ISDS Conference

Since October 2004, the Indiana State Health Department and the Marion County Health Department have been developing and using a syndromic surveillance system based on emergency department admission data. The system currently receives standards-based HL7 emergency department visit data, including free-text chief complaints from 72 hospitals throughout the state. Fourteen of these hospitals are in Marion County, which serves the Indianapolis metropolitan region (population 865,000).

Objective

This paper describes how a syndromic surveillance system based on emergency department data may be leveraged for other public health uses.

In the spring of 2005, the ISDH began using Electronic Surveillance System for the Early Notification of Community-based Epidemics  (ESSENCE) application to analyze emergency department (ED) chief complaint data for syndromic surveillance purposes.  While granting hospitals and local health departments access to their data through ESSENCE has been desirable since the start of the PHESS project, an aggressive timeline made it necessary to direct all resource capacity toward first establishing hospital ED data connections.  The Marion County Health Department (Indianapolis) was the only LHD in the state with access to its 14 hospitals through ESSENCE.

However, because hospitals and local health departments (except Marion County) did not have access to their data through ESSENCE, any syndromic alert follow-up conducted by the ISDH was accomplished primarily by telephone.   This method, while feasible, was inefficient.  The ISDH felt that alert data follow-up could be greatly facilitated if hospitals and LHDs could view these data through ESSENCE just as the ISDH was doing.

Objective

This paper describes how the Indiana State Department of Health (ISDH) improved response capability by increasing local health department (LHD) and hospital access to syndromic surveillance data as part of the stateís evolving Public Health Emergency Surveillance System (PHESS).

The North Carolina Bioterrorism and Emerging Infection Prevention System (NC BEIPS) serves public health users across North Carolina at the local, regional and state levels, providing syndromic surveillance capabilities.  At the state level, our primary users are in the General Communicable Disease Control Branch of the NC Division of Public Health.  NC BEIPS currently receives daily data from the North Carolina Emergency Department Database (NCEDD), Carolina Poison Control Center (CPC), Prehospital Medical Information System (PreMIS) and the Piedmont Wildlife Center (PWC). Future data sources will include the North Carolina State University College of Veterinary Medicine Laboratories.  The PWC is a non-profit organization dedicated to wildlife rehabilitation, education, and scientific study of health and disease in wildlife populations.  PWC admits approximately 3,000 animals annually, including mammals, birds, and reptiles, the majority of which are from 21 counties in central North Carolina.  

Objective

This poster will illustrate how a novel data source, wildlife health center data, is being incorporated and used in a syndromic surveillance system.

Current state-of-the-art outbreak detection methods [1-3] combine spatial, temporal, and other covariate information from multiple data streams to detect emerging clusters of disease.  However, these approaches use fixed methods and models for analysis, and cannot improve their performance over time.   Here we consider two methods for overcoming this limitation, learning a prior over outbreak regions and learning outbreak models from user feedback, using the recently proposed multivariate Bayesian scan statistic (MBSS) framework [1]. Given a set of outbreak types {Ok}, set of space-time regions S, and the multivariate dataset D, MBSS computes the posterior probability Pr(H1(S, Ok) | D) of each outbreak type in each region, using Bayes’ Theorem to combine the prior probabilities Pr(H1(S, Ok)) and the data likelihoods Pr(D | H1(S, Ok)). Each outbreak type can have a different prior distribution over regions, as well as a different model for its effects on the multiple streams.  The set of outbreak types, as well as the region priors and outbreak models for each type, can be learned incrementally from labeled data or user feedback.

Objective

We argue that the incorporation of machine learning algorithms is a natural next step in the evolution and improvement of disease surveillance systems. We consider how learning can be incorporated into one recently proposed multivariate detection method, and demonstrate that learning can enable systems to substantially improve detection performance over time.

In epidemiology, contact tracing is a process to control the spread of an infectious disease and identify individuals who were previously exposed to patients with the disease. After the emergence of AIDS, SNA was demonstrated to be a good supplementary tool for contact tracing [1]. Traditionally, social networks for disease investigation are constructed only with personal contacts since personal contacts are the most identifiable paths for disease transmission. However, for diseases which transmit not only through personal contacts, incorporating geographical contacts into SNA has been demonstrated to reveal potential contacts among patients [2][3].

Objective

In this research, we aim to investigate the necessity of incorporating geographical contacts into Social Network Analysis (SNA) for contact tracing in epidemiology and explore the strengths of multi-mode networks with patients and geographical locations in network visualization for disease spread investigation.

Developing and evaluating outbreak detection is challenging for many reasons.  A central difficulty is that the data the detection algorithms are “trained” on are often relatively short historical samples and thus do not represent the full range of possible background scenarios.  Once developed, the same dearth of historical data complicates evaluation.  In systems where only a count of cases is provided, plausible synthetic data are relatively easy to generate.  When precise location data is available, simple approaches to generating hypothetical cases is more difficult.

Advances in epidemiological modeling have allowed for increasingly realistic simulations of infectious disease spread in highly detailed synthetic populations. These agent-based simulations are capable of better representing real-world stochastic disease transmission process and thus show highly variable results even under identical initial conditions. Due to their ability to mimic a wide range outcomes and more fully represent the unknowns in a system, models of this class have become increasingly used to help inform decisions about public policies about hypothetical situations (eg pandemic influenza [1]).  This characteristic also makes them a powerful tool to represent the processes that create surveillance information.

Objective

Developing and evaluating detection algorithms in noisy surveillance data is complicated by a lack of realistic noise, meaning the surveillance data stream when nothing of public health interest is happening. These jobs are even more complex when data on the precise location of cases is available. This paper describes a methodology for plausible generation of such noise using agent-based models of infectious disease transmission based on highly resolved dynamic social networks.

The lack of a standardized vocabulary for recording CC complicates the collection, aggregation, and analysis of CC for any purpose, but especially for real-time surveillance of patterns of illness and injury. The need for a controlled CC vocabulary has been articulated by national groups and a plan proposed for developing such a vocabulary. To date there has been no comparison of published CC lists.  This study lays the groundwork for a controlled ED CC vocabulary by comparing selected terms from several published ED CC lists.

Objective

The purpose of this study was to compare the most common chief complaints (CC) from a national emergency department (ED) survey, with four published CC lists in order to identify issues relevant to the creation of a controlled ED CC vocabulary.

Text-based syndrome case definitions published by the Center for Disease Control (CDC)1 form the basis for the syndrome queries used by the North Carolina Disease Event Tracking and Epidemiologic Collection Tool (NC DETECT). Keywords within these case definitions were identified by public health epidemiologists for use as search terms with the goal of capturing symptom complexes from free-text chief complaint and triage note data for the purpose of early event detection and situational awareness. Initial attempts at developing SQL queries incorporating these search terms resulted in the return of many unwanted records due to the inability to control for certain terms imbedded within unrelated free text strings. For example, a query containing the search term “h/a”, a common abbreviation for headache, also returns false positives such as “cough/asthma”, “skin rash/allergic reaction” or “psych/anxiety”.  Simple abbreviations without punctuation, such as “ha”, were even more problematic.  Global wildcards ('%') indicate that zero or more characters of any type may substitute for the wildcard.2 The term “ha” as a synonym for "headache" appears frequently in the data, but searching this term bracketed by global wildcards returns any instance where the two letters appear together (e.g. pharyngitis, hand, hallucinations, toothache). Using global wild cards to search for common symptoms such as headache using simple abbreviations, with or without specialized punctuation, results in the return of many unwanted false positive records. We describe here the advanced application of SQL character set wildcards to address this problem.

Objective

This paper describes a novel approach to the construction of syndrome queries written in Structured Query Language (SQL). Through the advanced application of character set wildcards, we are able to increase the number of valid records identified by our queries while simultaneously decreasing the number of false positives.

In order to assess the use of rabies post-exposure prophylaxis in Indiana, the Communicable Disease Reporting Rule, adopted October 11, 2000, requires the reporting of rabies PEP administration.

Indiana is a “home rule” state; that is to say, local (county) health departments (LHD) are responsible for health issues within their jurisdiction.  Reportable diseases are passively reported to the ISDH through local health department by hospitals and physicians.   Often this results in under-reporting of things such as rabies PEP.

While the primary purpose of the PHESS is to enable early detection of acts of bioterrorism, naturally occurring outbreaks, and as a situational awareness tool, PHESS staff have continually worked to find other practical public health applications for the syndromic data.  The Epidemiology Resource Center at the ISDH houses subject matter experts in many areas of public health, including veterinary epidemiology.  Until fall of 2006, the veterinary epidemiologist received all reports of rabies PEP via hard copy.

Objective

The purpose of this paper is to describe how the Indiana State Department of Health (ISDH) leverages syndromic surveillance data to improve statewide rabies post-exposure prophylaxis (PEP) reporting by hospitals. The Public Health Emergency Surveillance System (PHESS) is Indianaís syndromic surveillance system and resides at the ISDH.

Emergency Department (ED) triage notes are clinical notes that expand upon the chief complaint, and are included in the AHIC minimum dataset for biosurveillance.1  Clinical notes can improve the accuracy of keyword-based syndromes but require processing that addresses negated terms.2,3  The North Carolina Disease Event Tracking and Epidemiologic Collection Tool (NC DETECT) syndrome classifier searches for keywords in free-text chief complaint and triage note data for the purpose of early event detection. Initial attempts to handle negation were included in the syndrome queries beginning in August 2005.  Query statements were written to identify and ignore select symptoms immediately following negated terms, such as denies fvr or no h/a.  Many  negated terms, however, were not addressed and continue to create false positive syndrome hits.  The purpose of this pilot was to address negation with NegEx (a negation tool)4, supplemented by selected modules from the Emergency Medical Text Processor (EMTP), a chief complaint pre-processor. 

Objective

The objective of this pilot study was to explore methods for addressing negation in triage notes.