Search

Uncertainty surrounding the risk of developing and dying from Thrombosis and Thromobocytopenia Syndrome (TTS) associated with the AstraZeneca (AZ) COVID-19 vaccine may contribute to vaccine hesitancy. A model is urgently needed to combine and effectively communicate the existing evidence on the risks versus benefits of the AZ vaccine. We developed a Bayesian network to consolidate the existing evidence on risks and benefits of the AZ vaccine, and parameterised the model using data from a range of empirical studies, government reports, and expert advisory groups. Expert judgement was used to interpret the available evidence and determine the structure of the model, relevant variables, data to be included, and how these data were used to inform the model.The model can be used as a decision support tool to generate scenarios based on age, sex, virus variant and community transmission rates, making it a useful for individuals, clinicians, and researchers to assess the chances of different health outcomes. Model outputs include the risk of dying from TTS following the AZ COVID-19 vaccine, the risk of dying from COVID-19 or COVID-19-associated atypical severe blood clots under different scenarios. Although the model is focused on Australia, it can be easily adaptable to international settings by re-parameterising it with local data. This paper provides detailed description of the model-building methodology, which can used to expand the scope of the model to include other COVID-19 vaccines, booster doses, comorbidities and other health outcomes (e.g., long COVID) to ensure the model remains relevant in the face of constantly changing discussion on risks versus benefits of COVID-19 vaccination.

Thrombosis and Thromobocytopenia Syndrome (TTS) has been associated with the AstraZencea (AZ) COVID-19 vaccine. Australia has reported low TTS incidence of 3000 deaths prevented under high transmission. Risks versus benefits varied significantly between age groups and transmission levels. Under high transmission, deaths prevented by AZ vaccine far exceed deaths from TTS (by 8 to >4500 times depending on age). Probability of dying from COVID-related atypical severe blood clots was 58-126 times higher (depending on age and sex) than dying from TTS. To our knowledge, this is the first example of the use of Bayesian networks for risk-benefit analysis for a COVID-19 vaccine. The model can be rapidly updated to incorporate new data, adapted for other countries, extended to other outcomes (e.g., severe disease), or used for other vaccines.HIGHLIGHTSAZ vaccination risk-benefit analysis must consider age/community transmission levelAZ vaccine benefits far outweigh risks in older age groups and during high transmissionAZ vaccine-associated TTS lower fatality than COVID-related atypical blood clotsBayesian networks utility for risk-benefit analysis of rapidly evolving situationsBNs allow integrating multiple data sources when large datasets are not available

Graphical abstract, Highlights • We compared logistic regression, recursive partitioning and Bayesian networks to identify risk factors for STH infection. • Logistic regression identified fewest variables associated with STH infections compared with the two alternative methods. • Recursive partitioning identified more demographic and WASH variables, and Bayesian networks more environmental variables. • Model performance was similar across all three statistical techniques. • Recursive partitioning can identify at-risk population subgroups, while Bayesian networks can run real-time scenarios., Soil-transmitted helminths (STHs) are parasitic intestinal worms that infect almost a fifth of the global population. Sustainable control of STHs requires understanding the complex interaction of factors contributing to transmission. Identifying risk factors has mainly relied on logistic regression models where the underlying assumption of independence between variables is not always satisfied. Previously demonstrated risk factors including water, sanitation and hygiene (WASH) access and behaviours, and socioeconomic status are intrinsically linked. Similarly, environmental factors including climate, soil and land attributes are often strongly correlated. Alternative methods such as recursive partitioning and Bayesian networks can handle correlated variables, but there are no published studies comparing these methods with logistic regression in the context of STH risk factor analysis. Baseline cross-sectional data from school-aged children in the (S)WASH-D for Worms study were used to compare risk factors identified from modelling the same data using three different statistical techniques. Outcomes of interest were infection with Ascaris spp. and any hookworm species (Necator americanus, Ancylostoma duodenale, and Ancylostoma ceylanicum). Mixed-effects logistic regression identified the fewest risk factors. Recursive partitioning identified the most WASH and demographic risk factors, while Bayesian networks identified the most environmental risk factors. Recursive partitioning produced classification trees that visualised potentially at-risk population sub-groups. Bayesian networks helped visualise relationships between variables and enabled interactive modelling of outcomes based on different scenarios for the predictor variables of interest. Model performance was similar across all techniques. Risk factors identified across all techniques were vegetation for Ascaris spp., and cleaning oneself with water after defecating for hookworm. This study adds to the limited body of evidence exploring alternative data modelling approaches in identifying risk factors for STH infections. Our findings suggest these approaches can provide novel insights for more robust interpretation.

Australia’s 2019–2020 mega-fires were exacerbated by drought, anthropogenic climate change and existing land-use management. Here, using a combination of remotely sensed data and species distribution models, we found these fires burnt ~97,000 km2 of vegetation across southern and eastern Australia, which is considered habitat for 832 species of native vertebrate fauna. Seventy taxa had a substantial proportion (>30%) of habitat impacted; 21 of these were already listed as threatened with extinction. To avoid further species declines, Australia must urgently reassess the extinction vulnerability of fire-impacted species and assist the recovery of populations in both burnt and unburnt areas. Population recovery requires multipronged strategies aimed at ameliorating current and fire-induced threats, including proactively protecting unburnt habitats. An assessment of the habitat of native vertebrate species burnt by the 2019–2020 Australian mega-fires shows that 70 taxa were severely affected.

Biodiversity conservation decisions are difficult, especially when they involve differing values, complex multidimensional objectives, scarce resources, urgency, and considerable uncertainty. Decision science embodies a theory about how to make difficult decisions and an extensive array of frameworks and tools that make that theory practical. We sought to improve conceptual clarity and practical application of decision science to help decision makers apply decision science to conservation problems. We addressed barriers to the uptake of decision science, including a lack of training and awareness of decision science; confusion over common terminology and which tools and frameworks to apply; and the mistaken impression that applying decision science must be time consuming, expensive, and complex. To aid in navigating the extensive and disparate decision science literature, we clarify meaning of common terms: decision science, decision theory, decision analysis, structured decision-making, and decision-support tools. Applying decision science does not have to be complex or time consuming; rather, it begins with knowing how to think through the components of a decision utilizing decision analysis (i.e., define the problem, elicit objectives, develop alternatives, estimate consequences, and perform trade-offs). This is best achieved by applying a rapid-prototyping approach. At each step, decision-support tools can provide additional insight and clarity, whereas decision-support frameworks (e.g., priority threat management and systematic conservation planning) can aid navigation of multiple steps of a decision analysis for particular contexts. We summarize key decision-support frameworks and tools and describe to which step of a decision analysis, and to which contexts, each is most useful to apply. Our introduction to decision science will aid in contextualizing current approaches and new developments, and help decision makers begin to apply decision science to conservation problems.Las decisiones sobre la conservación de la biodiversidad son difíciles de tomar, especialmente cuando involucran diferentes valores, objetivos multidimensionales complejos, recursos limitados, urgencia y una incertidumbre considerable. Las ciencias de la decisión incorporan una teoría sobre cómo tomar decisiones difíciles y una variedad extensa de marcos de trabajo y herramientas que transforman esa teoría en práctica. Buscamos mejorar la claridad conceptual y la aplicación práctica de las ciencias de la decisión para ayudar al órgano decisorio a aplicar estas ciencias a los problemas de conservación. Nos enfocamos en las barreras para la aceptación de las ciencias de la decisión, incluyendo la falta de capacitación y de conciencia por estas ciencias; la confusión por la terminología común y cuáles herramientas y marcos de trabajo aplicar; y la impresión errónea de que la aplicación de estas ciencias consume tiempo y debe ser costosa y compleja. Para asistir en la navegación de la literatura extensa y dispar de las ciencias de la decisión, aclaramos el significado de varios términos comunes: ciencias de la decisión, teoría de la decisión, análisis de decisiones, toma estructurada de decisiones y herramientas de apoyo para las decisiones. La aplicación de las ciencias de la decisión no tiene que ser compleja ni debe llevar mucho tiempo; de hecho, todo comienza con saber cómo pensar detenidamente en los componentes de una decisión mediante el análisis de decisiones (es decir, definir el problema, producir objetivos, desarrollar alternativas, estimar consecuencias y realizar compensaciones). Lo anterior se logra de mejor manera mediante la aplicación de una estrategia prototipos rápidos. En cada paso, las herramientas de apoyo para las decisiones pueden proporcionar visión y claridad adicionales, mientras que los marcos de apoyo para las decisiones (p.ej.: gestión de amenazas prioritarias y planeación sistemática de la conservación) pueden asistir en la navegación de los diferentes pasos de un análisis de decisiones para contextos particulares. Resumimos los marcos de trabajo y las herramientas más importantes de apoyo para las decisiones y describimos el paso, y el contexto, del análisis de decisiones para el que es más útil aplicarlos. Nuestra introducción a las ciencias de la decisión apoyará en la contextualización de las estrategias actuales y los nuevos desarrollos, y ayudarán al órgano decisorio a comenzar a aplicar estas ciencias en los problemas de conservación.

IntroductionThe increase in international travel brought about by globalisation has enabled the rapid spread of emerging pathogens with epidemic and pandemic potential. While travel connectivity-based assessments may help understand patterns of travel network-mediated epidemics, such approaches are rarely carried out in sufficient detail for Oceania where air travel is the dominant method of transportation between countries.DesignTravel data from the Australian Bureau of Statistics, Stats NZ and the United Nations World Tourism Organization websites were used to calculate travel volumes in 2018 within Oceania and between Oceania and the rest of the world. The Infectious Disease Vulnerability Index (IDVI) was incorporated into the analysis as an indicator of each country’s capacity to contain an outbreak. Travel networks were developed to assess the spread of infectious diseases (1) into and from Oceania, (2) within Oceania and (3) between each of the Pacific Island Countries and Territories (PICTs) and their most connected countries.ResultsOceania was highly connected to countries in Asia, Europe and North America. Australia, New Zealand and several PICTs were highly connected to the USA and the UK (least vulnerable countries for outbreaks based on the IDVI), and to China (intermediate low vulnerable country). High variability was also observed between the PICTs in the geographical distribution of their international connections. The PICTs with the highest number of international connections were Fiji, French Polynesia, Guam and Papua New Guinea.ConclusionTravel connectivity assessments may help to accurately stratify the risk of infectious disease importation and outbreaks in countries depending on disease transmission in other parts of the world. This information is essential to track future requirements for scaling up and targeting outbreak surveillance and control strategies in Oceania.

Lymphatic filariasis (LF) is a major public health problem globally and in the Pacific Region. The Global Programme to Eliminate LF has made great progress but LF is persistent and resurgent in some Pacific countries and territories. Samoa remains endemic for LF despite elimination efforts through multiple two-drug mass drug administrations (MDA) since 1965, including renewed elimination efforts started in 1999 under the Pacific Programme for Elimination of LF (PacELF). Despite eight rounds of national and two rounds of subnational MDA under PacELF, Samoa failed transmission assessment surveys (TAS) in all three evaluation units in 2017. In 2018, Samoa was the first to distribute countrywide triple-drug MDA using ivermectin, diethylcarbamazine (DEC), and albendazole. This paper provides a review of MDAs and historical survey results from 1998 to 2017 in Samoa and highlights lessons learnt from LF elimination efforts, including challenges and potential ways to overcome them to successfully achieve elimination.

The Global Programme to Eliminate Lymphatic Filariasis has made considerable progress but is experiencing challenges in meeting targets in some countries. Recent World Health Organization guidelines have recommended two rounds of triple-drug therapy with ivermectin, diethylcarbamazine (DEC), and albendazole (IDA), in areas where mass drug administration (MDA) results with two drugs (DEC and albendazole) have been suboptimal, as is the case in Samoa. In August 2018, Samoa was the first country in the world to implement countrywide triple-drug MDA. This paper aims to describe Samoa’s experience with program coverage and adverse events (AEs) in the first round of triple-drug MDA. We conducted a large cross-sectional community survey to assess MDA awareness, reach, compliance, coverage and AEs in September/October 2018, 7–11 weeks after the first round of triple-drug MDA. In our sample of 4420 people aged ≥2 years (2.2% of the population), age-adjusted estimates indicated that 89.0% of the eligible population were offered MDA, 83.9% of the eligible population took MDA (program coverage), and 80.2% of the total population took MDA (epidemiological coverage). Overall, 83.8% (2986/3563) reported that they did not feel unwell at all after taking MDA. Mild AEs (feeling unwell but able to do normal everyday things) were reported by 13.3% (476/3563) and moderate or severe AEs (feeling unwell and being unable to do normal everyday activities such as going to work or school) by 2.9% (103/3563) of participants. This study following the 2018 triple-drug MDA in Samoa demonstrated a high reported program awareness and reach of 90.8% and 89.0%, respectively. Age-adjusted program coverage of 83.9% of the total population showed that MDA was well accepted and well tolerated by the community., Author summary Lymphatic filariasis is a disease caused by worms and transmitted between humans by mosquitoes. A global program underway aims to eliminate lymphatic filariasis as a public health problem by distributing two deworming drugs to the whole population once a year for at least five years. In some countries, including Samoa, this strategy has not been sufficient to eliminate transmission. A new drug has been added, and in 2018, Samoa was the first country in the world to distribute country-wide triple-drug mass drug administration (MDA) using ivermectin, diethylcarbamazine, and albendazole. This study reports on the coverage achieved (percentage of people who reported taking the drugs) and adverse events after taking the drugs. The study was a large community survey of over 4000 people, done 7–11 weeks after the distribution of the first round of triple-drug MDA. We found that the program reached and offered MDA to approximately 90% of the whole population, and approximately 80% of the whole population swallowed the drugs. We also collected data on whether people felt unwell after taking the pills, and found that 84% of people taking the pills did not feel unwell at all, while 13% reported feeling unwell afterwards but were able to do normal everyday activities, and 3% reported that they had felt unwell and that it stopped them doing normal everyday activities, such as going to school or work.

Evaluating the effectiveness of management actions for threatened species recovery is critical for adaptive management. However, decision makers frequently lack the resources and time to develop data‐driven models for rigorous monitoring and evaluation. Expert knowledge can be useful in such situations, but can be challenging to translate into specific expectations about system responses. We describe a case study of the Saving Our Species program in New South Wales, Australia, showing how an integrated process drawing from structured decision making, conceptual modeling and structured expert elicitation can help species' managers formalize how a species is expected to respond to management over time and the assumptions about the mechanisms driving that response. The process described uses step‐by‐step guidelines to assist managers in defining the scope, documenting factors that influence outcomes and identifying indicators and target values for adaptive management plans. This case study demonstrates a robust, yet practical process for estimating species response to site management in situations where empirical monitoring data is absent or incomplete. In doing so it provides a useful tool that helps species' managers in data poor situations to take a more evidence‐informed approach to conservation.

The global elimination of lymphatic filariasis (LF) is a major focus of the World Health Organization. One key challenge is locating residual infections that can perpetuate the transmission cycle. We show how a targeted sampling strategy using predictions from a geospatial model, combining random forests and geostatistics, can improve the sampling efficiency for identifying locations with high infection prevalence. Predictions were made based on the household locations of infected persons identified from previous surveys, and environmental variables relevant to mosquito density. Results show that targeting sampling using model predictions would have allowed 52% of infections to be identified by sampling just 17.7% of households. The odds ratio for identifying an infected individual in a household at a predicted high risk compared to a predicted low risk location was 10.2 (95% CI 4.2–22.8). This study provides evidence that a ‘one size fits all’ approach is unlikely to yield optimal results when making programmatic decisions based on model predictions. Instead, model assumptions and definitions should be tailored to each situation based on the objective of the surveillance program. When predictions are used in the context of the program objectives, they can result in a dramatic improvement in the efficiency of locating infected individuals.

The impact of environmental factors on human health outcomes is well established. It is therefore not surprising that interventions aimed at improving human health are often environmental-based, such as restoring riparian vegetation for flood mitigation, with a view to reducing associated infectious disease transmission. Yet the risks and benefits of these interventions on the environment itself are rarely measured, or weighed up against potential health gains. One of the challenges with such an evaluation is the requirement for cross-sectoral support from decision makers in both the health and environmental sectors. To facilitate this support, cross-sectoral models are required that simultaneously estimate the impact of proposed environmental interventions on both sectors. Despite their obvious value, a systematic search of the peer-reviewed literature did not identify any model that concurrently models the impact of environmental intervention on both environmental and human infectious disease related outcomes. In this paper, we conceptually explore potential approaches for designing such a model, using leptospirosis as a case study to highlight the various data sources, spatial scales, temporal scales and required system behaviour that would need to be integrated for a cross-sectoral model of this complexity. By comparing these system requirements against the strengths and limitations of individual modelling techniques, we demonstrate the potential benefits of a hybrid-ensemble approach that uses component models from different frameworks. By combining the strengths of the different techniques to tackle this wicked problem, such a modelling approach supports the prioritisation of environmental interventions that optimise the overall benefit by considering impacts on both human health and the environment.

Background Samoa conducted eight nationwide rounds of mass drug administration (MDA) for lymphatic filariasis (LF) between 1999 and 2011, and two targeted rounds in 2015 and 2017 in North West Upolu (NWU), one of three evaluation units (EUs). Transmission Assessment Surveys (TAS) were conducted in 2013 (failed in NWU) and 2017 (all three EUs failed). In 2018, Samoa was the first in the world to distribute nationwide triple-drug MDA using ivermectin, diethylcarbamazine, and albendazole. Surveillance and Monitoring to Eliminate LF and Scabies from Samoa (SaMELFS Samoa) is an operational research program designed to evaluate the effectiveness of triple-drug MDA on LF transmission and scabies prevalence in Samoa, and to compare the usefulness of different indicators of LF transmission. This paper reports results from the 2018 baseline survey and aims to i) investigate antigen (Ag) prevalence and spatial epidemiology, including geographic clustering; ii) compare Ag prevalence between two different age groups (5–9 years versus ≥10 years) as indicators of areas of ongoing transmission; and iii) assess the prevalence of limb lymphedema in those aged ≥15 years. Methods A community-based cluster survey was conducted in 30 randomly selected and five purposively selected clusters (primary sampling units, PSUs), each comprising one or two villages. Participants were recruited through household surveys (age ≥5 years) and convenience surveys (age 5–9 years). Alere Filariasis Test Strips (FTS) were used to detect Ag, and prevalence was adjusted for survey design and standardized for age and gender. Adjusted Ag prevalence was estimated for each age group (5–9, ≥10, and all ages ≥5 years) for random and purposive PSUs, and by region. Intraclass correlation (ICC) was used to quantify clustering at regions, PSUs, and households. Results A total of 3940 persons were included (1942 children aged 5–9 years, 1998 persons aged ≥10 years). Adjusted Ag prevalence in all ages ≥5 years in randomly and purposively selected PSUs were 4.0% (95% CI 2.8–5.6%) and 10.0% (95% CI 7.4–13.4%), respectively. In random PSUs, Ag prevalence was lower in those aged 5–9 years (1.3%, 95% CI 0.8–2.1%) than ≥10 years (4.7%, 95% CI 3.1–7.0%), and poorly correlated at the PSU level (R-square = 0.1459). Adjusted Ag prevalence in PSUs ranged from 0% to 10.3% (95% CI 5.9–17.6%) in randomly selected and 3.8% (95% CI 1.3–10.8%) to 20.0% (95% CI 15.3–25.8%) in purposively selected PSUs. ICC for Ag-positive individuals was higher at households (0.46) compared to PSUs (0.18) and regions (0.01). Conclusions Our study confirmed ongoing transmission of LF in Samoa, in accordance with the 2017 TAS results. Ag prevalence varied significantly between PSUs, and there was poor correlation between prevalence in 5–9 year-olds and older ages, who had threefold higher prevalence. Sampling older age groups would provide more accurate estimates of overall prevalence, and be more sensitive for identifying residual hotspots. Higher prevalence in purposively selected PSUs shows local knowledge can help identify at least some hotspots., Author summary Lymphatic filariasis (LF), a disease caused by infection with worms transmitted by mosquitoes, has long been present in Samoa. Since the 1960s, Samoa has attempted to control the disease through many rounds of annual administration of two deworming drugs to the whole population. However, Samoa recently observed that LF transmission was still occurring, prompting mass drug administration (MDA) with three drugs in 2018. Here, we report the baseline survey of an operational research program to evaluate the triple-drug MDA. The survey assessed prevalence and geographical distribution of LF in the population, compared prevalence by age groups, and investigated the burden of elephantiasis (swollen limbs caused by long-term LF infection). The study confirmed ongoing transmission, with 4% of those aged ≥5 years showing antigen in their blood as evidence of infection. Antigen prevalence was more than three times higher in those aged ≥10 years (4.7%) than in 5–9 year-old children (1.3%). Infection was highly clustered within households and villages, with up to 20% of residents infected in known hotspot villages. Future surveillance strategies should consider that i) testing older age groups would provide more accurate indication of LF transmission, and ii) local knowledge can help identify transmission hotspots.

Background: The increase in international travel brought about by globalisation has enabled the rapid spread of emerging pathogens with epidemic and pandemic potential. While travel connectivity-based assessments may help understand patterns of travel network-mediated epidemics, such approaches are rarely carried out in sufficient detail for Oceania where air travel is the dominant method of transportation between countries. Methods: Travel data from the Australian Bureau of Statistics, Stats NZ, and the United Nations World Tourism Organization websites were used to calculate travel volumes in 2018 within Oceania and between Oceania and the rest of the world. The Infectious Disease Vulnerability Index (IDVI) was incorporated into the analysis as an indicator of each country’s capacity to contain an outbreak. Travel networks were developed to assess the introduction of infectious diseases to and from Oceania, the spread of infectious diseases within Oceania and the spread of infectious diseases between each PICT and their most connected countries. Findings: Oceania and countries within the continent were highly connected to countries in Asia, Europe, and North America. Australia, New Zealand and several PICTs were highly connected to the United States of America and the United Kingdom (least vulnerable countries for outbreaks based on the IDVI), and to China (intermediate low vulnerable country). High variability was also observed between the PICTS in the geographical distribution of their international connections. The PICTs with the highest number of such connections were Fiji, French Polynesia, Guam, and Papua New Guinea. Interpretation: Travel connectivity assessments may help to stratify more accurately outbreak risks in countries depending on disease transmission in other parts of the world. This information is essential to track future requirements for scaling up and targeting outbreak surveillance and control strategies in Oceania. Funding Statement: LFK was supported by Australian National Health and Medical Research Council Early Career Fellowships (APP1158469). Declaration of Interests: The authors declare no competing interests.

Governments in Australia and internationally are experimenting with policy instruments to facilitate the adoption of farming practices with reduced environmental impacts. The Great Barrier Reef (Australia) is one such case, where sustained efforts over 20 years have yielded insufficient progress towards targets to reduce the impacts of agriculture on water quality in downstream marine ecosystems. We present a critical review of policy instruments as implemented in Great Barrier Reef catchments. We catalogue the evolving mix of policy instruments employed in reef programs, and examine evidence of the effectiveness of agricultural extension, financial incentives, and direct regulation of farming practices. There is little robust evidence to assess instrument effectiveness, in part due to the evolving mix of the instruments employed, weak program evaluation and heterogeneity of agricultural enterprises. We identify the need to improve the understanding of instrument fit to landholders and enterprises. We recommend a modelling approach to clarify pathways to impact and guide improved policy evaluation.

Pooled testing (also known as group testing), where diagnostic tests are performed on pooled samples, has broad applications in the surveillance of diseases in animals and humans. An increasingly common use case is molecular xenomonitoring (MX), where surveillance of vector-borne diseases is conducted by capturing and testing large numbers of vectors (e.g. mosquitoes). The R package PoolTestR was developed to meet the needs of increasingly large and complex molecular xenomonitoring surveys but can be applied to analyse any data involving pooled testing. PoolTestR includes simple and flexible tools to estimate prevalence and fit fixed- and mixed-effect generalised linear models for pooled data in frequentist and Bayesian frameworks. Mixed-effect models allow users to account for the hierarchical sampling designs that are often employed in surveys, including MX. We demonstrate the utility of PoolTestR by applying it to a large synthetic dataset that emulates a MX survey with a hierarchical sampling design.

Regression models are the standard approaches used in infectious disease epidemiology, but have limited ability to represent causality or complexity. We explore Bayesian networks (BNs) as an alternative approach for modelling infectious disease transmission, using leptospirosis as an example. Data were obtained from a leptospirosis study in Fiji in 2013. We compared the performance of naive versus expert-structured BNs for modelling the relative importance of animal species in disease transmission in different ethnic groups and residential settings. For BNs of animal exposures at the individual/household level, R-2 for predicted versus observed infection rates were 0.59 for naive and 0.75-0.93 for structured models of ethnic groups; and 0.54 for naive and 0.93-1.00 for structured models of residential settings. BNs provide a promising approach for modelling infectious disease transmission under complex scenarios. The relative importance of animal species varied between subgroups, with important implications for more targeted public health control strategies. (C) 2017 Elsevier Ltd. All rights reserved.

The range and quality of freely available geo-referenced datasets is increasing. We evaluate the usefulness of free datasets for deforestation prediction by comparing generalised linear models and generalised linear mixed models (GLMMs) with a variety of machine learning models (Bayesian networks, artificial neural networks and Gaussian processes) across two study regions. Freely available datasets were able to generate plausible risk maps of deforestation using all techniques for study zones in both Mexico and Madagascar. Artificial neural networks outperformed GLMMs in the Madagascan (average AUC 0.83 vs 0.80), but not the Mexican study zone (average AUC 0.81 vs 0.89). In Mexico and Madagascar, Gaussian processes (average AUC 0.89, 0.85) and structured Bayesian networks (average AUC 0.88, 0.82) performed at least as well as GLMMs (average AUC 0.89, 0.80). Bayesian networks produced more stable results across different sampling methods. Gaussian processes performed well (average AUC 0.85) with fewer predictor variables. Freely available datasets have proven valuable in predicating deforestation.Machine learning techniques are a reliable alternative to statistics.Gaussian processes are suggested as an alternative to artificial neural networks.Bayesian networks were more stable across sample methods.

Introduction Leptospirosis is a zoonotic disease responsible for over 1 million severe cases and 60,000 deaths annually. The wide range of animal hosts and complex environmental drivers of transmission make targeted interventions challenging, particularly when restricted to regression-based analyses which have limited ability to deal with complexity. In Fiji, important environmental and socio-demographic factors include living in rural areas, poverty, and livestock exposure. This study aims to examine drivers of transmission under different scenarios of environmental and livestock exposures. Methods Spatial Bayesian networks (SBN) were used to analyse the influence of livestock and poverty on the risk of leptospirosis infection in urban compared to rural areas. The SBN models used a combination of spatially-explicit field data from previous work and publically available census information. Predictive risk maps were produced for overall risk, and for scenarios related to poverty, livestock, and urban/rural setting. Results While high, rather than low, commercial dairy farm density similarly increased the risk of infection in both urban (12% to 18%) and rural areas (70% to 79%), the presence of pigs in a village had different impact in rural (43% to 84%) compared with urban areas (4% to 24%). Areas with high poverty rates were predicted to have 26.6% and 18.0% higher probability of above average seroprevalence in rural and urban areas, respectively. In urban areas, this represents >300% difference between areas of low and high poverty, compared to 43% difference in rural areas. Conclusions Our study demonstrates the use of SBN to provide valuable insights into the drivers of leptospirosis transmission under complex scenarios. By estimating the risk of leptospirosis infection under different scenarios, such as urban versus rural areas, these subgroups or areas can be targeted with more precise interventions that focus on the most relevant key drivers of infection., Author summary Leptospirosis is a zoonotic disease responsible for over 60,000 deaths annually and is transmitted from mammal hosts to humans through contact with infected urine. The range of possible hosts and complex environmental factors related to transmission make targeted interventions challenging. We used spatial Bayesian Networks applied to a case study in Fiji to show that livestock exposure and poverty affect the probability of infection differently in rural compared to urban areas. This work illustrates the complexity of leptospirosis transmission drivers in Fiji, and shows how they are affected by the interactions between livestock exposure and other environmental and socio-demographic factors. In doing so, we support previous findings linking the risk of leptospirosis to poverty.

There are a many reasons for creating quantitative models of deforestation, supported by a variety of modelling techniques for doing so. We examine the suitability of four different modelling techniques for predicting deforestation; Bayesian networks (BNs), artificial neural networks (ANNs), Gaussian processes (GPs), and generalised linear mixed models (GLMMs). The analysis is provided in the context of the Verified Carbon Standard Approved Methodology for Avoided Unplanned Deforestation to illustrate specific examples of scenarios where each technique would be suitable. ANNs potentially improve predictions over GLMMs, while GPs were better able to generalise to other geographic areas. ANNs were sensitive to design decisions, which may limit their applications where results are required within short time frames. BNs easily facilitate scenario analysis, but are also sensitive to design decisions. With machine learning becoming easier each day, we encourage researchers to consider carefully which technique is suitable for their situation prior to beginning implementation.

BACKGROUND: Leptospirosis is a globally important zoonotic disease, with complex exposure pathways that depend on interactions between human beings, animals, and the environment. Major drivers of outbreaks include flooding, urbanisation, poverty, and agricultural intensification. The intensity of these drivers and their relative importance vary between geographical areas; however, non-spatial regression methods are incapable of capturing the spatial variations. This study aimed to explore the use of geographically weighted logistic regression (GWLR) to provide insights into the ecoepidemiology of human leptospirosis in Fiji. METHODS: We obtained field data from a cross-sectional community survey done in 2013 in the three main islands of Fiji. A blood sample obtained from each participant (aged 1-90 years) was tested for anti-Leptospira antibodies and household locations were recorded using GPS receivers. We used GWLR to quantify the spatial variation in the relative importance of five environmental and sociodemographic covariates (cattle density, distance to river, poverty rate, residential setting [urban or rural], and maximum rainfall in the wettest month) on leptospirosis transmission in Fiji. We developed two models, one using GWLR and one with standard logistic regression; for each model, the dependent variable was the presence or absence of anti-Leptospira antibodies. GWLR results were compared with results obtained with standard logistic regression, and used to produce a predictive risk map and maps showing the spatial variation in odds ratios (OR) for each covariate. FINDINGS: The dataset contained location information for 2046 participants from 1922 households representing 81 communities. The Aikaike information criterion value of the GWLR model was 1935·2 compared with 1254·2 for the standard logistic regression model, indicating that the GWLR model was more efficient. Both models produced similar OR for the covariates, but GWLR also detected spatial variation in the effect of each covariate. Maximum rainfall had the least variation across space (median OR 1·30, IQR 1·27-1·35), and distance to river varied the most (1·45, 1·35-2·05). The predictive risk map indicated that the highest risk was in the interior of Viti Levu, and the agricultural region and southern end of Vanua Levu. INTERPRETATION: GWLR provided a valuable method for modelling spatial heterogeneity of covariates for leptospirosis infection and their relative importance over space. Results of GWLR could be used to inform more place-specific interventions, particularly for diseases with strong environmental or sociodemographic drivers of transmission. FUNDING: WHO, Australian National Health & Medical Research Council, University of Queensland, UK Medical Research Council, Chadwick Trust.

There are many reasons why we study deforestation including predicting at risk areas, predicting deforestation rate and informing the development of conservation policies and programs. Each study will have its own set of objectives to meet (such as setting a deforestation baseline or advising on forest protection policies) and constraints to work within (such as time and data constraints and access to experts). This thesis develops a framework for helping to decide which of several statistical and machine learning methodologies; generalised linear models (GLMs), generalised linear mixed models (GLMMs), artificial neural networks (ANNs), Bayesian networks (BNs) and Gaussian processes (GPs) might be suitable for a given deforestation study. One common constraint on deforestation studies is data availability, as it is often not possible to acquire all the datasets that would ideally be included. High resolution demographic or socio-economic information can be costly, and obtaining the value of dynamic variables such as road location for the correct point in time may be difficult. By using either freely available or low cost datasets to generate the variables for this thesis, it was possible to evaluate the usefulness of these data in predicting deforestation and identifying its predisposing factors. Their proven utility demonstrates that they could provide effective substitutes in those cases where the ideal datasets are not available. The main datasets used were the Conservation International land use change data for southern Mexico and north-eastern Madagascar, which are raster datasets at 30m resolution showing forest loss for either two (Mexico) or three (Madagascar) time steps. Predictor variables were also generated from the World Database on Protected Areas, the NASA Landsat digital elevation model and several Natural Earth datasets on city and river location. Random samples were generated across the forested areas of the study zones and models were then trained to predict whether there would be any deforestation within a 500m x 500m zone around each sample. Models were implemented using either R or Matlab and their performance was evaluated using sensitivity, specificity, true skill statistic and the area under the receiver operating curve. The results of the best performing model designs for each methodology were mapped to examine whether the predicted high risk areas were close to where actual deforestation occurred. Separate maps were produced for the predicted results at a 50% probability cut off, as well as across all probabilities to produce a map of predicted high risk areas. Additional maps were created showing the predictions after correcting for the rate of expected deforestation. When applied to complex problems such as deforestation analysis, machine learning (ML) techniques have several theoretical and practical advantages over classical statistics, primarily the ability to take into account non-linear relationships. The ML models were therefore expected to outperform the simpler statistical versions, however the results showed that this was not always the case. While the GLMMs outperformed ANNs in two of the three study zones, ML techniques did offer improvements with BNs scoring higher on the true skill statistic than GLMMs when trained on standard, rather than stratified data and GPs improving performance when fewer variables are available. Most models showed promising results for predicting the location of high risk areas, although this was dependent in some cases on using stratified data to boost the number of positive deforestation samples that the models could learn from. Outside of predictive performance, other methodology attributes, such as interpretability and ease of implementation can also dictate a model’s suitability in meeting the objectives and constraints of a study. Taking this into account, several recommendations are presented based on the findings in this thesis. Firstly, freely available datasets should be considered as a valuable source of data for deforestation studies. In terms of the application of ML methodologies to deforestation studies, when generalised linear models are used, performance may be improved by modelling the spatial dimension as a random effect (in this case the X and Y coordinates were used). When examining the drivers of deforestation, if p-values are not required, Bayesian networks offer better interpretability than the statistical models with no decrease in predictive performance. When predicting location, the Gaussian processes implemented in this thesis outperformed the artificial neural networks and were easier to design, although each model took longer to run. The recommendations derived from this research go some way towards providing guidance to environmental management practitioners on which ML methodologies can improve on the classical statistical techniques that are frequently employed. This contributes to closing the gap between the disciplines and opening up new tools and datasets that can help with the ongoing challenges facing those attempting to curb deforestation and reverse the trends of environmental degradation associated with it.

The UQ CrocaRoos entered the RoboCup Simulation League for the first time in 2001. The team demonstrated many of the basic functions required to play the simulated game: vision, action and multi-agent planning. The design of the team extends many of the concepts developed for the real robot leagues. This paper describes the fundamental architecture of the team, and the principles for developing the team in the future.