Issue 8 March 2003 - HIGHLIGHTS FROM THE 1ST NATIONAL SYNDROMIC SURVEILLANCE CONFERENCE IN NEW YORK CITY

________________________________________________________________________________

HIGHLIGHTS FROM THE 1ST NATIONAL SYNDROMIC SURVEILLANCE CONFERENCE IN NEW YORK CITY
Sponsored by: CDC, NYC Dept Health (NYCDH), NY Academy Medicine, and the Sloan Foundation

Underscoring the importance of Syndromic Surveillance since 9/11/01, the first ever national syndromic surveillance conference was held at the NY Academy of Medicine in New York City, September 23-24, 2002. Attended by "over 400 public health practitioners, academics, physicians, computer scientists, vendors, statisticians and lawyers from over the world, this conference provided a forum to define the evolving science of syndromic surveillance, review and evaluate systems, discuss methodologies, opportunities and challenges in this emerging field". (1)

Topics covered: (1)

* Syndromic Surveillance in the context of public health
* Draft CDC Framework for evaluating syndromic surveillance systems
* Temporal and Temporal-spatial outbreak detection
* Potential non-traditional data sources
* Data Transfer, warehousing and information technology
* Legal mandate and confidentiality
* Investigation of syndromic alarms

Notable Points

Mike Osterholm, PhD, Special Advisor to Tommy G. Thompson, US Secretary Health & Human Services, led off the conference with a warning to all health departments to start implementing syndromic surveillance now. "We're all waiting for the other shoe to drop and there's no time for that".

Farzad Mostashari, MD, Assistant Commissioner of the Division of Epidemiology at the NYCDH, is a strong proponent of syndromic surveillance and believes that early detection will decrease morbidity and mortality in public health outbreaks as well as strengthen traditional reporting. Since September 2001, NYC has been using a homegrown SAS program to import ASCII Chief Complaint data from Emergency Departments (ED) around the city. Their current data sources are: ED logs (Chief Complaint) and ambulance dispatches. Ambulance data is used as a form of redundancy to confirm what is being seen in the ED.

Claire Broome, CDC's Senior Advisor for Integrated Health Information, spoke on the issue of confidentiality. She indicated that most states have rules that require investigation of clusters. Therefore, the health departments should be able to collect identifiable data for syndromic surveillance. HIPAA does not pre-empt state public health law. However, state public health rules do not protect academics or researchers, many of whom have research-based Syndromic Surveillance systems. These individuals need a “data use agreement” to collect even a limited amount of confidential data.

It’s not easy to collect good data

The general consensus of the conference was that "Drop-in", web-based surveillance systems don't work in the long run because they require dedicated data entry personnel. Several drop-in systems were tested in the Salt Lake City area during the Olympics. These web-based systems were deemed to be too labor- intensive to be practical on a routine basis. In fact, even during the brief time frame of the Olympics, data entry personnel quickly lost interest, entering less and less data each week. Organizers had to encourage them with doughnuts and other enticements to enter the required data every day. There were also complaints about web-based data entry in general, and how difficult it was to correct data entry mistakes. 

While everyone wants “real-time” data, it turns out “real-time” is hard to define. Sometimes “real-time” is used to describe web-based data entry; but web data may not be as "instantaneous" as one might imagine, considering: 1) Is someone really entering data "real-time" on all 3 shifts? 2) Is the Internet connection really up 24 hours/day? 3) Is the data queued before being sent outside the hospital firewall? 4) Is the data quarantined before being added to the master files? 

HL-7 data is sometimes called “real-time” because it can be transmitted 24 hours/day. However, if the data needs to be input into a computer and/or ICD-9 coded, there may be a 3- or 4-day delay before the HL-7 data is available to be transmitted. This choice was also considered impractical because of the cost, lack of availability, and inconsistency in different versions of HL-7. For example, it was reported to cost $20,000 per hospital to set up an outbound HL-7 data server. The Seattle King County HD, in conjunction with the University of Washington is collecting data from 3 hospitals; none of which have outbound HL-7 capabilities.

At the end of the conference, there was a consensus that the most practical method of performing syndromic surveillance is to use existing ED data. That is, to translate via natural language processing the ED chief complaint and other available symptom text data into symptoms and/or syndromes. These data can be analyzed to determine clusters of symptoms by time and space. Various examples were given of clusters that were investigated to determine if an outbreak actually occurred and whether it was BT-related.

References and Helpful Websites 
Check out these links for additional information

1. www.nyam.org/events/syndromicconference/index.shtml
More details of the SS conference and links to Poster Sessions

2. www.tulsabiowatch.com/public_surveillance_info/city-county_surveillance.htm 
Tulsa's latest syndromic surveillance information and syndrome graphs

3. www.cs.cmu.edu/~awm/antiterror
Data Mining Information, Technologies, and Algorithms

Back to the top