Syndromic Surveillance Picks up Terror Signals’

Establish thresholds, investigate anomalies

The wave of the future in bioterrorism surveillance is computer-based systems for syndromic reports that can serve as an early tripwire if an attack is under way.

Comparing selected clinical indicators against baseline thresholds, syndromic surveillance systems are increasingly being applied in health care systems and offered in the private sector.

But surveillance is nothing new to infection control professionals. The increasingly sophisticated syndromic systems don’t replace time-honored gumshoe epidemiology, said Adi Gundlapalli, MD, PhD, an epidemiologist at University of Utah Medical Center in Salt Lake City.

"Computer-based surveillance systems are not really new, they are just novel applications," he said recently in San Antonio at the annual conference of the Association for Professionals in Infection Control and Epidemiology. "There are many surveillance systems we use almost everyday for hospital-acquired infections. It can be as simple as flagging resistant organisms [in lab reports]."

Gundlapalli helped develop and monitor a syndrome-based surveillance system to detect bioterrorism incidents during the 2002 Winter Olympics in Salt Lake City. "As professionals in infection control and hospital epidemiology, we have a unique expertise. We actually liaison between the hospital and public health," he said. "That turns out to be very important, especially when there is an event. You have to have pre-established relationships."

Held in February 2002, the Winter Olympics was one of the first major international events held in the wake of 9/11 and in the shadow of the anthrax attacks. "There was a week or two in Salt Lake City that people were not even sure if the games were going to be held," Gundlapalli said. "Bioterrorism had become a reality. There was no question that this had changed from the academic papers."

Deciding What to Track

To meet the threat, Gundlapalli and colleagues developed the Advanced Logic for Event Detection in Real Time (Alert) system. The system monitored patients seen in the University Hospital’s emergency department, outpatient clinics, and a special clinic at the Olympic Village. "What we get out of it, for lack of a better word, are signals," he said. "Each data element is like a signal to us. The first question was what would be of interest to infection control personnel. The answer, of course, was everything. But you have to be able to handle it, so we sat down and talked about specific signals."

The surveillance indicators selected include emergency department admissions and orders for cultures, X-rays, and diagnostic tests. As specific clinical indicators were winnowed out, the clinicians also set threshold levels based on baseline data. "At the University of Utah hospital, we do 30 X-rays a day for various reasons," Gundlapalli said. "So on any day, if you are doing 60, then you know that there is something else going on. It may indicate that somebody is worried about something going on in the lungs."

By the same token, baseline levels and thresholds for response were set for blood cultures, stool cultures, and orders for diagnostic tests for specific diseases. "If someone orders a nasal swab for anthrax, you want to know about it," he said. "Even one—you want to know about." Other tracked indicators were patient’s primary complaint, prescription of antibiotics, and use of any antidotes. "This is an attempt to look at patterns, to try and infer diseases or syndromes. I stress the point infer’ because I don’t think we can ever be sure just putting together syndromes," he added.

In the Alert system, patients are categorized on a 1 to 5 scale indicating severity of illness. The number increases as additional indicators are added to a patient’s record. "We have a user interface that can be navigated," Gundlapalli told APIC attendees. "We are able to drill down. If I have a patient who comes up positive as an infectious pneumonia signal, then I could click on that and review their electronic medical record."

Beginning Feb. 1, 2002, the system was monitored via computer three or four times a day for four weeks. "It took about 30 to 45 minutes per session," he said. "You can imagine if I pulled up the grid and saw infectious pneumonia for five Level 4 or 5 patients, I would then go to the electronic medical records for each patient." The medical records were so complete, Gundlapalli only made two follow-up phone calls during the period to the physicians treating the actual patients. "We saw that most signals were within thresholds. The ER visits increased, on occasion, hospital admissions increased, and a lot of blood cultures were ordered during that time, which worried me a little bit. But really, no specific pattern was detected."

On one occasion, when the 75-patient threshold for daily emergency department visits hit 115, a surveillance colleague actually went to the hospital to get a first hand look at the situation, he said. "It was just a bad day for the ER. There was nothing specific. It shows you the power of this, but it also shows that you do need someone continuously monitoring this. You cannot have the computer make the decision."

The only real signal of a sharp disease increase was for influenza, which struck the Olympic Village and surrounding areas, he noted. Syndromic surveillance can be done without such elaborate measures, he said, urging ICPs to be, above all, "smart observers." Use the resources you have and the triggers you can read within the limits of your program, Gundlapalli advised.