programmes [6]. For example, in the global databases compiled by UNICEF, information on the use of oral rehydration therapy for children with diarrhoea comes from DHS and MICS surveys for 98% of the countries with available data. For care seeking for pneumonia, the comparable figure is 93%. DHS and MICS data have also made a major contribution to the scientific literature. According to a recent study and data from the DHS website, more than 1,100 articles based entirely or primarily on DHS data have been published in 346 peer-reviewed journals [7]. Both programmes provide free public access to survey reports and datasets. Other standardized household survey programmes that have generated coverage data on selected indicators include the CDC-supported Reproductive Health Surveys conducted mostly in Latin America, Eastern Europe, and Central Asia between 1975 and 2009 [8], the Pan-Arab Population and Family Health Surveys (PAPFAM), supported by the Arab League and conducted in the Arab region [9], and the WHO-supported World Health Survey conducted from 2002–2004 [10]. There are also standard household surveys focused on individual diseases or intervention programmes such as the Malaria Indicator Surveys [11], the AIDS Indicator Surveys [12], and the Standardized Monitoring and Assessment of Relief and Transitions (SMART) surveys conducted in many sub-Saharan African countries in the early 2000s to provide information related to child nutrition, some of which are still being conducted on an annual or semi-annual basis [13]. Also, a few countries have mounted their own nationally representative surveys, usually based on adapted versions of the DHS and MICS protocols. Over time and as a result of arduous consultations at the global level, a set of standard indicators and “gold-standard" methodologies have emerged that are incorporated in DHS and MICS surveys, as well as in some more specialized surveys. The methodological challenges of measuring RMNCH coverage received relatively little attention in the literature until recently, in contrast to the measurement of mortality [14]. Much of the testing of alternative coverage indicators, questions, and analytical techniques is available only in internal reports of work conducted by DHS or in the heads of the technical experts who have conducted the surveys. In this review, which is part of the PLOS Medicine “Measuring Coverage in MNCH" Collection, we draw on the DHS and MICS experience to highlight key methodological principles and challenges in using household survey data to measure RMNCH coverage. For more details, we direct readers to resource documents on survey design including sampling, questionnaires, data cleaning, and analysis as well as reports [5],[6],[15]. Other reviews and research articles in this collection will focus on measurement challenges in tracking trends in coverage for interventions targeting specific health conditions [16]–[21] and cross-cutting methodological issues [22]. Here, we complement this content by providing insights to help improve survey measurements of RMNCH coverage and to promote the informed use of coverage results to improve programmes. Overview of the DHS and MICS Survey Programmes The DHS programme has been operating since 1984 with core funding from USAID and substantial contributions from other donors and participating countries. The programme is coordinated by ICF International. Its aim is to provide high-quality nationally representative data on health and population trends, with emphasis on fertility, family planning, mortality, reproductive health, child health, gender-related issues such as domestic violence, HIV/AIDS, malaria, and nutrition. The MICS programme was developed by UNICEF in 1995 in response to the World Summit for Children and has expanded over time to measure progress towards the Millennium Development Goals and other international targets for women and children. MICS surveys provide key information on mortality, health, nutrition, education, HIV/AIDS, and child protection for use in programme decision making, advocacy, and national and global reporting. Figure 1 shows the number of DHS and MICS surveys conducted annually since 1984. About ten to 15 DHS surveys have been conducted annually since 1995; consultative processes have led to major revisions in the core questionnaires every 5 years. MICS surveys were conducted in “rounds," every 5 years until 2007 and every 3 years thereafter, with about 60 surveys in each round. 10.1371/journal.pmed.1001391.g001 Figure 1 Number of DHS and MICS surveys by year. The core DHS [23] and MICS [24] questionnaires have expanded over time, in content and complexity, to respond to country and global needs and the growing number of effective RMNCH interventions. Both surveys now include an increasing number of optional modules and complementary data collection tools for use in individual countries. Table 1 summarizes the characteristics of both survey programmes, including questionnaire content, which is typically decided through collaboration among government agencies, donors, key stakeholders, and DHS/MICS. Table 2 lists the major differences between DHS and MICS surveys and Box 1 highlights a particularly important difference between DHS and MICS surveys—the way in which they handle information on orphans and foster children. 10.1371/journal.pmed.1001391.t001 Table 1 Characteristics of the DHS and MICS survey programmes. Characteristics DHS MICS Content of “core" questionnaires and modules (2012) Both surveys Fertility and family planning; infant and child mortality; maternal mortality; antenatal care (number of visits, provider, components of antenatal care, intermittent preventive treatment for malaria during pregnancy); delivery care (place of birth, delivery assistance, cesarean section, birth weight, birth size); postnatal care (postnatal care visits, timing of visits, type of provider); child protection (birth registration, child marriage); child feeding practices (prelacteal feed, breastfeeding, diet); child immunisation coverage; childhood fever, acute respiratory infections, diarrhoea (prevalence, care-seeking behaviour, place and type of treatment); children's living arrangements; malaria (ownership and use of mosquito nets, treatment of fever, indoor residual spraying against mosquitoes, malaria diagnosis); HIV (knowledge of transmission and prevention, prior testing, stigma, and discrimination); sexual behaviour; female genital cutting; environmental health (water, sanitation, handwashing, disposal of children's stools, cooking fuel); biomarkers (height, weight) See DHS Primarily one survey Vitamin A supplementation, iron supplementation, sexually transmitted infections other than HIV (self reports, symptoms), exposure to second-hand smoke, biomarkers including tests for anaemia, HIV, and malaria, timing of antenatal care visits, domestic violence, fistula, women's empowerment Child labour, child discipline, early child development, knowledge of danger signs for child illness Complementary protocols (2012) MIS, AIS, SPA surveys, KIS Guidelines for survey implementation Length of interviewer training, including field practice 4 weeks 3 weeks Composition of field teams Supervisor, field editor, and four interviewers who are the same sex as the respondents See DHS Health technician(s) for biomarker testing Separate measurer for anthropometry Software package used for primary data processing CSPro CSPro Imputation and data analysis CSPro CSPro→SPSS Preparation of report In-country report writing workshop Regional workshops, in-country support Technical assistance Technical assistance visits by ICF International Regional workshops, in-country support, regional coordinators Characteristics of nationally representative surveys conducted in 2011 Typical duration of fieldwork 3–6 months 2–4 months Mean number of households Around 15,000 households Around 10,000 households Average time between completion of data collection and release of the report 3 months for Preliminary Report, 10–12 months for Final Report 12–13 months for Final Report Other characteristics Free public access to datasets www.measuredhs.com www.childinfo.org Easy access to survey results STATcompiler MICS Compiler AIS, AIDS Indicator Surveys; KIS, Key Indicator Surveys; MIS, Malaria Indicator Surveys; SPA, Service Provision Assessment Surveys. 10.1371/journal.pmed.1001391.t002 Table 2 Differences between standard DHS and MICS protocols and their potential implications for coverage measurement. Characteristics DHS MICS Potential Implications for Coverage Measurement Sampling and survey design Sample size per cluster Rural: 30–40 women; Urban: 20–25 women 15–30 households — Construction of household rosters All usual members of the household plus visitors who spent the previous night in the household. DHS tables on coverage measurement are based on de facto persons in the household (that is, persons who stayed in the household the previous night). All usual members of the household (de jure household members) included. De facto approach gives better representation of mobile populations. De jure approach is more consistent with selection probabilities based on censuses. Unlikely to lead to any bias, since response rates remain very high in both approaches. Respondents for information about children less than 5 years of age Biological mothers only except for anthropometric indicators and anaemia, which are collected for all children. Mothers or primary caregivers of children under 5 living in the household. Inclusion of caregivers means orphans and foster children are included in the samples for MNCH coverage estimates for MICS, and not for DHS. See Box 1 for implications for coverage measurement. Reference periods for selected MNCH coverage indicators Skilled attendance at delivery All births during the past 5 years Last birth during the past 2 years The advantage of a shorter reference period is that the coverage estimates refer to a more recent date; on the other hand, the sample size is reduced when the reference period is shorter, which increases the confidence intervals. Antenatal care Last birth during the past 5 years Last birth during the past 2 years See above Tetanus toxoid Last birth during the past 5 years Last birth during the past 2 years See above Initial Breastfeeding Last birth in the past 5 years Last birth during the past 2 years See above Exclusive breastfeeding Youngest child age 0–4 years living with the mother All living children age 0–4 years See above Postnatal care Last birth during the past 5 years Last birth during the past 2 years See above Birth weight All births in the past 5 years Last birth during the past 2 years See above MNCH, maternal, newborn, and child health. Box 1. Orphans and Foster Children An important difference between MICS and DHS surveys is in the collection of information on children under 5. In MICS surveys, information on children under 5 is collected from mothers or primary caregivers of children under 5 in the household, making it possible to collect information on all children under 5 (including orphans and foster children), regardless of whether their biological mothers are in the same household. In DHS surveys, the bulk of information on children under 5 is collected from their biological mothers in the Woman's Questionnaire. Therefore, information on some coverage indicators is not collected for children under 5 who are orphaned or not living with their biological mothers. An exception is the collection of anthropometric measurements and biomarkers for all children under 5 in the household, regardless of the survival status or whereabouts of their biological mothers. In a recent analysis of 12 DHS and MICS surveys in 12 countries (unpublished), we found that the inclusion or exclusion of orphaned and fostered children does not have a substantial influence on national estimates of undernutrition. In three of the surveys, there was more than a 5 percentage point difference in the prevalence of underweight between children living with their biological mothers and other children. However, in seven of the surveys, the difference was less than 2 percentage points. In all 12 surveys, the difference between the national estimate of the prevalence of underweight for all children under 5 and for children whose mothers were interviewed was negligible. The differentials may vary for other indicators, so additional research would be desirable. Since their inception, DHS and MICS surveys have played an important role in shaping the global agenda on tracking coverage and in populating global databases. They have also influenced policies and intervention strategies. For example, DHS/MICS data are often used to establish targets in national economic and social development plans, to provide advocacy for programmes to improve women's and children's health, and to assist programmes in identifying target groups in most need of interventions. The role that these data play at the national and international level make it imperative that data quality is the foremost consideration when designing surveys and providing estimates of key indicators. In the following sections, therefore, we draw on DHS and MICS experience to highlight the challenges associated with measuring coverage through household surveys. Basic Principles and Survey Design Valid measurement of coverage requires, first and foremost, representative population samples based on scientific probability sampling. We will discuss this essential aspect of coverage measurement in the next section of our review but first we will present three other key considerations that need to be taken into account when planning and conducting household surveys and when using their results. First, some information is simply not amenable to collection through a household survey, because respondents do not know or cannot recall the required information. For example, a parent cannot know a child's birthweight if the child was not weighed at birth and is unlikely to remember the exact timing of tetanus toxoid vaccinations received over a lifetime. Decisions about what questions respondents can answer with acceptable accuracy are generally made by survey designers in consultation with technical experts; the research papers in this collection provide some of the first rigorous research assessing the validity of self-reported exposure to RMNCH interventions [17],[18],[25]–[28]. For coverage questions that cannot be answered through household surveys, alternative methods should be explored. Second, survey interview length needs to be taken into account. In principle, a survey interview should be long enough to engage the respondent and obtain complete answers to the survey questions without rushing, but not so long that respondents become bored or frustrated. Completion of all questionnaires in a household averages around 2 hours for DHS and MICS surveys. Technical staff at DHS and MICS are concerned that the current protocols—especially when optional modules, biomarkers, and country-specific questions are included—are approaching lengths that adversely affect data quality. The need to keep survey interviews to a reasonable length is an important source of tension when discussing the addition of new questions with stakeholders. Third, response rates are a concern in household surveys as low response rates can adversely affect the representativeness of the interviewed sample. Efforts need to be made to ensure that non-response is as low as possible. Non-response is well below 10% in DHS and MICS surveys for most countries. Even though DHS and MICS protocols specifically prohibit the substitution of households, response rates remain high thanks to training protocols emphasizing multiple revisits to households and close monitoring of response rates by field staff. Survey Sampling The sample design determines the representativeness of household survey results, which is required to produce coverage estimates for the general population. Textbooks have been written about survey sampling (e.g., [29],[30]) and we will not attempt to reduce them to a few paragraphs here. We recommend strongly that all survey planning teams should include an experienced technical sampling expert. Moreover, it is important not to underestimate the time involved in doing sampling properly. DHS [31] and MICS [32] surveys adhere to the fundamentals of scientific sampling, including complete coverage of the target population, use of suitable sample sizes, the need to conduct a new household listing and pre-selection of sample households, and preparation of appropriate sample documentation. However, deviations from standard procedures are sometimes required owing to cost limitations and practical considerations, including security concerns. Importantly, coverage levels in household surveys that use uncontrolled non-probability sampling (such as quota sampling or purposive sampling) do not provide valid estimates of population coverage. DHS and MICS sampling frames are limited to the population residing in fixed households, and exclude populations living in group quarters (such as military barracks, hospitals, and hotels) and persons living on the street. This is a potential source of bias, because those populations are likely to have different characteristics and health conditions from those living in households. One very practical lesson learned through the DHS and MICS experience is that sample households should be selected in the central office, not in the field, if possible. Only preselected households should be eligible for interviewing, with no substitutions allowed during the fieldwork. This prevents interviewing teams from reducing their workload by avoiding listing or interviewing large or more remote households. A key sampling question is how many households/individuals should be interviewed per sample cluster. DHS recommends a sample “take" of about 30–40 women per rural cluster and 20–25 women per urban cluster. MICS recommends 15–30 households per cluster, although detailed analyses of available budgets, logistical limitations, survey content, and information on sampling errors from previous MICS and DHS surveys are also undertaken on a case-by-case basis. Currently, DHS and MICS surveys include around 15,000 and 10,000 households respectively, which is a sufficient sample size to produce statistically reliable estimates of most indicators at the national, urban–rural, and regional levels, but not at lower administrative levels, such as districts, slums, and small population groups. However, both survey programmes can oversample population groups or geographical areas to produce statistically valid estimates, when needed. Some DHS and MICS surveys have large enough sample sizes to produce estimates at lower administrative levels, but the added value of producing coverage estimates at these administrative levels with larger samples needs to be weighed against the added logistical and management challenges. Unfortunately, the use of sampling weights is often misunderstood, and consequently misused or ignored. Some strata (such as urban areas or selected regions or provinces) are often oversampled to ensure that the final dataset includes enough cases to produce reliable results. In these cases, sampling weights need to be applied to account for varying design weights and non-response levels. Generally, analyses of survey data should use sampling weights calculated for each interviewed household or individual respondent. There is a lack of consensus, however, about whether weights should be used in multivariate analyses, and the decision is often based on the purpose of the analysis [33],[34]. Comparability of Measurement over Time and between DHS and MICS Tracking changes in RMNCH coverage over time is a valuable way to assess progress and is one of the most widespread applications of DHS/MICS data. But there are many pitfalls, and often coverage estimates are taken from survey reports and plotted to show changes without sufficient attention being paid to comparability. Notably, RMNCH coverage indicators change over time in response to modifications in policies and programmes and as lessons are learned about measurement and interpretation. Other papers in this collection report on multiple changes in indicators of diarrhoea case management since 1990 [19], more recent changes in treatment of childhood malaria [18], and postnatal care for mothers and infants [35]. Rather than abstracting coverage data from published survey reports, users should access the UNICEF childinfo.org database, in which all standard global indicators have been checked for comparability and recalculated where necessary. If users obtain DHS and MICS data files themselves to recalculate coverage indicators over time, care must be taken to use standard indicator definitions and appropriate sampling weights. MICS and DHS collaborate closely and work through interagency processes to ensure that their survey tools are harmonized and comparable and their data can, therefore, be combined in global databases covering a large majority of developing countries. The article in this collection by Requejo, Newby, and Bryce highlights the importance of using standard methods to produce comparable coverage data across countries [36]. Differences between DHS and MICS surveys that may affect RMNCH coverage estimates are presented in Table 2, which also summarizes the evidence on the possible magnitude of the effects of these differences. Finally, as mentioned earlier, Box 1 highlights a particularly important difference between DHS and MICS surveys—their inclusion or exclusion of information on orphans and foster children. Challenges in Survey Implementation Accumulated experience from the two survey programmes underscores the importance of incorporating quality control mechanisms at every step in the survey process. Some data quality steps are described elsewhere [36]; here we focus on several common field problems that could affect the quality of RMNCH coverage estimates. One of the major field problem concerns is related to interviewer training and supervision. Table 1 presents information about how certain aspects of survey implementation are dealt with in MICS and DHS surveys. Both programmes have minimum requirements for selecting interviewers (at least a high school diploma). Moreover, interviewers are not directly involved in the management/provision of health services to avoid potential conflicts of interest. Importantly, the MICS programme was originally intended to be a relatively quick and light exercise at the country level, with limited in-country technical support. The MICS programme was designed to produce the limited data needed to monitor progress towards the World Summit for Children Goals. Over time, however, concerns about quality issues associated with the expanded questionnaire length and increased sample sizes have led MICS to develop a technical support system that emphasizes the duration and content of interviewer training and field supervision, but also includes rigorous controls on data entry and checking. These changes bring the quality of MICS data more in line with that of DHS (which has always included substantial centralized technical support). Another important quality assurance measure for both survey programmes concerns the entry of data for paper-based questionnaires simultaneously with the fieldwork. For both paper-based questionnaires and surveys conducted with personal data assistants (PDAs) or tablets, field check tables are produced while the fieldwork teams are still in the field, thereby making it possible to spot systematic errors in data collection and take measures to improve data quality. However, despite meticulous attention to quality control, all household surveys are prone to quality problems, the most important of which we will now highlight. One of the most challenging aspects of implementing DHS and MICS protocols is how to ensure that these protocols are appropriately adapted to the country context. Despite the availability of technical guidelines and assistance, our experience suggests that problems frequently occur in this area. One clear example is the adaptation of response categories for diarrhoea treatment at home. The standard response category for “government-recommended home fluids" is often not customized to reflect country-specific recommendations, in many cases because of the lack of clear national policies or the long list of recommended fluids [19]. Another example is described by Hazel and colleagues in this collection [21] who show that coverage gains attributable to recently adopted strategies for treating childhood illness at the community level cannot be assessed unless the response options for care seeking are adapted to reflect the country-specific providers of care at the community level. Ensuring that standard survey protocols are adapted to specific country contexts is a continuing challenge, especially given the numerous topics included in each survey and the time and resources needed to adapt questionnaires in each country. Another implementation challenge relates to what demographers working on child mortality refer to as the tendency for reports of child deaths to “heap" at certain ages, notably 12 months [37]. If this tendency is also true for reports of service utilisation and receipt of interventions, it could affect the accuracy of coverage estimates with time-bounded effectiveness, such as the timing of antenatal or postnatal care. We are not aware of any research that investigates the effects of age heaping in coverage measurement, and believe that this warrants further attention. Finally, a related problem is the tendency for interviewers to “transfer" children to age groups (especially over 5 years of age) that exclude them from lengthy portions of the interview and therefore reduce interviewer workload [37]. Even if only some interviewers modify children's ages in this way occasionally, it can affect coverage estimates, especially if the children whose ages are changed systematically have different characteristics from other children. This is a continuing problem in some surveys despite serious attempts to minimize abuses. Age displacement may have more of an effect on fertility and mortality estimates than RMNCH estimates, but displacement remains a matter of concern overall. Conclusions Large-scale, nationally representative household surveys are the primary source of data on RMNCH coverage. Despite efforts to improve routine information systems, surveys are likely to remain the primary source of data for many years to come. As we discuss in this review, it is essential that survey planners and consumers of survey results understand the strengths, limitations, and constraints of coverage measurements generated through household surveys and also appreciate the technical issues involved in sampling and quality control. Moreover, as our review and the other articles in this collection highlight, conducting household surveys that generate valid and reliable information on coverage is a complex exercise. We believe that the findings of current and future experimental studies will help to inform continuing efforts to improve coverage measurement in household surveys, particularly in the areas of improved question wording and interviewer training. Finally, we stress that calls for more and better data on coverage must be accompanied by sufficient resources and by an ongoing research programme to continue to improve and refine methods and analytical techniques. Key Points DHS and MICS surveys are the principal source of national-level data on maternal, newborn, and child health indicators in low- and middle-income countries. Despite efforts to improve routine information systems, household surveys are likely to remain the primary source of population data for the foreseeable future. To analyse coverage estimates over time and across countries from household survey data, it is essential that similar survey methods and questions be employed to ensure comparability. The strengths and weaknesses of all data collection efforts need to be transparent and well understood by data users. DHS and MICS will continue to benefit from research findings to make further improvements in the collection of reproductive, maternal and child health coverage data.
- Record: found
- Abstract: found
- Article: not found
Artesunate combinations for treatment of malaria: meta-analysis.
P Olliaro, Jonathan Taylor, Spencer N. White … (2004)
Addition of artemisinin derivatives to existing drug regimens for malaria could reduce treatment failure and transmission potential. We assessed the evidence for this hypothesis from randomised controlled trials. We undertook a meta-analysis of individual patients' data from 16 randomised trials (n=5948) that studied the effects of the addition of artesunate to standard treatment of Plasmodium falciparum malaria. We estimated odds ratios (OR) of parasitological failure at days 14 and 28 (artesunate combination compared with standard treatment) and calculated combined summary ORs across trials using standard methods. For all trials combined, parasitological failure was lower with 3 days of artesunate at day 14 (OR 0.20, 95% CI 0.17-0.25, n=4504) and at day 28 (excluding new infections, 0.23, 0.19-0.28, n=2908; including re-infections, 0.30, 0.26-0.35, n=4332). Parasite clearance was significantly faster (rate ratio 1.98, 95% CI 1.85-2.12, n=3517) with artesunate. In participants with no gametocytes at baseline, artesunate reduced gametocyte count on day 7 (OR 0.11, 95% CI 0.09-0.15, n=2734), with larger effects at days 14 and 28. Adding artesunate for 1 day (six trials) was associated with fewer failures by day 14 (0.61, 0.48-0.77, n=1980) and day 28 (adjusted to exclude new infections 0.68, 0.53-0.89, n=1205; unadjusted including reinfections 0.77, 0.63-0.95, n=1958). In these trials, gametocytes were reduced by day 7 (in participants with no gametocytes at baseline 0.11, 0.09-0.15, n=2734). The occurrence of serious adverse events did not differ significantly between artesunate and placebo. The addition of 3 days of artesunate to standard antimalarial treatments substantially reduce treatment failure, recrudescence, and gametocyte carriage.
Author and article information
Contributors
Adam Bennett
Donal Bisanzio
Joshua O Yukich
Bonnie Mappin
Cristin A Fergus
Michael Lynch
Richard E Cibulskis
Samir Bhatt
Daniel J Weiss
Ewan Cameron
Peter W Gething
Thomas P Eisele: Role: Prof
Journal
Journal ID (nlm-journal-id): 101613665
Journal ID (pubmed-jr-id): 42402
Journal ID (nlm-ta): Lancet Glob Health
Journal ID (iso-abbrev): Lancet Glob Health
Title:
The Lancet. Global health
ISSN
(Electronic):
2214-109X
Publication date Nihms-submitted: 5
April
2017
Publication date
(Print):
April
2017
Publication date PMC-release: 01
April
2018
Volume: 5
Issue: 4
Pages: e418-e427
Affiliations
Malaria Elimination Initiative, Global Health Group, University of San Francisco,
San Francisco, CA, USA
Oxford Big Data Institute, Li Ka Shing Centre for Health Information and Discovery,
Nuffield Department of Medicine, University of Oxford, Oxford, UK
Center for Applied Malaria Research and Evaluation, Tulane University School of Public
Health and Tropical Medicine, New Orleans, LA, USA
Oxford Big Data Institute, Li Ka Shing Centre for Health Information and Discovery,
Nuffield Department of Medicine, University of Oxford, Oxford, UK
Oxford Big Data Institute, Li Ka Shing Centre for Health Information and Discovery,
Nuffield Department of Medicine, University of Oxford, Oxford, UK
Oxford Big Data Institute, Li Ka Shing Centre for Health Information and Discovery,
Nuffield Department of Medicine, University of Oxford, Oxford, UK
Oxford Big Data Institute, Li Ka Shing Centre for Health Information and Discovery,
Nuffield Department of Medicine, University of Oxford, Oxford, UK
Author notes
Correspondence to: Dr Adam Bennett, Malaria Elimination Initiative, Global Health
Group, University of San Francisco, San Francisco, CA 94158, USA,
adam.bennett@
123456ucsf.edu
Article
Manuscript ID: HHSPA860032DOI: 10.1016/S2214-109X(17)30076-1
PMC ID: 5450656
PubMed ID: 28288746
SO-VID: aeb6a194-fcb8-4ea4-93af-c31367c8be47
License:
This is an Open Access article under the CC BY license.
History
Categories
Subject:
Article
Data availability: