Ever since the threat of an outbreak of anthrax or avian flu entered the national consciousness, state health departments have been working overtime to develop early warning disease-surveillance systems.
But technical and policy gaps remain in the patchwork of systems that send data about disease outbreaks to national repositories, such as the Centers for Disease Control and Prevention’s National Electronic Disease Surveillance System (NEDSS).
A report released in December underscored those gaps. The fifth annual study by the nonprofit Trust for America’s Health, titled “Ready or Not? Protecting the Public’s Health from Disease, Disasters and Bioterrorism,” listed 12 states whose surveillance systems are incompatible with NEDSS. Those states include Alaska, Arizona, California, Connecticut and Minnesota.
Health information technology planners say the problem is not a lack of monitoring capabilities at the local level but a question of how to efficiently package relevant data into easily shared formats. There are also practical and cultural issues, including concerns about privacy and the tension that flares whenever the federal government imposes partially funded requirements on state officials.
“The feds are preparing for [health emergencies], but it’s the locals who, if the deal goes down, will have the spotlight on them,” said David Siegrist, a senior research fellow at the Potomac Institute, a think tank that focuses on technology and national security.
That approach might be appropriate because states own the playing field. “I call it the ‘Harold Hill hypothesis’ from ‘The Music Man,’” Siegrist said. “You’ve got to know the territory. Especially for early detection and how to interpret [data], authority should devolve to the local public health person who actually knows the territory.”
Some states have launched sophisticated surveillance systems that detect unusual health problems or diseases occurring in unexpectedly high volumes. The systems sift through data from labs, hospital emergency rooms, physician practices and pharmacies looking for anomalies.
In Indiana, Health Data Center officials don’t question whether to provide data to the federal government. They routinely share medical complaint and related hospital information with CDC to help coordinate regional response efforts. “We understand not all outbreaks stay within our borders,” said Roland Gamache, the center’s director.
Instead, Indiana health officials ask how they can best share data that arrives on a variety of forms and in many formats.
The state’s Public Health Emergency Surveillance System draws on technology developed at Johns Hopkins University to sort through information from emergency departments and physician offices to determine whether a higher number of people than expected have certain symptoms.
The software accounts for variations based on the day of the week and the season in an attempt to uncover early evidence of a disease outbreak. Epidemiologists then perform an analysis using Health Level 7-
formatted data from physicians’ appointment scheduling systems and from hospital admission, discharge and transfer systems.
But although HL7, as a health care-specific data format, provides a common method for reliably exchanging information, it doesn’t solve all interoperability problems. For example, it works well for symptom and geographic information, Gamache said. But other data, such as lab results, often start out with proprietary codes based on the preferences of individual software vendors.
“We could transfer [the proprietary codes] in HL7, but [the recipients] wouldn’t understand what [they] were getting,” he said.
Instead, the state translates the codes into universal formats, such as Logical Observation Identifiers Names and Codes, a standard for representing clinical results. Unfortunately, the initial task of translating lab codes to LOINC formats largely relies on time-consuming human interpretation rather than automated systems, he added.
Similarly, the symptom information Indiana receives from hospital emergency departments originates as text based on the discussions patients had with their doctors. Formal, easily coded classifications come much later in the diagnostic process.
“You have to wait for the lab results and everything else to come back to confirm” the complaint, Gamache said. “We are trying to get something that happens a lot faster, so we are taking the text information and trying to put that into a symptom-surveillance system.”
Questions also arise about whether to notify local or federal health authorities first when an outbreak occurs. Gamache advocates for the traditional approach of alerting local authorities first, then the national ones.
Local preference “makes sense because of the relationships that have been built for years,” he said. “We are in the same community, so we know who to call.”
Indiana officials also struggle with what level of detail to share with federal agencies. “What information is really needed at a national level in order for them to do national surveillance?” Gamache asked. “The idea is to share enough granularity in the data to do a thorough analysis but not share so much that you put an individual’s data at risk. Any time you give additional information, you increase that risk.”
However, systems that analyze and share data are starting to show concrete benefits, said Dr. J. Marc Overhage, director of medical informatics at the Regenstrief Institute, a health research organization in Indiana.
“When we were first starting this work in around 2000, we noticed a couple more cases of Shigella,” he said, referring to a group of bacteria that cause shigellosis, a disease characterized by diarrhea, fever and stomach cramps. “It became a large outbreak in the community, affecting over 500 children, closing day care centers, making parents take time off work. I was just amazed at what outbreaks like that ended up costing the community.”
Because the state was in the early stages of its disease-surveillance system, it could do little but watch the outbreak grow. Five years later, however, with a more fully developed system, Indiana’s health department was able to stem a wider outbreak after only five more people than normal showed up in emergency rooms with Shigella infections.
“The state epidemiologists noticed this [rise], then the Department of Health did its usual good shoe-leather epidemiology,” said Overhage, who is also president and chief executive officer of the Indiana Health Information Exchange. “It followed up with these people who were showing up in the emergency department and found out that there was a grocery store that had poor food-handling practices. It addressed those practices within 24 hours of the trend being noticed.”
With data comes responsibility
But efficiency can backfire. “If you increase the number of cases that you identify through electronic surveillance versus traditional spontaneous reporting, where are the resources going to come fr