On Watch
Written by Peter Buxbaum
MMT 2010 Volume: 14 Issue: 6 (September)
Biosurveillance is a multifaceted enterprise undertaken for multiple purposes by several United States government departments and agencies. At the national level, agencies such as the Centers for Disease Control seek to monitor potential outbreaks of diseases such as pandemic influenza in order to inform and protect the U.S. population as a whole. A more recent wrinkle to the national mission has been scouting for possible terror-related diseases such as anthrax.
The Department of Defense has its own biosurveillance programs that are related both to the national mission as well as to a more narrow departmental focus. DoD clearly has an interest in protecting the health of its millions of uniformed personnel, civilian employees and their dependents for several reasons, military readiness, efficiency and cost controls among them. Because defense employees and families are so numerous and are spread out across the country, DoD data also provides a window on public health generally.
ESSENCE
DoD’s Military Health System operates a surveillance system that monitors medical records at some 450 non-combat military treatment facilities worldwide. The system, known as the Electronic Surveillance System for Early Notification of Communitybased Epidemics (ESSENCE), was developed by DoD in conjunction with the Johns Hopkins Advanced Physics Laboratory and applies software algorithms to detect disease outbreaks. ESSENCE is a Web-based application that monitors and provides alerting for rapid or unusual increases in the occurrence of infectious diseases and biological outbreaks.
ESSENCE works by capturing the numeric diagnostic codes entered into DoD electronic health records. The software automatically collates that data among 10 categories of syndromes so ESSENCE users can look out for outbreaks of diseases like pneumonia, influenza and food-borne illnesses.
The system calculates expected rates of infectious disease outbreaks and provides reporting, analysis and alerting functions. The users—several hundred epidemiologists and public health officers within each of the armed services—access ESSENCE’s databank to obtain medical situational awareness and investigate disease events.
“If the focus is on influenza, for example, the system will identify the number of cases in a specific location that showed up the day before and going back four weeks,” explained Dr. Ken Cox, director of the Department of Defense/Department of Veterans Affairs Program Coordination Office. “Then it looks at those data and applies a statistical approach to determine whether the number of cases is more than would otherwise be expected and assigns a statistical value to it.”
If the system comes up with a Level Red designation, it has determined the likelihood of an outbreak of the malady in question. But DoD does not rely on statistics alone to assess medical situations. “The statistics are always verified at the local level to determine whether it is a true outbreak or not,” said Cox.
There are several technologies that are integrated in ESSENCE. Informatica provides an extract, transform and load (ETL) package that makes sure that data from various sources loaded into the ESSENCE databases are clean and in the proper format.
ESSENCE runs three separate data bases so that different processes—ETL, detection, and querying and recording— can each work independently and simultaneously. The Unix databases are managed with the IBM DB2 software environment.
Cognos provides business intelligence tools for querying and reporting, ESRI provides a geospatial component to the analysis, while Johns Hopkins provides its proprietary algorithms for processing of data and detection of patterns. ESSENCE is managed and maintained by MHS within an environment hosted by the Defense Information Systems Agency.
Gathering data from multiple sources for central processing, as ESSENCE does, presents a number of challenges. The data must be in the proper format and include all fields necessary for further processing. The data must also comply with agency-specific business rules, which, in the case of sharing health data often requires de-identification of certain fields or records.
These are the types of issues are handled by Informatica’s ETL tools. “We help bring together lots of different data sources to be able to identify particular types of outbreaks,” said Mike Applebaum, general manager of the company’s complex event processing unit. “The Informatica technology handles everything from identity resolution for name matching all the way to data quality, making sure the data we are processing actually is what we expect it to be.”
An Informatica tool called B2B is able to extract, transform and standardize unstructured data contained in document, spreadsheet and messaging files and combine them with the structured data extracted from electronic medical records. Another Informatica tool, PowerCenter, is able to monitor databases to extract new information automatically.
The geospatial tools provided by ESRI help analysts detect patterns from among thousands, or even millions, of seemingly random points of data. ESRI’s ArcGIS product puts latitude and longitude data on top of geographically referenced information, such as a patient’s home address. This data can be displayed in tabular format to facilitate statistical analysis or graphically, by displaying the data on map.
“Both the tables and the maps can help statisticians determine where concentrations of symptoms have occurred,” said Bill Davenhall, the company’s global marketing manager for health and human service solutions. “You always have to apply statistical analysis to determine whether the data show cases that would exceed the number expected in a given geographic area. Without the geographical information you don’t know whether you are seeing an outbreak or something random.”
Four Versions Are Planned
There are three versions of ESSENCE that are working or in the works at this moment, while yet a fourth is being planned. Block One of ESSENCE is what is currently running. Block Two, which will be introduced later this year, will be detecting new syndromes in addition to the 10 currently tracked by ESSENCE. The new version will be tracking evidence of flu-like illnesses in addition to influenza itself by tracking data such as the prescribing of anti-viral medications. Block Two will also expand ESSENCE beyond its view of diagnostic codes to include laboratory orders and the prescription of selected medications.
“This is not a complacent team,” said Robert Bell, director of health surveillance for the Defense Health Services Systems. “There is a comprehensive list of future improvements for which we will be asking increased funding.”
Block Three, which is currently in development, will add an analysis of the patient’s chief complaint into the mix, a task that will present some challenges, according to Cox. “This is not all captured as computable data,” he said. “We will be using text parsers and natural language processors to look for key words and match those up with our categories.” Block Three will also move ESSENCE from ESRI’s ArcGIS desktop application to ArcGIS Server, an enterprise application.
The move to the server application will provide noticeable improvements to ESSENCE’s performance, according to Davenhall. “Up to now the data has been run in batch and has been near real time,” he explained. “The server version will allow data to be immediately geo-coded in real time. There is going to be a big jump in the quality and timeliness of the data.”
A projected Block Four will analyze laboratory results to determine whether they are normal or abnormal. “That will lead us to where we are heading,” said Cox, “the development of a fused detector algorithm instead of applying algorithms to separate data. We’ll be able to sort data and post it on a dashboard which will reduce what users have to do with it.”
Future iterations of ESSENCE could benefit from a new tool from Informatica called complex event processing. “Complex event processing makes automated correlations of data based on user defined rules,” said Applebaum. DoD already uses complex event processing for maritime domain awareness and other applications.
In the case of biosurveillance, complex event processing could be used to write a rule for the persistent monitoring of incoming data form locations where disease outbreaks are suspected “without always having to eyeball the data,” said Applebaum. “Then automatic alerts can be generated” if patterns of interest are detected.
DoD has an ongoing and growing collaboration with the Department of Veterans Affairs on biosurveillance. Although originally developed by DoD, VA, through its Office of Public Health Surveillance and Research (OPHSR), has also implemented its own version of ESSENCE at 153 VA medical facilities, largely in an effort to protect its own population of beneficiaries.
Recently, DoD and VA have joined forces to demonstrate ESSENCE’S ability to perform medical surveillance on a combined beneficiary population. A pilot project was established at the joint DoD/ VA North Chicago medical facility and there are plans to roll the collaboration out to other areas with large veteran and military populations, such as San Diego and Norfolk, and eventually, nationally. The project includes the eventual development of a DoD/VHA biosurveillance backup data repository so the departments can serve as disaster recovery sites for each other. The dual-departmental Health Executive Council approved funding of $4.8 million for the two-year project.
“The pilot work in North Chicago showed that combining data in one ESSENCE platform makes biosurveillance more robust,” said Dr. Mark Holodniy, the OPHSR director. “Our patient population tends to be older and sicker and overwhelmingly male,” he explained. “Our average patient is in his 50s. DoD data includes younger dependents, bringing the average age of the population we are looking at to the 30s. It also brings in 20 to 25 percent women. That gives us a better picture of what is going on in the community.”
Instances of the H1N1 pandemic flu virus, for example, have occurred much more heavily among young children and young adults than among older folks. “It would be more difficult to pick up an outbreak by just looking at VA populations,” said Holodniy.
The joint project is being rolled out to harvest the “lowest hanging fruit” first, said Holodniy. “Next year, we will go on to create a system that will have a bidirectional linkage among all VA and DoD facilities stateside.”
That development will be stepping stone, he added, to an “envisioned national biosurveillance system. ♦