Upload
others
View
2
Download
0
Embed Size (px)
Citation preview
T-Cube Web Interface for Real-time Biosurveillance
2009 ISDS 8th Annual Conference, Miami
in Sri Lanka and Tamil Nadu
Maheshkumar Sabhnani1, Artur Dubrawski1 and Nuwan Waidyanatha2
1 The Auton LabC i ll i iCarnegie Mellon University
2 LIRNEAsiaColombo, Sri Lanka
T-Cube Web Interface for Real-time Biosurveillance in Sri Lanka
Objective:To present an on-going effort to establish a system for real-time disease surveillance in a developing countryin a developing country
Agenda:1. Current disease reporting system in Sri Lanka2. Concept of the Real-Time Biosurveillance Program3. TCWI: An interactive surveillance component of RTBP
Slide 2 of 14 Copyright © 2009 Carnegie Mellon University and LIRNEasia
3. TCWI: An interactive surveillance component of RTBPa. Short demo (video)b. Examples of evaluations performed so far
4. Conclusion
The Need for Real-Time Biosurveillance
Slide 3 of 14 Copyright © 2009 Carnegie Mellon University and LIRNEasia
Source: DARPA
We have built an early warning system for
outbreaks of diseases (with U.Pitt.)...
Real-time Monitoring of Emergency Department Patient Data during 2002 Winter Olympics
Slide 4 of 14 Copyright © 2009 Carnegie Mellon University and LIRNEasiaFebruary 5, 2002
Interactive analyticsAstrophysics Learning Locomotion
Bio-surveillance
Food safety
United Nations CTBTO
Slide 5 of 14 Copyright © 2009 Carnegie Mellon University and LIRNEasia
Food safety
Nuclear threat detection
Safety of agriculture
Fleet prognostics
Sri Lanka Disease Notification: Current System
Black arrows represent the current manual paper/postal system for health data collection and reporting
• It can take up to 10 days p yfor Epidemiology Unit to receive data
• It only records cases of a few reportable diseases and no syndrome or diagnosis information
• No direct alerting
Slide 6 of 14 Copyright © 2009 Carnegie Mellon University and LIRNEasia
capability; alerts distributed by phone throughout organizational hierarchy (Epi Unit Regional Epis Medical Officers of Health Hospitals, Providers, etc )
SL Disease Notification: Opportunity for Improvement
Red lines: RTBP mobile phone communication system for heath data collection and reporting
M d l i• Many developing countries, including Sri Lanka, have pretty good cell phone coverage
• RTPB leverages that to mitigate reporting latencies
• RTBP adds automated
Slide 7 of 14 Copyright © 2009 Carnegie Mellon University and LIRNEasia
• RTBP adds automated alerting and advanced analytic capabilities
• It also aims at deepening and widening the scope of what is being reported
RTBP: Actors, Processes, Data Flow
Slide 8 of 14 Copyright © 2009 Carnegie Mellon University and LIRNEasia
Improved Detail and Resolution of Data
RTBP:• More dimensions• More values per dimension
distinct value count 17 9case date location disease
Min 2009-12-16 Bihalpola Dengue_feverMax 2009-07-10 Bingiriya Dysentery
Bopitiya EncephalitisDambadeniya Enteric_feverDunakadeniya Food_poisoningHettipola Human_rabiesHorathapola Leptospirosisrr
ent
distinct value count 33 140 3 10 36 31possible choices single single single single multiple multiple
case date location disease gender age group sign symptom
• More values per dimension• Not just reportable diseases• Daily resolution and
collection cycles• Finer spatial resolution
Kandanegedara Typhus_feverKattimahana Viral_HepatitisKatupothaKoshenaKuliyapitiyaMinuwangatteMunamaldeniyaNarammalaSandalankawaWariyapola
Cu
r
Slide 9 of 14 Copyright © 2009 Carnegie Mellon University and LIRNEasia
Min 2009-09-03 Ammanamulla Abdominal pain Female 0-1 Abdominal Pain Coated tongueMax 2009-11-30 Bogahapitiya Abscesses Male 1-5 Anal itching Crepitation
Bopitiya Accident Unknown 6-14 Backache CyanosisDeegalla Acne 15-19 Body ache DehydrationDunakadeniya Acute Diarrheal Disease 20-24 Burning sensation in the stomach DeliriumHavenegedara Acute Gastroenteritis 25-29 Chest pain Deviation of the tongueHorathapola Allergy 30-34 Cold Difficulty in breathingHorombawa Anemia 35-39 Constipation Heart murmurKadapathwehera Angina 40-45 Cough Increase of respiratory rateKandanegedara Anxiety Above_45 Deviation of the tongue Low Blood pressure
… … … … …
New
T-Cube Web Interface in RTBP
• Efficient representation of data substantially speeds-up analyses and user interactions
• Implements statistical pevent detection techniques
• Enables data visualization along temporal and spatial dimensions at interactive speeds
• Automates alerts
Slide 10 of 14 Copyright © 2009 Carnegie Mellon University and LIRNEasia
Video: Demo of the RTBP T-Cube Web Interface
Slide 11 of 14 Copyright © 2009 Carnegie Mellon University and LIRNEasia
RTBP: TCWI Evaluations Performed So Far
24 Health Sub Centers; 4 Public Health Centers Thirupathurblock Sivaganga
17 Hospitals/Central Dispensaries16 Sarvodaya Suwadana Centers
• Data is being collected under the new schema only for a few months now• Its amount & consistency do not warrant objective evaluations just yet
block, SivagangaDistrict, state of Tamil Nadu, India
16 Sarvodaya Suwadana Centers Kurunegala District, NorthWestern Province, Sri Lanka
Slide 12 of 14 Copyright © 2009 Carnegie Mellon University and LIRNEasia
• Therefore, so far, evaluations rely on semi-synthetic data:• Use actual historically reported counts per region and per reportable disease• Other attributes in RTBP schema are then probabilistically synthesized• We test the ability to detect known events, the usability in tracking of
emerging events and in revealing systematic patterns of events.
8/14/1710/30
Example Evaluation: Leptospirosis: Deadly Pile-up during Q4_2008--Q1_2009
• Over 360 deaths• Peak in Sept/08
originally noticed much later later
• RTBP would have• RTBP would have picked up leading waves much earlier than Sept/08
10/30 4/17 8/1
Slide 13 of 14 Copyright © 2009 Carnegie Mellon University and LIRNEasia
An elevated global score noted
4/14Spatial scan global score4/15
Situation in Kandy escalates, regions to the W
4/24
South begins to see
5/28
South continues to escalate, while the
Example Evaluation: Progression of Dengue Fever Outbreak (Spring 2009)
score noted, centered in Kandy
eg o s to t eand SW from it follow suit
increased counts
esca ate, e t emagnitudes in other regions subside
Slide 14 of 14 Copyright © 2009 Carnegie Mellon University and LIRNEasia
• Note apparent seasonality of Dengue fever
• It seems to
Aug 30,2007
Example Evaluation: Dengue Fever: Seasonal and Spatial Patterns
It seems to be moving South and West
May 1,2007 May 21,2008
April 15,2009 May 28,2009
Slide 15 of 14 Copyright © 2009 Carnegie Mellon University and LIRNEasia
Summary and Conclusion
1. RTBP deepens and widens coverage of reported health data while significantly reducing latency of reporting
2. T-Cube Web Interface (TCWI) supports RTBP by enabling: a. Automated comprehensive searches for events of interest
through large collections of data b. Interactive data navigation and visualization c. Automated explanation of detected patterns
3. A fielded RTBP-like system can qualitatively yet affordably improve ability to detect and mitigate bio-medical threats in
Slide 16 of 14 Copyright © 2009 Carnegie Mellon University and LIRNEasia
p y gcountries with limited resources and infrastructure
4. The pilot studies are ongoing in Sri Lanka and India, but we are on the outlook for additional challenges.
This work has been supported in part by the International Development Research Centre of Canada (105130), and by CDC (R01-PH000028) and NSF (IIS-0911032).
Acknowledgements
We have also received support from numerous collaboratorscollaborators not specifically named here!
Slide 17 of 14 Copyright © 2009 Carnegie Mellon University and LIRNEasia
Partners in RTBP: • LIRNEAsia• Indian Institute of Technology Madras, India• National Center for Biological Sciences, India• Sarvodaya Shramadana Society, Sri Lanka• Respere Lanka (Private) Limited, Sri Lanka• University of Alberta, Canada• Carnegie Mellon University Auton Lab, USA
Backup Slides
Slide 18 of 14 Copyright © 2009 Carnegie Mellon University and LIRNEasia
m-HealthSurvey Midlet [IITM's-RTBI]
(a) Main menu
(b) Profile registration
(c) Retrieve locations
(d) Patient record screen I(d) Patient record screen I
(e) Patient record screen II
Slide 19 of 14 Copyright © 2009 Carnegie Mellon University and LIRNEasia
Messaging/Alerting CAP/EDXL Broker [Sahana]Single input multiple output engine; Able to channel through multiple technologies
Manage publisher /subscribers and SOP
Adopt PHIN Communication and Alerting Guidelines for EDXL/CAP
Challenge of creating the EDXL/CAP template editor
Relating the template editor with the SMS/Email Messaging module
Slide 20 of 14 Copyright © 2009 Carnegie Mellon University and LIRNEasia
Understanding direct and cascading alert from a regional jurisdictional prospective
Designing the short and long text messages
Addressing in multiple languages