Susan Huang, MD MPH

1

• (Susan Huang, MD MPH)
• Ken Kleinman, ScD
• Department of Ambulatory Care Prevention
• Harvard Medical School and Harvard Pilgrim
  Healthcare

• National Institute of General Medical Sciences
• National Institutes of Health
• Department of Health and Human Services

2
Hospital Outbreak Detection

• Required of every hospital by the Joint
  Commission
• Critical elements are mostly familiar
• Correct assessment
• Timely identification
• Tracking of containment (I.e. When is an outbreak
  over?)

3
Current Issues in Hospital Outbreak Detection

• Outbreaks can involve
• Any of hundreds of organisms
• Any hospital unit
• Any clinical service
• Medical equipment

4
Current Issues in Hospital Outbreak Detection

• Currently Incomplete ascertainment
• Limited surveillance of organisms, patients
• Mostly based on clinician report
• Routine tracking of only a few organisms, e.g.
• MRSA
• VRE
• Other highly resistant bacteria
• Aspergillus

5
Current Issues in Hospital Outbreak Detection

• Example of current surveillance MRSA
• Common current alert method 3 incident cases in
  a unit within two weeks outbreak
• Simulation study showed 3-7 alerts per year,
  24-110 days in alert with this method, based on
  observed unit characteristics and assuming
  constant risk of infection
• Close to observed outbreak data

6
WHONET-SaTScan A MIDAS product

• Uses lab data directly instead of clinician
  report
• Links microbiologic analysis to statistical
  analysis
• Enables hospital outbreak detection
• Hospital-wide
• By unit and related unit groups
• By service and related service groups
• By antibiotic resistance pattern

7
WHONET

• WHO sponsored free software
• Describes microbiologic data
• Management
• Analysis
• 1200 laboratories 80 countries17 languages
• Automatically imports data from labs

www.who.int/drugresistance/whonetsoftware
8
Patients with Staphylococcus aureus Isolates
9
www.satscan.org
10
SaTScan Analysis

• Assesses temporal trends
• Compares rates across organisms
• Assesses organism-specific rates
• Using prior baseline in past year
• Stratified by unit, service, organism, antibiotic
  profile
• Provides daily alerts

11
Outbreak AlertSaTScan Parameters

• First signal for an event has a maximum length of
  60 days
• Statistical threshold (novel, in this setting!)
• One false alert per year per comparison
• recurrence interval of 1 in 365
• p0.0027

12
WHONET SaTScan Report

• Signal Alerts
• Daily report of all new alerts
• Repeat alert of same cluster if cases increase
• Alert Data
• Type of alert (organism, unit, abx resistance)
• 1st alert date
• 1st culture date
• Observed cases in outbreak
• Expected cases in outbreak
• Recurrence Interval

13
Example WHONET Clusters
Dates are fictitious
Confidential
14
Acinetobacter baumanii Cluster
Dates are fictitious
Confidential
15
Acinetobacter baumanii Isolates
Start date 6/1/05
End date 8/1/05
Dates are fictitious
16
Acinetobacter baumanii Suspicious Susceptibility
Pattern
Start 6/1/05
End 7/28/05
Dates are fictitious
17
Outbreak Detection viaWHONET-SaTScan Example

• Dataset Brigham and Womens Hosp.
• 2001-6 microbiology lab data
• All organisms
• First ever per patient (new acquisition)
• Isolated gt2 calendar days from admission
• Elements
• Patient identifiers Clinical service
• Organism Antibiotic profile
• Date of culture Location of culture

18
Evaluation2001-6 SaTScan Alerts

• Median 15 alerts per year
• Outbreak Size
• 2 patients 24
• 3-5 patients 46
• 6-10 patients 16
• gt10 patients 14
• Outbreak Type
• Hospitalwide 12
• Service 10
• Unit 25
• Antibiotic Profile 53

19
Evaluation Sensitivity

• SaTScan found 5 of 5 of Infection
  Control-determined definitive (large) outbreaks
  2001-2006
• I.e. no false negatives (treating infection
  control as the gold standard)

20
Evaluation False positives?
Non-identification of outbreaks in real-time IC
surveillance does not imply no outbreak Two
readers (infection control program directors)
were given output of SaTScan as if live, answered
questionnaire
21
Concordance

• Data 2 Years, 23 outbreak alerts
• Level of concern
• True/false not known face validity only
• Recall more data available here than in original
  surveillance

22
Conclusions and Next Steps

• Meaningful mining of micro data with SaTScan
• Detection of all epidemiologically confirmed
  outbreaks
• Frequency, importance of alerts are reasonable
• Next steps
• Survey assessment of additional years
• Chart review assessment of WHONET-SaTScan
  outbreaks not previously known to IC
• Development of user-friendly interface

23
Collaborators

• Deborah S Yokoe, MD MPH
• John Stelling, MD MPH
• Martin Kulldorff, PhD
• Hilary Placzek, MPH
• Michael Calderwood, MD
• Thomas F. OBrien, MD
• and Richard Platt, MD MS

24
WHONET Use in the World

• African Regional Office of WHO (AFRO)
• Algeria, Kenya, Namibia, South Africa, Tanzania,
  Zambia
• Eastern Mediterranean Regional Office of WHO
  (EMRO)
• Jordan, Kuwait, Lebanon, Libya, Morocco, Oman,
  Pakistan, Saudi Arabia, Tunisia
• European Regional Office of WHO (EURO)
• Austria, Belgium, Bulgaria, Croatia, Czech
  Republic, Denmark, Estonia, Finland, France,
  Georgia, Germany, Greece, Iceland, Ireland,
  Israel, Italy, Latvia, Luxembourg, Malta,
  Netherlands, Norway, Poland, Portugal, Romania,
  Russia, Slovakia, Slovenia, Spain, Sweden,
  Ukraine, United Kingdom
• Pan-American Health Organization (PAHO)
• Argentina, Bolivia, Brazil, Chile, Colombia,
  Costa Rica, Cuba, Dominican Republic, Ecuador, El
  Salvador, Guatemala, Mexico, Nicaragua, Panama,
  Paraguay, Peru, United States, Uruguay, Venezuela
• South-East Asian Regional Office of WHO (SEARO)
• India, Indonesia, Sri Lanka, Thailand
• Western Pacific Regional Office of WHO (WPRO)
• China, Hong Kong (China), Japan, Republic of
  Korea, Malaysia, Philippines, Singapore, Taiwan,
  Viet Nam

25
(No Transcript)
26
RIS Histograms Ps. aeruginosa