Experimental Fully Automatic Syndromic Surveillance in Japan

1
Experimental Fully Automatic Syndromic
Surveillance in Japan

• Yasushi Ohkusa, Ph.D.1), Tamie Suger, Ph.D. 1),
  
• Hiroaki Sugiura,M.D. 2),Kazuo Kodama,M.D. 3),
• Takushi Horie,M.D. 4),Kiyoshi Kikuchi,M.D.,Ph.D.5)
  ,
• Kiyosu Taniguchi,M.D.,Ph.D.1), Nobuhiko
  Okabe,M.D.,Ph.D.1)
• 1)National Institute of Infectious Diseases,
  2)SugiuraClinic,
• 3)Kodama Clinic,4)Chiimiya-Horie Clinic,5)
  Department of Pediatrics, Shimane Prefectual
  Central Hospital,

2
Situation in Japan

• Infection Control Law asks doctors to cooperate
  in syndromic surveillance for pandemic flu and
  smallpox
• But doctors have to report by typing the number
  of patients on the web site or sending a fax to
  the local public health center
• It imposes the heavy burden of reporting, and
  thus it has not worked yet
• We need a fully automatic syndromic surveillance
  system

3
Situation in Japan

• Medical services for outpatients are well
  developed due to universal public health
  insurance
• Even patients who has mild symptoms can visit a
  clinic freely in Japan
• Especially, 70 of patients visit clinic, not
  hospital
• Electronic Medical Record (EMR) have not had
  much penetration, less than 10
• Moreover, almost nobody use HL7 or other standard
  for EMR
• We have to develop the system for each type of
  EMR

4
Situation in Japan

• Privacy is a much bigger concern in Japan, in
  comparison with the USA
• Zip codes which are more specific than just the
  city are not permitted to be used
• We have to search keywords for complaints in EMR
  in the clinic or hospital
• Then we collect information only about the number
  of patients with each symptom from the clinics
  or hospitals

5
A Clinic
Electronic Medical Record
Fever, Respiratory Symptoms, Diarrhea, Vomiting,
Rash
Keyword Search

Excluding Negative Meaning
Full Automatic
Counting of Patients by age and gender
Statistical Analysis
Outbreak Detections
Number of Patient Outbreak Detections


Secure Internet
6
Local Government Public Health Center NIID
B Hospital
Electronic Medical Record
Fever, Respiratory Symptoms, Diarrhea, Vomiting,
Rash
Keyword Search
Excluding Negative Meaning
Counting of Patients by age and gender
Statistical Analysis
Outbreak Detections
Number of Patient Outbreak Detection

C Clinic
Electronic Medical Record
Fever, Respiratory Symptoms, Diarrhea, Vomiting,
Rash
Keyword Search
Excluding Negative Meaning
Secure
Counting of Patients by age and gender
Statistical Analysis
Outbreak Detections
Internet
Number of Patient Outbreak Detection

7
Detection Algorithm

• We adopt multiple regression model to detect
  outbreaks
• If more than two years of data is available, it
  can separate regular unusual events from seasonal
  patterns
• We regress the number of patients in each day
  using dummies for the week number(1-52,53) , the
  day of the week (Sunday-Saturday) , post-holiday,
  and time trends such that
• Number of cases t
• aSißi (Week No)i
• Sj?j (Day-of-the-Week)j
• ?(the Day after Holiday)?tdt2et

8
Multiple Regression Model (cont.)

• Three criterions are used for alert
• Low level If the observation is higher than
  prediction by more than three times of standard
  deviation
• Medium Level four times
• High Level five time

9
Real Time Syndromic Surveillance for
Outpatients Feedback Homepage through Internet
Server
Number of Patients and Outbreak Detections in
this clinic/hospital
Outbreak Detections in this community
Real Time
Real Time Syndromic Surveillance
2006 October, 2006( Monday)
A Clinic
Outbreak Detection in this aria
Number of Patients and Outbreak Detections in
this clinic/hospital
Number of Patients and Outbreak Detections in
this clinic/hospital
Outbreak Detections in this aria
Alert level
Number of Patient
Male
Fever
Female
Respiratory Symptoms
Male
Female
Diarrhea
Vomiting
Rash
A Clinic
10
4th September, 2007 (Tuesday)
Outbreak Detection in the Community
Outbreak Detection in the Clinic
Fever
Respiratory Symptoms
Diarrhea
Vomiting
Rash
Convulsion
Fever Respiratory Symptom
11
Outbreak Detection in this Clinic
Number of Patients in Vomiting in this six months
of Patients
Date
12
Outbreak Detection in this Clinic
Number of Patients in Vomiting in all periods
of Patients
Date
13
Outbreak Detection in the Community
Outbreak in Vomiting in this six months
Recommendation from local public health center to
schools
Proportion()
Date
14
Outbreak Detection in the Community
Outbreak in Vomiting in all period
Proportion()
Date
15
(No Transcript)
16
This system detected

• Epidemic of Noro virus in the last November.
• Moreover, clinics and hospital were able to
  respond quickly in the last late influenza
  epidemic by
• Announcing that an "Influenza epidemic is coming
  ... to patients with poster in clinics, earlier
  than public report.
• Purchasing additional masks and rapid test kits.
• Canceling the leave plan of nurses or clerks.
• Having doctors visit patients' house
• Informed neighborhood pharmacies that an
  Influenza epidemic is coming ...

17
Current Situation of Fully Automatic Syndromic
Surveillance through Electronic Medical Record
Working ?3 cities (16 Clinics 1 Hospitals )?
Under construction ?2 cities (2 Clinics 1
Hospitals )?
Planning ?3 Cities (3 Hospitals )?
Clinics
1 Clinics
Hospitals
2 Clinics
9 Clinics
1 Hospitals
5 Clinics 1 Hospitals
1 Hospitals
1 Clinics
1 Hospitals
1 Hospitals
18
Challenging
OTC Sales
Electronic Medical Record
Fire Department
Prescription of Drug
1200 Pharmacies
In the next 3 years
1000 Clinics 50 Hospitals
3000 Pharmacies
10 Cities
Local Syndromic Surveillance System
Feed back
Technical Support
Local Government Public Health
Center Response Evaluation
NIID Imposing statistical algorithm
Checking sensitivity / specificity
19
Thank you!