ArticlePDF Available

Improving Imperfect Data from Health Management Information Systems in Africa Using Space–Time Geostatistics

PLOS
PLOS Medicine
Authors:
Peter Gething at University of Oxford
Peter Gething
Abdisalan Noor at KEMRI-Wellcome Trust Research Programme
Abdisalan Noor
Priscilla Gikandi at King Saud University
Priscilla Gikandi
Esther A A Ogara
Esther A A Ogara
Esther A A Ogara
  • This person is not on ResearchGate, or hasn't claimed this research yet.
Show all 8 authorsHide
Download full-text PDFDownload full-text PDFDownload full-text PDFDownload full-text PDF
Read full-text
Download citation

Abstract and Figures

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. 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. 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.
Schematic Diagram of the Modelling Framework Four stages were used to predict the count of outpatients treated for malaria (MC) for each facility-month with missing data: (1) MMTC was estimated for each facility using both existing and predicted values of TC; (2) existing MC data at each facility were standardised by the corresponding MMTC value to create SMC values; (3) STK was used to predict all missing values of SMC; and (4) MMTC values were used to back-transform the predicted SMC values in order to obtain final predictions of MC. DOI: 10.1371/journal.pmed.0030271.g001
Schematic Diagram of the Modelling Framework Four stages were used to predict the count of outpatients treated for malaria (MC) for each facility-month with missing data: (1) MMTC was estimated for each facility using both existing and predicted values of TC; (2) existing MC data at each facility were standardised by the corresponding MMTC value to create SMC values; (3) STK was used to predict all missing values of SMC; and (4) MMTC values were used to back-transform the predicted SMC values in order to obtain final predictions of MC. DOI: 10.1371/journal.pmed.0030271.g001
… 
. Summary of Government Health Facilities in Kenya and Their Reporting Behaviour during the 84-mo Study Period January 1996 to December 2002
. Summary of Government Health Facilities in Kenya and Their Reporting Behaviour during the 84-mo Study Period January 1996 to December 2002
… 
Percentage of Government Health Facilities in Each Kenyan District (Fourth Level Administrative Unit) Submitting a Monthly Outpatient Morbidity Report to the HMIS The 2 mo shown are (A) the most complete (February 1996) and (B) the least complete (December 1997) during the 84-mo study period January 1996– December 2002. DOI: 10.1371/journal.pmed.0030271.g002
Percentage of Government Health Facilities in Each Kenyan District (Fourth Level Administrative Unit) Submitting a Monthly Outpatient Morbidity Report to the HMIS The 2 mo shown are (A) the most complete (February 1996) and (B) the least complete (December 1997) during the 84-mo study period January 1996– December 2002. DOI: 10.1371/journal.pmed.0030271.g002
… 
. Predicted Mean Annual Counts of Outpatients Treated for Malaria at All Kenyan Government Hospitals, Health Centres, and Dispensaries for the Period 1996-2002
. Predicted Mean Annual Counts of Outpatients Treated for Malaria at All Kenyan Government Hospitals, Health Centres, and Dispensaries for the Period 1996-2002
… 
Number of Outpatients Treated for Malaria at Government Facilities Predicted mean annual totals for each district for the period 1996-2002. Values represent the combined sum of existing and predicted values. DOI: 10.1371/journal.pmed.0030271.g003
+1
Number of Outpatients Treated for Malaria at Government Facilities Predicted mean annual totals for each district for the period 1996-2002. Values represent the combined sum of existing and predicted values. DOI: 10.1371/journal.pmed.0030271.g003
… 
Figures - uploaded by Priscilla Gikandi
Author content
All figure content in this area was uploaded by Priscilla Gikandi
Content may be subject to copyright.
Gething et al. (200
5).pdf
Content uploaded by Priscilla Gikandi
Author content
All content in this area was uploaded by Priscilla Gikandi on Feb 12, 2015
Content may be subject to copyright.
Improving Imperfect Data from Health Management Information Systems in Africa Using Space–Time Geostatisti
cs.pdf
Available via license: CC BY
Content may be subject to copyright.
Improving Imperfect Data from Health Management Information Systems in Africa Using Space–Time Geostatisti
cs.pdf
Available via license: CC BY
Content may be subject to copyright.
pmed.00302
71.pdf
Content uploaded by Abdisalan Noor
Author content
All content in this area was uploaded by Abdisalan Noor on Jul 28, 2014
Content may be subject to copyright.
A preview of the PDF is not available

Supplementary resources (3)

Protocol S1
Data
June 2006
Peter W Gething · Abdisalan Noor · Priscilla Gikandi · Esther A. A Ogara · Peter M Atkinson
Figure S1
Data
June 2006
Peter W Gething · Abdisalan Noor · Priscilla Gikandi · Esther A. A Ogara · Peter M Atkinson
Improving imperfect data from health management information systems in Africa using space–time geostatistics - Supplementary Information
Data
June 2006
Peter Gething · Abdisalan Noor · Priscilla Gikandi · Esther A A Ogara · Peter M Atkinson
... Further, these data are freely available to researchers across the world (Figure 2.15), and therefore represent a promising avenue for obtaining spatial demographic data, particularly amongst developing countries. Lastly, many countries across Africa rely on country-wide health management information systems (HMIS) for routine health data and disease surveillance (Gething et al., 2006). These systems, when regularly updated and uniformly applied, can result in reliable and timely information that can be used to robustly monitor and evaluate a variety of public health conditions. ...
... In a highly functioning HMIS setting, this results in large datasets which can be analysed at both high spatial resolutions and across fine temporal scales, such as monthly estimates. Despite this, the reality of HMIS settings within Africa is less than ideal, with many countries facing competing priorities, resulting in poor data coverage and incomplete routine surveillance efforts (Gething et al., 2006;Health Metrics Network, 2008). Regardless, HMIS remain a robust and underutilised source of spatial health data, holding promise for further strengthening over the coming years. ...
Mapping spatial and temporal inequalities in utilisation of maternal and newborn care in five East African countries
Thesis
  • Jan 2020
Historically, maternal and newborn health (MNH) outcomes used to monitor progress in achieving global and national targets have been measured at an aggregate level, showing vast inequalities between and within countries. To ensure no one is left behind in improving health, researchers have called for the spatial and temporal disaggregation of MNH data. This thesis aims to generate high spatial resolution data over time that can be used to monitor progress in reducing inequalities amongst utilisation of key MNH services in the East African Community (EAC) region, including Burundi, Kenya, Rwanda, Tanzania, and Uganda. Following a ‘three-paper’ format, the first paper in this thesis employs a hierarchical mixed effects logistic regression framework, to estimate the odds of: 1) skilled birth attendance (SBA), 2) receiving 4+ antenatal care (ANC) visits, and 3) receiving a postnatal health check-up (PNC) within 48 hours of delivery. Model results are applied to an 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. Across all outcomes, decreasing wealth and education levels are associated with lower odds of obtaining MNH care, while increasing geographic inaccessibility scores are associated with the strongest effect in lowering odds of obtaining care observed across outcomes, with the widest disparities observed among skilled birth attendance. The second paper explores temporal trends in absolute and relative spatial inequalities in utilisation of these MNH services between 1990 and 2015. A Bayesian framework is employed to generate sub-national estimates of utilisation of SBA, ANC, and PNC over several time points. Absolute change in estimates over time is reported, as well as relative change in ratios of the best- to-worst performing districts per country. Across all countries, the greatest spatial equality is observed among ANC, while SBA and PNC tend to have greater spatial variability. Lastly, while progress has been made to reduce coverage gaps between districts, improvement in PNC coverage has stagnated and should be monitored closely over the coming decades. The final paper comprising this work explores the trade-off between increasing spatial resolution in model inputs and resulting model uncertainty, with aims of understanding the optimal spatial resolution to report health outcomes. Prevalence of childbirth via c-section is estimated in Tanzania, using geospatial covariates at varying levels of spatial coarseness within a Bayesian model framework. Uncertainty in posterior outcomes is reported as the distribution of 95% credible intervals at each spatial resolution, and visualised at the spatial resolution with the greatest model precision. Overall, higher spatial resolution input increases model uncertainty, while model precision is best approximated at the highest spatial resolution, suggesting an important policy trade-off between identifying concealed spatial heterogeneities in health indicators. This thesis makes substantive contributions to the literature by outlining where spatial inequalities in key MNH services are occurring within the EAC region and how these disparities are evolving over time. This work also makes methodological contributions by demonstrating how spatial approaches can be used to monitor health indicators, as well as exploring uncertainty in the application of these techniques, with important implications in communicating results to policy makers. These techniques can be applied across health and development outcomes, notably across Sustainable Development Goal indictors, ensuring “no one left behind” by 2030.
View
... The study of mineral resource management has therefore extensively used geostatistics (44,45). With the wide range of applications offered through geostatistics, it has been also successfully employed in studies of medical (46) and health sciences (47). A part from that geostatistics also plays crucial vitals in the study of COVID-19. ...
Spatial statistical methodologies in COVID-19 Studies: A systematic review
Article
Full-text available
  • Sep 2023
Objective: As of August 2023, COVID-19 had claimed 7 million lives, making it the pandemic with the highest mortality rate. Therefore, The use of cutting-edge technologies and methods is essential when battling the COVID-19 epidemic. This paper aims to systematically review and synthetize applications of spatial statistical methodologies in the analysis of COVID-19. Material and Methods: 55 articles in total were screened from four main digital databases including Web of Science, SCOPUS, PubMed/MEDLINE, and Google schoolar. Three distinct concerns with the use of spatial statistical techniques in the analysis of COVID-19 are discussed, namely (i) applications of spatial regressions in the evaluation of COVID-19's effects, (ii) COVID-19 mapping using of hotspots and spatial clustering analyses, and (iii) applications of interpolation and geostatistics on COVID-19 studies, respectively. Results: Spatial regressions can support the assessment of the COVID-19 impacts on social-economy and environment. Whereas, hotspots and spatial clustering analysis can help effectively on COVID-19 mapping. Last but not least, geostatistics and interpolation are crucial for predicting COVID-19. Conclusion: This review not only emphasises the significance of spatial statistical techniques in COVID-19 studies, but it also sheds light on the practical applications of spatial statistics in COVID-19 research.
View
... The spatial epidemiology of some diseases often renders surveillancebased methods problematic for estimating the population at risk of infection (Murray et al., 2004;Gething et al., 2006), while the diseases are transmitted (Riley, 2007;Kubiak et al., 2010). By approaching diseases cartographically and using spatial mathematical models, it is possible to better approximate them (Brooker et al., 2002;Tsatsaris et al., 2005;Ferguson et al., 2005;Hay et al., 2010). ...
Geoinformatics, spatial epidemiology, and public health
Chapter
  • Jun 2023
Medical Geography, sometimes called health geography, is a field of medical research that incorporates geographical parameters into the study of health and the spread of disease. In addition, medical geography studies the effects of all spatial variables (e.g., climate) and the position of an individual with respect to his/her health as well as the distribution of health services. Spatial Epidemiology is an important field because it aims to understand health problems and improve the health of people worldwide based on the various geographical factors that affect them. The Geography of Health Care mostly focuses on the location of facilities, in terms of accessibility and utilisation, while the Geography of Diseases focuses on the distribution of diseases in geographical space and the analysis and description of disease correlations, with existing or potential environmental factors. This chapter is an extensive summary of work related to Medical Geography, Spatial Epidemiology using new technologies such as Geographic Information Systems (GIS), Remote Sensing, etc.
View
... Second, we introduced a further source of uncertainty via the imputation of zero-incidence samples due to extremely low ascertainment. This replaced months during the malaria season reporting zero cases with predictions from a spatio-temporal model, mirroring established methods (Gething et al. 2006), and replaced a small fraction of the overall data (8.77%). Finally, our method does not include a step to disaggregate consultation rates to a finer spatial scale than that reported by the PHC, often a major limiting step to accessing disease incidence data at a fine spatial scale. ...
The Zero-Corrected, Gravity-Model Multiplier (ZERO-G): A novel method to estimate disease dynamics at the community-scale from passive surveillance data
Preprint
Full-text available
  • Mar 2023
Data on population health are vital to evidence-based decision making by public health officials, but are rarely adequately localized, particularly in rural areas where barriers to healthcare can result in extremely low ascertainment of cases by the health system. Here, we demonstrate a new method to estimate disease incidence at the community level from passive surveillance data collected at primary health centers. The zero-corrected, gravity-based multiplier (ZERO-G) method explicitly models sampling intensity as a function of health facility characteristics and statistically accounts for extremely low rates of ascertainment, resulting in an unbiased, standardized estimate of disease incidence at a spatial resolution nearly ten times finer than typically reported by the facility-based passive surveillance system. We assessed the robustness of this method by applying it to a case study of malaria incidence in a rural health district in southeastern Madagascar. ZERO-G decreased geographic and financial bias in the dataset by over 64% and doubled the agreement rate between spatial patterns in malaria incidence and prevalence rates. ZERO-G can be applied to other infectious diseases and settings, increasing the availability of long-term, high quality surveillance datasets at the community scale.
View
... 1. RACD (standard of care control arm): rapid diagnostic testing and treatment of positives with artemether-lumefantrine (AL) and single dose primaquine of individuals residing within a 500 m radius of a recent passively detected index case 2. rfMDA: presumptive treatment with artemetherlumefantrine (AL) of individuals residing within a 500 m radius of a recent passively detected index case 3. RAVC and RACD combined: indoor residual spraying (IRS) using pirimiphos-methyl, administered to households of individuals residing within a 500 m radius of a recent passively detected index case, plus standard of care RACD as described above 4. rfMDA and RAVC combined: indoor residual spraying (IRS) using pirimiphos-methyl, administered to households of individuals residing within a 500 m radius of a recent passively detected index case, plus rfMDA as described above All clusters received routine annual IRS before the start of the malaria season using dichloro-diphenyl-trichloroethane (DDT) conducted as part of standard malaria control activities by the Namibian Ministry of Health and Social Services (MoHSS). ...
Serological evaluation of the effectiveness of reactive focal mass drug administration and reactive vector control to reduce malaria transmission in Zambezi Region, Namibia: Results from a secondary analysis of a cluster randomised trial
Article
Full-text available
  • Feb 2022
Background Due to challenges in measuring changes in malaria at low transmission, serology is increasingly being used to complement clinical and parasitological surveillance. Longitudinal studies have shown that serological markers, such as Etramp5.Ag1, can reflect spatio-temporal differences in malaria transmission. However, these markers have yet to be used as endpoints in intervention trials. Methods Based on data from a 2017 cluster randomised trial conducted in Zambezi Region, Namibia, evaluating the effectiveness of reactive focal mass drug administration (rfMDA) and reactive vector control (RAVC), this study conducted a secondary analysis comparing antibody responses between intervention arms as trial endpoints. Antibody responses were measured on a multiplex immunoassay, using a panel of eight serological markers of Plasmodium falciparum infection - Etramp5.Ag1, GEXP18, HSP40.Ag1, Rh2.2030, EBA175, PfMSP119, PfAMA1, and PfGLURP.R2. Findings Reductions in sero-prevalence to antigens Etramp.Ag1, PfMSP119, Rh2.2030, and PfAMA1 were observed in study arms combining rfMDA and RAVC, but only effects for Etramp5.Ag1 were statistically significant. Etramp5.Ag1 sero-prevalence was significantly lower in all intervention arms. Compared to the reference arms, adjusted prevalence ratio (aPR) for Etramp5.Ag1 was 0.78 (95%CI 0.65 – 0.91, p = 0.0007) in the rfMDA arms and 0.79 (95%CI 0.67 – 0.92, p = 0.001) in the RAVC arms. For the combined rfMDA plus RAVC intervention, aPR was 0.59 (95%CI 0.46 – 0.76, p < 0.0001). Significant reductions were also observed based on continuous antibody responses. Sero-prevalence as an endpoint was found to achieve higher study power (99.9% power to detect a 50% reduction in prevalence) compared to quantitative polymerase chain reaction (qPCR) prevalence (72.9% power to detect a 50% reduction in prevalence). Interpretation While the observed relative reduction in qPCR prevalence in the study was greater than serology, the use of serological endpoints to evaluate trial outcomes measured effect size with improved precision and study power. Serology has clear application in cluster randomised trials, particularly in settings where measuring clinical incidence or infection is less reliable due to seasonal fluctuations, limitations in health care seeking, or incomplete testing and reporting. Funding This study was supported by Novartis Foundation (A122666), the Bill & Melinda Gates Foundation (OPP1160129), and the Horchow Family Fund (5,300,375,400).
View
... The capacity of data analysis, interpretation, and action is extremely limited. The diagnosis of PKDL and co-infections is of variable quality or absent as in the case of malaria [9,10]. However, disease surveillance is considered as one of the principal activity of public health systems. ...
Improved kala-azar case management through implementation of health facility-based sentinel sites surveillance system in Bihar, India
Article
Full-text available
Background Visceral leishmaniasis (VL), also known as kala-azar (KA), is a neglected vector-borne disease, targeted for elimination, but several affected blocks of Bihar are posing challenges with the high incidence of cases, and moreover, the disease is spreading in newer areas. High-quality kala-azar surveillance in India, always pose great concern. The complete and accurate patient level data is critical for the current kala-azar management information system (KMIS). On the other side, no accurate data on the burden of post kala-azar dermal leishmaniasis (PKDL) and co-infections are available under the current surveillance system, which might emerge as a serious concern. Additionally, in low case scenario, sentinel surveillance may be useful in addressing post-elimination activities and sustaining kala-azar (KA) elimination. Health facility-based sentinel site surveillance system has been proposed, first time to do a proper accounting of KA, PKDL and co-infection morbidity, mortality, diagnosis, case management, hotspot identification and monitoring the impact of elimination interventions.Methodology/principal findingsKala-azar sentinel site surveillance was established and activated in thirteen health facilities of Bihar, India, using stratified sampling technique during 2011 to 2014. Data were collected through specially designed performa from all patients attending the outpatient departments of sentinel sites. Among 20968 symptomatic cases attended sentinel sites, 2996 cases of KA and 53 cases of PKDL were registered from 889 endemic villages. Symptomatic cases meant a person with fever of more than 15 days, weight loss, fatigue, anemia, and substantial swelling of the liver and spleen (enlargement of spleen and liver).The proportion of new and old cases was 86.1% and 13.9% respectively. A statistically significant difference was observed for reduction in KA incidence from 4.13/10000 in 2011 to 1.75/10000 in 2014 (p
View
... 15,18 As a result, information on the burden of schistosomiasis for most areas, including Milola Ward in Lindi District, has been based mainly on hospital reports. 11,15 Such information is liable to inaccuracy and unreliability due to poor recording and lack of random sampling, 19 and the data may not be useful for designing effective disease control programmes. 20 Brooker and colleagues 15 conducted a countrywide survey on the distribution of schistosomiasis in Tanzania from 1980 to 2009 and indicated that schistosomiasis was most likely endemic to Lindi Region, although no field assessment was made on the local distribution of the disease in the region. ...
Demographic Factors Driving Schistosomiasis and Soil-Transmitted Helminthiases in Milola Ward, Lindi District, Tanzania: A Useful Guide for Launching Intervention Programmes
Article
Full-text available
  • Nov 2018
Background: Current information on the distribution of and risk factors for schistosomiasis and soil-transmitted helminthiases is scarce for most areas of southern Tanzania, including Milola Ward in Lindi District. This study was initiated to establish the status of these infections in Milola Ward and to assess how they vary with demographic factors. Methods: From September to October 2014, 2 sets of stool and urine samples were collected from residents of Milola Ward. The Kato-Katz technique was used to examine stool samples for faecal-borne parasites, and the filtration technique was used to examine urine for urinary schistosomes. A total of 195 individuals aged 5 to 90 years were enrolled in the study; 190 (97%) participants submitted adequate urine samples, of whom 107 (56%) were female and 83 (43%) were male. Of the 195 participants who took part in the initial sampling exercise, 158 (81%) provided adequate stool samples; 121 (77%) of these were adults, and the rest (n=37, 23%) were children. Only 53 urine and 26 faecal samples were obtained in the second round of sampling, and due to marked inconsistencies, these have been excluded from the analysis. Mean parasite abundance was analysed for its association with demographic factors, such as age and sex. Results: Three varieties of parasite were detected, namely, Schistosoma haematobium in 44 (23%) of 190 urine samples, hookworms in 12 (8%) of 158 stool samples, and Trichuris trichiura in 6 (4%) of 158 stool samples. The difference in S. haematobium prevalence between male and female participants (27 of 107 females, 25% vs 17 of 83 males, 20%) was not statistically significant (Kruskal-Wallis test, P=.47). Linear regression analysis of S. haematobium infection with age showed a significant association, with children having higher infection intensities than adults (P<.001). S. haematobium prevalence and intensity did not vary significantly between villages (intensity [Kruskal-Wallis test], P=.95; prevalence, P=.88). Discussion: These data confirm that in this setting, the mean age of peak helminthiasis prevalence decreases as transmission pressure increases, with non-school children below 18 years old being most at risk of acquiring parasitic infections. This was the first baseline survey of parasitic infections in Milola Ward, so the results will be crucial for guiding control efforts against parasitic diseases in the area.
View
Mange: Epidemiology and ivermectin efficacy in goats and application of zero-inflated negative binomial regression in Uba Debre Tsehay, southern Ethiopia
Article
  • Dec 2022
Mites are one of the most common and widely distributed ectoparasites of goats in Ethiopia, contributing to major burdens in livestock productivity in the country. Between February 2021 and July 2021, this study was conducted to estimate the prevalence of mange mites, assess the potential risk factors, identify the species infesting goats, and evaluate the efficacy of ivermectin in naturally infested goats in the Uba Debere Tsehay district of Gofa Zone, Southern Ethiopia. A cross-sectional study, longitudinal field efficacy, and questionnaire survey were conducted. A total of 384 goats suspected of having mange were scraped for mite prevalence and count. The mite count data were analyzed using zero-inflated negative binomial (ZINB) models with explanatory variables. The ZINB models indicated that a substantial proportion of the observed zero mite count reflected a failure to detect mites in suspected goats, meaning that the estimated true prevalence was much higher than the apparent prevalence as calculated using a simple proportion of nonzero mite counts. Overall prevalence of mange was 21.87% (84/384) in the study areas. Sarcoptes species (21.09%) and Demodex species (0.78%) were the mite genera identified in this study. Among goats with poor, medium, and good body conditions, mite prevalence was 36.3%, 12.3%, and 10.9%, respectively. Both the prevalence and intensity of infestation were significantly associated with body condition scores, but other risk factors were not. The questionnaire survey indicated that 85.94% of the participants preferred to use modern treatment options (ivermectin 1%, injection) and 76.56% (98/128) respondents replied that ivermectin treatment is effective. Wilcoxon rank-sum test analysis shows that there was significantly (P < 0.05) fewer mites counted on goats treated with ivermectin than on untreated goats at each count up to day 56 after treatment. No live mites were found on any treated animal on days 28 and 56. Mixed ANOVA indicated that there was a significant difference within treatment groups. This study showed that mites are one of the constraints to goat production in the study area and ivermectin was highly effective against Demodex and Sarcoptes mites in goats. Hence, there is a need to create awareness about the impact of mange on goat production, and appropriate ivermectin treatment against mites should be implemented.
View
A comparison of modelling the spatio-temporal pattern of disease: a case study of schistosomiasis japonica in Anhui Province, China
Article
  • Dec 2021
The construction of spatio-temporal models can be either descriptive or dynamic. In this study we aim to evaluate the differences in model fitting between a descriptive model and a dynamic model of the transmission for intestinal schistosomiasis caused by Schistosoma japonicum in Guichi, Anhui Province, China. The parasitological data at the village level from 1991 to 2014 were obtained by cross-sectional surveys. We used the fixed rank kriging (FRK) model, a descriptive model, and the integro-differential equation (IDE) model, a dynamic model, to explore the space–time changes of schistosomiasis japonica. In both models, the average daily precipitation and the normalized difference vegetation index are significantly positively associated with schistosomiasis japonica prevalence, while the distance to water bodies, the hours of daylight and the land surface temperature at daytime were significantly negatively associated. The overall root mean square prediction error of the IDE and FRK models was 0.0035 and 0.0054, respectively, and the correlation reflected by Pearson's correlation coefficient between the predicted and observed values for the IDE model (0.71; p<0.01) was larger than that for the FRK model (0.53; p=0.02). The IDE model fits better in capturing the geographic variation of schistosomiasis japonica. Dynamic spatio-temporal models have the advantage of quantifying the process of disease transmission and may provide more accurate predictions.
View
View
Health information systems: The foundations of public health
Article
Full-text available
  • Aug 2005
  • B WORLD HEALTH ORGAN
Public health decision-making is critically dependent on the timely availability of sound data. The role of health information systems is to generate, analyse and disseminate such data. In practice, health information systems rarely function systematically. The products of historical, social and economic forces, they are complex, fragmented and unresponsive to needs. International donors in health are largely responsible for the problem, having prioritized urgent needs for data over longer-term country capacity-building. The result is painfully apparent in the inability of most countries to generate the data needed to monitor progress towards the Millennium Development Goals. Solutions to the problem must be comprehensive; money alone is likely to be insufficient unless accompanied by sustained support to country systems development coupled with greater donor accountability and allocation of responsibilities. The Health Metrics Network, a global collaboration in the making, is intended to help bring such solutions to the countries most in need
View
Geostatistics for Natural Resource Evaluation
Chapter
Full-text available
  • Jan 1997
  • TECHNOMETRICS
The main text of this book is divided into seven chapters, covering the most important areas of geostatistical methodology: exploratory data analysis; the random function model; inference and modelling; local estimation - accounting for a single attribute; local estimation - accounting for secondary information; assessment of local uncertainty; and assessment of spatial uncertainty. The presentation follows the typical steps of geostatistical analysis, introducing tools for description, quantitative modelling of spatial continuity, spatial prediction and uncertainty assessment. This book uses the notation of the GSLIB guidebook, and the various tools are illustrated using a multivariate soil data set related to heavy metal pollution of a region in the Swiss Jura.
View
Local Prediction of Spatio-Temporal Process with An Application to Wet Sulfate Deposition
Article
  • Dec 1995
A prediction method is given for a first- and second-order nonstationary spatio-temporal process. The predictor uses local data only and consists of a two-stage generalized regression estimate of the local drift at the prediction location added to a kriging prediction of the residual process at that location. This predictor is applied to observations on seasonal, rainfall-deposited sulfate over the conterminous United States between summer 1986 and summer 1992. Analyses suggest that predictions and estimated prediction standard errors have negligible to small biases, there is spatially heterogeneous temporal drift, and temporal covariance is negligible.
View
View
Geostatistical space–time simulation model for air quality prediction
Article
  • Jun 2005
  • ENVIRONMETRICS
In recent decades there has been a growing interest in modelling and simulating natural phenomena, in environmental sciences, motivated by the large amount of data from monitoring activities and by the necessity of answering important environmental problems. Most natural phenomena can be seen as a space–time process, and space–time modelling has recently become an active area of research. However, and probably because these space–time relationships have usually quite different characteristics and levels of uncertainty, the simultaneous integration of space and time has not been easy, and further developments are clearly required. The main goal of this study is to present a simulation model for the characterization of space–time dispersion of air pollutants. A space–time methodology, based on the simultaneous integration of space and time air pollutant dispersion characteristics, is applied to assess the particle contamination of the Setúbal Peninsula (South of Lisbon, Portugal)—a study that is part of a project for the evaluation of regional air quality risk maps and air quality control and management. Copyright
View
Geostatistical Space-Time Simulation Model for Characterization of Air Quality
Chapter
  • Jan 2004
The characterization of spatial uncertainty has been addressed in earth sciences using spatial models, based on stochastic simulation algorithms. Dynamic processes are characterized by two components — space and time. These usually have quite different levels of uncertainty: on the one hand, the heterogeneity of the static component — normally related to the space — can sometimes not be compared with the complexity of the dynamic part of the process; on the other hand, the available knowledge is usually quite different for these two components. This is possibly the main reason why the development of simulation algorithms for spatial processes with a time component is still at an early stage. The main goal of this study is to present a simulation model for the characterization of space-time dispersion of air pollutants. The objective of this model is to predict critical scenarios to support air quality control and management. This space-time simulation approach is applied to assess the particles contamination of Setúbal Peninsula (South of Lisbon — Portugal); a study, that is part of a project for the evaluation of regional air quality risk maps.
View
Geostatistical Space–Time Models: A Review
Article
  • Aug 1999
Geostatistical space–time models are used increasingly for addressing environmental problems, such as monitoring acid deposition or global warming, and forecasting precipitation or stream flow. Each discipline approaches the problem of joint space–time modeling from its own perspective, a fact leading to a significant amount of overlapping models and, possibly, confusion. This paper attempts an annotated survey of models proposed in the literature, stating contributions and pinpointing shortcomings. Stochastic models that extend spatial statistics (geostatistics) to include the additional time dimension are presented with a common notation to facilitate comparison. Two conceptual viewpoints are distinguished: (1) approaches involving a single spatiotemporal random function model, and (2) approaches involving vectors of space random functions or vectors of time series. Links between these two viewpoints are then revealed; advantages and shortcomings are highlighted. Inference from space–time data is revisited, and assessment of joint space–time uncertainty via stochastic imaging is suggested.
View
GSLIB. Geostatistical Software Library and User's guide, 2nd ed.
Chapter
  • Feb 1995
The primary goal of this work is to present a geostatistical software library known as GSLIB. An important prerequisite to geostatical case studies and research is the availability of flexible and understandable software. Flexibility is achieved by providing the original FORTRAN source code. A detailed description of the theoretical background along with specific application notes allows the algorithms to be understood and used as the basis for more advanced customized programs. The three main chapters of this guidebook are based on the three major problem areas of geostatistics: quantifying spatial variability (variograms), generalized linear regression techniques (kriging), and stochastic simulation. Additional utility programs and problem sets with partial solutions are given to allow a full exploration of GSLIB and to check new software installations.
View
Soil Water Content Interpolation Using Spatio-Temporal Kriging with External Drift
Article
  • Mar 2003
In this study, two techniques for spatio-temporal (ST) kriging of soil water content are compared. The first technique, spatio-temporal ordinary kriging, is the simplest of the two, and uses only information about soil water content. The second technique, spatio-temporal kriging with external drift, uses also the relationship between soil water content and net-precipitation to aid the interpolation. It is shown that the behaviour of the soil water content predictions is physically more realistic when using spatio-temporal kriging with external drift. Also, the prediction uncertainties are slightly smaller. The data used in this study consist of Time Domain Reflectometry (TDR) measurements from a 30-day irrigation experiment on a 60×60-m grassland in the Netherlands.
View
FORTRAN programs for space-time modeling
Article
  • Mar 2002
  • COMPUT GEOSCI-UK
Modified GSLIB FORTRAN 77 routines are given in this paper for estimating and modeling space-time variograms. Two general families of models are incorporated in the programs: these are the product model and the product-sum model, both based on the decomposition of the space-time covariance in terms of a space covariance and a time covariance. The GSLIB kriging program has also been modified to incorporate these space-time models. One of the programs detects and removes temporal periodicities in the data. The program removes them and generates residuals for all monitoring stations, in order to estimate and model the spatial-temporal variogram using residuals. The modified kriging program also allows the use of cross-validation in conjunction with fitting of space-time variogram models. The trend component and the residual variogram model can be used for prediction. To illustrate the use of the programs, hourly averages of NO2 for the first ten months of 1998 in Lombardy were used.
View