Data Sharing

The North Carolina Division of Public Health (NC DPH) has been collecting emergency department data in collaboration with the Carolina Center for Health Informatics in the UNC Department of Emergency Medicine since 1999. As of August 2011, there are 113 of 115 emergency departments sending data electronically at least once daily to NC DETECT. Data elements include disposition, initial vital signs, up to 11 ICD-9-CM final diagnosis codes, up to five external cause of injury codes (E-codes),as well as the arrival date and time, patient sex and age, patient zip and county, and chief complaint. As of January 2008, NC DETECT emergency department data covered 99% of the NC population and captures approximately 4.5 million ED visits each year. As a result, requests for data from researchers continue to increase. Use of the data for public health purposes is covered by the mandate requiring hospitals to submit their emergency department data to NC DPH.

Objective

To describe the process by which researchers request access to data sets of emergency department data from NC DETECT,the history of this process,and the resulting best practices and lessons learned.

Public health surveillance relies on multiple systems and methodologies for data collection, analysis, and interpretation. Each component provides only part of the picture, such as detection of possible outbreaks or events of concern; geographic profiles or time courses of disease activity; or indicators of clinical severity by age, risk factors, etc. Novel, unstructured data sources like Twitter feeds and aggregated news reports are growing as a source of information about health and disease. What and where are the contributions of these nontraditional, often non-specific, data types to BSV? The answer will depend on the purpose and target population. Different data streams often have greater utility for one BSV function (e.g., outbreak detection) than another (e.g., situation awareness). Furthermore, public health agencies at different levels need and use data differently, as determined by their priorities for public health. New types of data can also be useful for disease prediction and forecasting, pandemic modeling, and developing analytic tools. Before any new data modality can be integrated into standards of surveillance practice, it needs to be evaluated for its contribution to understanding disease activity and the value added when compared to other sources of data with regard to validity, timeliness, accuracy, representativeness, and positive and negative predictive values. Furthermore, questions remain about when novel, unstructured, or nontraditional data sources are acceptable evidence to inform decision-making and public health actions. To address this, the strengths and weaknesses of different types of data for various surveillance functions need to be discussed among stakeholders that bring various perspectives from surveillance research, practice, and policy.

Objective

To gather thought leaders in informatics, public health practice, surveillance research, and strategic decision-making to provide their insights into where and how to effectively integrate novel data streams, such as social media, into biosurveillance (BSV) systems and standards of public health surveillance practice.

Collaborative relationships between academicians and public health practitioners are necessary to ensure that methodologies created in the research setting translate into practice. One barrier to forging these collaborations is restrictions on the sharing and availability of public health surveillance data; therefore, most academics with expertise in method development cannot access 'real world' surveillance data with which to evaluate their approaches. The ISDS Technical Conventions Committee was established in 2013 to facilitate and expedite the development, evaluation, and implementation of technical methods for public health surveillance. The purpose of the committee is to bridge a long-standing gap between technical challenges in public health practice and solution developers needing both understanding of these challenges and representative data.

Objective

The purpose of this panel is to facilitate the dissemination of surveillance-related use cases by public health practitioners with accompanying benchmark datasets to method developers. The panel will present practitioners' experiences with preparing patient-level emergency department data sets to accompany a use case submitted to the ISDS Technical Conventions Committee.

The aerosol release of a pathogen during a bioterrorist incident may not always be caught on environmental sensors - it may be too small, may consist of a preparation that is coarse and heavy (and consequently precipitates quickly) or may simply have occurred in an uninstrumented location. In such a case, the first intimation of an event is the first definitive diagnosis of a patient. Being able to infer the size of the attack, its time, and the dose received has important ramifications for planning a response. Estimates drawn from such a short observation period will have limited accuracy, and hence establishing confidence levels (i.e., error bounds) on these estimates is an major concern. These estimates of outbreak characteristics can be also be used as initial conditions for epidemic models to predict the evolution of disease (along with error bounds in the predictions), in particular, communicable diseases in which the contagious period starts soon after infection (e.g., plague).

In this paper, we will consider anthrax and smallpox as our model pathogens. Since the contagious period of smallpox usually starts after the long incubation period (7–17 days), and the early epoch will consist only of index cases, we will model it as a non-contagious disease. Inputs will be obtained from simulated outbreaks as well as from the Sverdlovsk anthrax outbreak of 1979.

Objective

This paper presents a method that infers the number of infected people, the time of infection and the dose received from an aerosol release of a pathogen during a bioterrorism incident. Inputs into the inference process are the number of new diagnosed patients showing symptoms each day as observed over a short duration (3–4 days) during the early epoch of the outbreak.

Developing effective data-driven algorithms and visualizations for disease surveillance hinges on the ability to provide application developers with realistic data. However, the sensitivity of the data creates a barrier to its distribution. We have created a tool that assists data providers with de-identifying their data in preparation for sharing. The functions in the tool help data providers comply with the HIPAA 'Safe Harbor' de-identification standard by removing or obscuring information such as names, geographic locations, and identifying numbers.

Objective

To develop a robust, flexible, and easy-to-use data de-identification tool that makes it easier for data providers to create data sets that are sharable with external collaborators.