Lombardo

Assessing Syndromic Surveillance:Costs, Benefits, Future(Comments from the JHU/APL ESSENCE Experience) : 

Assessing Syndromic Surveillance: Costs, Benefits, Future (Comments from the JHU/APL ESSENCE Experience) Joe Lombardo Howard Burkom The Johns Hopkins University Applied Physics Laboratory Joe.Lombardo@jhuapl.edu

The Questions?: 

The Questions? What is “Syndromic Surveillance”? What is “Situational Awareness”? Can an automated surveillance tool reliability provide for early warning? Can automated surveillance tool provide situational awareness? System costs versus the benefits? Where do we go from here?

Decomposition of Automated Surveillance Tools : 

Decomposition of Automated Surveillance Tools Automated surveillance tools have several functions (automated data capture, archiving, text parsing, processing to detect signals, data mining, analysis, display, and alerting are just a few). Automated surveillance tools provide users with the ability to acquire data which would otherwise require labor and time to acquire. Syndromic Surveillance functions include grouping signs and symptoms into large categories (RESP, GI, etc.) and trying to detect the leading edge of a signal in noise. Modern surveillance tools can query for a specific case definition, thereby reducing the noise and improving detection performance. Modern surveillance tools provide the user with a variety of data mining and display functions. Modern surveillance tools have both early event and situational awareness functions.

The Most Critical Component of a Surveillance System: 

The Most Critical Component of a Surveillance System The most critical component of any surveillance tool is the data monitor! A less than ideal automated surveillance tool in the hands of an experienced and energetic data monitor can be very valuable. The best surveillance tool in the world is of little value to a public health organization that lacks resources or commitment for routine monitoring. The data monitors must be actively engaged in using their tools. The performance of the data monitors must be considered when evaluating a system. Training, exercises, and investigation of small local outbreaks are all critical for effective surveillance.

Examples of Early Alerting: 

Examples of Early Alerting Daily Counts for GI at MCRD 2002 vs. 2004 Courtesy of LTC Julie Pavlin Walter Reed Army Institute of Research ESSENCE IB 60 days ESSENCE IV 90 days

A Syndromic Surveillance Success from 2005Small Norovirus Outbreak Found with ESSENCE: 

A Syndromic Surveillance Success from 2005 Small Norovirus Outbreak Found with ESSENCE Detection of Norovirus outbreak is California High School trip to the Capital. CDC investigating similar incident for Canadian group after students got sick on plane. Both groups had similar itineraries. Airlines Isolated group on their trip back to California. Taken from: Detection of an Outbreak of Gastroenteritis in Montgomery County Expedites Public Health Response - Kathy J. Hurt-Mullen, MPH, Montgomery County Department of Health and Human Services

Can an Automated System Provide Situational Awareness?: 

Can an Automated System Provide Situational Awareness? On January 6, 2005, two freight trains collided in Graniteville, South Carolina (approximately 10 miles northeast of Augusta, Georgia), releasing an estimated 11,500 gallons of chlorine gas, which caused nine deaths and sent at least 529 persons seeking medical treatment for possible chlorine exposure. What is the impact on Ft Gordon?

Slide8: 

Location of Patients Showing up at EDs

Facility Monitoring: 

Facility Monitoring Event Event

ISDS Public Health Practice Committee(Examples): 

ISDS Public Health Practice Committee (Examples) Don Olson (NYCHD) – Improvements to flu morbidity using NYC surveillance data Amy Ising (UNC) – Event detection using NCDETECT (Norovirus in Chapel Hill, meningitis in Pitt County, meningococcemia in Rowan County) Tigi Ward (Lubbock HD) – Monitoring for plague after samples of organism were stolen, early influenza testing using SYRIS Brian Fowler (Ohio HD) – Rash outbreak in elementary school (RODS) Bill Stephenson (Tarrant County HD) – Early recognition of influenza (ESSENCE) Joe Gibson (Marion County) – CO Poisoning, gastroenteritis from food handling procedures at local store, tracking tornado casualties (ESSENCE) Jeff Schlegelmilch (Boston PHC) – Tracking the spread of Norovirus in 2005-6

E-mail from Atar Baer, Seattle King County: 

E-mail from Atar Baer, Seattle King County Public Health - Seattle & King County has had a syndromic surveillance system in place since 1999. The chief benefits have been: (1) Monitoring for diseases of public health importance following planned events (e.g., WTO Ministerial in Seattle in 1999). (2) Surveillance after natural disasters, where information was available via emergency department-based syndromic surveillance much more quickly than by obtaining information manually via active surveillance. For example, in December 2006, following a severe windstorm in western Washington, there was a surge in patients presenting to EDs with carbon monoxide poisoning; the system was used not only for case-finding, but also to describe the outbreak with regard to person/place/time, and monitor for increases in secondary illnesses, such as gastrointestinal complaints due to loss of refrigeration and consumption of potentially spoiled food products. (3) In conjunction with other traditional data sources (e.g., laboratory data), to aid in identifying the "start" of the flu season, and to help monitor the severity of the season compared with previous years as reflected in the distribution of cases by age and by disposition. This is mostly a "subjective" benefit of the system and has not been formally evaluated. (4) Assist with case-finding to help with outbreak containment (e.g., measles outbreak in 2004) and as routine (daily) monitoring for diseases of public health significance that have not been reported to the health department (e.g., animal bites or exposures that require rabies PEP).

Situational Awareness has Many Features,Varying by Institution & Jurisdiction : 

Situational Awareness has Many Features, Varying by Institution & Jurisdiction Reported benefits include: Corroboration of clinical suspicions Tracking spread of outbreaks, identifying likely population at risk Assessing the likely disease burden of seasonal influenza Monitoring of large public events Tracking effects of natural disasters Explaining away/ruling out/rumor control e.g., determining that a school-based outbreak is not spreading Investigating effects of severe weather

System Value When There are “No” Alerts : 

System Value When There are “No” Alerts Was there any value to the DEW Line operated by NORAD during the cold war? What is the value to the country to know that those who defend it are healthy? What is the value to health departments to know that the populations they serve are not at risk? What is the value to their communities?

Key Findings: 

Key Findings Courtesy of Marc Paladini NYCHD

Costs Breakdown : 

Costs Breakdown Typical Cost Components Hardware & Software Data Acquisition IT Maintenance Software Upgrades System Monitors Alert Follow-ups Training and Exercises System costs are a function of: Number of data sources Quantity of data Mechanism of acquiring data Number of servers Amount of analysis performed Number of personnel needed etc. National systems will cost more than local systems. Current methods may not be the most cost effective.

Redefining Syndromic Surveillance : 

Redefining Syndromic Surveillance Old vision: large, fixed syndrome classification groups Used to filter daily data records to create data time series for control charts and spatial cluster detection Most emphasis on detection of leading edge of outbreak signal Institutions finding utility in biosurveillance systems are changing this vision Situational awareness is an everyday benefit More “dual use” capabilities, not focused only on bioterrorism Migration toward smaller, adjustable syndrome groups Mirroring the evolution of epidemiological case definitions More dependent on case linkage, clinical findings Challenges to informatics: better symptom-based query capability Challenges to analytics: more agile, robust algorithms with fewer alerts

Summary: 

Summary Early alerting is possible with syndromic surveillance, but information systems that permit analysis by an initial case definitions have to potential to provide increased early detection performance. Automated collection of surveillance data can be used to provide situational awareness before, during and after outbreaks. Surveillance tools are only as good as the individuals who use those tools. System costs can vary widely depending on the approach used to acquire, host and monitor surveillance data.