Spatio-temporal modelling of routine health facility data for malaria risk micro-stratification in mainland Tanzania

As malaria transmission declines, the need to monitor the heterogeneity of malaria risk at finer scales becomes critical to guide community-based targeted interventions. Although routine health facility (HF) data can provide epidemiological evidence at high spatial and temporal resolution, its incomplete nature of information can result in lower administrative units without empirical data. To overcome geographic sparsity of data and its representativeness, geo-spatial models can leverage routine information to predict risk in un-represented areas as well as estimate uncertainty of predictions. Here, a Bayesian spatio-temporal model was applied on malaria test positivity rate (TPR) data for the period 2017-2019 to predict risks at the ward level, the lowest decision-making unit in mainland Tanzania. To quantify the associated uncertainty, the probability of malaria TPR exceeding programmatic threshold was estimated. Results showed a marked spatial heterogeneity in malaria TPR across wards. 17.7 million people resided in areas where malaria TPR was high (≥ 30; 90% certainty) in the North-West and South-East parts of Tanzania. Approximately 11.7 million people lived in areas where malaria TPR was very low (< 5%; 90% certainty). HF data can be used to identify different epidemiological strata and guide malaria interventions at micro-planning units in Tanzania. These data, however, are imperfect in many settings in Africa and often require application of geo-spatial modelling techniques for estimation.

Geographic accessibility and hospital competition for emergency blood transfusion services in Bungoma, Western Kenya

BACKGROUND

Estimating accessibility gaps to essential health interventions helps to allocate and prioritize health resources. Access to blood transfusion represents an important emergency health requirement. Here, we develop geo-spatial models of accessibility and competition to blood transfusion services in Bungoma County, Western Kenya.

METHODS

Hospitals providing blood transfusion services in Bungoma were identified from an up-dated geo-coded facility database. AccessMod was used to define care-seeker's travel times to the nearest blood transfusion service. A spatial accessibility index for each enumeration area (EA) was defined using modelled travel time, population demand, and supply available at the hospital, assuming a uniform risk of emergency occurrence in the county. To identify populations marginalized from transfusion services, the number of people outside 1-h travel time and those residing in EAs with low accessibility indexes were computed at the sub-county level. Competition between the transfusing hospitals was estimated using a spatial competition index which provided a measure of the level of attractiveness of each hospital. To understand whether highly competitive facilities had better capacity for blood transfusion services, a correlation test between the computed competition metric and the blood units received and transfused at the hospital was done.

RESULTS

15 hospitals in Bungoma county provide transfusion services, however these are unevenly distributed across the sub-counties. Average travel time to a blood transfusion centre in the county was 33 min and 5% of the population resided outside 1-h travel time. Based on the accessibility index, 38% of the EAs were classified to have low accessibility, representing 34% of the population, with one sub-county having the highest marginalized population. The computed competition index showed that hospitals in the urban areas had a spatial competitive advantage over those in rural areas.

CONCLUSION

The modelled spatial accessibility has provided an improved understanding of health care gaps essential for health planning. Hospital competition has been illustrated to have some degree of influence in provision of health services hence should be considered as a significant external factor impacting the delivery, and re-design of available services.

Quality of maternal healthcare and travel time influence birthing service utilisation in Ghanaian health facilities: a geographical analysis of routine health data

OBJECTIVES

To investigate how the quality of maternal health services and travel times to health facilities affect birthing service utilisation in Eastern Region, Ghana.

DESIGN

The study is a cross-sectional spatial interaction analysis of birth service utilisation patterns. Routine birth data were spatially linked to quality care, service demand and travel time data.

SETTING

131 Health facilities (public, private and faith-based) in 33 districts in Eastern Region, Ghana.

PARTICIPANTS

Women who gave birth in health facilities in the Eastern Region, Ghana in 2017.

OUTCOME MEASURES

The count of women giving birth, the quality of birthing care services and the geographic coverage of birthing care services.

RESULTS

As travel time from women's place of residence to the health facility increased up to two2 hours, the utilisation rate markedly decreased. Higher quality of maternal health services haves a larger, positive effect on utilisation rates than service proximity. The quality of maternal health services was higher in hospitals than in primary care facilities. Most women (88.6%) travelling via mechanised transport were within two2 hours of any birthing service. The majority (56.2%) of women were beyond the two2 -hour threshold of critical comprehensive emergency obstetric and newborn care (CEmONC) services. Few CEmONC services were in urban centres, disadvantaging rural populations.

CONCLUSIONS

To increase birthing service utilisation in Ghana, higher quality health facilities should be located closer to women, particularly in rural areas. Beyond Ghana, routinely collected birth records could be used to understand the interaction of service proximity and quality.

Insecticide-treated net distribution in Western Kenya: impacts related to COVID-19 and health worker strikes

We examined the impact of coronavirus disease (COVID) mitigation, supply and distribution interruptions on the delivery of long-lasting insecticide-treated nets (LLINs) in Western Kenya. The median monthly distribution of LLINs declined during COVID mitigation strategies (March-July 2020) and during the health worker strikes (December 2020-February 2021). Recovery periods followed initial declines, indicative of a 'catching up' on missed routine distribution. Mass community campaigns were delayed by 10 months. These observations offer encouragement for routine net distribution resilience, but complete interruptions of planned mass distributions require alternate strategies during pandemics.

Unmet need for COVID-19 vaccination coverage in Kenya

COVID-19 has impacted the health and livelihoods of billions of people since it emerged in 2019. Vaccination for COVID-19 is a critical intervention that is being rolled out globally to end the pandemic. Understanding the spatial inequalities in vaccination coverage and access to vaccination centres is important for planning this intervention nationally. Here, COVID-19 vaccination data, representing the number of people given at least one dose of vaccine, a list of the approved vaccination sites, population data and ancillary GIS data were used to assess vaccination coverage, using Kenya as an example. Firstly, physical access was modelled using travel time to estimate the proportion of population within 1 hour of a vaccination site. Secondly, a Bayesian conditional autoregressive (CAR) model was used to estimate the COVID-19 vaccination coverage and the same framework used to forecast coverage rates for the first quarter of 2022. Nationally, the average travel time to a designated COVID-19 vaccination site (n = 622) was 75.5 min (Range: 62.9 - 94.5 min) and over 87% of the population >18 years reside within 1 hour to a vaccination site. The COVID-19 vaccination coverage in December 2021 was 16.70% (95% CI: 16.66 - 16.74) - 4.4 million people and was forecasted to be 30.75% (95% CI: 25.04 - 36.96) - 8.1 million people by the end of March 2022. Approximately 21 million adults were still unvaccinated in December 2021 and, in the absence of accelerated vaccine uptake, over 17.2 million adults may not be vaccinated by end March 2022 nationally. Our results highlight geographic inequalities at sub-national level and are important in targeting and improving vaccination coverage in hard-to-reach populations. Similar mapping efforts could help other countries identify and increase vaccination coverage for such populations.

Understanding factors associated with attending secondary school in Tanzania using household survey data

BACKGROUND

Sustainable Development Goal (SDG) 4 aims to ensure inclusive and equitable access for all by 2030, leaving no one behind. One indicator selected to measure progress towards achievement is the participation rate of youth in education (SDG 4.3.1). Here we aim to understand drivers of school attendance using one country in East Africa as an example.

METHODS

Nationally representative household survey data (2015-16 Tanzania Demographic and Health Survey) were used to explore individual, household and contextual factors associated with secondary school attendance in Tanzania. These included, age, head of household's levels of education, gender, household wealth index and total number of children under five. Contextual factors such as average pupil to qualified teacher ratio and geographic access to school were also tested at cluster level. A two-level random intercept logistic regression model was used in exploring association of these factors with attendance in a multi-level framework.

RESULTS

Age of household head, educational attainments of either of the head of the household or parent, child characteristics such as gender, were important predictors of secondary school attendance. Being in a richer household and with fewer siblings of lower age (under the age of 5) were associated with increased odds of attendance (OR = 0.91, CI 95%: 0.86; 0.96). Contextual factors were less likely to be associated with secondary school attendance.

CONCLUSIONS

Individual and household level factors are likely to impact secondary school attendance rates more compared to contextual factors, suggesting an increased focus of interventions at these levels is needed. Future studies should explore the impact of interventions targeting these levels. Policies should ideally promote gender equality in accessing secondary school as well as support those families where the dependency ratio is high. Strategies to reduce poverty will also increase the likelihood of attending school.

Malaria hospitalisation in East Africa: age, phenotype and transmission intensity

BACKGROUND

Understanding the age patterns of disease is necessary to target interventions to maximise cost-effective impact. New malaria chemoprevention and vaccine initiatives target young children attending routine immunisation services. Here we explore the relationships between age and severity of malaria hospitalisation versus malaria transmission intensity.

METHODS

Clinical data from 21 surveillance hospitals in East Africa were reviewed. Malaria admissions aged 1 month to 14 years from discrete administrative areas since 2006 were identified. Each site-time period was matched to a model estimated community-based age-corrected parasite prevalence to provide predictions of prevalence in childhood (PfPR2-10). Admission with all-cause malaria, severe malaria anaemia (SMA), respiratory distress (RD) and cerebral malaria (CM) were analysed as means and predicted probabilities from Bayesian generalised mixed models.

RESULTS

52,684 malaria admissions aged 1 month to 14 years were described at 21 hospitals from 49 site-time locations where PfPR2-10 varied from < 1 to 48.7%. Twelve site-time periods were described as low transmission (PfPR2-10 < 5%), five low-moderate transmission (PfPR2-10 5-9%), 20 moderate transmission (PfPR2-10 10-29%) and 12 high transmission (PfPR2-10 ≥ 30%). The majority of malaria admissions were below 5 years of age (69-85%) and rare among children aged 10-14 years (0.7-5.4%) across all transmission settings. The mean age of all-cause malaria hospitalisation was 49.5 months (95% CI 45.1, 55.4) under low transmission compared with 34.1 months (95% CI 30.4, 38.3) at high transmission, with similar trends for each severe malaria phenotype. CM presented among older children at a mean of 48.7 months compared with 39.0 months and 33.7 months for SMA and RD, respectively. In moderate and high transmission settings, 34% and 42% of the children were aged between 2 and 23 months and so within the age range targeted by chemoprevention or vaccines.

CONCLUSIONS

Targeting chemoprevention or vaccination programmes to areas where community-based parasite prevalence is ≥10% is likely to match the age ranges covered by interventions (e.g. intermittent presumptive treatment in infancy to children aged 2-23 months and current vaccine age eligibility and duration of efficacy) and the age ranges of highest disease burden.

Plasmodium falciparum parasite prevalence in East Africa: Updating data for malaria stratification

The High Burden High Impact (HBHI) strategy for malaria encourages countries to use multiple sources of available data to define the sub-national vulnerabilities to malaria risk, including parasite prevalence. Here, a modelled estimate of Plasmodium falciparum from an updated assembly of community parasite survey data in Kenya, mainland Tanzania, and Uganda is presented and used to provide a more contemporary understanding of the sub-national malaria prevalence stratification across the sub-region for 2019. Malaria prevalence data from surveys undertaken between January 2010 and June 2020 were assembled form each of the three countries. Bayesian spatiotemporal model-based approaches were used to interpolate space-time data at fine spatial resolution adjusting for population, environmental and ecological covariates across the three countries. A total of 18,940 time-space age-standardised and microscopy-converted surveys were assembled of which 14,170 (74.8%) were identified after 2017. The estimated national population-adjusted posterior mean parasite prevalence was 4.7% (95% Bayesian Credible Interval 2.6-36.9) in Kenya, 10.6% (3.4-39.2) in mainland Tanzania, and 9.5% (4.0-48.3) in Uganda. In 2019, more than 12.7 million people resided in communities where parasite prevalence was predicted ≥ 30%, including 6.4%, 12.1% and 6.3% of Kenya, mainland Tanzania and Uganda populations, respectively. Conversely, areas that supported very low parasite prevalence (<1%) were inhabited by approximately 46.2 million people across the sub-region, or 52.2%, 26.7% and 10.4% of Kenya, mainland Tanzania and Uganda populations, respectively. In conclusion, parasite prevalence represents one of several data metrics for disease stratification at national and sub-national levels. To increase the use of this metric for decision making, there is a need to integrate other data layers on mortality related to malaria, malaria vector composition, insecticide resistance and bionomic, malaria care-seeking behaviour and current levels of unmet need of malaria interventions.

Model building and assessment of the impact of covariates for disease prevalence mapping in low-resource settings: to explain and to predict

This paper provides statistical guidance on the development and application of model-based geostatistical methods for disease prevalence mapping. We illustrate the different stages of the analysis, from exploratory analysis to spatial prediction of prevalence, through a case study on malaria mapping in Tanzania. Throughout the paper, we distinguish between predictive modelling, whose main focus is on maximizing the predictive accuracy of the model, and explanatory modelling, where greater emphasis is placed on understanding the relationships between the health outcome and risk factors. We demonstrate that these two paradigms can result in different modelling choices. We also propose a simple approach for detecting over-fitting based on inspection of the correlation matrix of the estimators of the regression coefficients. To enhance the interpretability of geostatistical models, we introduce the concept of domain effects in order to assist variable selection and model validation. The statistical ideas and principles illustrated here in the specific context of disease prevalence mapping are more widely applicable to any regression model for the analysis of epidemiological outcomes but are particularly relevant to geostatistical models, for which the separation between fixed and random effects can be ambiguous.

Malaria micro-stratification using routine surveillance data in Western Kenya

BACKGROUND

There is an increasing need for finer spatial resolution data on malaria risk to provide micro-stratification to guide sub-national strategic plans. Here, spatial-statistical techniques are used to exploit routine data to depict sub-national heterogeneities in test positivity rate (TPR) for malaria among patients attending health facilities in Kenya.

METHODS

Routine data from health facilities (n = 1804) representing all ages over 24 months (2018-2019) were assembled across 8 counties (62 sub-counties) in Western Kenya. Statistical model-based approaches were used to quantify heterogeneities in TPR and uncertainty at fine spatial resolution adjusting for missingness, population distribution, spatial data structure, month, and type of health facility.

RESULTS

The overall monthly reporting rate was 78.7% (IQR 75.0-100.0) and public-based health facilities were more likely than private facilities to report ≥ 12 months (OR 5.7, 95% CI 4.3-7.5). There was marked heterogeneity in population-weighted TPR with sub-counties in the north of the lake-endemic region exhibiting the highest rates (exceedance probability > 70% with 90% certainty) where approximately 2.7 million (28.5%) people reside. At micro-level the lowest rates were in 14 sub-counties (exceedance probability < 30% with 90% certainty) where approximately 2.2 million (23.1%) people lived and indoor residual spraying had been conducted since 2017.

CONCLUSION

The value of routine health data on TPR can be enhanced when adjusting for underlying population and spatial structures of the data, highlighting small-scale heterogeneities in malaria risk often masked in broad national stratifications. Future research should aim at relating these heterogeneities in TPR with traditional community-level prevalence to improve tailoring malaria control activities at sub-national levels.

Routine data for malaria morbidity estimation in Africa: challenges and prospects

BACKGROUND

The burden of malaria in sub-Saharan Africa remains challenging to measure relying on epidemiological modelling to evaluate the impact of investments and providing an in-depth analysis of progress and trends in malaria response globally. In malaria-endemic countries of Africa, there is increasing use of routine surveillance data to define national strategic targets, estimate malaria case burdens and measure control progress to identify financing priorities. Existing research focuses mainly on the strengths of these data with less emphasis on existing challenges and opportunities presented.

CONCLUSION

Here we define the current imperfections common to routine malaria morbidity data at national levels and offer prospects into their future use to reflect changing disease burdens.

Severe-malaria infection and its outcomes among pregnant women in Burkina Faso health-districts: Hierarchical Bayesian space-time models applied to routinely-collected data from 2013 to 2018

Fine-scale hotspots detection is crucial for optimum delivery of essential health-services for reducing severe malaria in pregnancy (MiP) and death cases in Burkina Faso. This study used hierarchical-Bayesian Spatio-temporal modeling to explore space-time patterns and pinpoint health-districts with an exceedance probability of severe MiP incidence and fatality rate. Study also assessed effect of health-district service delivery (readiness) on severe-MiP outcomes.

Severe-MiP fatality rate declined considerably while its incidence rate remained unchanged between January-2013 and December-2018. Severe-MiP cases persisted throughout the year with peaks between August and November. These peaks increased 2.5-fold the fatality rate. Furthermore, severe-MiP fatality was higher in health-districts classified as low-readiness (IRR = 2.469, 95%CrI: 1.632–3.738). However, the fatality rate decreased significantly with proper coverage with three doses for intermittent-preventive-treatment with sulphadoxine-pyrimethamine. Severe-MiP burden was heterogeneous spatially and temporally. The study suggested that health-programs should increase health-districts readiness and optimize resource allocation in high burden areas and months.

Estimating hospital catchments from in-patient admission records: a spatial statistical approach applied to malaria

Admission records are seldom used in sub-Saharan Africa to delineate hospital catchments for the spatial description of hospitalised disease events. We set out to investigate spatial hospital accessibility for severe malarial anaemia (SMA) and cerebral malaria (CM). Malaria admissions for children between 1 month and 14 years old were identified from prospective clinical surveillance data recorded routinely at four referral hospitals covering two complete years between December 2015 to November 2016 and November 2017 to October 2018. These were linked to census enumeration areas (EAs) with an age-structured population. A novel mathematical-statistical framework that included EAs with zero observations was used to predict hospital catchment for malaria admissions adjusting for spatial distance. From 5766 malaria admissions, 5486 (95.14%) were linked to specific EA address, of which 272 (5%) were classified as cerebral malaria while 1001 (10%) were severe malaria anaemia. Further, results suggest a marked geographic catchment of malaria admission around the four sentinel hospitals although the extent varied. The relative rate-ratio of hospitalisation was highest at <1-hour travel time for SMA and CM although this was lower outside the predicted hospital catchments. Delineation of catchments is important for planning emergency care delivery and in the use of hospital data to define epidemiological disease burdens. Further hospital and community-based studies on treatment-seeking pathways to hospitals for severe disease would improve our understanding of catchments.

A spatial database of health facilities managed by the public health sector in sub Saharan Africa

Health facilities form a central component of health systems, providing curative and preventative services and structured to allow referral through a pyramid of increasingly complex service provision. Access to health care is a complex and multidimensional concept, however, in its most narrow sense, it refers to geographic availability. Linking health facilities to populations has been a traditional per capita index of heath care coverage, however, with locations of health facilities and higher resolution population data, Geographic Information Systems allow for a more refined metric of health access, define geographic inequalities in service provision and inform planning. Maximizing the value of spatial heath access requires a complete census of providers and their locations. To-date there has not been a single, geo-referenced and comprehensive public health facility database for sub-Saharan Africa. We have assembled national master health facility lists from a variety of government and non-government sources from 50 countries and islands in sub Saharan Africa and used multiple geocoding methods to provide a comprehensive spatial inventory of 98,745 public health facilities.

Mapping vaccination coverage to explore the effects of delivery mechanisms and inform vaccination strategies

The success of vaccination programs depends largely on the mechanisms used in vaccine delivery. National immunization programs offer childhood vaccines through fixed and outreach services within the health system and often, additional supplementary immunization activities (SIAs) are undertaken to fill gaps and boost coverage. Here, we map predicted coverage at 1 × 1 km spatial resolution in five low- and middle-income countries to identify areas that are under-vaccinated via each delivery method using Demographic and Health Surveys data. We compare estimates of the coverage of the third dose of diphtheria-tetanus-pertussis-containing vaccine (DTP3), which is typically delivered through routine immunization (RI), with those of measles-containing vaccine (MCV) for which SIAs are also undertaken. We find that SIAs have boosted MCV coverage in some places, but not in others, particularly where RI had been deficient, as depicted by DTP coverage. The modelling approaches outlined here can help to guide geographical prioritization and strategy design.

Using parasite genetic and human mobility data to infer local and cross-border malaria connectivity in Southern Africa

Local and cross-border importation remain major challenges to malaria elimination and are difficult to measure using traditional surveillance data. To address this challenge, we systematically collected parasite genetic data and travel history from thousands of malaria cases across northeastern Namibia and estimated human mobility from mobile phone data. We observed strong fine-scale spatial structure in local parasite populations, providing positive evidence that the majority of cases were due to local transmission. This result was largely consistent with estimates from mobile phone and travel history data. However, genetic data identified more detailed and extensive evidence of parasite connectivity over hundreds of kilometers than the other data, within Namibia and across the Angolan and Zambian borders. Our results provide a framework for incorporating genetic data into malaria surveillance and provide evidence that both strengthening of local interventions and regional coordination are likely necessary to eliminate malaria in this region of Southern Africa.

Spatial distribution and determinants of asymptomatic malaria risk among children under 5 years in 24 districts in Burkina Faso

BACKGROUND

In malaria endemic countries, asymptomatic cases constitute an important reservoir of infections sustaining transmission. Estimating the burden of the asymptomatic population and identifying areas with elevated risk is important for malaria control in Burkina Faso. This study analysed the spatial distribution of asymptomatic malaria infection among children under 5 in 24 health districts in Burkina Faso and identified the determinants of this distribution.

METHODS

The data used in this study were collected in a baseline survey on "evaluation of the impact of pay for performance on the quality of care" conducted in 24 health districts in Burkina Faso, between October 2013 and March 2014. This survey involved 7844 households and 1387 community health workers. A Bayesian hierarchical logistic model that included spatial dependence and covariates was implemented to identify the determinants of asymptomatic malaria infection. The posterior probability distribution of a parameter from the model was summarized using odds ratio (OR) and 95% credible interval (95% CI).

RESULTS

The overall prevalence of asymptomatic malaria infection in children under 5 years of age was estimated at 38.2%. However, significant variation was observed between districts ranging from 11.1% in the district of Barsalgho to 77.8% in the district of Gaoua. Older children (48-59 vs < 6 months: OR: 6.79 [5.62, 8.22]), children from very poor households (Richest vs poorest: OR: 0.85 [0.74-0.96]), households located more than 5 km from a health facility (< 5 km vs  ≥ 5 km: OR: 1.14 [1.04-1.25]), in localities with inadequate number of nurses (< 3 vs  ≥ 3: 0.72 [0.62, 0.82], from rural areas (OR: 1.67 [1.39-2.01]) and those surveyed in high transmission period of asymptomatic malaria (OR: 1.27 [1.10-1.46]) were most at risk for asymptomatic malaria infection. In addition, the spatial analysis identified the following nine districts that reported significantly higher risks: Batié, Boromo, Dano, Diébougou, Gaoua, Ouahigouya, Ouargaye, Sapouy and Toma. The district of Zabré reported the lowest risk.

CONCLUSION

The analysis of spatial distribution of infectious reservoir allowed the identification of risk areas as well as the identification of individual and contextual factors. Such national spatial analysis should help to prioritize areas for increased malaria control activities.

Temporal trends in spatial inequalities of maternal and newborn health services among four east African countries, 1999-2015

BACKGROUND

Sub-Saharan Africa continues to account for the highest regional maternal mortality ratio (MMR) in the world, at just under 550 maternal deaths per 100,000 live births in 2015, compared to a global rate of 216 deaths. Spatial inequalities in access to life-saving maternal and newborn health (MNH) services persist within sub-Saharan Africa, however, with varied improvement over the past two decades. While previous research within the East African Community (EAC) region has examined utilisation of MNH care as an emergent property of geographic accessibility, no research has examined how these spatial inequalities have evolved over time at similar spatial scales.

METHODS

Here, we analysed temporal trends of spatial inequalities in utilisation of antenatal care (ANC), skilled birth attendance (SBA), and postnatal care (PNC) among four East African countries. Specifically, we used Bayesian spatial statistics to generate district-level estimates of these services for several time points using Demographic and Health Surveys data in Kenya, Tanzania, Rwanda, and Uganda. We examined temporal trends of both absolute and relative indices over time, including the absolute difference between estimates, as well as change in performance ratios of the best-to-worst performing districts per country.

RESULTS

Across all countries, we found the greatest spatial equality in ANC, while SBA and PNC tended to have greater spatial variability. In particular, Rwanda represented the only country to consistently increase coverage and reduce spatial inequalities across all services. Conversely, Tanzania had noticeable reductions in ANC coverage throughout most of the country, with some areas experiencing as much as a 55% reduction. Encouragingly, however, we found that performance gaps between districts have generally decreased or remained stably low across all countries, suggesting countries are making improvements to reduce spatial inequalities in these services.

CONCLUSIONS

We found that while the region is generally making progress in reducing spatial gaps across districts, improvement in PNC coverage has stagnated, and should be monitored closely over the coming decades. This study is the first to report temporal trends in district-level estimates in MNH services across the EAC region, and these findings establish an important baseline of evidence for the Sustainable Development Goal era.

A spatial regression model for the disaggregation of areal unit based data to high-resolution grids with application to vaccination coverage mapping

The growing demand for spatially detailed data to advance the Sustainable Development Goals agenda of ‘leaving no one behind’ has resulted in a shift in focus from aggregate national and province-based metrics to small areas and high-resolution grids in the health and development arena. Vaccination coverage is customarily measured through aggregate-level statistics, which mask fine-scale heterogeneities and ‘coldspots’ of low coverage. This paper develops a methodology for high-resolution mapping of vaccination coverage using areal data in settings where point-referenced survey data are inaccessible. The proposed methodology is a binomial spatial regression model with a logit link and a combination of covariate data and random effects modelling two levels of spatial autocorrelation in the linear predictor. The principal aspect of the model is the melding of the misaligned areal data and the prediction grid points using the regression component and each of the conditional autoregressive and the Gaussian spatial process random effects. The Bayesian model is fitted using the INLA-SPDE approach. We demonstrate the predictive ability of the model using simulated data sets. The results obtained indicate a good predictive performance by the model, with correlations of between 0.66 and 0.98 obtained at the grid level between true and predicted values. The methodology is applied to predicting the coverage of measles and diphtheria-tetanus-pertussis vaccinations at 5 × 5 km² in Afghanistan and Pakistan using subnational Demographic and Health Surveys data. The predicted maps are used to highlight vaccination coldspots and assess progress towards coverage targets to facilitate the implementation of more geographically precise interventions. The proposed methodology can be readily applied to wider disaggregation problems in related contexts, including mapping other health and development indicators.

Geospatial mapping of access to timely essential surgery in sub-Saharan Africa

INTRODUCTION

Despite an estimated one-third of the global burden of disease being surgical, only limited estimates of accessibility to surgical treatment in sub-Saharan Africa exist and these remain spatially undefined. Geographical metrics of access to major hospitals were estimated based on travel time. Estimates were then used to assess need for surgery at country level.

METHODS

Major district and regional hospitals were assumed to have capability to perform bellwether procedures. Geographical locations of hospitals in relation to the population in the 47 sub-Saharan countries were combined with spatial ancillary data on roads, elevation, land use or land cover to estimate travel-time metrics of 30 min, 1 hour and 2 hours. Hospital catchment was defined as population residing in areas less than 2 hours of travel time to the next major hospital. Travel-time metrics were combined with fine-scale population maps to define burden of surgery at hospital catchment level.

RESULTS

Overall, the majority of the population (92.5%) in sub-Saharan Africa reside in areas within 2 hours of a major hospital catchment defined based on spatially defined travel times. The burden of surgery in all-age population was 257.8 million to 294.7 million people and was highest in high-population density countries and lowest in sparsely populated or smaller countries. The estimated burden in children <15 years was 115.3 million to 131.8 million and had similar spatial distribution to the all-age pattern.

CONCLUSION

The study provides an assessment of accessibility and burden of surgical disease in sub-Saharan Africa. Yet given the optimistic assumption of adequare surgical capability of major hospitals, the true burden of surgical disease is expected to be much greater. In-depth health facility assessments are needed to define infrastructure, personnel and medicine supply for delivering timely and safe affordable surgery to further inform the analysis.

High resolution age-structured mapping of childhood vaccination coverage in low and middle income countries

BACKGROUND

The expansion of childhood vaccination programs in low and middle income countries has been a substantial public health success story. Indicators of the performance of intervention programmes such as coverage levels and numbers covered are typically measured through national statistics or at the scale of large regions due to survey design, administrative convenience or operational limitations. These mask heterogeneities and 'coldspots' of low coverage that may allow diseases to persist, even if overall coverage is high. Hence, to decrease inequities and accelerate progress towards disease elimination goals, fine-scale variation in coverage should be better characterized.

METHODS

Using measles as an example, cluster-level Demographic and Health Surveys (DHS) data were used to map vaccination coverage at 1 km spatial resolution in Cambodia, Mozambique and Nigeria for varying age-group categories of children under five years, using Bayesian geostatistical techniques built on a suite of publicly available geospatial covariates and implemented via Markov Chain Monte Carlo (MCMC) methods.

RESULTS

Measles vaccination coverage was found to be strongly predicted by just 4-5 covariates in geostatistical models, with remoteness consistently selected as a key variable. The output 1 × 1 km maps revealed significant heterogeneities within the three countries that were not captured using province-level summaries. Integration with population data showed that at the time of the surveys, few districts attained the 80% coverage, that is one component of the WHO Global Vaccine Action Plan 2020 targets.

CONCLUSION

The elimination of vaccine-preventable diseases requires a strong evidence base to guide strategies and inform efficient use of limited resources. The approaches outlined here provide a route to moving beyond large area summaries of vaccination coverage that mask epidemiologically-important heterogeneities to detailed maps that capture subnational vulnerabilities. The output datasets are built on open data and methods, and in flexible format that can be aggregated to more operationally-relevant administrative unit levels.

Access to emergency hospital care provided by the public sector in sub-Saharan Africa in 2015: a geocoded inventory and spatial analysis

BACKGROUND

Timely access to emergency care can substantially reduce mortality. International benchmarks for access to emergency hospital care have been established to guide ambitions for universal health care by 2030. However, no Pan-African database of where hospitals are located exists; therefore, we aimed to complete a geocoded inventory of hospital services in Africa in relation to how populations might access these services in 2015, with focus on women of child bearing age.

METHODS

We assembled a geocoded inventory of public hospitals across 48 countries and islands of sub-Saharan Africa, including Zanzibar, using data from various sources. We only included public hospitals with emergency services that were managed by governments at national or local levels and faith-based or non-governmental organisations. For hospital listings without geographical coordinates, we geocoded each facility using Microsoft Encarta (version 2009), Google Earth (version 7.3), Geonames, Fallingrain, OpenStreetMap, and other national digital gazetteers. We obtained estimates for total population and women of child bearing age (15-49 years) at a 1 km2 spatial resolution from the WorldPop database for 2015. Additionally, we assembled road network data from Google Map Maker Project and OpenStreetMap using ArcMap (version 10.5). We then combined the road network and the population locations to form a travel impedance surface. Subsequently, we formulated a cost distance algorithm based on the location of public hospitals and the travel impedance surface in AccessMod (version 5) to compute the proportion of populations living within a combined walking and motorised travel time of 2 h to emergency hospital services.

FINDINGS

We consulted 100 databases from 48 sub-Saharan countries and islands, including Zanzibar, and identified 4908 public hospitals. 2701 hospitals had either full or partial information about their geographical coordinates. We estimated that 287 282 013 (29·0%) people and 64 495 526 (28·2%) women of child bearing age are located more than 2-h travel time from the nearest hospital. Marked differences were observed within and between countries, ranging from less than 25% of the population within 2-h travel time of a public hospital in South Sudan to more than 90% in Nigeria, Kenya, Cape Verde, Swaziland, South Africa, Burundi, Comoros, São Tomé and Príncipe, and Zanzibar. Only 16 countries reached the international benchmark of more than 80% of their populations living within a 2-h travel time of the nearest hospital.

INTERPRETATION

Physical access to emergency hospital care provided by the public sector in Africa remains poor and varies substantially within and between countries. Innovative targeting of emergency care services is necessary to reduce these inequities. This study provides the first spatial census of public hospital services in Africa.

FUNDING

Wellcome Trust and the UK Department for International Development.

Malaria prevalence metrics in low- and middle-income countries: an assessment of precision in nationally-representative surveys

Background

One pillar to monitoring progress towards the Sustainable Development Goals is the investment in high quality data to strengthen the scientific basis for decision-making. At present, nationally-representative surveys are the main source of data for establishing a scientific evidence base, monitoring, and evaluation of health metrics. However, little is known about the optimal precisions of various population-level health and development indicators that remains unquantified in nationally-representative household surveys. Here, a retrospective analysis of the precision of prevalence from these surveys was conducted.

Methods

Using malaria indicators, data were assembled in nine sub-Saharan African countries with at least two nationally-representative surveys. A Bayesian statistical model was used to estimate between- and within-cluster variability for fever and malaria prevalence, and insecticide-treated bed nets (ITNs) use in children under the age of 5 years. The intra-class correlation coefficient was estimated along with the optimal sample size for each indicator with associated uncertainty.

Findings

Results suggest that the estimated sample sizes for the current nationally-representative surveys increases with declining malaria prevalence. Comparison between the actual sample size and the modelled estimate showed a requirement to increase the sample size for parasite prevalence by up to 77.7% (95% Bayesian credible intervals 74.7–79.4) for the 2015 Kenya MIS (estimated sample size of children 0–4 years 7218 [7099–7288]), and 54.1% [50.1–56.5] for the 2014–2015 Rwanda DHS (12,220 [11,950–12,410]).

Conclusion

This study highlights the importance of defining indicator-relevant sample sizes to achieve the required precision in the current national surveys. While expanding the current surveys would need additional investment, the study highlights the need for improved approaches to cost effective sampling.

Electronic supplementary material

The online version of this article (10.1186/s12936-017-2127-y) contains supplementary material, which is available to authorized users.

Mapping poverty using mobile phone and satellite data

Poverty is one of the most important determinants of adverse health outcomes globally, a major cause of societal instability and one of the largest causes of lost human potential. Traditional approaches to measuring and targeting poverty rely heavily on census data, which in most low- and middle-income countries (LMICs) are unavailable or out-of-date. Alternate measures are needed to complement and update estimates between censuses. This study demonstrates how public and private data sources that are commonly available for LMICs can be used to provide novel insight into the spatial distribution of poverty. We evaluate the relative value of modelling three traditional poverty measures using aggregate data from mobile operators and widely available geospatial data. Taken together, models combining these data sources provide the best predictive power (highest r² = 0.78) and lowest error, but generally models employing mobile data only yield comparable results, offering the potential to measure poverty more frequently and at finer granularity. Stratifying models into urban and rural areas highlights the advantage of using mobile data in urban areas and different data in different contexts. The findings indicate the possibility to estimate and continually monitor poverty rates at high spatial resolution in countries with limited capacity to support traditional methods of data collection.

Treatment-seeking behaviour in low- and middle-income countries estimated using a Bayesian model

BACKGROUND

Seeking treatment in formal healthcare for uncomplicated infections is vital to combating disease in low- and middle-income countries (LMICs). Healthcare treatment-seeking behaviour varies within and between communities and is modified by socio-economic, demographic, and physical factors. As a result, it remains a challenge to quantify healthcare treatment-seeking behaviour using a metric that is comparable across communities. Here, we present an application for transforming individual categorical responses (actions related to fever) to a continuous probabilistic estimate of fever treatment for one country in Sub-Saharan Africa (SSA).

METHODS

Using nationally representative household survey data from the 2013 Demographic and Health Survey (DHS) in Namibia, individual-level responses (n = 1138) were linked to theoretical estimates of travel time to the nearest public or private health facility. Bayesian Item Response Theory (IRT) models were fitted via Markov Chain Monte Carlo (MCMC) simulation to estimate parameters related to fever treatment and estimate probability of treatment for children under five years. Different models were implemented to evaluate computational needs and the effect of including predictor variables such as rurality. The mean treatment rates were then estimated at regional level.

RESULTS

Modelling results suggested probability of fever treatment was highest in regions with relatively high incidence of malaria historically. The minimum predicted threshold probability of seeking treatment was 0.3 (model 1: 0.340; 95% CI 0.155-0.597), suggesting that even in populations at large distances from facilities, there was still a 30% chance of an individual seeking treatment for fever. The agreement between correctly predicted probability of treatment at individual level based on a subset of data (n = 247) was high (AUC = 0.978), with a sensitivity of 96.7% and a specificity of 75.3%.

CONCLUSION

We have shown how individual responses in national surveys can be transformed to probabilistic measures comparable at population level. Our analysis of household survey data on fever suggested a 30% baseline threshold for fever treatment in Namibia. However, this threshold level is likely to vary by country or endemicity. Although our focus was on fever treatment, the methodology outlined can be extended to multiple health seeking behaviours captured in routine national survey data and to other infectious diseases.

Spatio-temporal analysis of malaria vector density from baseline through intervention in a high transmission setting

Background

An increase in effective malaria control since 2000 has contributed to a decline in global malaria morbidity and mortality. Knowing when and how existing interventions could be combined to maximise their impact on malaria vectors can provide valuable information for national malaria control programs in different malaria endemic settings. Here, we assess the effect of indoor residual spraying on malaria vector densities in a high malaria endemic setting in eastern Uganda as part of a cohort study where the use of long-lasting insecticidal nets (LLINs) was high.

Methods

Anopheles mosquitoes were sampled monthly using CDC light traps in 107 households selected randomly. Information on the use of malaria interventions in households was also gathered and recorded via a questionnaire. A Bayesian spatio-temporal model was then used to estimate mosquito densities adjusting for climatic and ecological variables and interventions.

Results

Anopheles gambiae (sensu lato) were most abundant (89.1%; n = 119,008) compared to An. funestus (sensu lato) (10.1%, n = 13,529). Modelling results suggest that the addition of indoor residual spraying (bendiocarb) in an area with high coverage of permethrin-impregnated LLINs (99%) was associated with a major decrease in mosquito vector densities. The impact on An. funestus (s.l.) (Rate Ratio 0.1508; 97.5% CI: 0.0144–0.8495) was twice as great as for An. gambiae (s.l.) (RR 0.5941; 97.5% CI: 0.1432–0.8577).

Conclusions

High coverage of active ingredients on walls depressed vector populations in intense malaria transmission settings. Sustained use of combined interventions would have a long-term impact on mosquito densities, limiting infectious biting.

Electronic supplementary material

The online version of this article (doi:10.1186/s13071-016-1917-3) contains supplementary material, which is available to authorized users.

Dynamic denominators: the impact of seasonally varying population numbers on disease incidence estimates

BACKGROUND

Reliable health metrics are crucial for accurately assessing disease burden and planning interventions. Many health indicators are measured through passive surveillance systems and are reliant on accurate estimates of denominators to transform case counts into incidence measures. These denominator estimates generally come from national censuses and use large area growth rates to estimate annual changes. Typically, they do not account for any seasonal fluctuations and thus assume a static denominator population. Many recent studies have highlighted the dynamic nature of human populations through quantitative analyses of mobile phone call data records and a range of other sources, emphasizing seasonal changes. In this study, we use mobile phone data to capture patterns of short-term human population movement and to map dynamism in population densities.

METHODS

We show how mobile phone data can be used to measure seasonal changes in health district population numbers, which are used as denominators for calculating district-level disease incidence. Using the example of malaria case reporting in Namibia we use 3.5 years of phone data to investigate the spatial and temporal effects of fluctuations in denominators caused by seasonal mobility on malaria incidence estimates.

RESULTS

We show that even in a sparsely populated country with large distances between population centers, such as Namibia, populations are highly dynamic throughout the year. We highlight how seasonal mobility affects malaria incidence estimates, leading to differences of up to 30 % compared to estimates created using static population maps. These differences exhibit clear spatial patterns, with likely overestimation of incidence in the high-prevalence zones in the north of Namibia and underestimation in lower-risk areas when compared to using static populations.

CONCLUSION

The results here highlight how health metrics that rely on static estimates of denominators from censuses may differ substantially once mobility and seasonal variations are taken into account. With respect to the setting of malaria in Namibia, the results indicate that Namibia may actually be closer to malaria elimination than previously thought. More broadly, the results highlight how dynamic populations are. In addition to affecting incidence estimates, these changes in population density will also have an impact on allocation of medical resources. Awareness of seasonal movements has the potential to improve the impact of interventions, such as vaccination campaigns or distributions of commodities like bed nets.

Correction: Equality in Maternal and Newborn Health: Modelling Geographic Disparities in Utilisation of Care in Five East African Countries

[This corrects the article DOI: 10.1371/journal.pone.0162006.].

Equality in Maternal and Newborn Health: Modelling Geographic Disparities in Utilisation of Care in Five East African Countries

BACKGROUND

Geographic accessibility to health facilities represents a fundamental barrier to utilisation of maternal and newborn health (MNH) services, driving historically hidden spatial pockets of localized inequalities. Here, we examine utilisation of MNH care as an emergent property of accessibility, highlighting high-resolution spatial heterogeneity and sub-national inequalities in receiving care before, during, and after delivery throughout five East African countries.

METHODS

We calculated a geographic inaccessibility score to the nearest health facility at 300 x 300 m using a dataset of 9,314 facilities throughout Burundi, Kenya, Rwanda, Tanzania and Uganda. Using Demographic and Health Surveys data, we utilised hierarchical mixed effects logistic regression to examine the odds of: 1) skilled birth attendance, 2) receiving 4+ antenatal care visits at time of delivery, and 3) receiving a postnatal health check-up within 48 hours of delivery. We applied model results onto the accessibility surface to visualise the probabilities of obtaining MNH care at both high-resolution and sub-national levels after adjusting for live births in 2015.

RESULTS

Across all outcomes, decreasing wealth and education levels were associated with lower odds of obtaining MNH care. Increasing geographic inaccessibility scores were associated with the strongest effect in lowering odds of obtaining care observed across outcomes, with the widest disparities observed among skilled birth attendance. Specifically, for each increase in the inaccessibility score to the nearest health facility, the odds of having skilled birth attendance at delivery was reduced by over 75% (0.24; CI: 0.19-0.3), while the odds of receiving antenatal care decreased by nearly 25% (0.74; CI: 0.61-0.89) and 40% for obtaining postnatal care (0.58; CI: 0.45-0.75).

CONCLUSIONS

Overall, these results suggest decreasing accessibility to the nearest health facility significantly deterred utilisation of all maternal health care services. These results demonstrate how spatial approaches can inform policy efforts and promote evidence-based decision-making, and are particularly pertinent as the world shifts into the Sustainable Goals Development era, where sub-national applications will become increasingly useful in identifying and reducing persistent inequalities.

Fine resolution mapping of population age-structures for health and development applications

The age-group composition of populations varies considerably across the world, and obtaining accurate, spatially detailed estimates of numbers of children under 5 years is important in designing vaccination strategies, educational planning or maternal healthcare delivery. Traditionally, such estimates are derived from population censuses, but these can often be unreliable, outdated and of coarse resolution for resource-poor settings. Focusing on Nigeria, we use nationally representative household surveys and their cluster locations to predict the proportion of the under-five population in 1 × 1 km using a Bayesian hierarchical spatio-temporal model. Results showed that land cover, travel time to major settlements, night-time lights and vegetation index were good predictors and that accounting for fine-scale variation, rather than assuming a uniform proportion of under 5 year olds can result in significant differences in health metrics. The largest gaps in estimated bednet and vaccination coverage were in Kano, Katsina and Jigawa. Geolocated household surveys are a valuable resource for providing detailed, contemporary and regularly updated population age-structure data in the absence of recent census data. By combining these with covariate layers, age-structure maps of unprecedented detail can be produced to guide the targeting of interventions in resource-poor settings.

Modelling the incidence of Plasmodium vivax and Plasmodium falciparum malaria in Afghanistan 2006-2009

BACKGROUND

Identifying areas that support high malaria risks and where populations lack access to health care is central to reducing the burden in Afghanistan. This study investigated the incidence of Plasmodium vivax and Plasmodium falciparum using routine data to help focus malaria interventions.

METHODS

To estimate incidence, the study modelled utilisation of the public health sector using fever treatment data from the 2012 national Malaria Indicator Survey. A probabilistic measure of attendance was applied to population density metrics to define the proportion of the population within catchment of a public health facility. Malaria data were used in a Bayesian spatio-temporal conditional-autoregressive model with ecological or environmental covariates, to examine the spatial and temporal variation of incidence.

FINDINGS

From the analysis of healthcare utilisation, over 80% of the population was within 2 hours' travel of the nearest public health facility, while 64.4% were within 30 minutes' travel. The mean incidence of P. vivax in 2009 was 5.4 (95% Crl 3.2-9.2) cases per 1000 population compared to 1.2 (95% Crl 0.4-2.9) cases per 1000 population for P. falciparum. P. vivax peaked in August while P. falciparum peaked in November. 32% of the estimated 30.5 million people lived in regions where annual incidence was at least 1 case per 1,000 population of P. vivax; 23.7% of the population lived in areas where annual P. falciparum case incidence was at least 1 per 1000.

CONCLUSION

This study showed how routine data can be combined with household survey data to model malaria incidence. The incidence of both P. vivax and P. falciparum in Afghanistan remain low but the co-distribution of both parasites and the lag in their peak season provides challenges to malaria control in Afghanistan. Future improved case definition to determine levels of imported risks may be useful for the elimination ambitions in Afghanistan.

The changing risk of Plasmodium falciparum malaria infection in Africa: 2000-10: a spatial and temporal analysis of transmission intensity

BACKGROUND

Over a decade ago, the Roll Back Malaria Partnership was launched, and since then there has been unprecedented investment in malaria control. We examined the change in malaria transmission intensity during the period 2000-10 in Africa.

METHODS

We assembled a geocoded and community Plasmodium falciparum parasite rate standardised to the age group 2-10 years (PfPR2-10) database from across 49 endemic countries and territories in Africa from surveys undertaken since 1980. The data were used within a Bayesian space-time geostatistical framework to predict PfPR2-10 in 2000 and 2010 at a 1 × 1 km spatial resolution. Population distribution maps at the same spatial resolution were used to compute populations at risk by endemicity class and estimate population-adjusted PfPR2-10 (PAPfPR2-10) for each of the 44 countries for which predictions were possible for each year.

FINDINGS

Between 2000 and 2010, the population in hyperendemic (>50% to 75% PfPR2-10) or holoendemic (>75% PfPR2-10) areas decreased from 218·6 million (34·4%) of 635·7 million to 183·5 million (22·5%) of 815·7 million across 44 malaria-endemic countries. 280·1 million (34·3%) people lived in areas of mesoendemic transmission (>10% to 50% PfPR2-10) in 2010 compared with 178·6 million (28·1%) in 2000. Population in areas of unstable or very low transmission (<5% PfPR2-10) increased from 131·7 million people (20·7%) in 2000 to 219·0 million (26·8%) in 2010. An estimated 217·6 million people, or 26·7% of the 2010 population, lived in areas where transmission had reduced by at least one PfPR2-10 endemicity class. 40 countries showed a reduction in national mean PAPfPR2-10. Only ten countries contributed 87·1% of the population living in areas of hyperendemic or holoendemic transmission in 2010.

INTERPRETATION

Substantial reductions in malaria transmission have been achieved in endemic countries in Africa over the period 2000-10. However, 57% of the population in 2010 continued to live in areas where transmission remains moderate to intense and global support to sustain and accelerate the reduction of transmission must remain a priority.

FUNDING

Wellcome Trust.

The demographics of human and malaria movement and migration patterns in East Africa

Introduction

The quantification of parasite movements can provide valuable information for control strategy planning across all transmission intensities. Mobile parasite carrying individuals can instigate transmission in receptive areas, spread drug resistant strains and reduce the effectiveness of control strategies. The identification of mobile demographic groups, their routes of travel and how these movements connect differing transmission zones, potentially enables limited resources for interventions to be efficiently targeted over space, time and populations.

Methods

National population censuses and household surveys provide individual-level migration, travel, and other data relevant for understanding malaria movement patterns. Together with existing spatially referenced malaria data and mathematical models, network analysis techniques were used to quantify the demographics of human and malaria movement patterns in Kenya, Uganda and Tanzania. Movement networks were developed based on connectivity and magnitudes of flow within each country and compared to assess relative differences between regions and demographic groups. Additional malaria-relevant characteristics, such as short-term travel and bed net use, were also examined.

Results

Patterns of human and malaria movements varied between demographic groups, within country regions and between countries. Migration rates were highest in 20–30 year olds in all three countries, but when accounting for malaria prevalence, movements in the 10–20 year age group became more important. Different age and sex groups also exhibited substantial variations in terms of the most likely sources, sinks and routes of migration and malaria movement, as well as risk factors for infection, such as short-term travel and bed net use.

Conclusion

Census and survey data, together with spatially referenced malaria data, GIS and network analysis tools, can be valuable for identifying, mapping and quantifying regional connectivities and the mobility of different demographic groups. Demographically-stratified HPM and malaria movement estimates can provide quantitative evidence to inform the design of more efficient intervention and surveillance strategies that are targeted to specific regions and population groups.

Estimation of malaria incidence in northern Namibia in 2009 using Bayesian conditional-autoregressive spatial-temporal models

As malaria transmission declines, it becomes increasingly important to monitor changes in malaria incidence rather than prevalence. Here, a spatio-temporal model was used to identify constituencies with high malaria incidence to guide malaria control. Malaria cases were assembled across all age groups along with several environmental covariates. A Bayesian conditional-autoregressive model was used to model the spatial and temporal variation of incidence after adjusting for test positivity rates and health facility utilisation. Of the 144,744 malaria cases recorded in Namibia in 2009, 134,851 were suspected and 9893 were parasitologically confirmed. The mean annual incidence based on the Bayesian model predictions was 13 cases per 1000 population with the highest incidence predicted for constituencies bordering Angola and Zambia. The smoothed maps of incidence highlight trends in disease incidence. For Namibia, the 2009 maps provide a baseline for monitoring the targets of pre-elimination.

Malaria control and the intensity of Plasmodium falciparum transmission in Namibia 1969-1992

BACKGROUND

Historical evidence of the levels of intervention scale up and its relationships to changing malaria risks provides important contextual information for current ambitions to eliminate malaria in various regions of Africa today.

METHODS

Community-based Plasmodium falciparum prevalence data from 3,260 geo-coded time-space locations between 1969 and 1992 were assembled from archives covering an examination of 230,174 individuals located in northern Namibia. These data were standardized the age-range 2 to less than 10 years and used within a Bayesian model-based geo-statistical framework to examine the changes of malaria risk in the years 1969, 1974, 1979, 1984 and 1989 at 5×5 km spatial resolution. This changing risk was described against rainfall seasons and the wide-scale use of indoor-residual house-spraying and mass drug administration.

RESULTS

Most areas of Northern Namibia experienced low intensity transmission during a ten-year period of wide-scale control activities between 1969 and 1979. As control efforts waned, flooding occurred, drug resistance emerged and the war for independence intensified the spatial extent of moderate-to-high malaria transmission expanded reaching a peak in the late 1980s.

CONCLUSIONS

Targeting vectors and parasite in northern Namibia was likely to have successfully sustained a situation of low intensity transmission, but unraveled quickly to a peak of transmission intensity following a sequence of events by the early 1990s.

The receptive versus current risks of Plasmodium falciparum transmission in northern Namibia: implications for elimination

BACKGROUND

Countries aiming for malaria elimination need to define their malariogenic potential, of which measures of both receptive and current transmission are major components. As Namibia pursues malaria elimination, the importation risks due to cross-border human population movements with higher risk neighboring countries has been identified as a major challenge. Here we used historical and contemporary Plasmodium falciparum prevalence data for Namibia to estimate receptive and current levels of malaria risk in nine northern regions. We explore the potential of these risk maps to support decision-making for malaria elimination in Namibia.

METHODS

Age-corrected geocoded community P. falciparum rate PfPR2-10 data from the period 1967-1992 (n = 3,260) and 2009 (n = 120) were modeled separately within a Bayesian model-based geostatistical (MBG) framework. A full Bayesian space-time MBG model was implemented using the 1967-1992 data to make predictions for every five years from 1969 to 1989. These maps were used to compute the maximum mean PfPR2-10 at 5 x 5 km locations in the northern regions of Namibia to estimate receptivity. A separate spatial Bayesian MBG was fitted to the 2009 data to predict current risk of malaria at similar spatial resolution. Using a high-resolution population map for Namibia, population at risk by receptive and current endemicity by region and population adjusted PfPR2-10 by health district were computed. Validations of predictions were undertaken separately for the historical and current risk models.

RESULTS

Highest receptive risks were observed in the northern regions of Caprivi, Kavango and Ohangwena along the border with Angola and Zambia. Relative to the receptive risks, over 90% of the 1.4 million people across the nine regions of northern Namibia appear to have transitioned to a lower endemic class by 2009. The biggest transition appeared to have occurred in areas of highest receptive risks. Of the 23 health districts, 12 had receptive PAPfPR2-10 risks of 5% to 18% and accounted for 57% of the population in the north. Current PAPfPR2-10 risks was largely <5% across the study area.

CONCLUSIONS

The comparison of receptive and current malaria risks in the northern regions of Namibia show health districts that are most at risk of importation due to their proximity to the relatively higher transmission northern neighbouring countries, higher population and modeled receptivity. These health districts should be prioritized as the cross-border control initiatives are rolled out.

The effect of an anti-malarial subsidy programme on the quality of service provision of artemisinin-based combination therapy in Kenya: a cluster-randomized, controlled trial

BACKGROUND

Many patients with suspected malaria in sub-Saharan Africa seek treatment from private providers, but this sector suffers from sub-standard medicine dispensing practices. To improve the quality of care received for presumptive malaria from the highly accessed private retail sector in western Kenya, subsidized pre-packaged artemether-lumefantrine (AL) was provided to private retailers, together with a one day training for retail staff on malaria diagnosis and treatment, job aids and community engagement activities.

METHODS

The intervention was assessed using a cluster-randomized, controlled design. Provider and mystery-shopper cross-sectional surveys were conducted at baseline and eight months post-intervention to assess provider practices. Data were analysed based on cluster-level summaries, comparing control and intervention arms.

RESULTS

On average, 564 retail outlets were interviewed per year. At follow-up, 43% of respondents reported that at least one staff member had attended the training in the intervention arm. The intervention significantly increased the percentage of providers knowing the first line treatment for uncomplicated malaria by 24.2% points (confidence interval (CI): 14.8%, 33.6%; adjusted p=0.0001); the percentage of outlets stocking AL by 31.7% points (CI: 22.0%, 41.3%; adjusted p=0.0001); and the percentage of providers prescribing AL for presumptive malaria by 23.6% points (CI: 18.7%, 28.6%; adjusted p=0.0001). Generally outlets that received training and job aids performed better than those receiving one or none of these intervention components.

CONCLUSION

Overall, subsidizing ACT and retailer training can significantly increase the percentage of outlets stocking and selling AL for the presumptive treatment of malaria, but further research is needed on strategies to improve the provision of counselling advice to retail customers.

Spatial modelling of healthcare utilisation for treatment of fever in Namibia

Background

Health care utilization is affected by several factors including geographic accessibility. Empirical data on utilization of health facilities is important to understanding geographic accessibility and defining health facility catchments at a national level. Accurately defining catchment population improves the analysis of gaps in access, commodity needs and interpretation of disease incidence. Here, empirical household survey data on treatment seeking for fever were used to model the utilisation of public health facilities and define their catchment areas and populations in northern Namibia.

Method

This study uses data from the Malaria Indicator Survey (MIS) of 2009 on treatment seeking for fever among children under the age of five years to characterize facility utilisation. Probability of attendance of public health facilities for fever treatment was modelled against a theoretical surface of travel times using a three parameter logistic model. The fitted model was then applied to a population surface to predict the number of children likely to use a public health facility during an episode of fever in northern Namibia.

Results

Overall, from the MIS survey, the prevalence of fever among children was 17.6% CI [16.0-19.1] (401 of 2,283 children) while public health facility attendance for fever was 51.1%, [95%CI: 46.2-56.0]. The coefficients of the logistic model of travel time against fever treatment at public health facilities were all significant (p < 0.001). From this model, probability of facility attendance remained relatively high up to 180 minutes (3 hours) and thereafter decreased steadily. Total public health facility catchment population of children under the age five was estimated to be 162,286 in northern Namibia with an estimated fever burden of 24,830 children. Of the estimated fevers, 8,021 (32.3%) were within 30 minutes of travel time to the nearest health facility while 14,902 (60.0%) were within 1 hour.

Conclusion

This study demonstrates the potential of routine household surveys to empirically model health care utilisation for the treatment of childhood fever and define catchment populations enhancing the possibilities of accurate commodity needs assessment and calculation of disease incidence. These methods could be extended to other African countries where detailed mapping of health facilities exists.

The impact of retail-sector delivery of artemether-lumefantrine on malaria treatment of children under five in Kenya: a cluster randomized controlled trial

BACKGROUND

It has been proposed that artemisinin-based combination therapy (ACT) be subsidised in the private sector in order to improve affordability and access. This study in western Kenya aimed to evaluate the impact of providing subsidized artemether-lumefantrine (AL) through retail providers on the coverage of prompt, effective antimalarial treatment for febrile children aged 3-59 months.

METHODS AND FINDINGS

We used a cluster-randomized, controlled design with nine control and nine intervention sublocations, equally distributed across three districts in western Kenya. Cross-sectional household surveys were conducted before and after the delivery of the intervention. The intervention comprised provision of subsidized packs of paediatric ACT to retail outlets, training of retail outlet staff, and community awareness activities. The primary outcome was defined as the proportion of children aged 3-59 months reporting fever in the past 2 weeks who started treatment with AL on the same day or following day of fever onset. Data were collected using structured questionnaires and analyzed based on cluster-level summaries, comparing control to intervention arms, while adjusting for other covariates. Data were collected on 2,749 children in the target age group at baseline and 2,662 at follow-up. 29% of children experienced fever within 2 weeks before the interview. At follow-up, the percentage of children receiving AL on the day of fever or the following day had risen by 14.6% points in the control arm (from 5.3% [standard deviation (SD): 3.2%] to 19.9% [SD: 10.0%]) and 40.2% points in the intervention arm (from 4.7% [SD: 3.4%] to 44.9% [SD: 11.7%]). The percentage of children receiving AL was significantly greater in the intervention arm at follow-up, with a difference between the arms of 25.0% points (95% confidence interval [CI]: 14.1%, 35.9%; unadjusted p = 0.0002, adjusted p = 0.0001). No significant differences were observed between arms in the proportion of caregivers who sought treatment for their child's fever by source, or in the child's adherence to AL.

CONCLUSIONS

Subsidizing ACT in the retail sector can significantly increase ACT coverage for reported fevers in rural areas. Further research is needed on the impact and cost-effectiveness of such subsidy programmes at a national scale.

TRIAL REGISTRATION

Current Controlled Trials ISRCTN59275137 and Kenya Pharmacy and Poisons Board Ethical Committee for Clinical Trials PPB/ECCT/08/07.

Increasing malaria hospital admissions in Uganda between 1999 and 2009

BACKGROUND

Some areas of Africa are witnessing a malaria transition, in part due to escalated international donor support and intervention coverage. Areas where declining malaria rates have been observed are largely characterized by relatively low baseline transmission intensity and rapid scaling of interventions. Less well described are changing patterns of malaria burden in areas of high parasite transmission and slower increases in control and treatment access.

METHODS

Uganda is a country predominantly characterized by intense, perennial malaria transmission. Monthly pediatric admission data from five Ugandan hospitals and their catchments have been assembled retrospectively across 11 years from January 1999 to December 2009. Malaria admission rates adjusted for changes in population density within defined catchment areas were computed across three time periods that correspond to periods where intervention coverage data exist and different treatment and prevention policies were operational. Time series models were developed adjusting for variations in rainfall and hospital use to examine changes in malaria hospitalization over 132 months. The temporal changes in factors that might explain changes in disease incidence were qualitatively examined sequentially for each hospital setting and compared between hospital settings

RESULTS

In four out of five sites there was a significant increase in malaria admission rates. Results from time series models indicate a significant month-to-month increase in the mean malaria admission rates at four hospitals (trend P < 0.001). At all hospitals malaria admissions had increased from 1999 by 47% to 350%. Observed changes in intervention coverage within the catchments of each hospital showed a change in insecticide-treated net coverage from <1% in 2000 to 33% by 2009 but accompanied by increases in access to nationally recommended drugs at only two of the five hospital areas studied.

CONCLUSIONS

The declining malaria disease burden in some parts of Africa is not a universal phenomena across the continent. Despite moderate increases in the coverage of measures to reduce infection and disease without significant coincidental increasing access to effective medicines to treat disease may not lead to severe disease burden reductions in high transmission areas of Africa. More data is needed from a wider range of malaria settings to provide an honest tracking progress of the impact of scaled intervention coverage in Africa.

Predicting the unmet need for biologically targeted coverage of insecticide-treated nets in Kenya

In some countries the biological targeting of universal malaria prevention may offer optimal impact on disease and significant cost-savings compared with approaches that presume universal risk. Spatially defined data on coverage of treated nets from recent national household surveys in Kenya were used within a Bayesian geostatistical framework to predict treated net coverage nationally. When combined with the distributions of malaria risk and population an estimated 8.1 million people were not protected with treated nets in 2010 in biologically defined priority areas. After adjusting for the proportion of nets in use that were not long lasting, an estimated 5.5 to 6.3 million long-lasting treated nets would be required to achieve universal coverage in 2010 in Kenya in at-risk areas compared with 16.4 to 18.1 million nets if not restricted to areas of greatest malaria risk. In Kenya, this evidence-based approach could save the national program at least 55 million US dollars.

Changing malaria intervention coverage, transmission and hospitalization in Kenya

BACKGROUND

Reports of declining incidence of malaria disease burden across several countries in Africa suggest that the epidemiology of malaria across the continent is in transition. Whether this transition is directly related to the scaling of intervention coverage remains a moot point.

METHODS

Paediatric admission data from eight Kenyan hospitals and their catchments have been assembled across two three-year time periods: September 2003 to August 2006 (pre-scaled intervention) and September 2006 to August 2009 (post-scaled intervention). Interrupted time series (ITS) models were developed adjusting for variations in rainfall and hospital use by surrounding communities to show changes in malaria hospitalization over the two periods. The temporal changes in factors that might explain changes in disease incidence were examined sequentially for each hospital setting, compared between hospital settings and ranked according to plausible explanatory factors.

RESULTS

In six out of eight sites there was a decline in Malaria admission rates with declines between 18% and 69%. At two sites malaria admissions rates increased by 55% and 35%. Results from the ITS models indicate that before scaled intervention in September 2006, there was a significant month-to-month decline in the mean malaria admission rates at four hospitals (trend P < 0.05). At the point of scaled intervention, the estimated mean admission rates for malaria was significantly less at four sites compared to the pre-scaled period baseline. Following scaled intervention there was a significant change in the month-to-month trend in the mean malaria admission rates in some but not all of the sites. Plausibility assessment of possible drivers of change pre- versus post-scaled intervention showed inconsistent patterns however, allowing for the increase in rainfall in the second period, there is a suggestion that starting transmission intensity and the scale of change in ITN coverage might explain some but not all of the variation in effect size. At most sites where declines between observation periods were documented admission rates were changing before free mass ITN distribution and prior to the implementation of ACT across Kenya.

CONCLUSION

This study provides evidence of significant within and between location heterogeneity in temporal trends of malaria disease burden. Plausible drivers for changing disease incidence suggest a complex combination of mechanisms, not easily measured retrospectively.

A high resolution spatial population database of Somalia for disease risk mapping

BACKGROUND

Millions of Somali have been deprived of basic health services due to the unstable political situation of their country. Attempts are being made to reconstruct the health sector, in particular to estimate the extent of infectious disease burden. However, any approach that requires the use of modelled disease rates requires reasonable information on population distribution. In a low-income country such as Somalia, population data are lacking, are of poor quality, or become outdated rapidly. Modelling methods are therefore needed for the production of contemporary and spatially detailed population data.

RESULTS

Here land cover information derived from satellite imagery and existing settlement point datasets were used for the spatial reallocation of populations within census units. We used simple and semi-automated methods that can be implemented with free image processing software to produce an easily updatable gridded population dataset at 100 × 100 meters spatial resolution. The 2010 population dataset was matched to administrative population totals projected by the UN. Comparison tests between the new dataset and existing population datasets revealed important differences in population size distributions, and in population at risk of malaria estimates. These differences are particularly important in more densely populated areas and strongly depend on the settlement data used in the modelling approach.

CONCLUSIONS

The results show that it is possible to produce detailed, contemporary and easily updatable settlement and population distribution datasets of Somalia using existing data. The 2010 population dataset produced is freely available as a product of the AfriPop Project and can be downloaded from: http://www.afripop.org.

Estimating the number of paediatric fevers associated with malaria infection presenting to Africa's public health sector in 2007

BACKGROUND

As international efforts to increase the coverage of artemisinin-based combination therapy in public health sectors gather pace, concerns have been raised regarding their continued indiscriminate presumptive use for treating all childhood fevers. The availability of rapid-diagnostic tests to support practical and reliable parasitological diagnosis provides an opportunity to improve the rational treatment of febrile children across Africa. However, the cost effectiveness of diagnosis-based treatment polices will depend on the presumed numbers of fevers harbouring infection. Here we compute the number of fevers likely to present to public health facilities in Africa and the estimated number of these fevers likely to be infected with Plasmodium falciparum malaria parasites.

METHODS AND FINDINGS

We assembled first administrative-unit level data on paediatric fever prevalence, treatment-seeking rates, and child populations. These data were combined in a geographical information system model that also incorporated an adjustment procedure for urban versus rural areas to produce spatially distributed estimates of fever burden amongst African children and the subset likely to present to public sector clinics. A second data assembly was used to estimate plausible ranges for the proportion of paediatric fevers seen at clinics positive for P. falciparum in different endemicity settings. We estimated that, of the 656 million fevers in African 0-4 y olds in 2007, 182 million (28%) were likely to have sought treatment in a public sector clinic of which 78 million (43%) were likely to have been infected with P. falciparum (range 60-103 million).

CONCLUSIONS

Spatial estimates of childhood fevers and care-seeking rates can be combined with a relational risk model of infection prevalence in the community to estimate the degree of parasitemia in those fevers reaching public health facilities. This quantification provides an important baseline comparison of malarial and nonmalarial fevers in different endemicity settings that can contribute to ongoing scientific and policy debates about optimum clinical and financial strategies for the introduction of new diagnostics. These models are made publicly available with the publication of this paper.

Serologic markers for detecting malaria in areas of low endemicity, Somalia, 2008

Areas in which malaria is not highly endemic are suitable for malaria elimination, but assessing transmission is difficult because of lack of sensitivity of commonly used methods. We evaluated serologic markers for detecting variation in malaria exposure in Somalia. Plasmodium falciparum or P. vivax was not detected by microscopy in cross-sectional surveys of samples from persons during the dry (0/1,178) and wet (0/1,128) seasons. Antibody responses against P. falciparum or P. vivax were detected in 17.9% (179/1,001) and 19.3% (202/1,044) of persons tested. Reactivity against P. falciparum was significantly different between 3 villages (p<0.001); clusters of seroreactivity were present. Distance to the nearest seasonal river was negatively associated with P. falciparum (p = 0.028) and P. vivax seroreactivity (p = 0.016). Serologic markers are a promising tool for detecting spatial variation in malaria exposure and evaluating malaria control efforts in areas where transmission has decreased to levels below the detection limit of microscopy.

Fever prevalence and management among three rural communities in the North West Zone, Somalia

Between March and August 2008 we undertook 2 cross-sectional surveys among 1375 residents of 3 randomly selected villages in the district of Gebiley in the North-West Zone, Somalia. We investigated for the presence of malaria infection and the period prevalence of self-reported fever 14 days prior to both surveys. All blood samples examined were negative for both species of Plasmodium. The period prevalence of 14-day fevers was 4.8% in March and 0.6% in August; the majority of fevers (84.4%) were associated with other symptoms including cough, running nose and sore throat; 48/64 cases had resolved by the day of interview (mean duration 5.4 days). Only 18 (37.5%) fever cases were managed at a formal health care facility: 7 within 24 hours and 10 within 24-72 hours of onset. None of the fevers were investigated for malaria; they were treated with antibiotics, antipyretics and vitamins.

Malaria paediatric hospitalization between 1999 and 2008 across Kenya

BACKGROUND

Intervention coverage and funding for the control of malaria in Africa has increased in recent years, however, there are few descriptions of changing disease burden and the few reports available are from isolated, single site observations or are of reports at country-level. Here we present a nationwide assessment of changes over 10 years in paediatric malaria hospitalization across Kenya.

METHODS

Paediatric admission data on malaria and non-malaria diagnoses were assembled for the period 1999 to 2008 from in-patient registers at 17 district hospitals in Kenya and represented the diverse malaria ecology of the country. These data were then analysed using autoregressive moving average time series models with malaria and all-cause admissions as the main outcomes adjusted for rainfall, changes in service use and populations-at-risk within each hospital's catchment to establish whether there has been a statistically significant decline in paediatric malaria hospitalization during the observation period.

RESULTS

Among the 17 hospital sites, adjusted paediatric malaria admissions had significantly declined at 10 hospitals over 10 years since 1999; had significantly increased at four hospitals, and remained unchanged in three hospitals. The overall estimated average reduction in malaria admission rates was 0.0063 cases per 1,000 children aged 0 to 14 years per month representing an average percentage reduction of 49% across the 10 hospitals registering a significant decline by the end of 2008. Paediatric admissions for all-causes had declined significantly with a reduction in admission rates of greater than 0.0050 cases per 1,000 children aged 0 to 14 years per month at 6 of 17 hospitals. Where malaria admissions had increased three of the four sites were located in Western Kenya close to Lake Victoria. Conversely there was an indication that areas with the largest declines in malaria admission rates were areas located along the Kenyan coast and some sites in the highlands of Kenya.

CONCLUSION

A country-wide assessment of trends in malaria hospitalizations indicates that all is not equal, important variations exist in the temporal pattern of malaria admissions between sites and these differences require more detailed investigation to understand what is required to promote a clinical transition across Africa.

The risks of malaria infection in Kenya in 2009

Background

To design an effective strategy for the control of malaria requires a map of infection and disease risks to select appropriate suites of interventions. Advances in model based geo-statistics and malaria parasite prevalence data assemblies provide unique opportunities to redefine national Plasmodium falciparum risk distributions. Here we present a new map of malaria risk for Kenya in 2009.

Methods

Plasmodium falciparum parasite rate data were assembled from cross-sectional community based surveys undertaken from 1975 to 2009. Details recorded for each survey included the month and year of the survey, sample size, positivity and the age ranges of sampled population. Data were corrected to a standard age-range of two to less than 10 years (PfPR_2-10) and each survey location was geo-positioned using national and on-line digital settlement maps. Ecological and climate covariates were matched to each PfPR_2-10 survey location and examined separately and in combination for relationships to PfPR_2-10. Significant covariates were then included in a Bayesian geostatistical spatial-temporal framework to predict continuous and categorical maps of mean PfPR_2-10 at a 1 × 1 km resolution across Kenya for the year 2009. Model hold-out data were used to test the predictive accuracy of the mapped surfaces and distributions of the posterior uncertainty were mapped.

Results

A total of 2,682 estimates of PfPR_2-10 from surveys undertaken at 2,095 sites between 1975 and 2009 were selected for inclusion in the geo-statistical modeling. The covariates selected for prediction were urbanization; maximum temperature; precipitation; enhanced vegetation index; and distance to main water bodies. The final Bayesian geo-statistical model had a high predictive accuracy with mean error of -0.15% PfPR_2-10; mean absolute error of 0.38% PfPR_2-10; and linear correlation between observed and predicted PfPR_2-10 of 0.81. The majority of Kenya's 2009 population (35.2 million, 86.3%) reside in areas where predicted PfPR_2-10 is less than 5%; conversely in 2009 only 4.3 million people (10.6%) lived in areas where PfPR_2-10 was predicted to be ≥40% and were largely located around the shores of Lake Victoria.

Conclusion

Model based geo-statistical methods can be used to interpolate malaria risks in Kenya with precision and our model shows that the majority of Kenyans live in areas of very low P. falciparum risk. As malaria interventions go to scale effectively tracking epidemiological changes of risk demands a rigorous effort to document infection prevalence in time and space to remodel risks and redefine intervention priorities over the next 10-15 years.

A spatial national health facility database for public health sector planning in Kenya in 2008

BACKGROUND

Efforts to tackle the enormous burden of ill-health in low-income countries are hampered by weak health information infrastructures that do not support appropriate planning and resource allocation. For health information systems to function well, a reliable inventory of health service providers is critical. The spatial referencing of service providers to allow their representation in a geographic information system is vital if the full planning potential of such data is to be realized.

METHODS

A disparate series of contemporary lists of health service providers were used to update a public health facility database of Kenya last compiled in 2003. These new lists were derived primarily through the national distribution of antimalarial and antiretroviral commodities since 2006. A combination of methods, including global positioning systems, was used to map service providers. These spatially-referenced data were combined with high-resolution population maps to analyze disparity in geographic access to public health care.

FINDINGS

The updated 2008 database contained 5,334 public health facilities (67% ministry of health; 28% mission and nongovernmental organizations; 2% local authorities; and 3% employers and other ministries). This represented an overall increase of 1,862 facilities compared to 2003. Most of the additional facilities belonged to the ministry of health (79%) and the majority were dispensaries (91%). 93% of the health facilities were spatially referenced, 38% using global positioning systems compared to 21% in 2003. 89% of the population was within 5 km Euclidean distance to a public health facility in 2008 compared to 71% in 2003. Over 80% of the population outside 5 km of public health service providers was in the sparsely settled pastoralist areas of the country.

CONCLUSION

We have shown that, with concerted effort, a relatively complete inventory of mapped health services is possible with enormous potential for improving planning. Expansion in public health care in Kenya has resulted in significant increases in geographic access although several areas of the country need further improvements. This information is key to future planning and with this paper we have released the digital spatial database in the public domain to assist the Kenyan Government and its partners in the health sector.