When the nation began hearing news about the influenza outbreak last fall, it came as no surprise to Ed Carubis. As CIO and associate commissioner of the New York City Department of Health and Mental Hygiene (DHMH), Carubis had access to an IT system that warned him and his staff weeks prior to the outbreak.
"We predicted the flu outbreak two weeks in advance of any traditional surveillance methods," said Carubis. "That two-week jump was very important -- it allowed us to notify the medical provider community and allowed them to take action to encourage those at risk to get a flu vaccine."
Carubis and his staff use a combination of information systems to keep them abreast of public health concerns throughout New York City. Chief among these is a syndromic surveillance system designed to provide early detection of disease clusters and serve as an early warning system for terrorism events involving biological, chemical or radiological agents. "During outbreaks, time is of the essence in the ability to control, mitigate or respond to an event," said Carubis.
Historically city disease and illness surveillance efforts relied on patient visits to doctors' offices and lab tests, but that often meant days or weeks before a pattern was detected. "Traditional public health reporting has a built-in delay between when a person first gets sick, and that diagnosis and reporting," said Rick Heffernan, director of data analysis in the bureau of communicable disease at the DHMH. "Syndromic surveillance is trying to move further upstream to capture some of the data collected early in a person's illness."
New York's syndromic surveillance system uses a wide variety of data sources -- 911 calls, emergency room data, pharmacy sales data -- to rapidly detect increased reports that may indicate an outbreak. The system tracks the reports both geographically and by number of cases.
In 1999, the DHMH began examining 911 call data. "We believed there were certain 911 calls that could be indicative of an event that's occurring, from a food-borne illness all the way up to a bio-terrorism event," said Carubis. "We were receiving 911 call data from the fire department routinely anyway, so it was an easy way to start."
After Sept. 11, the DHMH looked at expanding the system, and started working with hospitals that collect data electronically and tried convincing them to share the information. "We didn't want to make any additional work for the individuals in the emergency rooms," Carubis said. "So it's very much been an IT-to-IT discussion of how to exchange information between systems."
Complaint data from patients at 40 different hospitals in the city now comes to the DHMH automatically through the syndromic surveillance system, which uses sophisticated data-mining software to collect and analyze the information.
Next, the DHMH worked with major pharmacy chains to gather certain over-the-counter drug data. The department wanted to survey data on citizens who didn't feel bad enough to call 911 or go to the emergency room, but were feeling poor enough to visit the pharmacy.
Most recently, the DHMH worked with major New York City employers that operate electronic attendance systems, hoping to detect a large rise in absenteeism that could indicate a disease outbreak in the city.
Mining the Data
Once all data is collected from the various sources in different formats, it's merged and transformed into a common format. The DHMH then runs an analysis on the data, and the system alerts them to conditions that may indicate an outbreak. "We send the results of those analyses to a group of epidemiologists who decide whether or not it warrants a response," said Heffernan.
The epidemiologists form hypotheses about what may be occurring and determine the information needed to confirm their hypotheses. They then conduct interviews with patients to determine exactly what they are dealing with.
The DHMH uses a health alert network (HAN) in conjunction with the syndromic surveillance system when a possible outbreak is detected. The HAN provides rapid notification, specific instructions and additional information to the health provider community via e-mail and fax. The only downfall is, because HAN is a subscriber-based system, it's up to the DHMH to get health providers to subscribe to make the best use of the system.
"We currently have 1,000 medical providers that are part of the HAN community," said Dale Rosenberg, HAN coordinator. "Our goal is to eventually get all medical providers in New York City to subscribe -- potentially 40,000 to 100,000 individuals."
The DHMH paid for both systems with federal funding designated for public health preparedness. The department recently received a grant to continue developing the application into an open source Web service, which could be used nationwide.
The syndromic surveillance system came together quickly, according to Carubis. "We basically put something in place between October 2001 and January 2002, but there was a lot of background work before that, and there's been continuous work since."
The DHMH continuously looks to improve and expand its syndromic surveillance system. "There's really no end to a data-mining system -- you're always looking for new ways to refine the model and take advantage of new data sets," said Carubis.
The department is building a contact-tracing component into the system, so it can manage not only individuals who have contracted an illness, but also people with whom they came into contact.
"When you try to do that on paper, it can become unmanageable very rapidly," said Carubis. "Speed is of the essence in managing an outbreak. Once there's repeated exposure, it's very difficult to contain, so information systems are critical."
The DHMH also is evaluating school absenteeism and ambulance transport data as possible sources of additional information. "We're not sure the school absenteeism will be worth adding to the system," Carubis said. "The problem is you know someone is absent but you don't know why, so that really reduces the value of the information."
Whether or not the DHMH expands its data sources, the system already provides them with critical clues to public health concerns. Shortly after the blackout in New York City in fall 2003, the system detected an increase in diarrheal illnesses. The agency immediately deployed public health messages encouraging people to dispose of food that might have spoiled during the blackout.
The system also detected Norovirus-like symptoms before the first cruise ship outbreak occurred, allowing the DHMH to notify medical providers in the city to look for more cases of the fast-spreading virus.
When individuals come to health-care providers with common symptoms, such as fever or cough, it can mean a lot of different things, Carubis said. "It really helps to know what we're seeing outside of the physician's office. Getting a head start by even a day or two in a communicable disease outbreak can significantly curtail the spread of disease."