An effective approach for gap-filling continental scale remotely sensed time-series

The archives of imagery and modeled data products derived from remote sensing programs with high temporal resolution provide powerful resources for characterizing inter- and intra-annual environmental dynamics. The impressive depth of available time-series from such missions (e.g., MODIS and AVHRR) affords new opportunities for improving data usability by leveraging spatial and temporal information inherent to longitudinal geospatial datasets. In this research we develop an approach for filling gaps in imagery time-series that result primarily from cloud cover, which is particularly problematic in forested equatorial regions. Our approach consists of two, complementary gap-filling algorithms and a variety of run-time options that allow users to balance competing demands of model accuracy and processing time. We applied the gap-filling methodology to MODIS Enhanced Vegetation Index (EVI) and daytime and nighttime Land Surface Temperature (LST) datasets for the African continent for 2000-2012, with a 1 km spatial resolution, and an 8-day temporal resolution. We validated the method by introducing and filling artificial gaps, and then comparing the original data with model predictions. Our approach achieved R² values above 0.87 even for pixels within 500 km wide introduced gaps. Furthermore, the structure of our approach allows estimation of the error associated with each gap-filled pixel based on the distance to the non-gap pixels used to model its fill value, thus providing a mechanism for including uncertainty associated with the gap-filling process in downstream applications of the resulting datasets.

Interpreting predictive maps of disease: highlighting the pitfalls of distribution models in epidemiology

The application of spatial modelling to epidemiology has increased significantly over the past decade, delivering enhanced understanding of the environmental and climatic factors affecting disease distributions and providing spatially continuous representations of disease risk (predictive maps). These outputs provide significant information for disease control programmes, allowing spatial targeting and tailored interventions. However, several factors (e.g. sampling protocols or temporal disease spread) can influence predictive mapping outputs. This paper proposes a conceptual framework which defines several scenarios and their potential impact on resulting predictive outputs, using simulated data to provide an exemplar. It is vital that researchers recognise these scenarios and their influence on predictive models and their outputs, as a failure to do so may lead to inaccurate interpretation of predictive maps. As long as these considerations are kept in mind, predictive mapping will continue to contribute significantly to epidemiological research and disease control planning.

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.

Modelling environmental factors correlated with podoconiosis: a geospatial study of non-filarial elephantiasis

INTRODUCTION

The precise trigger of podoconiosis - endemic non-filarial elephantiasis of the lower legs - is unknown. Epidemiological and ecological studies have linked the disease with barefoot exposure to red clay soils of volcanic origin. Histopathology investigations have demonstrated that silicon, aluminium, magnesium and iron are present in the lower limb lymph node macrophages of both patients and non-patients living barefoot on these clays. We studied the spatial variation (variations across an area) in podoconiosis prevalence and the associated environmental factors with a goal to better understanding the pathogenesis of podoconiosis.

METHODS

Fieldwork was conducted from June 2011 to February 2013 in 12 kebeles (administrative units) in northern Ethiopia. Geo-located prevalence data and soil samples were collected and analysed along with secondary geological, topographic, meteorological and elevation data. Soil data were analysed for chemical composition, mineralogy and particle size, and were interpolated to provide spatially continuous information. Exploratory, spatial, univariate and multivariate regression analyses of podoconiosis prevalence were conducted in relation to primary (soil) and secondary (elevation, precipitation, and geology) covariates.

RESULTS

Podoconiosis distribution showed spatial correlation with variation in elevation and precipitation. Exploratory analysis identified that phyllosilicate minerals, particularly clay (smectite and kaolinite) and mica groups, quartz (crystalline silica), iron oxide, and zirconium were associated with podoconiosis prevalence. The final multivariate model showed that the quantities of smectite (RR = 2.76, 95% CI: 1.35, 5.73; p = 0.007), quartz (RR = 1.16, 95% CI: 1.06, 1.26; p = 0.001) and mica (RR = 1.09, 95% CI: 1.05, 1.13; p < 0.001) in the soil had positive associations with podoconiosis prevalence.

CONCLUSIONS

More quantities of smectite, mica and quartz within the soil were associated with podoconiosis prevalence. Together with previous work indicating that these minerals may influence water absorption, potentiate infection and be toxic to human cells, the present findings suggest that these particles may play a role in the pathogenesis of podoconiosis and acute adenolymphangitis, a common cause of morbidity in podoconiosis patients.

Exploring the spatio-temporal relationship between two key aeroallergens and meteorological variables in the United Kingdom

Constructing accurate predictive models for grass and birch pollen in the air, the two most important aeroallergens, for areas with variable climate conditions such as the United Kingdom, require better understanding of the relationships between pollen count in the air and meteorological variables. Variations in daily birch and grass pollen counts and their relationship with daily meteorological variables were investigated for nine pollen monitoring sites for the period 2000-2010 in the United Kingdom. An active pollen count sampling method was employed at each of the monitoring stations to sample pollen from the atmosphere. The mechanism of this method is based on the volumetric spore traps of Hirst design (Hirst in Ann Appl Biol 39(2):257-265, 1952). The pollen season (start date, finish date) for grass and birch were determined using a first derivative method. Meteorological variables such as daily rainfall; maximum, minimum and average temperatures; cumulative sum of Sunshine duration; wind speed; and relative humidity were related to the grass and birch pollen counts for the pre-peak, post peak and the entire pollen season. The meteorological variables were correlated with the pollen count data for the following temporal supports: same-day, 1-day prior, 1-day mean prior, 3-day mean prior, 7-day mean prior. The direction of influence (positive/negative) of meteorological variables on pollen count varied for birch and grass, and also varied when the pollen season was treated as a whole season, or was segmented into the pre-peak and post-peak seasons. Maximum temperature, sunshine duration and rainfall were the most important variables influencing the count of grass pollen in the atmosphere. Both maximum temperature (pre-peak) and sunshine produced a strong positive correlation, and rain produced a strong negative correlation with grass pollen count in the air. Similarly, average temperature, wind speed and rainfall were the most important variables influencing the count of birch pollen in the air. Both wind speed and rain produced a negative correlation with birch pollen count in the air and average temperature produced a positive correlation.

Bayesian spatial modelling and the significance of agricultural land use to scrub typhus infection in Taiwan

Scrub typhus is transmitted by the larval stage of trombiculid mites. Environmental factors, including land cover and land use, are known to influence breeding and survival of trombiculid mites and, thus, also the spatial heterogeneity of scrub typhus risk. Here, a spatially autoregressive modelling framework was applied to scrub typhus incidence data from Taiwan, covering the period 2003 to 2011, to provide increased understanding of the spatial pattern of scrub typhus risk and the environmental and socioeconomic factors contributing to this pattern. A clear spatial pattern in scrub typhus incidence was observed within Taiwan, and incidence was found to be significantly correlated with several land cover classes, temperature, elevation, normalized difference vegetation index, rainfall, population density, average income and the proportion of the population that work in agriculture. The final multivariate regression model included statistically significant correlations between scrub typhus incidence and average income (negatively correlated), the proportion of land that contained mosaics of cropland and vegetation (positively correlated) and elevation (positively correlated). These results highlight the importance of land cover on scrub typhus incidence: mosaics of cropland and vegetation represent a transitional land cover type which can provide favourable habitats for rodents and, therefore, trombiculid mites. In Taiwan, these transitional land cover areas tend to occur in less populated and mountainous areas, following the frontier establishment and subsequent partial abandonment of agricultural cultivation, due to demographic and socioeconomic changes. Future land use policy decision-making should ensure that potential public health outcomes, such as modified risk of scrub typhus, are considered.

The dispersal ecology of Rhodesian sleeping sickness following its introduction to a new area

Tsetse-transmitted human and animal trypanosomiasis are constraints to both human and animal health in sub-Saharan Africa, and although these diseases have been known for over a century, there is little recent evidence demonstrating how the parasites circulate in natural hosts and ecosystems. The spread of Rhodesian sleeping sickness (caused by Trypanosoma brucei rhodesiense) within Uganda over the past 15 years has been linked to the movement of infected, untreated livestock (the predominant reservoir) from endemic areas. However, despite an understanding of the environmental dependencies of sleeping sickness, little research has focused on the environmental factors controlling transmission establishment or the spatially heterogeneous dispersal of disease following a new introduction. In the current study, an annually stratified case-control study of Rhodesian sleeping sickness cases from Serere District, Uganda was used to allow the temporal assessment of correlations between the spatial distribution of sleeping sickness and landscape factors. Significant relationships were detected between Rhodesian sleeping sickness and selected factors, including elevation and the proportion of land which was "seasonally flooding grassland" or "woodlands and dense savannah." Temporal trends in these relationships were detected, illustrating the dispersal of Rhodesian sleeping sickness into more 'suitable' areas over time, with diminishing dependence on the point of introduction in concurrence with an increasing dependence on environmental and landscape factors. These results provide a novel insight into the ecology of Rhodesian sleeping sickness dispersal and may contribute towards the implementation of evidence-based control measures to prevent its further spread.

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.

An exploratory GIS-based method to identify and characterise landscapes with an elevated epidemiological risk of Rhodesian human African trypanosomiasis

BACKGROUND

Specific land cover types and activities have been correlated with Trypanosoma brucei rhodesiense distributions, indicating the importance of landscape for epidemiological risk. However, methods proposed to identify specific areas with elevated epidemiological risk (i.e. where transmission is more likely to occur) tend to be costly and time consuming. This paper proposes an exploratory spatial analysis using geo-referenced human African trypanosomiasis (HAT) cases and matched controls from Serere hospital, Uganda (December 1998 to November 2002) to identify areas with an elevated epidemiological risk of HAT.

METHODS

Buffers 3 km from each case and control were used to represent areas in which village inhabitants would carry out their daily activities. It was hypothesised that the selection of areas where several case village buffers overlapped would enable the identification of locations with increased risk of HAT transmission, as these areas were more likely to be frequented by HAT cases in several surrounding villages. The landscape within these overlap areas should more closely relate to the environment in which transmission occurs as opposed to using the full buffer areas. The analysis was carried out for each of four annual periods, for both cases and controls, using a series of threshold values (number of overlapping buffers), including a threshold of one, which represented the benchmark (e.g. use of the full buffer area as opposed to the overlap areas).

RESULTS

A greater proportion of the overlap areas for cases consisted of seasonally flooding grassland and lake fringe swamp, than the control overlap areas, correlating well with the preferred habitat of the predominant tsetse species within the study area (Glossina fuscipes fuscipes). The use of overlap areas also resulted in a greater difference between case and control landscapes, when compared with the benchmark (using the full buffer area).

CONCLUSIONS

These results indicate that the overlap analysis has enabled the selection of areas more likely to represent epidemiological risk zones than similar analyses using full buffer areas. The identification of potential epidemiological risk zones using this method requires fewer data than other proposed methods and further development may provide vital information for the targeting of control measures.

Geographical access to care at birth in Ghana: a barrier to safe motherhood

BACKGROUND

Appropriate facility-based care at birth is a key determinant of safe motherhood but geographical access remains poor in many high burden regions. Despite its importance, geographical access is rarely audited systematically, preventing integration in national-level maternal health system assessment and planning. In this study, we develop a uniquely detailed set of spatially-linked data and a calibrated geospatial model to undertake a national-scale audit of geographical access to maternity care at birth in Ghana, a high-burden country typical of many in sub-Saharan Africa.

METHODS

We assembled detailed spatial data on the population, health facilities, and landscape features influencing journeys. These were used in a geospatial model to estimate journey-time for all women of childbearing age (WoCBA) to their nearest health facility offering differing levels of care at birth, taking into account different transport types and availability. We calibrated the model using data on actual journeys made by women seeking care.

RESULTS

We found that a third of women (34%) in Ghana live beyond the clinically significant two-hour threshold from facilities likely to offer emergency obstetric and neonatal care (EmONC) classed at the 'partial' standard or better. Nearly half (45%) live that distance or further from 'comprehensive' EmONC facilities, offering life-saving blood transfusion and surgery. In the most remote regions these figures rose to 63% and 81%, respectively. Poor levels of access were found in many regions that meet international targets based on facilities-per-capita ratios.

CONCLUSIONS

Detailed data assembly combined with geospatial modelling can provide nation-wide audits of geographical access to care at birth to support systemic maternal health planning, human resource deployment, and strategic targeting. Current international benchmarks of maternal health care provision are inadequate for these purposes because they fail to take account of the location and accessibility of services relative to the women they serve.

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.

Elucidating the spatially varying relation between cervical cancer and socio-economic conditions in England

BACKGROUND

Geographically weighted Poisson regression (GWPR) was applied to the relation between cervical cancer disease incidence rates in England and socio-economic deprivation, social status and family structure covariates. Local parameters were estimated which describe the spatial variation in the relations between incidence and socio-economic covariates.

RESULTS

A global (stationary) regression model revealed a significant correlation between cervical cancer incidence rates and social status. However, a local (non-stationary) GWPR model provided a better fit with less spatial correlation (positive autocorrelation) in the residuals. Moreover, the GWPR model was able to represent local variation in the relations between cervical cancer incidence and socio-economic covariates across space, whereas the global model represented only the overall (or average) relation for the whole of England. The global model could lead to misinterpretation of the relations between cervical cancer incidence and socio-economic covariates locally.

CONCLUSIONS

Cervical cancer incidence was shown to have a non-stationary relationship with spatially varying covariates that are available through national datasets. As a result, it was shown that if low social status sectors of the population are to be targeted preferentially, this targeting should be done on a region-by-region basis such as to optimize health outcomes. While such a strategy may be difficult to implement in practice, the research does highlight the inequalities inherent in a uniform intervention approach.

Analysis of CDC social control measures using an agent-based simulation of an influenza epidemic in a city

Background

The transmission of infectious disease amongst the human population is a complex process which requires advanced, often individual-based, models to capture the space-time details observed in reality.

Methods

An Individual Space-Time Activity-based Model (ISTAM) was applied to simulate the effectiveness of non-pharmaceutical control measures including: (1) refraining from social activities, (2) school closure and (3) household quarantine, for a hypothetical influenza outbreak in an urban area.

Results

Amongst the set of control measures tested, refraining from social activities with various compliance levels was relatively ineffective. Household quarantine was very effective, especially for the peak number of cases and total number of cases, with large differences between compliance levels. Household quarantine resulted in a decrease in the peak number of cases from more than 300 to around 158 for a 100% compliance level, a decrease of about 48.7%. The delay in the outbreak peak was about 3 to 17 days. The total number of cases decreased to a range of 3635-5403, that is, 63.7%-94.7% of the baseline value.

When coupling control measures, household quarantine together with school closure was the most effective strategy. The resulting space-time distribution of infection in different classes of activity bundles (AB) suggests that the epidemic outbreak is strengthened amongst children and then spread to adults. By sensitivity analysis, this study demonstrated that earlier implementation of control measures leads to greater efficacy. Also, for infectious diseases with larger basic reproduction number, the effectiveness of non-pharmaceutical measures was shown to be limited.

Conclusions

Simulated results showed that household quarantine was the most effective control measure, while school closure and household quarantine implemented together achieved the greatest benefit. Agent-based models should be applied in the future to evaluate the efficacy of control measures for a range of disease outbreaks in a range of settings given sufficient information about the given case and knowledge about the transmission processes at a fine scale.

Bayesian geostatistical analysis and prediction of Rhodesian human African trypanosomiasis

Background

The persistent spread of Rhodesian human African trypanosomiasis (HAT) in Uganda in recent years has increased concerns of a potential overlap with the Gambian form of the disease. Recent research has aimed to increase the evidence base for targeting control measures by focusing on the environmental and climatic factors that control the spatial distribution of the disease.

Objectives

One recent study used simple logistic regression methods to explore the relationship between prevalence of Rhodesian HAT and several social, environmental and climatic variables in two of the most recently affected districts of Uganda, and suggested the disease had spread into the study area due to the movement of infected, untreated livestock. Here we extend this study to account for spatial autocorrelation, incorporate uncertainty in input data and model parameters and undertake predictive mapping for risk of high HAT prevalence in future.

Materials and Methods

Using a spatial analysis in which a generalised linear geostatistical model is used in a Bayesian framework to account explicitly for spatial autocorrelation and incorporate uncertainty in input data and model parameters we are able to demonstrate a more rigorous analytical approach, potentially resulting in more accurate parameter and significance estimates and increased predictive accuracy, thereby allowing an assessment of the validity of the livestock movement hypothesis given more robust parameter estimation and appropriate assessment of covariate effects.

Results

Analysis strongly supports the theory that Rhodesian HAT was imported to the study area via the movement of untreated, infected livestock from endemic areas. The confounding effect of health care accessibility on the spatial distribution of Rhodesian HAT and the linkages between the disease's distribution and minimum land surface temperature have also been confirmed via the application of these methods.

Conclusions

Predictive mapping indicates an increased risk of high HAT prevalence in the future in areas surrounding livestock markets, demonstrating the importance of livestock trading for continuing disease spread. Adherence to government policy to treat livestock at the point of sale is essential to prevent the spread of sleeping sickness in Uganda.

The tsetse transmitted parasites, Trypanosoma brucei rhodesiense and Trypanosoma brucei gambiense, cause the fatal disease human African trypanosomiasis (HAT); the clinical progression, as well as the preferred diagnostic and treatment methods differ between the two types. Currently, the two do not overlap, although recent spread of Rhodesian HAT in Uganda has raised concerns over a potential future overlap. A recent study using geo-referenced HAT case records suggested that the most recent spread of Rhodesian HAT may have been due to movements of infected, untreated livestock (the main reservoir of the parasite). Here, the initial analysis has been extended by explicitly accounting for spatial locations and their proximity to one another, providing improved accuracy. The results provide strengthened evidence of the significance of livestock movements for the continued spread of Rhodesian HAT within Uganda, despite the introduction of cattle treatment regulations which were implemented in an effort to curb the disease's spread. The application of predictive mapping indicates an increased risk of HAT in areas surrounding livestock markets, demonstrating the importance of livestock trading for continuing disease spread. This robust evidence can be used for the targeting of disease control efforts within Uganda to prevent further spread of Rhodesian HAT.

Spatial predictions of Rhodesian Human African Trypanosomiasis (sleeping sickness) prevalence in Kaberamaido and Dokolo, two newly affected districts of Uganda

The continued northwards spread of Rhodesian sleeping sickness or Human African Trypanosomiasis (HAT) within Uganda is raising concerns of overlap with the Gambian form of the disease. Disease convergence would result in compromised diagnosis and treatment for HAT. Spatial determinants for HAT are poorly understood across small areas. This study examines the relationships between Rhodesian HAT and several environmental, climatic and social factors in two newly affected districts, Kaberamaido and Dokolo. A one-step logistic regression analysis of HAT prevalence and a two-step logistic regression method permitted separate analysis of both HAT occurrence and HAT prevalence. Both the occurrence and prevalence of HAT were negatively correlated with distance to the closest livestock market in all models. The significance of distance to the closest livestock market strongly indicates that HAT may have been introduced to this previously unaffected area via the movement of infected, untreated livestock from endemic areas. This illustrates the importance of the animal reservoir in disease transmission, and highlights the need for trypanosomiasis control in livestock and the stringent implementation of regulations requiring the treatment of cattle prior to sale at livestock markets to prevent any further spread of Rhodesian HAT within Uganda.

Issues of scale and uncertainty in the global remote sensing of disease

Scale and uncertainty are important issues for the global prediction of disease. Disease mapping over the entire surface of the Earth usually involves the use of remotely sensed imagery to provide environmental covariates of disease risk or disease vector density. It further implies that the spatial resolution of such imagery is relatively coarse (e.g., 8 or 1km). Use of a coarse spatial resolution limits the information that can be extracted from imagery and has important effects on the results of epidemiological analyses. This paper discusses geostatistical models for (i) characterizing the scale(s) of spatial variation in data and (ii) changing the scale of measurement of both the data and the geostatistical model. Uncertainty is introduced, highlighting the fact that most epidemiologists are interested in accuracy, aspects of which can be estimated with measurable quantities. This paper emphasizes the distinction between data- and model-based methods of accuracy assessment and gives examples of both. The key problem of validating global maps is considered.

The use of Dispholidus typus venom in elucidating the cause of a low prothrombin index

A simple and rapid in vitro test was devised to detect the presence of the prothrombin precursor molecule circulating in the plasma of 55 out of 56 patients receiving oral vitamin K antagonists and in 4 out of 5 patients with obstructive jaundice, using either Dispholidus typus or Echis carinatus venoms. The absence of the molecule from the plasma of jaundiced patients was suggestive of hepatocellular rather than obstructive jaundice. The test is based on the clotting of aluminium hydroxide gel adsorbed plasma by the venoms. It is more sensitive and easier to perform than measuring antigenic prothrombin levels.

Developing geostatistical space-time models to predict outpatient treatment burdens from incomplete national data

Basic health system data such as the number of patients utilising different health facilities and the types of illness for which they are being treated are critical for managing service provision. These data requirements are generally addressed with some form of national Health Management Information System (HMIS) which coordinates the routine collection and compilation of data from national health facilities. HMIS in most developing countries are characterised by widespread under-reporting. Here we present a method to adjust incomplete data to allow prediction of national outpatient treatment burdens. We demonstrate this method with the example of outpatient treatments for malaria within the Kenyan HMIS. Three alternative modelling frameworks were developed and tested in which space-time geostatistical prediction algorithms were used to predict the monthly tally of treatments for presumed malaria cases (MC) at facilities where such records were missing. Models were compared by a cross-validation exercise and the model found to most accurately predict MC incorporated available data on the total number of patients visiting each facility each month. A space-time stochastic simulation framework to accompany this model was developed and tested in order to provide estimates of both local and regional prediction uncertainty. The level of accuracy provided by the predictive model, and the accompanying estimates of uncertainty around the predictions, demonstrate how this tool can mitigate the uncertainties caused by missing data, substantially enhancing the utility of existing HMIS data to health-service decision-makers.

Individual Space – Time Activity-Based Model: A Model for the Simulation of Airborne Infectious-Disease Transmission by Activity-Bundle Simulation

Activity bundle (AB) simulation is a method for obtaining a specific contact network (specific to target infectious disease) from the space–time dynamics of individuals constrained both by their social activity and by the physical condition of the space. Taking advantage of AB simulation, an individual space–time activity-based model (ISTAM) is presented which integrates the infectious-disease evolution process, individual activity patterns, and stochastic infection model. ISTAM was applied to the University of Southampton in order to simulate a hypothetical influenza epidemic. The results show that the model behaviour is approximately consistent with expectations.

Information for decision making from imperfect national data: tracking major changes in health care use in Kenya using geostatistics

BACKGROUND

Most Ministries of Health across Africa invest substantial resources in some form of health management information system (HMIS) to coordinate the routine acquisition and compilation of monthly treatment and attendance records from health facilities nationwide. Despite the expense of these systems, poor data coverage means they are rarely, if ever, used to generate reliable evidence for decision makers. One critical weakness across Africa is the current lack of capacity to effectively monitor patterns of service use through time so that the impacts of changes in policy or service delivery can be evaluated. Here, we present a new approach that, for the first time, allows national changes in health service use during a time of major health policy change to be tracked reliably using imperfect data from a national HMIS.

METHODS

Monthly attendance records were obtained from the Kenyan HMIS for 1 271 government-run and 402 faith-based outpatient facilities nationwide between 1996 and 2004. A space-time geostatistical model was used to compensate for the large proportion of missing records caused by non-reporting health facilities, allowing robust estimation of monthly and annual use of services by outpatients during this period.

RESULTS

We were able to reconstruct robust time series of mean levels of outpatient utilisation of health facilities at the national level and for all six major provinces in Kenya. These plots revealed reliably for the first time a period of steady nationwide decline in the use of health facilities in Kenya between 1996 and 2002, followed by a dramatic increase from 2003. This pattern was consistent across different causes of attendance and was observed independently in each province.

CONCLUSION

The methodological approach presented can compensate for missing records in health information systems to provide robust estimates of national patterns of outpatient service use. This represents the first such use of HMIS data and contributes to the resurrection of these hugely expensive but underused systems as national monitoring tools. Applying this approach to Kenya has yielded output with immediate potential to enhance the capacity of decision makers in monitoring nationwide patterns of service use and assessing the impact of changes in health policy and service delivery.

Modelling the effect of urbanization on the transmission of an infectious disease

This paper models the impact of urbanization on infectious disease transmission by integrating a CA land use development model, population projection matrix model and CA epidemic model in S-Plus. The innovative feature of this model lies in both its explicit treatment of spatial land use development, demographic changes, infectious disease transmission and their combination in a dynamic, stochastic model. Heuristically-defined transition rules in cellular automata (CA) were used to capture the processes of both land use development with urban sprawl and infectious disease transmission. A population surface model and dwelling distribution surface were used to bridge the gap between urbanization and infectious disease transmission. A case study is presented involving modelling influenza transmission in Southampton, a dynamically evolving city in the UK. The simulation results for Southampton over a 30-year period show that the pattern of the average number of infection cases per day can depend on land use and demographic changes. The modelling framework presents a useful tool that may be of use in planning applications.

Improving imperfect data from health management information systems in Africa using space-time geostatistics

BACKGROUND

Reliable and timely information on disease-specific treatment burdens within a health system is critical for the planning and monitoring of service provision. Health management information systems (HMIS) exist to address this need at national scales across Africa but are failing to deliver adequate data because of widespread underreporting by health facilities. Faced with this inadequacy, vital public health decisions often rely on crudely adjusted regional and national estimates of treatment burdens.

METHODS AND FINDINGS

This study has taken the example of presumed malaria in outpatients within the largely incomplete Kenyan HMIS database and has defined a geostatistical modelling framework that can predict values for all data that are missing through space and time. The resulting complete set can then be used to define treatment burdens for presumed malaria at any level of spatial and temporal aggregation. Validation of the model has shown that these burdens are quantified to an acceptable level of accuracy at the district, provincial, and national scale.

CONCLUSIONS

The modelling framework presented here provides, to our knowledge for the first time, reliable information from imperfect HMIS data to support evidence-based decision-making at national and sub-national levels.

Modelling distances travelled to government health services in Kenya

OBJECTIVE

To systematically evaluate descriptive measures of spatial access to medical treatment, as part of the millennium development goals to reduce the burden of HIV/AIDS, tuberculosis and malaria.

METHODS

We obtained high-resolution spatial and epidemiological data on health services, population, transport network, topography, land cover and paediatric fever treatment in four Kenyan districts to develop access and use models for government health services in Kenya. Community survey data were used to model use of government health services by febrile children. A model based on the transport network was then implemented and adjusted for actual use patterns. We compared the predictive accuracy of this refined model to that of Euclidean distance metrics. RESULTS Higher-order facilities were more attractive to patients (54%, 58% and 60% in three scenarios) than lower-order ones. The transport network model, adjusted for competition between facilities, was most accurate and selected as the best-fit model. It estimated that 63% of the population of the study districts were within the 1 h national access benchmark, against 82% estimated by the Euclidean model.

CONCLUSIONS

Extrapolating the results from the best-fit model in study districts to the national level shows that approximately six million people are currently incorrectly estimated to have access to government health services within 1 h. Simple Euclidean distance assumptions, which underpin needs assessments and against which millennium development goals are evaluated, thus require reconsideration.

Urbanization, malaria transmission and disease burden in Africa

Many attempts have been made to quantify Africa's malaria burden but none has addressed how urbanization will affect disease transmission and outcome, and therefore mortality and morbidity estimates. In 2003, 39% of Africa's 850 million people lived in urban settings; by 2030, 54% of Africans are expected to do so. We present the results of a series of entomological, parasitological and behavioural meta-analyses of studies that have investigated the effect of urbanization on malaria in Africa. We describe the effect of urbanization on both the impact of malaria transmission and the concomitant improvements in access to preventative and curative measures. Using these data, we have recalculated estimates of populations at risk of malaria and the resulting mortality. We find there were 1,068,505 malaria deaths in Africa in 2000 - a modest 6.7% reduction over previous iterations. The public-health implications of these findings and revised estimates are discussed.