AN EARLY WARNING SYSTEM FOR DISEASES IN NEW YORK
04 Apr 2003
Source: New York Times, April 4, 2003
An Early Warning System for Diseases in New York
By RICHARD PEREZ-PENA
In mid-March, a corner of Queens suffered a sudden, sharp increase in the number of people with fever and trouble breathing, turning up mostly at one hospital's emergency room. Where a statistical analysis said there should have been 7 such cases over three days, there were 23. The next day, there were 47.
Suspecting an outbreak of the new Asian respiratory disease, a team of city health workers fanned out, questioning doctors, nurses and patients about symptoms, travels and recoveries, and making plans to isolate them if necessary. It turned out to be a false alarm, a statistical anomaly brought on by conditions as disparate as kidney stones and asthma.
What was remarkable about that episode was not so much the city's response, but that it was detected at all. No one at the hospital noticed it. (City officials would not name the hospital or the neighborhood.)
Rather, the increase was caught by a computer at the New York City Department of Health and Mental Hygiene's headquarters in Lower Manhattan, where a statistical analysis program warned that there was only a 1 in 1,000 chance of such a thing being a random occurrence.
This was the work of the city's Syndromic Surveillance System, which public health experts call the most advanced early warning system for possible disease outbreaks in the country. Amid growing national concern about bioterrorism and the spread of exotic diseases, a number of cities have become interested in this field, and many have called New York, hoping to learn how it is done.
"New York City is clearly at the cutting edge of this, the model for others to follow," said Paula J. Olsiewski, a program director of the Alfred P. Sloan Foundation, a philanthropic group that has financed the development of syndromic surveillance software. "By using the software every day, improving it, making it easier to use, regularly trying new things, New York City is making it possible for other agencies around the country that are not as big or technically sophisticated to learn from their experience."
The nearest analogy may be to Compstat, the Police Department's vaunted crime-fighting computer system, used since the early 1990's to sort mounds of data every day so that the police can respond to problems as soon as they crop up.
The Health Department built its system in fits and starts over several years, but the effort did not really take off until after the World Trade Center and anthrax attacks in 2001. Now, the system, which costs about $1 million a year in labor and computer expenses, analyzes more than 50,000 pieces of information daily, including 911 calls, emergency room visits and drugstore purchases, sifting them by symptom, time and place for patterns that might escape human notice.
"Our system routinely picks up patterns that were never caught before," said Dr. Don Weiss, medical director of the department's communicable disease surveillance unit. "What's sobering is that almost every time we call a hospital and say, `Hey, your area has had a big increase in this or that,' they're completely unaware of it."
Other cities have similar efforts under way, including Boston, Los Angeles and Washington, but none is as well developed as New York City's, experts say. New York State officials are considering building a statewide system, modeled on the city's.
Now, New York City officials who spent years proselytizing about the virtues of syndromic surveillance find themselves warning against inflated expectations.
"This system is probably going to be far more useful in detecting natural outbreaks than man-made ones," said Dr. Farzad Mostashari, assistant commissioner for epidemiology services for the Health Department. "There is no guarantee that it will detect even a modest-sized bioterror attack, or that it will detect that attack before an astute clinician would. This is still in its infancy. It's very much a work in progress."
The anthrax attacks and the West Nile virus illustrate perfectly the limitations, city officials say. Each episode began with a mere handful of cases, most of them nonfatal, involving mostly run-of-the-mill symptoms like fevers. Neither case would have offered enough evidence to register against the background noise of a city of eight million people and their ailments — at least, not until long after some sharp-eyed doctors had figured out what was happening.
Still, the system is a significant departure from traditional epidemiology, which tracks and responds to patterns of disease. Syndromic surveillance focuses not on diseases, but on syndromes — symptoms — that might offer signs of a disease a day or even a week before the disease itself becomes apparent.
About once a week, New York City's system turns up something that officials decide warrants a second look. Most turn out to be nothing, but not all. The system spots the onset of flu season earlier than the traditional reliance on anecdotal reports from doctors, and it detected a surge in sales of nicotine patches immediately after a big increase in cigarette taxes took effect last year.
Last fall, a wave of people sought treatment for vomiting and diarrhea, symptoms that turned out to be early signs of the outbreak of a Norwalk-type virus, the same outbreak that tore through several cruise ships. That gave city officials a few days' head start to warn doctors of the contagion and tell them to take extra care in handling highly contagious body fluids.
Much of the interest in syndromic surveillance dates to 1993, when Milwaukee's water supply was contaminated with cryptosporidium, a microscopic parasite, sickening most of the city's population, sending more than 4,000 people to hospitals, and killing more than 100.
No one understood the magnitude of the outbreak until weeks after it had begun. Afterward, health officials found that there had been several early warning signs: more deaths in nursing homes, more people buying Pepto-Bismol at neighborhood stores, more children absent from school.
If only it were possible, they said, to see that evidence early on, to collect it and analyze it before it was obvious to everyone that there was a problem. But it would be several years before enough records were computerized, and the right hardware and software were available.
In the late 1990's, Dr. Mostashari began trying to gather that kind of information from 911 calls in New York City, which go through the city's computerized central dispatch system. The dispatch system began sending a computer file to the Health Department each day, with an entry on each of the medical calls for the previous 24 hours, an average of 3,000 a day. Each entry notes the patient's age, sex and ZIP code, and lists the caller's chief complaint in one of 52 broad categories, like difficulty breathing or abdominal pain.
When the trade center was destroyed, health officials worried about early detection of health effects, so on Sept. 13, 2001, the department began closely monitoring some major emergency rooms.
"For those first weeks, we actually had people sitting there, taking down notes on every case that came into the E.R.," said Richard Heffernan, director of the department's data analysis unit. "That was completely unsustainable."
But the urgency to collect the information only grew with the anthrax attacks and the fear that symptoms could easily be overlooked as the infection spread. For a while, emergency rooms faxed daily reports to the department, where someone typed them into a computer, but that, too, was untenable.
Eventually, a system was developed in which a hospital sends the department a daily computer file with a short entry on each emergency room patient from the previous day, with more information than in the 911 reports. Participation is up to 40 hospitals, and counting, that report on 7,000 visits daily, about three-quarters of the city's total.
After hospitals, the effort expanded to pharmacies. A large drugstore chain — the city will not say which one — sends daily reports on its sales of certain medicines, about 8,000 prescription entries and 40,000 over-the-counter items. The city is trying to enlist other drugstores.
Recently, the system added reports on how many people call in sick at some city agencies, and officials hope to add other agencies and attendance reports from schools.
At first, the department relied on people to go through all that information, a mind-numbing, time-consuming task. Then, last year, city health officials, statisticians at the University of Connecticut, and the New York Academy of Medicine, with a Sloan Foundation grant, developed a computer program to sort through the material. The program sorts the reports by ZIP codes and by groups of neighboring ZIP codes; it matches medical terms with their synonyms; it accounts for known outbreaks like influenza or allergies; and it adjusts for variations like the fact that fewer people visit emergency rooms on weekends.
"It's come a long way, through trial and error," Mr. Heffernan said. "We're constantly tweaking it, trying to improve it. It's pretty good, but we still have a lot to learn."