Colin Robertson

Review of methods for space-time disease surveillance.

Abstract A review of some methods for analysis of space–time disease surveillance data is presented. Increasingly, surveillance systems are capturing spatial and temporal data on disease and health outcomes in a variety of public health contexts. A vast and growing suite of methods exists for detection of outbreaks and trends in surveillance data and the selection of appropriate methods in a given surveillance context is not always clear. While most reviews of methods focus on algorithm performance, in practice, a variety of factors determine what methods are appropriate for surveillance. In this review, we focus on the role of contextual factors such as scale, scope, surveillance objective, disease characteristics, and technical issues in relation to commonly used approaches to surveillance. Methods are classified as testing-based or model-based approaches. Reviewing methods in the context of factors other than algorithm performance highlights important aspects of implementing and selecting appropriate disease surveillance methods.

Mobile phone-based infectious disease surveillance system, Sri Lanka.

Because many infectious diseases are emerging in animals in low-income and middle-income countries, surveillance of animal health in these areas may be needed for forecasting disease risks to humans. We present an overview of a mobile phone–based frontline surveillance system developed and implemented in Sri Lanka. Field veterinarians reported animal health information by using mobile phones. Submissions increased steadily over 9 months, with ≈4,000 interactions between field veterinarians and reports on the animal population received by the system. Development of human resources and increased communication between local stakeholders (groups and persons whose actions are affected by emerging infectious diseases and animal health) were instrumental for successful implementation. The primary lesson learned was that mobile phone–based surveillance of animal populations is acceptable and feasible in lower-resource settings. However, any system implementation plan must consider the time needed to garner support for novel surveillance methods among users and stakeholders.

Review of software for space-time disease surveillance

Disease surveillance makes use of information technology at almost every stage of the process, from data collection and collation, through to analysis and dissemination. Automated data collection systems enable near-real time analysis of incoming data. This context places a heavy burden on software used for space-time surveillance. In this paper, we review software programs capable of space-time disease surveillance analysis, and outline some of their salient features, shortcomings, and usability. Programs with space-time methods were selected for inclusion, limiting our review to ClusterSeer, SaTScan, GeoSurveillance and the Surveillance package for R. We structure the review around stages of analysis: preprocessing, analysis, technical issues, and output. Simulated data were used to review each of the software packages. SaTScan was found to be the best equipped package for use in an automated surveillance system. ClusterSeer is more suited to data exploration, and learning about the different methods of statistical surveillance.

A Multi-Scale Approach to Exploring Urban Places in Geotagged Photographs

User-generated content (UGC) that contains spatial references, often referred to by the more bounded concept of Volunteered Geographic Information (VGI), is often touted as a potentially revolutionary data source for geographical research. This paper explores the capacity of one increasingly prevalent source of these data, geographically encoded photographs, to capture spatial expressions of place in an urban environment. Geotagged photographs were obtained from the Flickr API to build a geographic database of photographs for the city of Vancouver, Canada from 2001-2012. These data were aggregated to multiple geographic units represented as hexagonal lattices. Spatial patterns of photo aggregation were examined for tessellations that ranged from 0.25 ha to 1024 ha. Tags associated with each photo were also explored through the notion of ‘tag-space’ at multiple resolutions, or “scales”, of analysis through local log-odds ratios. Results indicate a significant interaction between tag-space semantics and spatial aggregation which suggests that consideration of scale effects should be integral to analysis of this type of tagged VGI for exploring citizens’ sensing of urban environments. The results indicate further that we may have to reconsider the interaction between encoded meaning, the methods used for extracting such meaning from tag-space, and exogenous and endogenous spatial scales of spatial UGC.

Comparing interpolation techniques for monthly rainfall mapping using multiple evaluation criteria and auxiliary data sources

Interpolating climatic variables such as rainfall is challenging due to the highly variable nature of meteorological processes, the effects of terrain and geography, and the difficulty in establishing a representative network of stations. While interpolation models are being adapted to include these effects, often the rainfall data contain significant gaps in coverage. In this paper, we evaluated rainfall data from an agro-ecological monitoring network for producing maps of total monthly rainfall in Sri Lanka. We compared four spatial interpolation techniques: inverse distance weighting, thin-plate splines, ordinary kriging, and Bayesian kriging. Error metrics were used to validate interpolations against independent data. Satellite data were used to assess the spatial pattern of rainfall. Results indicated that Bayesian kriging and splines performed best in low and high rainfall, respectively. Rainfall maps generated from the agro-ecological network were found to have accuracies consistent with previous studies in Sri Lanka. Rainfall data from an agro-ecological monitoring network were evaluated for producing maps of monthly rainfall in Sri Lanka.Inverse distance weighting, thin-plate splines, ordinary kriging, and Bayesian kriging were compared.Error metrics and the structural similarity index were employed to validate interpolations against independent data.Bayesian kriging and splines predicted the most accurately in low and high rainfall conditions, respectively.Interpolated rainfall predictions were found to be as accurate as previous studies in Sri Lanka.

Spatial epidemiology of suspected clinical leptospirosis in Sri Lanka.

Leptospirosis is one of the most widespread zoonoses in the world. A large outbreak of suspected human leptospirosis began in Sri Lanka during 2008. This study investigated spatial variables associated with suspected leptospirosis risk during endemic and outbreak periods. Data were obtained for monthly numbers of reported cases of suspected clinical leptospirosis for 2005-2009 for all of Sri Lanka. Space-time scan statistics were combined with regression modelling to test associations during endemic and outbreak periods. The cross-correlation function was used to test association between rainfall and leptospirosis at four locations. During the endemic period (2005-2007), leptospirosis risk was positively associated with shorter average distance to rivers and with higher percentage of agriculture made up of farms <0·20 hectares. Temporal correlation analysis of suspected leptospirosis cases and rainfall revealed a 2-month lag in rainfall-case association during the baseline period. Outbreak locations in 2008 were characterized by shorter distance to rivers and higher population density. The analysis suggests the possibility of household transmission in densely populated semi-urban villages as a defining characteristic of the outbreak. The role of rainfall in the outbreak remains to be investigated, although analysis here suggests a more complex relationship than simple correlation.

Comparative Spatial Dynamics of Japanese Encephalitis and Acute Encephalitis Syndrome in Nepal

Japanese Encephalitis (JE) is a vector-borne disease of major importance in Asia. Recent increases in cases have spawned the development of more stringent JE surveillance. Due to the difficulty of making a clinical diagnosis, increased tracking of common symptoms associated with JE—generally classified as the umbrella term, acute encephalitis syndrome (AES) has been developed in many countries. In Nepal, there is some debate as to what AES cases are, and how JE risk factors relate to AES risk. Three parts of this analysis included investigating the temporal pattern of cases, examining the age and vaccination status patterns among AES surveillance data, and then focusing on spatial patterns of risk factors. AES and JE cases from 2007–2011 reported at a district level (n = 75) were examined in relation to landscape risk factors. Landscape pattern indices were used to quantify landscape patterns associated with JE risk. The relative spatial distribution of landscape risk factors were compared using geographically weighted regression. Pattern indices describing the amount of irrigated land edge density and the degree of landscape mixing for irrigated areas were positively associated with JE and AES, while fragmented forest measured by the number of forest patches were negatively associated with AES and JE. For both JE and AES, the local GWR models outperformed global models, indicating spatial heterogeneity in risks. Temporally, the patterns of JE and AES risk were almost identical; suggesting the relative higher caseload of AES compared to JE could provide a valuable early-warning signal for JE surveillance and reduce diagnostic testing costs. Overall, the landscape variables associated with a high degree of landscape mixing and small scale irrigated agriculture were positively linked to JE and AES risk, highlighting the importance of integrating land management policies, disease prevention strategies and promoting healthy sustainable livelihoods in both rural and urban-fringe developing areas.

Refining spatial neighbourhoods to capture terrain effects

IntroductionSpatially explicit ecological research has increased substantially in the past 20 years. Most spatial approaches require the definition of a spatial neighbourhood or the region over which spatial relationships are modelled or assessed. Spatial neighbourhood definitions impact analysis results, and there are benefits in considering neighbourhood definitions that better capture ecological processes. The goal of this research is to present a simple and flexible approach in constraining ecological spatial neighbourhoods using terrain data.MethodsUsing watershed boundaries, we can restrict spatial neighbourhoods from combining populations or processes that should be separated by terrain effects. We demonstrate the need for ecological constraints by way of a simulation study and highlight our approach with a case study examining mountain pine beetle (Dendroctonus ponderosae, Coleoptera; Hopkins) infestation hot spots.ResultsOur results demonstrate how failure to constrain neighbourhoods can lead to errors when the spatial signals from unrelated populations are mixed. Also, unconstrained spatial neighbourhoods can unintentionally detect spatial relationships across many scales.ConclusionsThere will be benefits to studies that develop new, ecology-based approaches in defining spatial neighbourhoods that better illuminate ecological function of phenomena under study.

A Hidden Markov Model for Analysis of Frontline Veterinary Data for Emerging Zoonotic Disease Surveillance

Surveillance systems tracking health patterns in animals have potential for early warning of infectious disease in humans, yet there are many challenges that remain before this can be realized. Specifically, there remains the challenge of detecting early warning signals for diseases that are not known or are not part of routine surveillance for named diseases. This paper reports on the development of a hidden Markov model for analysis of frontline veterinary sentinel surveillance data from Sri Lanka. Field veterinarians collected data on syndromes and diagnoses using mobile phones. A model for submission patterns accounts for both sentinel-related and disease-related variability. Models for commonly reported cattle diagnoses were estimated separately. Region-specific weekly average prevalence was estimated for each diagnoses and partitioned into normal and abnormal periods. Visualization of state probabilities was used to indicate areas and times of unusual disease prevalence. The analysis suggests that hidden Markov modelling is a useful approach for surveillance datasets from novel populations and/or having little historical baselines.

Bumps and bruises in the digital skins of cities: unevenly distributed user-generated content across US urban areas

ABSTRACT As momentum and interest build to leverage new forms of user-generated content that contains geographical information, classical issues of data quality remain significant research challenges. In this article we explore issues of representativeness for one form of user-generated content, geotagged photographs in US urban centers. Generalized linear models were developed to associate photograph distribution with underlying socioeconomic descriptors at the city-scale, and examine intra-city variation in relation to income inequality. We conclude our analyses with a detailed examination of Dallas, Seattle, and New Orleans. Our findings add to the growing volume of evidence outlining uneven representativeness in user-generated data, and our approach contributes to the stock of methods available to investigate geographic variations in representativeness. In addition to city-scale variables relating to distribution of user-generated content, variability remains at localized scales that demand an individual and contextual understanding of their form and nature. The findings demonstrate that careful analysis of representativeness at both macro and micro scales can simultaneously provide important insights into the processes giving rise to user-generated data sets and potentially shed light on their embedded biases and suitability as inputs to analysis.