Paula Moraga

Spatial and spatio-temporal log-gaussian cox processes: Extending the geostatistical paradigm

In this paper we first describe the class of log-Gaussian Cox processes (LGCPs) as models for spatial and spatio-temporal point process data. We discuss inference, with a particular focus on the computational challenges of likelihood-based inference. We then demonstrate the usefulness of the LGCP by describing four applications: estimating the intensity surface of a spatial point process; investigating spatial segregation in a multi-type process; constructing spatially continuous maps of disease risk from spatially discrete data; and real-time health surveillance. We argue that problems of this kind fit naturally into the realm of geostatistics, which traditionally is defined as the study of spatially continuous processes using spatially discrete observations at a finite number of locations. We suggest that a more useful definition of geostatistics is by the class of scientific problems that it addresses, rather than by particular models or data formats.

Spatiotemporal Determinants of Urban Leptospirosis Transmission: Four-Year Prospective Cohort Study of Slum Residents in Brazil

Background Rat-borne leptospirosis is an emerging zoonotic disease in urban slum settlements for which there are no adequate control measures. The challenge in elucidating risk factors and informing approaches for prevention is the complex and heterogeneous environment within slums, which vary at fine spatial scales and influence transmission of the bacterial agent. Methodology/Principal Findings We performed a prospective study of 2,003 slum residents in the city of Salvador, Brazil during a four-year period (2003–2007) and used a spatiotemporal modelling approach to delineate the dynamics of leptospiral transmission. Household interviews and Geographical Information System surveys were performed annually to evaluate risk exposures and environmental transmission sources. We completed annual serosurveys to ascertain leptospiral infection based on serological evidence. Among the 1,730 (86%) individuals who completed at least one year of follow-up, the infection rate was 35.4 (95% CI, 30.7–40.6) per 1,000 annual follow-up events. Male gender, illiteracy, and age were independently associated with infection risk. Environmental risk factors included rat infestation (OR 1.46, 95% CI, 1.00–2.16), contact with mud (OR 1.57, 95% CI 1.17–2.17) and lower household elevation (OR 0.92 per 10m increase in elevation, 95% CI 0.82–1.04). The spatial distribution of infection risk was highly heterogeneous and varied across small scales. Fixed effects in the spatiotemporal model accounted for the majority of the spatial variation in risk, but there was a significant residual component that was best explained by the spatial random effect. Although infection risk varied between years, the spatial distribution of risk associated with fixed and random effects did not vary temporally. Specific “hot-spots” consistently had higher transmission risk during study years. Conclusions/Significance The risk for leptospiral infection in urban slums is determined in large part by structural features, both social and environmental. Our findings indicate that topographic factors such as household elevation and inadequate drainage increase risk by promoting contact with mud and suggest that the soil-water interface serves as the environmental reservoir for spillover transmission. The use of a spatiotemporal approach allowed the identification of geographic outliers with unexplained risk patterns. This approach, in addition to guiding targeted community-based interventions and identifying new hypotheses, may have general applicability towards addressing environmentally-transmitted diseases that have emerged in complex urban slum settings.

Modelling the distribution and transmission intensity of lymphatic filariasis in sub-Saharan Africa prior to scaling up interventions: integrated use of geostatistical and mathematical modelling

BackgroundLymphatic filariasis (LF) is one of the neglected tropical diseases targeted for global elimination. The ability to interrupt transmission is, partly, influenced by the underlying intensity of transmission and its geographical variation. This information can also help guide the design of targeted surveillance activities. The present study uses a combination of geostatistical and mathematical modelling to predict the prevalence and transmission intensity of LF prior to the implementation of large-scale control in sub-Saharan Africa.MethodsA systematic search of the literature was undertaken to identify surveys on the prevalence of Wuchereria bancrofti microfilaraemia (mf), based on blood smears, and on the prevalence of antigenaemia, based on the use of an immuno-chromatographic card test (ICT). Using a suite of environmental and demographic data, spatiotemporal multivariate models were fitted separately for mf prevalence and ICT-based prevalence within a Bayesian framework and used to make predictions for non-sampled areas. Maps of the dominant vector species of LF were also developed. The maps of predicted prevalence and vector distribution were linked to mathematical models of the transmission dynamics of LF to infer the intensity of transmission, quantified by the basic reproductive number (R0).ResultsThe literature search identified 1267 surveys that provide suitable data on the prevalence of mf and 2817 surveys that report the prevalence of antigenaemia. Distinct spatial predictions arose from the models for mf prevalence and ICT-based prevalence, with a wider geographical distribution when using ICT-based data. The vector distribution maps demonstrated the spatial variation of LF vector species. Mathematical modelling showed that the reproduction number (R0) estimates vary from 2.7 to 30, with large variations between and within regions.ConclusionsLF transmission is highly heterogeneous, and the developed maps can help guide intervention, monitoring and surveillance strategies as countries progress towards LF elimination.

Gaussian component mixtures and CAR models in Bayesian disease mapping

Hierarchical Bayesian models involving conditional autoregression (CAR) components are commonly used in disease mapping. An alternative model to the proper or improper CAR is the Gaussian component mixture (GCM) model. A review of CAR and GCM models is provided in univariate settings where only one disease is considered, and also in multivariate situations where in addition to the spatial dependence between regions, the dependence among multiple diseases is analyzed. A performance comparison between models using a set of simulated data to help illustrate their respective properties is reported. The results show that both in univariate and multivariate settings, both models perform in a comparable way under a wide range of conditions. GCM and CAR models are applied for estimating the relative risk of low birth weight in Georgia, USA, in the year 2000.

Detection of spatial variations in temporal trends with a quadratic function

Methods for the assessment of spatial variations in temporal trends (SVTT) are important tools for disease surveillance, which can help governments to formulate programs to prevent diseases, and measure the progress, impact, and efficacy of preventive efforts already in operation. The linear SVTT method is designed to detect areas with unusual different disease linear trends. In some situations, however, its estimation trend procedure can lead to wrong conclusions. In this article, the quadratic SVTT method is proposed as alternative of the linear SVTT method. The quadratic method provides better estimates of the real trends, and increases the power of detection in situations where the linear SVTT method fails. A performance comparison between the linear and quadratic methods is provided to help illustrate their respective properties. The quadratic method is applied to detect unusual different cervical cancer trends in white women in the United States, over the period 1969 to 1995.

An investigation of the disparity in estimates of microfilaraemia and antigenaemia in lymphatic filariasis surveys

Background The diagnosis of lymphatic filariasis (LF) is based typically on either microfilaraemia as assessed by microscopy or filarial antigenaemia using an immuno-chromatographic test. While it is known that estimates of antigenaemia are generally higher than estimates of microfilaraemia, the extent of the difference is not known. Methods This paper presents the results of an extensive literature search for surveys that estimated both microfilaraemia and antigenaemia in order to better understand the disparity between the two measures. Results and Conclusions In some settings there was a very large disparity, up to 40–70%, between estimates of microfilaraemia and antigenaemia. Regression analysis was unable to identify any predictable relationship between the two measures. The implications of findings for risk mapping and surveillance of LF are discussed.

Spatial variation in cancer incidence and survival over time across Queensland, Australia

Interpreting changes over time in small-area variation in cancer survival, in light of changes in cancer incidence, aids understanding progress in cancer control, yet few space-time analyses have considered both measures. Bayesian space-time hierarchical models were applied to Queensland Cancer Registry data to examine geographical changes in cancer incidence and relative survival over time for the five most common cancers (colorectal, melanoma, lung, breast, prostate) diagnosed during 1997-2004 and 2005-2012 across 516 Queensland residential small-areas. Large variation in both cancer incidence and survival was observed. Survival improvements were fairly consistent across the state, although small for lung cancer. Incidence changes varied by location and cancer type, ranging from lung and colorectal cancers remaining relatively constant over time, to prostate cancer dramatically increasing across the entire state. Reducing disparities in cancer-related outcomes remains a health priority, and space-time modelling of different measures provides an important mechanism by which to monitor progress.

Model-based imputation of missing data from the 122 Cities Mortality Reporting System (122 CMRS)

National estimates of the all-cause and pneumonia and influenza (P&I) mortality burden derived from U.S. influenza surveillance data treat all missing or unreported values as zero counts. The effect of this methodological decision is to undercount influenza deaths, thus biasing estimates downward and producing underestimates of the true mortality burden. In this paper, a regression-based procedure is proposed to impute missing values and thus produce a more accurate estimate of mortality. Several model specifications are considered and evaluated to predict weekly death counts by city, calendar week, calendar year and age group. Revised all-cause, P&I and excess mortality estimates are calculated by imputing the missing data. The impact of the treatment of unreported mortality data on national estimates is evaluated by comparing the estimates obtained using data with and without imputation. This comparison reflects some differences in mortality burden, excess deaths, and trends over time. The model presented is a useful approach to impute missing counts and improve inference in situations with modest occurrence of missing data.

SpatialEpiApp: A Shiny web application for the analysis of spatial and spatio-temporal disease data

During last years, public health surveillance has been facilitated by the existence of several packages implementing statistical methods for the analysis of spatial and spatio-temporal disease data. However, these methods are still inaccesible for many researchers lacking the adequate programming skills to effectively use the required software. In this paper we present SpatialEpiApp, a Shiny web application that integrate two of the most common approaches in health surveillance: disease mapping and detection of clusters. SpatialEpiApp is easy to use and does not require any programming knowledge. Given information about the cases, population and optionally covariates for each of the areas and dates of study, the application allows to fit Bayesian models to obtain disease risk estimates and their uncertainty by using R-INLA, and to detect disease clusters by using SaTScan. The application allows user interaction and the creation of interactive data visualizations and reports showing the analyses performed.