Poster Session Abstracts

Hepatic Actinomycosis: A case report
Mythily Meda MD, Debapriya De MD, Uday Kanakadandi MD, Harminder Chani MD, Veterans Affairs Hospital, Danville

Purpose: We present a case of a 60-year-old male with actinomycosis in the liver, which is a rare site to be primarily affected with actinomycosis. Primary hepatic actinomycosis is a rare infection that can be confused with hepatic pyogenic abscesses or neoproliferative processes. To date about 75 cases of primary hepatic actinomycosis have been reported in the literature. It is usually cryptogenic and is more common among immunocompetent individuals and males.

Methods: Chart review and review of literature using Medline and bibliographies of published articles.

Results: A 60-year-old male presented with complaints of generalized weakness for six weeks, anorexia and a twelve-pound weight loss. On admission, he had a temperature of 101 F, pallor and mild right upper quadrant tenderness. Past medical history was pertinent for a normal colonoscopy one-month prior. Laboratory studies revealed a normocytic normochromic anemia and mild elevations in AST, ALT and alkaline phosphatase. During hospitalization, he continued to spike daily fevers for over three days despite broad-spectrum antibiotics. An abdominal ultrasound and CT scan of abdomen suggested a complex mass measuring 10 cm x 8 cm in the right lobe of the liver with mild dilatation of intrahepatic biliary tree. These findings were consistent with an abscess or a tumor. Serum alpha-fetoprotein was normal. Subsequent CT-guided aspiration biopsy revealed infiltration of neutrophils and characteristic sulfur granules by light microscopy. Actinomyces was cultured from liver aspirate. The patient was successfully treated with percutaneous catheter drainage and an extended course of intravenous and oral penicillin.

Conclusion: This relatively rare case illustrates the need for high index of suspicion, early diagnosis and appropriate management of actinomycosis as it can be cured with long-term antibiotic therapy.

LAHVA: Linked Animal Human Visual Analytics for Healthcare Surveillance, Management, and Response
Ross Maciejewski, Benjamin Tyner, Yun Jang, Hazem Elmeleegy, Ilya Figotin, Cheng Zheng, David Ebert, William Cleveland, Ahmed Elmagarmid, Mourad Ouzzani, Nita Glickman, Larry Glickman, Shaun Grannis

Coordinated animal-human health monitoring can provide an early warning system with fewer false alarms for naturally occurring disease outbreaks, biological attacks, and chemical attacks. This monitoring requires the integration of varied source data, analyzing said data with tools that are feasible and useful at the scale of the incoming stream, and employing advanced analytic tools that enable monitoring the occurrence levels of indicators, trends, and diagnoses in many categories. We have developed an initial integrated visual analytic tool that provides a unique capability of linking temporally varying geospatial visualization of animal and human patient health information with advanced statistical analysis displays of these multisource data, as well as factor analysis. Our interface provides a factor specification/filtering component to allow exploration of causal factors and spread patterns. For proof of concept development, we are investigating influenza-like illness and gastrointestinal problems using data from the Indiana Network for Patient Care Emergency Departments and "Banfield the Pet Hospital" data for avian, feline, and canine patients.

The MIDAS Research Network
Philip C. Cooley and Diglio A. Simoni, RTI International, Research Triangle Park, NC; Diane K. Wagener, RTI International, Washington, DC

MIDAS is a research partnership between the National Institutes of Health (NIH) and the scientific community whose goal is to help policymakers, public health workers, and researchers make informed decisions about emerging infectious diseases—both man-made and naturally occurring. MIDAS is a collaborative network of scientists. The major aim of their research is to develop mathematical models that represent infectious disease outbreaks. The overall objective of this effort is to provide feasible responses to hypothetical outbreaks that are based on the best available science and data sources that support the science. MIDAS consists of seven research groups and one centralized informatics group. The Informatics group consists of researchers from RTI and IBM.

The research groups focus on models that address research questions that span many aspects of spatiotemporal/biological issues. These include host-pathogen relationships, disease epidemiology, disease surveillance methods and pandemic response strategies. The research groups focus on information-driven research rather than hypothesis-driven investigations.

The informatics group manages computational resources (e.g., Linux clusters and associated data storage systems) and information resources (e.g., information on personal transportation characteristics, demographics, epidemiological information, molecular biology data, and geospatial databases). The informatics group also conducts research on validation, organization, analysis and distribution of tools and models, and implements systems to support MIDAS research.

MIDAS model developers use real or simulated data that is widely available through the MIDAS web site. The informatics group developed and maintains the web site, and manages the information and data resources that modelers may access.

The nature of providing a computational and informational resource for developing national models has demanded significant effort in many areas. These include cluster design, networking, security, software maintenance, system administration, and new methods for facilitating management and streamlining access to scientific data. This is in indicative of the level of commitment that is necessary to support contemporary epidemiological research.

The National Microbial Pathogen Data Resource (NMPDR): A Bioinformatics Platform for Research Support
Leslie McNeil and Claudia Reich, National Center for Supercomputing Applications, University of Illinois at Urbana-Champaign

The National Microbial Pathogen Data Resource (NMPDR) is one of eight Bioinformatics Resource Centers funded by the National Institute of Allergy and Infectious Disease (NIAID) to provide the comprehensive bioinformatics environment needed to support research in biodefense, emerging infectious diseases, and re-emerging pathogens. The NMPDR focus organisms, all NIAID Category B priority pathogens, include the food- and water-borne diarrheagenic bacteria Campylobacter jejuni, Vibrio cholerae, V. parahaemolyticus, V. vulnificus, and Listeria monocytogenes. Also included are the re-emerging, antibiotic-resistant, nosocomial pathogens Staphylococcus aureus, Streptococcus pneumoniae and S. pyogenes (Group A Strep).

NMPDR contains the genomes of nearly 50 strains of these pathogenic bacteria and more than 400 other genomes that provide a broad context for comparative analysis across the three phylogenetic domains. NMPDR integrates complete public genomes with expertly curated biological subsystems to provide the most consistent genome annotations. Subsystems are sets of functional roles related by a biologically meaningful organizing principle, which are built over large collections of genomes; they provide researchers with consistent functional assignments in a biologically structured context. Investigators can browse subsystems and reactions to develop accurate reconstructions of metabolic networks, pathogenicity markers, virulence factors, etc.

Organism summary pages contain a variety of information about the focus pathogens, from textbook descriptions to recently published findings. The organism summary pages are designed to provide user services and a collaborative environment for communities of investigators. A literature aggregator provides access to recent scientific developments. Epidemiological data will be added. NMPDR provides a comprehensive bioinformatics platform, with tools and viewers for genome analysis. Widely used bioinformatics tools and specially designed functionalities are integrated. Results of precomputed gene clustering analyses can be retrieved in tabular or graphic format with one-click tools. NMPDR's unique tools include Signature Genes, which finds the set of genes in common or that differentiates two groups of organisms. Essentiality data collated from genome-wide studies have been curated. Drug target identification and high-throughput, in silico, compound screening are in development.

OpenGeoDa — A Tool for Geovisualization and Spatial Exploration
Luc Anselin, Spatial Analysis Laboratory, Department of Geography, and National Center for Supercomputing Applications, University of Illinois at Urbana-Champaign

OpenGeoDa is the open source and cross-platform successor to the widely adopted GeoDa software package. It is a program for the visualization, exploration and regression analysis of spatial data. It deals with the analysis of information for which the location of the observations is important. The emphasis is on the analysis of discrete geospatial data, such as values associated with points or areas on a map. OpenGeoDa is written in C++ and based on the cross-platform wxWidgets graphical toolkit. It runs identically on Windows, MacOSX and Linux platforms with a native look and feel. This poster contains screen shots to illustrate the application of the methods implemented in OpenGeoDa to infectious disease surveillance. Examples used are the spatial distribution of West Nile Virus across Illinois counties and the incidence of malaria at the municipality level in Colombia.

The OpenGeoDa architecture is based on the powerful concepts of linking and brushing, developed in the dynamic graphics literature. Linking and brushing tie together all the open windows (i.e., all the maps and graphs) in the user interface. Linking is such that an observation selected in any of the windows is simultaneously highlighted in all. The dynamic extension of this, brushing, provides an effective method for interactive exploration of the data. While OpenGeoDa includes a range of standard mapping tools and data exploration methods (such as various types of choropleth maps, histogram, box plot, scatter plot, parallel coordinate plot, 3D scatter plot), the focus of the program is on the analysis of spatial autocorrelation. This functionality allows for the detection and visualization of distinct patterns in the data, such as spatial clusters and spatial outliers. The illustrations highlighted in the poster include the visualization of smoothed disease rates in linked maps and graphs, bivariate map animation, multivariate exploratory data analysis of disease rates and potential correlates, spatial autocorrelation analysis by means of a Moran scatter plot, and cluster analysis based on local indicators of spatial autocorrelation.

Pandemic Influenza Surveillance and Response Systems in Developing Countries: Framework and Pilot Application
SL Lewis and WA Loschen, Johns Hopkins University Applied Physics Laboratory, Laurel, MD; JP Chretien, Department of Defense Global Emerging Infections Surveillance & Response System, Silver Spring, MD; HS Burkom, Johns Hopkins University Applied Physics Laboratory, Laurel, MD; JS Glass, US Naval Medical Research Unit-2, Jakarta, Indonesia; JS Lombardo, Johns Hopkins University Applied Physics Laboratory, Laurel, MD

A pandemic caused by influenza A/H5N1 or another novel strain could kill millions of people and devastate economies worldwide. Recent computer simulations suggest that an emerging influenza pandemic might be contained in Southeast Asia through rapid detection, antiviral distribution, and other interventions. To facilitate containment, the World Health Organization (WHO) has established large, global antiviral stockpiles and called on countries to develop rapid pandemic detection and response protocols. However, developing countries in Southeast Asia would face significant challenges in containing an emerging pandemic. Limited surveillance coverage and diagnostic capabilities; poor communication and transportation infrastructure; and lack of resources to investigate outbreaks could cause critical delays in pandemic recognition. Wealthy countries have committed substantial funds to improve pandemic detection and response in developing countries, but tools to guide system planning, evaluation, and enhancement in such places are lacking.

The Early Warning Outbreak Recognition System (EWORS) is a hospital-based syndromic surveillance system that has been implemented in several Southeast Asian countries. This project is a systems research evaluation with the goal of determining the characteristics of an effective, affordable surveillance/response system for pandemic influenza and other emerging infectious diseases in developing countries. The team will conduct an on-site, end-to-end evaluation of existing EWORS capabilities utilizing CDC's framework as the basis for further expansion. Additional work in modeling / simulation will allow the team to quantitatively study the effects of current and enhanced surveillance/response system configuration on outcomes of interest. At the conclusion of the project, the team will have developed a list of general characteristics for successful surveillance/response systems in developing countries, suggested enhancements to EWORS (including cost/benefit analysis), and a surveillance system evaluation framework expanded for the context of underdeveloped countries.

Precipitation and West Nile virus infection: Implications for disease surveillance and modeling
Marilyn O. Ruiz, William Brown, and Julie Clennon, Department of Pathobiology, University of Illinois at Urbana-Champaign

Since its introduction in 1999, West Nile virus (WNV) has emerged as a major pathogen in the continental U.S. The Chicago area has been especially hard hit by WNV. Cook and DuPage County, in Illinois, are the focus of a three-year research project with the aim to develop a local spatial model for WNV transmission given the climatic conditions of a particular year and the mosquito, avian and urban environment conditions of a particular place. In 2002, human infection rates in Illinois were high, with 884 cases of human illness. In the Chicago region, these cases exhibited a clustered spatial pattern and a strong association with certain housing and environmental characteristics. The years 2003 and 2004 had less activity, but in 2005, Illinois reported 252 human illness cases with most in the Chicago area. This project considers the role of precipitation relative to the WNV mosquito infection rate in 2005 in order to advance the knowledge of the relationship between rainfall and the risk of WNV illness in the urban environment.

Data from 75 weather stations in and around Chicago provided daily rainfall amounts for 2005. Daily rainfall was summarized by week and spatially interpolated across the region. These interpolated maps were then summarized by 59 regions that constitute natural drainage areas, or watersheds. The 8-week rainfall accumulations were also measured for each week. Mosquito pool test results from a variety of institutions that test for WNV were processed to make a comparable weekly dataset of the mosquito infection rate (MIR) summarized for each watershed. A correlation analysis revealed that mosquito infection rates throughout the mosquito season were related to low accumulated rainfall during the spring season. For example, r = -0.633 for Wk 15 accumulated precipitation and Wk 28 MIE. In addition, areas in the north part of the region with higher MIR during later weeks (e.g. Wk 38) had more rain in relatively dry week 31 (r = 0.811) and week 35 (r = 0.761). These complex relationships will be analyzed for other years and compared to other environmental conditions to better predict WNV risk in the future.

Using Prediction Markets to Forecast Influenza
Philip M. Polgreen, MD, Carver College of Medicine, University of Iowa, Iowa City, IA, University of Iowa, Iowa City, IA; Forrest D. Nelson, PhD and George R. Neumann, PhD, Tippie College of Business, Department of Economics, University of Iowa, Iowa City, IA

Background: The extent and timing of local and regional influenza (FLU) epidemics vary yearly. Advance notice of FLU activity could greatly assist preparedness. Diverse information available to individual health care professionals (HCPs, e.g.: physicians, pharmacists, nurses, microbiologists) might predict FLU activity if it could be aggregated in real time. Prediction markets have been used to predict other outcomes (e.g., election results) with more accuracy than statistical forecasting methods and surveys.

Methods: We conducted a pilot FLU prediction market in Iowa from 10/04-4/05. Traders included a diverse mix of Iowa HCPs. Each trader was provided a $100 educational grant with which to trade. The contracts traded were based on the CDC's color-coded system of FLU activity. Each contract represented one color and one week of FLU activity. Traders bought and sold the color contracts depending on their beliefs about FLU activity (up to 8 weeks into the future). Prices can be interpreted as the consensus probability of future FLU activity for a given week. We compared the prediction market results with actual CDC surveillance outcomes for Iowa.

Results: At least three weeks in advance, the market correctly anticipated the future level of influenza activity within one level of activity 90% of the time. In general, as the length of prediction time decreased, the accuracy progressively increased. Regardless of the week-to-week accuracy, our market predicted the epidemic curve: the start, the peak and the end of the season at least two weeks in advance.