Uncertainty quantification for epidemiological forecasts of COVID-19 through combinations of model predictions.

Scientific advice to the UK government throughout the COVID-19 pandemic has been informed by ensembles of epidemiological models provided by members of the Scientific Pandemic Influenza group on Modelling. Among other applications, the model ensembles have been used to forecast daily incidence, deaths and hospitalizations. The models differ in approach (e.g. deterministic or agent-based) and in assumptions made about the disease and population. These differences capture genuine uncertainty in the understanding of disease dynamics and in the choice of simplifying assumptions underpinning the model. Although analyses of multi-model ensembles can be logistically challenging when time-frames are short, accounting for structural uncertainty can improve accuracy and reduce the risk of over-confidence in predictions. In this study, we compare the performance of various ensemble methods to combine short-term (14-day) COVID-19 forecasts within the context of the pandemic response. We address practical issues around the availability of model predictions and make some initial proposals to address the shortcomings of standard methods in this challenging situation.

Modeling the factors that influence exposure to SARS-CoV-2 on a subway train carriage.

We propose the Transmission of Virus in Carriages (TVC) model, a computational model which simulates the potential exposure to SARS-CoV-2 for passengers traveling in a subway rail system train. This model considers exposure through three different routes: fomites via contact with contaminated surfaces; close-range exposure, which accounts for aerosol and droplet transmission within 2 m of the infectious source; and airborne exposure via small aerosols which does not rely on being within 2 m distance from the infectious source. Simulations are based on typical subway parameters and the aim of the study is to consider the relative effect of environmental and behavioral factors including prevalence of the virus in the population, number of people traveling, ventilation rate, and mask wearing as well as the effect of model assumptions such as emission rates. Results simulate generally low exposures in most of the scenarios considered, especially under low virus prevalence. Social distancing through reduced loading and high mask-wearing adherence is predicted to have a noticeable effect on reducing exposure through all routes. The highest predicted doses happen through close-range exposure, while the fomite route cannot be neglected; exposure through both routes relies on infrequent events involving relatively few individuals. Simulated exposure through the airborne route is more homogeneous across passengers, but is generally lower due to the typically short duration of the trips, mask wearing, and the high ventilation rate within the carriage. The infection risk resulting from exposure is challenging to estimate as it will be influenced by factors such as virus variant and vaccination rates.

Characteristics and mortality rates among patients requiring intermediate care: a national cohort study using linked databases.

BACKGROUND: Adults increasingly live and die with chronic progressive conditions into advanced age. Many live with multimorbidity and an uncertain illness trajectory with points of marked decline, loss of function and increased risk of end of life. Intermediate care units support mainly older adults in transition between hospital and home to regain function and anticipate and plan for end of life. This study examined the patient characteristics and the factors associated with mortality over 1 year post-admission to an intermediate care unit to inform priorities for care. METHODS: A national cohort study of adults admitted to intermediate care units in England using linked individual-level Hospital Episode Statistics and death registration data. The main outcome was mortality within 1 year from admission. The cohort was examined as two groups with significant differences in mortality between main diagnosis of a non-cancer condition and cancer. Data analysis used Kaplan-Meier curves to explore mortality differences between the groups and a time-dependant Cox proportional hazards model to determine mortality risk factors. RESULTS: The cohort comprised 76,704 adults with median age 81 years (IQR 70-88) admitted to 220 intermediate care units over 1 year in 2016. Overall, 28.0% died within 1 year post-admission. Mortality varied by the main diagnosis of cancer (total n = 3680, 70.8% died) and non-cancer condition (total n = 73,024, 25.8% died). Illness-related factors had the highest adjusted hazard ratios [aHRs]. At 0-28 days post-admission, risks were highest for non-cancer respiratory conditions (pneumonia (aHR 6.17 [95%CI 4.90-7.76]), chronic obstructive pulmonary disease (aHR 5.01 [95% CI 3.78-6.62]), dementia (aHR 5.07 [95% CI 3.80-6.77]) and liver disease (aHR 9.75 [95% CI 6.50-14.6]) compared with musculoskeletal disorders. In cancer, lung cancer showed largest risk (aHR 1.20 [95%CI 1.04-1.39]) compared with cancer 'other'. Risks increased with high multimorbidity for non-cancer (aHR 2.57 [95% CI 2.36-2.79]) and cancer (aHR 2.59 [95% CI 2.13-3.15]) (reference: lowest). CONCLUSIONS: One in four patients died within 1 year. Indicators for palliative care assessment are respiratory conditions, dementia, liver disease, cancer and rising multimorbidity. The traditional emphasis on rehabilitation and recovery in intermediate care units has changed with an ageing population and the need for greater integration of palliative care.

Quantifying the incidence of severe-febrile-illness hospital admissions in sub-Saharan Africa.

Severe-febrile-illness (SFI) is a common cause of morbidity and mortality across sub-Saharan Africa (SSA). The burden of SFI in SSA is currently unknown and its estimation is fraught with challenges. This is due to a lack of diagnostic capacity for SFI in SSA, and thus a dearth of baseline data on the underlying etiology of SFI cases and scant SFI-specific causative-agent prevalence data. To highlight the public health significance of SFI in SSA, we developed a Bayesian model to quantify the incidence of SFI hospital admissions in SSA. Our estimates indicate a mean population-weighted SFI-inpatient-admission incidence rate of 18.4 (6.8-31.1, 68% CrI) per 1000 people for the year 2014, across all ages within areas of SSA with stable Plasmodium falciparum transmission. We further estimated a total of 16,200,337 (5,993,249-27,321,779, 68% CrI) SFI hospital admissions. This analysis reveals the significant burden of SFI in hospitals in SSA, but also highlights the paucity of pathogen-specific prevalence and incidence data for SFI in SSA. Future improvements in pathogen-specific diagnostics for causative agents of SFI will increase the abundance of SFI-specific prevalence and incidence data, aid future estimations of SFI burden, and enable clinicians to identify SFI-specific pathogens, administer appropriate treatment and management, and facilitate appropriate antibiotic use.

The contribution of non-malarial febrile illness co-infections to Plasmodium falciparum case counts in health facilities in sub-Saharan Africa.

BACKGROUND: The disease burden of Plasmodium falciparum malaria illness is generally estimated using one of two distinct approaches: either by transforming P. falciparum infection prevalence estimates into incidence estimates using conversion formulae; or through adjustment of counts of recorded P. falciparum-positive fever cases from clinics. Whilst both ostensibly seek to evaluate P. falciparum disease burden, there is an implicit and problematic difference in the metric being estimated. The first enumerates only symptomatic malaria cases, while the second enumerates all febrile episodes coincident with a P. falciparum infection, regardless of the fever's underlying cause. METHODS: Here, a novel approach was used to triangulate community-based data sources capturing P. falciparum infection, fever, and care-seeking to estimate the fraction of P. falciparum-positive fevers amongst children under 5 years of age presenting at health facilities that are attributable to P. falciparum infection versus other non-malarial causes. A Bayesian hierarchical model was used to assign probabilities of malaria-attributable fever (MAF) and non-malarial febrile illness (NMFI) to children under five from a dataset of 41 surveys from 21 countries in sub-Saharan Africa conducted between 2006 and 2016. Using subsequent treatment-seeking outcomes, the proportion of MAF and NMFI amongst P. falciparum-positive febrile children presenting at public clinics was estimated. RESULTS: Across all surveyed malaria-positive febrile children who sought care at public clinics across 41 country-years in sub-Saharan Africa, P. falciparum infection was estimated to be the underlying cause of only 37.7% (31.1-45.4, 95% CrI) of P. falciparum-positive fevers, with significant geographical and temporal heterogeneity between surveys. CONCLUSIONS: These findings highlight the complex nature of the P. falciparum burden amongst children under 5 years of age and indicate that for many children presenting at health clinics, a positive P. falciparum diagnosis and a fever does not necessarily mean P. falciparum is the underlying cause of the child's symptoms, and thus other causes of illness should always be investigated, in addition to prescribing an effective anti-malarial medication. In addition to providing new large-scale estimates of malaria-attributable fever prevalence, the results presented here improve comparability between different methods for calculating P. falciparum disease burden, with significant implications for national and global estimation of malaria burden.

Mapping the global endemicity and clinical burden of Plasmodium vivax, 2000-17: a spatial and temporal modelling study.

BACKGROUND: Plasmodium vivax exacts a significant toll on health worldwide, yet few efforts to date have quantified the extent and temporal trends of its global distribution. Given the challenges associated with the proper diagnosis and treatment of P vivax, national malaria programmes-particularly those pursuing malaria elimination strategies-require up to date assessments of P vivax endemicity and disease impact. This study presents the first global maps of P vivax clinical burden from 2000 to 2017. METHODS: In this spatial and temporal modelling study, we adjusted routine malariometric surveillance data for known biases and used socioeconomic indicators to generate time series of the clinical burden of P vivax. These data informed Bayesian geospatial models, which produced fine-scale predictions of P vivax clinical incidence and infection prevalence over time. Within sub-Saharan Africa, where routine surveillance for P vivax is not standard practice, we combined predicted surfaces of Plasmodium falciparum with country-specific ratios of P vivax to P falciparum. These results were combined with surveillance-based outputs outside of Africa to generate global maps. FINDINGS: We present the first high-resolution maps of P vivax burden. These results are combined with those for P falciparum (published separately) to form the malaria estimates for the Global Burden of Disease 2017 study. The burden of P vivax malaria decreased by 41·6%, from 24·5 million cases (95% uncertainty interval 22·5-27·0) in 2000 to 14·3 million cases (13·7-15·0) in 2017. The Americas had a reduction of 56·8% (47·6-67·0) in total cases since 2000, while South-East Asia recorded declines of 50·5% (50·3-50·6) and the Western Pacific regions recorded declines of 51·3% (48·0-55·4). Europe achieved zero P vivax cases during the study period. Nonetheless, rates of decline have stalled in the past five years for many countries, with particular increases noted in regions affected by political and economic instability. INTERPRETATION: Our study highlights important spatial and temporal patterns in the clinical burden and prevalence of P vivax. Amid substantial progress worldwide, plateauing gains and areas of increased burden signal the potential for challenges that are greater than expected on the road to malaria elimination. These results support global monitoring systems and can inform the optimisation of diagnosis and treatment where P vivax has most impact. FUNDING: Bill & Melinda Gates Foundation and the Wellcome Trust.

Mapping the global prevalence, incidence, and mortality of Plasmodium falciparum, 2000-17: a spatial and temporal modelling study.

BACKGROUND: Since 2000, the scale-up of malaria control interventions has substantially reduced morbidity and mortality caused by the disease globally, fuelling bold aims for disease elimination. In tandem with increased availability of geospatially resolved data, malaria control programmes increasingly use high-resolution maps to characterise spatially heterogeneous patterns of disease risk and thus efficiently target areas of high burden. METHODS: We updated and refined the Plasmodium falciparum parasite rate and clinical incidence models for sub-Saharan Africa, which rely on cross-sectional survey data for parasite rate and intervention coverage. For malaria endemic countries outside of sub-Saharan Africa, we produced estimates of parasite rate and incidence by applying an ecological downscaling approach to malaria incidence data acquired via routine surveillance. Mortality estimates were derived by linking incidence to systematically derived vital registration and verbal autopsy data. Informed by high-resolution covariate surfaces, we estimated P falciparum parasite rate, clinical incidence, and mortality at national, subnational, and 5 × 5 km pixel scales with corresponding uncertainty metrics. FINDINGS: We present the first global, high-resolution map of P falciparum malaria mortality and the first global prevalence and incidence maps since 2010. These results are combined with those for Plasmodium vivax (published separately) to form the malaria estimates for the Global Burden of Disease 2017 study. The P falciparum estimates span the period 2000-17, and illustrate the rapid decline in burden between 2005 and 2017, with incidence declining by 27·9% and mortality declining by 42·5%. Despite a growing population in endemic regions, P falciparum cases declined between 2005 and 2017, from 232·3 million (95% uncertainty interval 198·8-277·7) to 193·9 million (156·6-240·2) and deaths declined from 925 800 (596 900-1 341 100) to 618 700 (368 600-952 200). Despite the declines in burden, 90·1% of people within sub-Saharan Africa continue to reside in endemic areas, and this region accounted for 79·4% of cases and 87·6% of deaths in 2017. INTERPRETATION: High-resolution maps of P falciparum provide a contemporary resource for informing global policy and malaria control planning, programme implementation, and monitoring initiatives. Amid progress in reducing global malaria burden, areas where incidence trends have plateaued or increased in the past 5 years underscore the fragility of hard-won gains against malaria. Efforts towards elimination should be strengthened in such areas, and those where burden remained high throughout the study period. FUNDING: Bill & Melinda Gates Foundation.

malariaAtlas: an R interface to global malariometric data hosted by the Malaria Atlas Project.

BACKGROUND: The Malaria Atlas Project (MAP) has worked to assemble and maintain a global open-access database of spatial malariometric data for over a decade. This data spans various formats and topics, including: geo-located surveys of malaria parasite rate; global administrative boundary shapefiles; and global and regional rasters representing the distribution of malaria and associated illnesses, blood disorders, and intervention coverage. MAP has recently released malariaAtlas, an R package providing a direct interface to MAP's routinely-updated malariometric databases and research outputs. METHODS AND RESULTS: The current paper reviews the functionality available in malariaAtlas and highlights its utility for spatial epidemiological analysis of malaria. malariaAtlas enables users to freely download, visualise and analyse global malariometric data within R. Currently available data types include: malaria parasite rate and vector occurrence point data; subnational administrative boundary shapefiles; and a large suite of rasters covering a diverse range of metrics related to malaria research. malariaAtlas is here used in two mock analyses to illustrate how this data may be incorporated into a standard R workflow for spatial analysis. CONCLUSIONS: malariaAtlas is the first open-access R-interface to malariometric data, providing a new and reproducible means of accessing such data within a freely available and commonly used statistical software environment. In this way, the malariaAtlas package aims to contribute to the environment of data-sharing within the malaria research community.

How long do rapid diagnostic tests remain positive after anti-malarial treatment?

BACKGROUND: Rapid diagnostic tests (RDTs) are increasingly becoming a paradigm for both clinical diagnosis of malaria infections and for estimating community parasite prevalence in household malaria indicator surveys in malaria-endemic countries. The antigens detected by RDTs are known to persist in the blood after treatment with anti-malarials, but reports on the duration of persistence (and the effect this has on RDT positivity) of these antigens post-treatment have been variable. METHODS: In this review, published studies on the persistence of positivity of RDTs post-treatment are collated, and a bespoke Bayesian survival model is fit to estimate the number of days RDTs remain positive after treatment. RESULTS: Half of RDTs that detect the antigen histidine-rich protein II (HRP2) are still positive 15 (5-32) days post-treatment, 13 days longer than RDTs that detect the antigen Plasmodium lactate dehydrogenase, and that 5% of HRP2 RDTs are still positive 36 (21-61) days after treatment. The duration of persistent positivity for combination RDTs that detect both antigens falls between that for HRP2- or pLDH-only RDTs, with half of RDTs remaining positive at 7 (2-20) days post-treatment. This study shows that children display persistent RDT positivity for longer after treatment than adults, and that persistent positivity is more common when an individual is treated with artemisinin combination therapy than when treated with other anti-malarials. CONCLUSIONS: RDTs remain positive for a highly variable amount of time after treatment with anti-malarials, and the duration of positivity is highly dependent on the type of RDT used for diagnosis. Additionally, age and treatment both impact the duration of persistence of RDT positivity. The results presented here suggest that caution should be taken when using RDT-derived diagnostic outcomes from cross-sectional data where individuals have had a recent history of anti-malarial treatment.

Quantifying the contribution of Plasmodium falciparum malaria to febrile illness amongst African children.

Suspected malaria cases in Africa increasingly receive a rapid diagnostic test (RDT) before antimalarials are prescribed. While this ensures efficient use of resources to clear parasites, the underlying cause of the individual's fever remains unknown due to potential coinfection with a non-malarial febrile illness. Widespread use of RDTs does not necessarily prevent over-estimation of clinical malaria cases or sub-optimal case management of febrile patients. We present a new approach that allows inference of the spatiotemporal prevalence of both Plasmodium falciparum malaria-attributable and non-malarial fever in sub-Saharan African children from 2006 to 2014. We estimate that 35.7% of all self-reported fevers were accompanied by a malaria infection in 2014, but that only 28.0% of those (10.0% of all fevers) were causally attributable to malaria. Most fevers among malaria-positive children are therefore caused by non-malaria illnesses. This refined understanding can help improve interpretation of the burden of febrile illness and shape policy on fever case management.

Mapping Plasmodium falciparum Mortality in Africa between 1990 and 2015.

BACKGROUND: Malaria control has not been routinely informed by the assessment of subnational variation in malaria deaths. We combined data from the Malaria Atlas Project and the Global Burden of Disease Study to estimate malaria mortality across sub-Saharan Africa on a grid of 5 km2 from 1990 through 2015. METHODS: We estimated malaria mortality using a spatiotemporal modeling framework of geolocated data (i.e., with known latitude and longitude) on the clinical incidence of malaria, coverage of antimalarial drug treatment, case fatality rate, and population distribution according to age. RESULTS: Across sub-Saharan Africa during the past 15 years, we estimated that there was an overall decrease of 57% (95% uncertainty interval, 46 to 65) in the rate of malaria deaths, from 12.5 (95% uncertainty interval, 8.3 to 17.0) per 10,000 population in 2000 to 5.4 (95% uncertainty interval, 3.4 to 7.9) in 2015. This led to an overall decrease of 37% (95% uncertainty interval, 36 to 39) in the number of malaria deaths annually, from 1,007,000 (95% uncertainty interval, 666,000 to 1,376,000) to 631,000 (95% uncertainty interval, 394,000 to 914,000). The share of malaria deaths among children younger than 5 years of age ranged from more than 80% at a rate of death of more than 25 per 10,000 to less than 40% at rates below 1 per 10,000. Areas with high malaria mortality (>10 per 10,000) and low coverage (<50%) of insecticide-treated bed nets and antimalarial drugs included much of Nigeria, Angola, and Cameroon and parts of the Central African Republic, Congo, Guinea, and Equatorial Guinea. CONCLUSIONS: We estimated that there was an overall decrease of 57% in the rate of death from malaria across sub-Saharan Africa over the past 15 years and identified several countries in which high rates of death were associated with low coverage of antimalarial treatment and prevention programs. (Funded by the Bill and Melinda Gates Foundation and others.).

Treatment-seeking rates in malaria endemic countries.

BACKGROUND: The proportion of individuals who seek treatment for fever is an important quantity in understanding access to and use of health systems, as well as for interpreting data on disease incidence from routine surveillance systems. For many malaria endemic countries (MECs), treatment-seeking information is available from national household surveys. The aim of this paper was to assemble sub-national estimates of treatment-seeking behaviours and to predict national treatment-seeking measures for all MECs lacking household survey data. METHODS: Data on treatment seeking for fever were obtained from Demographic and Health Surveys, Malaria Indicator Surveys and Multiple Cluster Indicator Surveys for every MEC and year that data were available. National-level social, economic and health-related variables were gathered from the World Bank as putative covariates of treatment-seeking rates. A generalized additive mixed model (GAMM) was used to estimate treatment-seeking behaviours for countries where survey data were unavailable. Two separate models were developed to predict the proportion of fever cases that would seek treatment at (1) a public health facility or (2) from any kind of treatment provider. RESULTS: Treatment-seeking data were available for 74 MECs and modelled for the remaining 24. GAMMs found that the percentage of pregnant women receiving prenatal care, vaccination rates, education level, government health expenditure, and GDP growth were important predictors for both categories of treatment-seeking outcomes. Treatment-seeking rates, which varied both within and among regions, revealed that public facilities were not always the primary facility type used. CONCLUSIONS: Estimates of treatment-seeking rates show how health services are utilized and help correct reported malaria case numbers to obtain more accurate measures of disease burden. The assembled and modelled data demonstrated that while treatment-seeking rates have overall increased over time, access remains low in some malaria endemic regions and utilization of government services is in some areas limited.

Coverage and system efficiencies of insecticide-treated nets in Africa from 2000 to 2017.

Insecticide-treated nets (ITNs) for malaria control are widespread but coverage remains inadequate. We developed a Bayesian model using data from 102 national surveys, triangulated against delivery data and distribution reports, to generate year-by-year estimates of four ITN coverage indicators. We explored the impact of two potential 'inefficiencies': uneven net distribution among households and rapid rates of net loss from households. We estimated that, in 2013, 21% (17%-26%) of ITNs were over-allocated and this has worsened over time as overall net provision has increased. We estimated that rates of ITN loss from households are more rapid than previously thought, with 50% lost after 23 (20-28) months. We predict that the current estimate of 920 million additional ITNs required to achieve universal coverage would in reality yield a lower level of coverage (77% population access). By improving efficiency, however, the 920 million ITNs could yield population access as high as 95%.

Standardizing Plasmodium falciparum infection prevalence measured via microscopy versus rapid diagnostic test.

BACKGROUND: Large-scale mapping of Plasmodium falciparum infection prevalence relies on opportunistic assemblies of infection prevalence data arising from thousands of P. falciparum parasite rate (PfPR) surveys conducted worldwide. Variance in these data is driven by both signal, the true underlying pattern of infection prevalence, and a range of factors contributing to 'noise', including sampling error, differing age ranges of subjects and differing parasite detection methods. Whilst the former two noise components have been addressed in previous studies, the effect of different diagnostic methods used to determine PfPR in different studies has not. In particular, the majority of PfPR data are based on positivity rates determined by either microscopy or rapid diagnostic test (RDT), yet these approaches are not equivalent; therefore a method is needed for standardizing RDT and microscopy-based prevalence estimates prior to use in mapping. METHODS: Twenty-five recent Demographic and Health surveys (DHS) datasets from sub-Saharan Africa provide child diagnostic test results derived using both RDT and microscopy for each individual. These prevalence estimates were aggregated across level one administrative zones and a Bayesian probit regression model fit to the microscopy- versus RDT-derived prevalence relationship. An errors-in-variables approach was employed to account for sampling error in both the dependent and independent variables. In addition to the diagnostic outcome, RDT type, fever status and recent anti-malarial treatment were extracted from the datasets in order to analyse their effect on observed malaria prevalence. RESULTS: A strong non-linear relationship between the microscopy and RDT-derived prevalence was found. The results of regressions stratified by the additional diagnostic variables (RDT type, fever status and recent anti-malarial treatment) indicate that there is a distinct and consistent difference in the relationship when the data are stratified by febrile status and RDT brand. CONCLUSIONS: The relationships defined in this research can be applied to RDT-derived PfPR data to effectively convert them to an estimate of the parasite prevalence expected using microscopy (or vice versa), thereby standardizing the dataset and improving the signal-to-noise ratio. Additionally, the results provide insight on the importance of RDT brands, febrile status and recent anti-malarial treatment for explaining inconsistencies between observed prevalence derived from different diagnostics.

Defining the relationship between infection prevalence and clinical incidence of Plasmodium falciparum malaria.

In many countries health system data remain too weak to accurately enumerate Plasmodium falciparum malaria cases. In response, cartographic approaches have been developed that link maps of infection prevalence with mathematical relationships to predict the incidence rate of clinical malaria. Microsimulation (or 'agent-based') models represent a powerful new paradigm for defining such relationships; however, differences in model structure and calibration data mean that no consensus yet exists on the optimal form for use in disease-burden estimation. Here we develop a Bayesian statistical procedure combining functional regression-based model emulation with Markov Chain Monte Carlo sampling to calibrate three selected microsimulation models against a purpose-built data set of age-structured prevalence and incidence counts. This allows the generation of ensemble forecasts of the prevalence-incidence relationship stratified by age, transmission seasonality, treatment level and exposure history, from which we predict accelerating returns on investments in large-scale intervention campaigns as transmission and prevalence are progressively reduced.

Malaria mapping: understanding the global endemicity of falciparum and vivax malaria.

The mapping of malaria risk has a history stretching back over 100 years. The last decade, however, has seen dramatic progress in the scope, rigour and sophistication of malaria mapping such that its global distribution is now probably better understood than any other infectious disease. In this minireview we consider the main factors that have facilitated the recent proliferation of malaria risk mapping efforts and describe the most prominent global-scale endemicity mapping endeavours of recent years. We describe the diversification of malaria mapping to span a wide range of related metrics of biological and public health importance and consider prospects for the future of the science including its key role in supporting elimination efforts.

Re-examining environmental correlates of Plasmodium falciparum malaria endemicity: a data-intensive variable selection approach.

BACKGROUND: Malaria risk maps play an increasingly important role in disease control planning, implementation, and evaluation. The construction of these maps using modern geospatial techniques relies on covariate grids: continuous surfaces quantifying environmental factors that partially explain spatial heterogeneity in malaria endemicity. Although crucial, past variable selection processes for this purpose have often been subjective and ad-hoc, with many covariates used in modeling with little quantitative justification. METHODS: This research consists of an extensive covariate construction and selection process for predicting Plasmodium falciparum parasite rates (PfPR) in Africa for years 2000-2012. First, a literature review was conducted to establish a comprehensive list of covariates used for malaria mapping. Second, a library of covariate data was assembled to reflect this list, a process that included the construction of multiple, temporally dynamic datasets. Third, the resulting set of covariates was leveraged to create more than 50 million possible covariate terms via factorial combinations of different spatial and temporal aggregations, transformations, and pairwise interactions. Fourth, the expanded set of covariates was reduced via successive selection criteria to yield a robust covariate subset that was assessed using an out-of-sample validation approach. RESULTS: The final covariate subset included predominately dynamic covariates and it substantially out-performed earlier sets used by the Malaria Atlas Project (MAP) for creating global malaria risk maps, with the pseudo-R(2) value for the out-of-sample validation increasing from 0.43 to 0.52. Dynamic covariates improved the model, with 17 of the 20 new covariates consisting of monthly or annual products, but the selected covariates were typically interaction terms that included both dynamic and synoptic datasets. Thus the interplay between normal (i.e., long-term averages) and immediate conditions may be key for characterizing environmental controls on parasite rate. CONCLUSIONS: This analysis represents the first effort to systematically audit covariate utility for malaria mapping and then derive an objective, empirically based set of environmental covariates for modeling PfPR. The new covariates produce more reliable representations of malaria risk patterns and how they are changing through time, and these covariates will be used to characterize spatially and temporally varying environmental conditions affecting PfPR within a geostatistical-modeling framework, thus building upon previous research by MAP that produced global malaria maps for 2007 and 2010.