Objective This presentation discusses the problem of detecting small-scale events in biosurveillance data that are relatively sparse in the sense that the median count of monitored time series values is zero. Research goals are to understand conditions when methods adapted for sparseness are warranted, to examine adaptations of control charts and other algorithms under these scenarios, and to compare the detection performance of these algorithms.
To determine the predictive abilities of systematic emergency room chief complaints for the detection of acute infectious diseases in the Winnipeg population.
To describe and evaluate the SOS Medecins Bordeaux syndromic surveillance system .
The North Carolina Disease Event Tracking and Epidemiologic Collection Tool (NC DETECT) receives a designated set of data elements electronically available from 110 emergency departments (EDs) (98%) on at least a daily basis via a third party data aggregator. While automated processes monitor for data quality problems such as improper file formats or missing required elements, data corruption can occur at several stages before receipt, and if undetected, data can appear reliable. Hospitals might map to standard codes incorrectly, data aggregators might manipulate text improperly, or updates might be confused with original records. These inaccuracies cause delays and oversights in identifying events of public health importance.
Objective
This study evaluates the validity of a subset of ED data collected in NC DETECT, as well as measures the effectiveness of the data quality processes in place for this surveillance system.
Research evaluating the use of spatial data for surveillance purposes is ongoing and evolving. As spatial methods evolve, it is important to characterize their effectiveness in real-world settings. Assessing the performance of surveillance systems has been difficult because there has been a paucity of data from real bioterrorism events. Recent efforts to assess surveillance system performance have focused on injecting synthetic outbreak data (signal) into actual background visit data. These studies focused on either temporal data, a single syndrome category, or a single bioterrorism agent. We are unaware of prior studies evaluating the performance of spatial outbreak detection for multiple syndrome categories in an operational surveillance system.
Objective
To characterize the performance of a spatial scan statistic, we used SaTScan to measure the sensitivity and positive predictive value for detecting simulated outbreaks having varying size, case density, and syndrome type.
In November of 2001 a syndromic surveillance system was established in Los Angeles (LA) County to analyze emergency department (ED) chief complaints in select hospitals. Chief complaints were analyzed and categorized into a syndrome (rash, respiratory, neurological, gastrointestinal), and an algorithm was developed to create a daily threshold for each category. Questions remain as to what events can be detected by the system in a timely manner. On the community level, of interest is whether an outbreak with a wide epidemiological curve would have the intensity of case visits needed to trigger a signal. On the individual level, of interest is the length of time it takes for a person with a given disease characteristic to seek medical attention, whether medical care is sought in the ED first, and how the syndromic system classifies them upon visiting the ED. To address these questions the 2004 LA County West Nile community-wide outbreak was selected for review, with a focus on the more severe neuro-invasive cases.
Objective
To evaluate the effectiveness of monitoring emergency room chief complaints as an indicator for a neuro-invasive disease outbreak.
The abattoir and the fallen stock surveys constitute the active surveillance component aimed at improving the detection of scrapie across the EU. Previous studies have suggested the occurrence of significant differences in the operation of the surveys throughout Europe. Del Rio Vilas et al assessed the presence of heterogeneity between the observed prevalence estimates of 18 EU countries by means of a meta-analysis and showed a large residual variability indicating an inconsistent approach to the surveys across the EU. The study of these differences merits attention as they inform discrepancies in the performance of the surveys between countries. In the absence of sufficient covariate information to explain the observed variability across countries, we can model, still under the general context of the meta-analysis, the unobserved heterogeneity in our data. Countries could be grouped into clusters representing the underlying subpopulations relative to the risk of scrapie between the two surveys in each country.
Objective
In the present study we assessed the standardisation of the active surveillance of scrapie throughout time across the EU and identified countries with similar underlying characteristics allowing comparisons between them.
The utility of syndromic surveillance systems to augment health departments’ traditional surveillance for naturally occurring disease has not been prospectively evaluated.
Objective
In this interim report we describe the signals detected by a real-time ambulatory care-based syndromic surveillance system and discuss their relationship to true outbreaks of illness.
There are many proposed methods of identifying outbreaks of disease in surveillance data. However, there is little agreement about appropriate ways to choose amongst them. One common basis for comparison is simulating outbreaks and adding the simu lated cases to real data streams (‘injected outbreaks’); competing statistical methods then attempt to detect the outbreak. The receiver operating characteristic (ROC) curve and the area beneath it are well-known approaches to evaluation. The ROC curve plots the sensitivity against 1 less the specificity for a range of decision thresholds. Unfortunately, defining ROC curves in this context is not straightforward. In the usual setting of screening, ROC curves are constructed based on individuals, not populations, and it is unclear how to extend the concept to surveillance. In addition, the sensitivity and specificity need to be supplemented by the timeliness: a method with perfect sensitivity and specificity that detects outbreaks too late is useless.
Objective
We developed metrics for evaluating tools used for outbreak detection, assuming simulated outbreaks.
In previous work, we described a non-disease-specific outbreak simulator for the evaluation of outbreak detection algorithms. This Template-Driven Simulator generates disease patterns using user-defined template functions. Estimation of a template function from real outbreak data would enable researchers to repetitively simulate outbreaks that resemble a single real outbreak. These simulated outbreaks can then be used to evaluate outbreak detection algorithms. To demonstrate template estimation, we employ BARD, a disease-specific outbreak model for outdoor aerosol release of B. anthracis. It uses epidemiological and atmospheric dispersion models in conjunction with geographical and meteorological data to generate anthrax cases. The home census block group and time of visit to an emergency department are available for each simulated case.
Objective
In previous work, we developed a Template-Driven Simulator, which is a non-disease specific outbreak simulator that uses templates to describe the temporal or spatial-temporal pattern of an outbreak. Here we address the problem of estimating the template from outbreak data. We then conduct a limited validation of the outbreak simulation model by estimating the template using outbreak data generated from BARD, a sophisticated state-of-the-art anthrax outbreak simulator and detector. This limited validation confirms that the outbreak simulator is capable of generating complicated disease outbreak patterns for evaluating outbreak detection algorithms.