Real-Time Biosurveillance Program Pilot - India & Sri Lanka

Presentation Title
Conference Theme
Month day, 2010
Location
Nuwan Waidyanatha
LIRNEasia
Email: nuwan@lirneasia.net
http://www.lirneasia.net/profiles/nuwan-waidyanatha
Mobile: +8613888446352 (cn) +94773710394 (lk)
This work was carried out with the aid of a grant from the
International Development Research Centre, Canada.

Early detection and mitigation of common diseases and pandemics
Real-Time Biosurveillance Program to Revolutionize
disease surveillance and notification
www.lirneasia.net

INTRODUCTION TO THE RESEARCH
 Synergies of RTBP and Early Warning Systems
 Research question and specific objectives
DISEASE INFORMATION REQUIREMENTS
 Determinants of notifiable diseases in India and Sri Lanka
 Cycle of data collection, analysis, and dissemination
COMMUNICATION SYSTEM EVALUATIONS
 Data collection :: mHealthSurvey mobile application
 Event detection :: T-Cube Web Interface
 Disseminations :: Sahana Messaging/Alerting Module
 Cost effectiveness, efficiencies, and sensitivity
CONCLUSIONS & REFERENCES
www.lirneasia.net
Outline

Synopsis of IDRC funded PANACeA projects
PAN Asian Collaboration for Evidence-based e-Health Adoption and Application
Initiative to generate evidence in the field of e-health within the Asian context, by forming a network of
researchers and research projects from developing Asian countries.
http://www.aku.edu/CHS/panacea/about.shtml
Some PANACeA Initiatives (more :: http://tinyurl.com/39ypljm )
❏ Outbreak Management System
❏ Systematic review of ICTs in disasters
❏ e-Health system for community health care recording and reporting
❏ Mobile telemedicine system for ambulance and movable community health care
❏ Mobile telemedicine kit for disaster relief
www.lirneasia.net

Other m-Health Initiatives on disease control
 Panacea THIRRA (http://thirra.primacare.org.my): Portable system from telehealth and health information
www.lirneasia.net
in rural & remote areas
 Cell-Life: preventing HIV/Aids (): monitoring and intervention programs by communicating data via mobile
phone GPRS connected workstations
 EpiSurveyor Software developed by DataDyne (): Desktop tool to develop “forms” for handheld mobile
devices; J2ME mobile phone solution, Tested on PDAs in Uganda, Tanzania; on “CDC Maternal Health
Evaluation Forms”;live stock development board Sri Lanka
 OpenRosa/JavaRosa (): consortium developing standards based tools for collecting data via mobile phones,
analyzing data, and reporting data via mobile phones, Free software downloads – “Gather” is a project
supporting the openrosa work and testing solutions in Africa
 D-Tree Offers Childcare solutions ():They use javarosa code base; e-MICI project in Uganda – household
questionnaire to monitor Child diseases; doing work with BRAC in Tanzania on gathering household health
info
 ComCare: XForms based solution for collecting community health information
 InSTEDD: create and advocate open source tools (): Social network approach, Tested in Mekong river basin,
Event base surveillance techniques
 Click Diagnostics (): collect patient household data, stakeholders to run custom analyze , program to target
disease areas and improve intervention (e.g. HIV/AIDS staging and regimen management, cervical cancer
screening, malaria surveillance, TB surveillance)

Disease Surveillance
www.lirneasia.net

Doctrine of Real-Time Biosurveillance (RTBP)
Sensor
Detection
www.lirneasia.net
RTBP?
Response
RTBP?
Decision
Physical World
Broker
Health Providers,
Relief Workers
Observe
Relevant Data
RTBP
m-HealthSurvey
Record and
Transmit Data GSMphone
network
RTBP
Server and
Database
Store Data
RTBP
Interactive
Visualization,
Analysis and
Event Detection
Software
Monitor Data
Affected
Population
Automated
Alerts
Interactive
Analytics
Internet,
GSM
network
Internet
Manage
Relief Effort
Analysts, Health Officials,
Epidemiologists, Decision Makers
Health
Disaster
Management
RTBP?
RTBP?
RTBP?

Evaluating a Real-Time Biosurveillance Program: Pilot
Research Question: “Can software programs that analyze health statistics
and mobile phone applications that send and receive the health information
potentially be effective in the early detection and mitigation of disease
outbreaks?”
www.lirneasia.net
Specific Objectives
Evaluating the effectiveness of the
m-Health RTBP for detecting and
reporting outbreaks
Evaluating the benefits and
efficiencies of communicating
disease information
Contribution of community
organization and gender
participation
Develop a Toolkit for assessing m-
Health RTBPs

Research Question: “Can software programs that analyze health statistics and
mobile phone applications that send and receive the health information potentially be
effective in the early detection and mitigation of disease outbreaks?”
Data Collection Event Detection Alerting
www.lirneasia.net
mHealthSurvey a data entry
software works on any
standard java mobile phone.
A typical record contains
the patient visitation date,
location, gender, age,
disease, symptoms, and
signs. Data is transmitted
over GPRS cellular
networks.
T-Cube Web Interface
(TCWI) is an Internet
browser based tool to
visualize and manipulate
large spatio-temporal data
sets. Epidemiologists can
pin down a potential
outbreak of, for instance, a
gastrointestinal disease
among children in the
Sevanipatti PHC health
division.
Sahana Alerting Broker
(SABRO) allows for the
generic dissemination of
localized and standardized
interoperable messages.
Selected groups of
recipients would receive the
single-entry of the message
via SMS, Email, and Web.

Problem the RTBP is trying to solve in Sri Lanka
Black arrows: current manual paper/postal system for health data collection and reporting
Red lines: RTBP mobile phone communication system for heath data collection and reporting
www.lirneasia.net
Reduce 07-15 day
delays to Minutes
Re-engineer the
limited disease
activated passive
surveillance to
Comprehensive
Active surveillance

Problem RTBP is trying to solve in India
Black arrows: current manual paper/postal system for health data collection and reporting
Red lines: RTBP mobile phone communication system for heath data collection and reporting
www.lirneasia.net
Reduce 07-15 day
delays to Minutes
Re-engineer the
limited disease
activated passive
surveillance to
Comprehensive
Active surveillance

Actors, processes, and information flow of the proposed data collection, event
detection, and situational-awareness/alerting real-time program
www.lirneasia.net
RTBP high level system diagram
Skip the paper
1. Health records digitized by health
workers in Thirupathur block using mobile
phones.
2. Disease, symptoms, and demographic
information transmitted across GSM
mobile network to central database.
3. Data analysed by trained staff at the
IDSP and PHC Departments.
4. Automated event detection algorithms
process a daily ranked set of possible
disease outbreaks, which are presented to
IDSP and PHC staff.
5. List of possible outbreaks examined by
IDSP and PHC staff to determine
likelihood of an adverse event.
6. Confirmed adverse events disseminated
to Medical Officers, HIs, nurses, and other
health officials, within affected geographic
area.
7. Condensed version of the alert pushed
through SMS to get immediate attention of
the recipients.
8. More descriptive message emailed and
published on the web (also accessible
through mobile phone).

Why use mobile for data collection and dissemination?
Phone ownership (% of BOP teleusers)
Bangladesh Pakistan India Sri Lanka Philippines Thailand
www.lirneasia.net
100%
80%
60%
40%
20%
0%
These charts with demand side statistics were taken from LIRNEasia's teleuse at
BOP 3 study - http://lirneasia.net/projects/2008-2010/bop-teleuse-3/
Fixed only
Mobile + fixed
Mobile only

Evaluation metric verticals and horizontals
www.lirneasia.net
 Three verticals – data
collection, event detection
and reporting
 Four layers – Institutional,
content, application,
Transport
 Arrows showing the
Interoperability between
layers and verticals
 Objectively assess by
calculating various
indicators: costs, efficiencies,
error rates, etc
 Subjectively assess through
interviews and simulations
Talk about
this

www.lirneasia.net
The pilot in India and Sri Lanka
 24 Health Sub Center Village
Nurses
 4 Public Health Center Sector
Health Nurses, Health
Inspectors, and Data Entry
Operators
 1 Integrated Disease
Surveillance Program Unit of the
Deputy Director of Health
Services
 Thirupathur Block, Sivagangai
District, Tamil Nadu, India
 12 District/Base Hospitals
and Clinics
 15 Sarvodaya Suwadana
Center Assistants
 4 Medical Officer of Health
divisions & 1 Regional
Epidemiology Unit
 Kurunegala District,
Wayamba Province, Sri
Lanka

Why digitize the frontline health data?
Existing RTBP
www.lirneasia.net
❒ Activated passive surveillance
❒ Data limited to 25 infectious diseases (avg
LK=70 TN=600 monthly health records per
district)
❒ Only 20% of the diagnosis are confirmed,
rest are probable and suspected, likely to
miss signs of emerging outbreak
❒ No data on other-communicable or chronic
diseases
❒ Trend analysis is based on < 0.05% of
patient population
❒ Planing and resource allocation is based
on expert opinion, departments are not
data driven
❒ Active surveillance with situational
awareness
❒ All data on diseases infectious and chronic
diseases (avg 50000 monthly health
records per district)
❒ Collect symptoms and signs for syndromic
surveillance;
❒ A richer dataset with more attributes
❒ Statistics are on actual patient visitations;
even data on snake bites and accidents
❒ Comprehensive data can be used for long
term planning and move towards data
driven departments

3
4
www.lirneasia.net
Data collection process
1
2
(1) Patient is received by the Nurse
(2) Nurse issues a diagnosis chit to patient fills in Name, Age, Gender,
and OPD No
(3) Medical Officer fills in the chit with diagnosis and treatment
(4) Patient presents chit to pharmacy to receive medication
(5) Data Entry Operator digitizes and submits the data
5

www.lirneasia.net
mHealthSurvey Midlet by IIT-M
 (a) Main menu
 (b) Profile registration
 (c) Retrieve locations
 (d) Patient record screen I
 (e) Patient record screen II

mHealthSurvy software design
J2ME: Built on Java 2 Micro edition
CDC: works with CDC 1.1 and above (JSR)
MIDP: works with MIDP 2.0 or above
GPRS: transport technology
Each record is 2kb and costs INR 0.01 or
LKR 0.02 (USD 0.0002) i.e < USD 10
Handset/Month
Mobile phone around US$ 100
Tested on Nokia3110c, Motorola SLVR L7,
Gionee v6900. Amoi A636, Sony Ericsson
s302c
www.lirneasia.net

mHealthSurvey Certification Exercise
Exercise conducted 2 months after training and use of the mobile health software
Exercise Benchmark Sri Lanka India
Part I – installation and configuration (min) 12.00 10.75 17.48
Part II – submit up to 6 records (min) 20.00 10.80 27.26
Par III – standard operating procedures (points) 50.00 20.43 15.00
Outcome categorization
Certified trainers ( > 90 points) 02 of 15 Nil
Certified Users (90 ≥ points > 70) 13 of 15 04 of 23
Uncertified (points ≤ 70) --- 19 of 23
www.lirneasia.net
Sri Lanka India Benchmark
100.00
90.00
80.00
70.00
60.00
50.00
40.00
30.00
20.00
10.00
0.00
Part III
Part II
Part I
Country
Scores (Min=0;Max=100)
Younger (age 18 – 35) Sri
Lankan data entry
assistants with no health
training and no prior
experience were quick to
adapt to the technology
The relatively older (age 35 –
55) trained health workers in
India with 10 – 25 yr
experience had difficulty
adapting to the software but
have improved after some
intervention and additional
training
Average Country Scores

Data quality = Signal to Noise Ratio (SNR); i.e.
number records with errors/records submitted
www.lirneasia.net
Quality of the digitized data
The 23% noisy data in
India subsided to less
than 4% after
informing the
consequences of false
detections (SNR for
sub intervals: 0.18,
0.40. 0.31. 0.04, 0.07)
INDIA
SRI LANKA
Assistants in Sri
Lanka with no
formal health
training and no
affiliation to the
hospitals/clinics had
no incentive to
correct the 45%
errors (SNR for sub
intervals: 0.58,
0.30, 0.53, 0.57,
0.17) 1 Low quantities of data received from Health Sub Centers
2 Volume of records were better after including Primary Health Centers
3 Holiday effect: no records received
4 Learning curve getting medical officers to adopt to the new procedures of writing the diagnosis
5 Release of mHealthSurvey v1.3 with better predictive text

Timeliness = submitting the patient’s record the
same day as the patient visitation
www.lirneasia.net
Timeliness of data submission
Finding time to complete
the records without
disrupting current work
flow was a significant
barrier for real- time data
submission (sub interval
delay rates 0.28, 0.09,
0.21, 0.38, 0.44, 0.48,
0.68)
Data entry assistants
have no other role
besides digitizing
records but see delays
proportional to the
patient visitation counts
(sub interval delay
rates: 0.10, 0.27, 0.25,
0.36, 0.53, 0.21).
INDIA
SRI LANKA
1 Users with dysfunctional phones where sharing and were sending data on the
weekends or when friends phone was available for borrowing

Digitizing problems that affect the categorical data
SNOMED-CT
www.lirneasia.net
LOINC
UMLS

Fix the data collection shortcomings: noisy and off-time
3000
2000
1000
week
Need
solutions
here
3000
2000
1000
Week
www.lirneasia.net
02
03
01
04
05
06
07
08
09
10
11
12
13
9000
6000
3000
0
Week
Noisy Clear
Record Counts
01 02 03 04 05 06 07 08 09 10 11 12 13
8000
6000
4000
2000
0
Week
Off-Time Real-Time
Record Counts
01 02 03 04 05 06 07 08 09 10 11 12 13
0
Noisy Clear
Record Counts
01 02 03 04 05 06 07 08 09 10 11 12 13
0
Off-Time Real-Time
Record Count
From: 01-Sep-2009
To: 30-May-2010
Case Records:
220000+
From: 01-Jun-2009
To: 30-May-2010
Case Records:
81000+
Noisy
vs
Clean
data
Real-
Time
vs
Off-
Time
data

Observations of the data digitizing uncertainties
www.lirneasia.net
No observations
for India in this
quadrant → Data
submitted by
health workers
in India is
consistent
Diseases with
higher counts
but occurring
only in a single
location; hence
suspected of
possible mis-coding
by Sri
Lankan
assistants The likelihood of a measles
outbreak emerging only in a
single location without
spreading to other areas,
given that it is a viral
disease, is highly unlikely.
The assistant entering the
data had submitted data for
“Toxide vaccine” as
Tetanus.
These diseases
occurred only
once in one
location
Uniformity of geographic distribution of disease cases
( low: concentrated in a few locations, high: spread over)
Fever greater than 7 days
concentrated in February
and March of 2010, mainly
from a single location,
during the non rainy season
Historical accumulated case counts

Data digitization: Some Feedback
“Integrated Disease Surveillance Program Data Entry Operator and Data Manager fear they will lose their
jobs if mHealthSurvey and TCWI are introduced. At present these staff members receive phone calls from
all health facilities and enter the data in spreadsheets of tabulation of weekly aggregates.” - Senior Project
Officer, RTBI, India (19.08.10).
“Data digitizing nurses in India and assistants in Sri Lanka invest their own resources to
repair and replace ill-fated mobile phones.” - Field Coordinator, Rural Technology and Business
Incubator, India, consulted 18-December-2010 and Field Coordinator, Sarvodaya, Sri Lanka,
consulted (26.04.10).
“In the present day setup in Sri Lanka, most of the surveillance data comes from Inward
admissions and it is important that the data is collection is expanded to the Outpatient
Departments as in the case of this project.” - Wayamba Provincial Director of Health Services, Sri
Lanka, consulted (05.04.10).
“Sarvodaya Suwadana Center (primary health center) assistants in Sri Lanka have formed a social
network to keep each other informed of escalating health situations in their communities” - Field
Coordinator, Kurunegala District, Sri Lanka, consulted (06.10.10).
“For notifiable disease cases, digitizing the patient’s name and address is important for house
investigations.” - Village Health Nurse (Keelsevalipatty), workshop report, Sivaganga District, India,
consulted (01.10.10).
“It was easier for central officials in Chennai and New Delhi to monitor our individual statistics
and performance opposed to scanning through paper or aggregated for the same; therefore, we
are afraid to digitize data.” - Village Health Nurses (Nerkupai), www.Sivaganaga lirneasia.District, net
India, consulted
(29.09.10).

www.lirneasia.net
mHealth dala collection lessons
 Nurses sending data
 Near zero noise because impacts
their work
 No time to enter data patient care
and routine work comes first
 Under reporting to avoid extra work
 Improvise mHealthSurvey for
collection and reporting of other
 Older slow to learn but will catchup
 No prior experience beyond voice
 Resolve technical problems on their
own relative to PCs
 Replaced handsets on their own
 Outsourced data entry clerks
 No incentive because 1) lack of
knowledge 2) not direct impact
 Data entry is their only job
 No strings attached with reporting
quantity
 Nothing like that
 Young were quick to learn
 Knew all capabilities of mobile
 Resolve technical problems on their
own relative to PCs
 Replaced handsets on their own

Evaluation metric verticals and horizontals
www.lirneasia.net
 Three verticals – data
collection, event detection
and reporting
 Four layers – social, content,
application, Transport
 Arrows showing the
Interoperability between
layers and verticals
 Objectively assess by
calculating various
indicators: costs, efficiencies,
error rates, etc
 Subjectively assess through
interviews and simulations
Talk about
this

T-Cube Web Interface (TCWI) by Auton Lab
 AD Tree data structure
 Trained Bayesian Networks
 Fast response to queries
 Statistical estimations techniques
 Data visualization over temporal and
spatial dimensions
 Automated alerts
www.lirneasia.net

Interactive Astrophysics analytics
Bio-surveillance
Learning
Locomotion
Food safety
Nuclear
threat detection
Safety
of agriculture
Fleet prognostics
Slide 31 of 24 Copyright © 2009 by Artur Dubrawski
United Nations
CTBTO
Saving sea
turtles

Pre-Screening using Massive Tempotal Scan
www.lirneasia.
net

T-Cube Web Interface – Spatio – Temporal Presentation
www.lirneasia.
net

Overview of the T-Cube data structure and computations
www.lirneasia.net

Evaluation of TCWI
 Replication study :: Sri Lankan Weekly Epidemiological Return
(WER) reports published at www.epid.gov.lk notifiable disease
counts tabulated by District was semi synthesized by distributing
the weekly counts as daily counts taking day-of-week effect, gender
distribution, and age representations.
 Study the reliability and effectiveness :: significant events
detected by T-Cube is compared with the ground truth and also
weighed on the response actions or inaction
 Competency exercise :: injected fake data over a period of 5 days
and the subjects, unaware of the prefabricated events, were asked to
detect most significant events
 T-Cube Acceptance :: a questionnaire was designed based on the
Technology Assessment Methodology (TAM) and was subject to
TCWI users as well as health official associated with T-Cube who
make decisions on whether or not to take action
 Cost analysis :: compare the economic efficiencies and cost
effectiveness between present detection/analyses system and T-Cube
www.lirneasia.net

Replication study using synthesized WER data
 We took 2007 – 2009 Weekly Epidemiological Returns publicly
available data - http://www.epid.gov.lk/
 Synthesized the data to match that similar to the RTBP dimensions
by distributing the district weekly aggregates
- day-of-week visitation densities (M - F)
- female to male ration
- age-groups (0-5, 6-14, 15-20, 21-45, 46-65, above 65)
www.lirneasia.net

Replication study using Sri Lanka WER data 2007 - 2009
Food poison spike as
detected by spatial
scan around Feb
15,2007 in
Nuwara_Eliya, which
was reported as
outbreak by health
department.
In addition TCWI
detected spikes in
Kandy and Vauvniya
areas
Spatial scan is run
by 7 days windows
size.
www.lirneasia.net

Another Food poison
spike as detected by
spatial scan around
June 17,2009 in
Nuwara_Eliya, the
same location.
Spatial scan is run
by 7 days windows
size.
www.lirneasia.net

www.lirneasia.net
Dengue Fever
Seasonal and spatial
pattern
May 1,2007
Aug 30,2007
May 21,2008 April 15,2009 May 28,2009

Progression of Dengue Fever outbreak in April – June 2009
www.lirneasia.net
4/14
First day an
elevated global
score noted,
lead by region
Kandy
Spatial
Scan
global
Score
4/15
Situation in
Kandy
intensified,
together other
regions
4/24
Southern
Regions began
to see
increased
cases
5/28
Southern
region
continue to
see
progression,
while other
region
subsides

A recent outbreak of Acute Diarrheal Disease in Thirukostiyur area
www.lirneasia.net

Most frequently occurring wide spreading infectious disease outbreaks
These findings are from TCWI's spatial scan algorithms
Common Cold,
Sivaganga District –
India, 18 outbreak
episodes to date with over
23,188 cases.
Worm Infestation,
Sivaganga District –
India, 13 outbreak
episodes to date with over
1,236 cases.
Dysentery, Sivaganga
District – India, 5
outbreak episodes to date
with over 1,541 cases.
Common cold is the most popular but gastrointestinal infectious are, relatively, the most visible
www.lirneasia.net
Cough, Kurnegala
District – Sri Lanka, 11
Respiratory Tract
Infection, Kurnegala
Tonsilitis, Kurnegala
with over 5.086 cases
Respiratory infectious diseases, a correlated with environmental factors, are the most common

Trends in selected Chronic disease
These findings are from TCWI's statistical estimation and pivot table analysis methods
Hypertension (High Blood Pressure) has a linearly increasing trend over the one year period in both countries with Females and
Males over 45 years of age showing to be the most vulnerable. The dtrend in India shows an unusual increase between March and
May 2010; while the reported cases are consistent throughout the year in Sri Lanka.
Diabetes-Mellitus has a linearly increasing trend over the one year period in both countries with Indians over 40 years of age and
Sri Lankan over 45 years of age to be the most vulnerable groups.
Given that the Male to Female ratios, approximately, in Tamil Nadu, India and Kurunegala, Sri Lanka are both 1 : 1;
statistics to date show females to be more susceptible to the above mentioned life style diseases.
www.lirneasia.net

Trends in selected Chronic disease
These findings are from TCWI's statistical estimation and pivot table analysis methods
Arthritis and Rheumatoid-Arthritis has a linearly stagnate trend over the one year period in both countries with Males over 45
years of age and Females over 35 years of age to be the most susceptible in India; similarly Males over 45 and Females over 31
years of age to be the most vulnerable groups.
Asthma has a linearly decreasing trend over the one year period in both countries; the dtrend shows the counts to increase during
the rainy season, India: Sept'09-Jan'10 and Sri Lanka: Nov '09-Jan '10. In India, only males over 45 years of age are affected but
females in all age groups are affected. Both Male and Female over 31 years of age are in Sri Lanka are equally vulnerable.
Given that the Male to Female ratios, approximately, in Tamil Nadu, India and Kurunegala, Sri Lanka are both 1 : 1;
statistics to date show females to be more susceptible to the above mentioned life style diseases.
www.lirneasia.net

TCWI Competency Assessments with Injected Synthetic data
Used “Epigrass” to generate synthetic data with a SEIR model
Susceptible Exposed Infected Recovered
www.lirneasia.net
With a Network Flow
Injected 3 sets of data
1) Notifiable disease :: Dysentery
2) Other-Communicable disease :: ADD
3) Syndrome :: Fever, Pain, RTI

www.lirneasia.net
TCWI Simulation results

TCWI Actual Usage by Health Departments
Day-of-Week TCWI is Utilized Time spent each time
3 of 14 potential users spend less than 30 minutes
each time once a week on detection analysis;
remaining 9 did claim to be too busy to use TCWI
Day-of-Week TCWI is Utilized Time spent each time
75% of the 9 Sri Lankan users
spend more than 30 minutes each
time every day of the week on
detection analysis.
www.lirneasia.net

www.lirneasia.net
TCWI Preferred functions
India health officials'
primary preferences
are screening for
fever, other-communicable
diseases, and using
the pivot table.
Sri Lanka health
officials' primary
preferences are
screening the notifiable,
fever, and other-communicable
diseases.

TCWI Technology Assessment Model scores
Technology Acceptance Model was applied to obtain these results on perceived ease of use, perceived
usefulness, behavioral interaction, attitude towards using, and psychological attachment
Users attitude towards using TCWI
INDIAN
This part of the
questionnaire was not
completed.
SRI LANKN
The personal feeling is such
that, all things considered,
TCWI in the job is - quite a
good idea, slightly beneficial,
quite a wise idea, and slightly
positive
www.lirneasia.net
STRONGLY AGREE
AGREE
IMPARTIAL
DISAGREE
STRONGLY DISAGREE
The TAM questionnaire was conducted with 14 Indian and 09 Sri Lankan users (health officials and health workers)

T-Cube Web Interface: Some Feedback
“We can use this rich and comprehensive dataset and analysis tools for our annual planning,
now our planning relies on professional perception and not necessary data.”
- Deputy Director Planning, Kurunegala District, Sri Lanka, Consulted (06.10.09)
“Epidemiologists want TCWI to facilitate the old ways of monitoring outbreaks based on thresholds
opposed to statistical significance. For example, a single case of Malaria is regarded as an
outbreak in India, which requires response actions.”
- Deputy Director of Health Services, Sivaganga District, India, Consulted (19.12.09).
“It is important to monitor escalating fever cases, notifiable disease cases, and common clusters
of symptoms.”
- Regional Epidemiologist, Kurunegala District, Sri Lanka, consulted (19.12.09).
“Medical Officers, Nurses, Health Educators, etc, who are interested in learning of outbreaks
see the benefit and are happy with TCWI detection analysis methods but the staff at the Integrated
Disease Surveillance Program are not ready to accept change and want to stick to the
traditional system unless state or national level Authorities mandate it.”
- Senior Project Officer, RTBI, India, consulted (19.08.10).
www.lirneasia.net

T-Cube Web Interface: Some Feedback
“Pharmacists’ perceptions are such that a separate computer should
be given for detection analysis and they do not want to share their computers, which are used
for medicine and birth information.”
- Senior Project Officer, RTBI, India, Consulted (08.07.2010).
“RTBP’s real-time biosurveillance capabilities will enhance the present day passive or non-active
passive surveillance to an active surveillance system.”
- Wayamba Provincial Director of Health Services, consulted (07.07.10).
“All cases can be viewed in TCWI in real-time for detecting outbreaks swiftly, which other-wise
would take several days before the hospitals/clinics send the notification paper forms, by
which time the patient may be dead or discharged.”
- Public Health Inspector, Wariyapola, Sri Lanka, consulted (26.04.10).
www.lirneasia.net

www.lirneasia.net
TCWI some early lessons
 Health departments unfamiliar with
statistical estimation methods for
detection analysis
 Requirements were to observe ::
Integrated Disease Surveillance
Program (IDSP) P/S disease list
 No incentives to use and usage is
almost nil
 Health departments unfamiliar with
statistical estimation methods for
detection analysis
 Requirements were to observe :: set
of Notifiable disease and Weekly
Epidemiological Returns report
 Have incentives and using due to
known delays in present system
Studying the TCWI Acceptance through TAM
RESULTS TO BE ANNOUNCED

Existing methods of receiving health alerts
Survey responses from 28 health workers from June 2009 to March 2010
At present health
workers learn of
adverse health
events through
MEDIA and WORD-OF-
MOUTH, in
some cases from
PEERS
No formal
Government
method for
sharing health risk
information with
health workers
Survey responses from 15 health workers from June 2009 to March 2010
www.lirneasia.net

How do we integrate the subscribers and publishers?
How do we deliver early warnings in local language?
How do we use existing market available technologies?
How do we disseminate alerts over multiple channels?
How do we inter-operate between incompatible systems?
How do we effectively communicate the optimal content?
How do we address the communication strategy?
How do we accommodate upstream-downstream alerting?
www.lirneasia.net
Problem to solve

Common Alerting Protocol Overview
□ All you want to know in “CAP Cookbook”
□ XML Schema and Document Object Model
□ Interoperable Emergency Communication Standard
□ Specifically geared for Communicating Complete
Alerts
□ Capability for Digital encryption and signature X.509
□ Developed by OASIS for “all-hazards”
www.lirneasia.net
communication
□ Adopted by ITU-T for Recommendations X.1303
□ Incubated by W3C Emergency Information
Interoperability Framework
□ Used by USA, USGS, WMO, Gov of CA
□ Can be used as a guide for structuring alerts

CAP Document Object Model
Bold elements are
mandatory
Bold elements in <Alert>
segment are qualifiers
Others elements are
optional
Profile may specify other
mandatory elements from
optional list
Single <Alert> segment
Multiple <Info> segments
inside <Alert> segment
Multiple <Area> and
<Resource> segments
inside a <info> segment
(*) indicates multiple
instances are permitted
Alert
identifier
sender
Sent
Status
msgType
Source
Scope
Restriction
Address
Code (handling code)
Note
References (Ref ID)
Incidents (Incident ID)
Info
Language
Category
Event*
responseType
Urgency
Severity
Certainty
Audience
eventCode*
Effective (datetime)
Onset (datetime)
Expires (datetime)
senderName
Headline
Description
Instruction
Web (InformationURL)
Contact (contact details)
Parameter*
Resource
resourceDesc
mimeType
Size
URI
derefURI
digest
Area
areaDesc
Polygon*
Circle*
Geocode*
Altitude
Ceiling
www.lirneasia.net
*
*
*

Predefined values
CAP Element Predefined Values
<Status> Actual, Exercise, System, Test, Draft
<msgType> Alert, Update, Cancel, Ack, Error
<Scope> Public, Restricted, Private
<Language> en, fr, si, tm, …| codes ISO 639-1
<Category> Geo, Met, Safety, Security, Rescue, Fire,
Health, Env, Transport, Infra, CNRNE, Other
<responseType> Shelter, Evacuate, Prepare, Execute, Monitor,
www.lirneasia.net
Assess, None
<Urgency> Immediate, Expected, Future, Past, unknown
<Severity> Extreme, Sever, Moderate, Minor, Unknown
<Certainty> Observed, Likely, Possible, Unlikely, Unknown
<Area> b-WGS 84

Prioritizing Messages in CAP
Priority <urgency> <severity> <certainty>
Urgent Immediate Extreme Observed
High Expected Severe Observed
Medium Expected Moderate Observed
Low Expected Unknown Likely
www.lirneasia.net
Select value
Auto populate

Sahana Alerting Broker (SABRO) Subsytems
❏Inputs can be manual or automated
❏Message creation & validation uses CAP v1.1 and EDXL 1.0 data standards
❏Access control (permissions) and user rules are governed through the
Organization Resource Manager (ORM)
❏Direct alerts are sent to end user recipients and Cascade alerts are a system-to-system
communication determined by the message distribution method
❏Long-text, Short-text, and Voice-text are different forms of full CAP message for
the ease of message delivery to various end-user terminal devices
❏Message acknowledgement logs the recipient messages confirming receipt
www.lirneasia.net

www.lirneasia.net
Overview of Sahana

Sahana Messaging/Alerting CAP/EDXL Broker by Respere
Single input multiple output
engine; channeled through
multiple technologies
Manage publisher /subscribers
and SOP
Adopt PHIN Communication
and Alerting Guidelines for
EDXL/CAP
Relating the template editor with
the SMS/Email Messaging
module
Do direct and cascading alert
from a regional jurisdictional
prospective
Designing short, long, and voice
text messages
Addressing in multi languages
www.lirneasia.net

Example of template for SMS alert
<headline> : <status>
<msgType> for <areaDesc> area with
<priority> priority <event> issued by
<senderName>.
Msg: <identifier> sent on <sent>
Desc: <description>
More details
Web: <web>
Call: <contact>
www.lirneasia.net

Examaple of Automated Standard Message
Escalating mumps in Kurunegala district for
Wariyapola-PHI area
A low priority notifiable disease outbreak issued
by Dr Hemachandra.
Msg : nwpdhs-1281246871 sent on 2010-08-08
11:08:57.
Desc : 2 cases of Mumps for 15-20 age group and
all genders were reported in Munamaldeniya.
More Details
Web www.scdmc.lk
Call 2395521
www.lirneasia.net

www.lirneasia.net
CAP XML → XSL → delivery method
Single Input Multiple Output Mass Messaging;
towards a publisher subscriber model

CAP short/long text Message delivery methods
www.lirneasia.net
Single Input Multiple Output Mass Messaging;
towards a publisher subscriber model
Community
Suwadana Health Centers
Government
Regional Epidemiology and
Medical Officer of Health
departments
Government
Regional Epidemiology and
Medical Officer of Health
departments

Steps for setting up a CAP Profile
- determining the policy and procedures -
Audience <Scope>
Alert First Responders only (i.e. closed user group)
Example: police, health workers, civil society, public servants
Alert Public (entire population)
Combination of First Responders and Public
step 1: alert First-Responders to give them time to prepare
Step 2: warn public
Geographical Descriptions <Area>
Country wide
Province or State
District
Other – Geocodes or GPS polygons
National <Languages>
English only or Chinese only or Malay only
English, Hindi, Chinese, and Malay
Communications Technology?
Mobile phones – SMS, Cell Broadcast, Email, Applet
TV – Text, Audio, Visual
AM/FM Radio - Text, Audio
VHF/UHF Radio - Audio
Internet – HTTP, Email, Webservices
Audience
Geography
Language
Technology
www.lirneasia.net

Downstream messaging structure - INDIA
IDSP PHC
Message Creator – IDSP staff member Message Creator – PHC staff member
Message Issuer - DE Message Issuer - MO
www.lirneasia.net
Action
alert
Mode of
delivery *
1 SMS
2 Short Email
3 Long Email
Awareness
message
Awareness
Message
Mode of
delivery*
1 SMS
2 Short Email
3 Long Email
Action
Alert
Recipients
BMO
MO
HI
SHN
VHN
Recipients
BMO
MO
HI
SHN
VHN
Other health
officials at IDSP
Recipients
BMO
MO
HI
SHN
VHN
Other health
officials at IDSP
Recipients
MO
SHN
HI
VHN
Event Detection

Downstream messaging structure – SRI LANKA
www.lirneasia.net

Messaging exercises with Sahana Alerting Broker
3 users in India and 5 users in Sri Lanka participated in the message dissemination exercises. Each user was presented with
four varying scenarios in relation to escalating cases of diseases identified through TCWI and other sources.
www.lirneasia.net
Percentage of
messages sent on-time
(benchmark time-to-
completion was 5
minutes)
The security policy of the
software, by default, is set to
expire the session after 5
minutes to prevent
unauthorized use, which
forced the user to restart.
Accuracy of creating
the messages with
populating the
common alerting
protocol attributes of
the software
Templates with pre-populated
values and a clear structure
helped the users with creating
the messages
Correctly selecting the
appropriate delivery
channels targeting the
intended recipients
It was easier to comprehend
issuing of alerts but not the
the same with issuing
situational awareness
messages such as the weekly
top 5 diseases reports.
30%
35%
10%
50%
65%
10%
35%
10% 55%
INDIA Exercises were incomplete; no results to discuss

CAP SMS Alert/Situ-aware comprehension exercises
3 2 1 0Error
100
50
0
Get more info?
Response actions?
Msg priority?
Msg Identifier?
Who issued?
Event?
Affected locations?
Msg received via?
Points
Assessment design
 Participants receive 4 SMS text with varying values of the CAP attributes
 India = 23 and Sri Lanka = 19 health workers participated in the exercise
www.lirneasia.net
No of subjects gave answer
Questions
3 2 1 0Error
100
50
0
Response actions?
Msg priority?
Msg Identifier?
Who issued?
Event?
Affected locations?
Msg received via?
Get more info?
Points
No of subjects gave answer
Questions
Outcomes
 Everyone did quite well in the exercises except for 1 or 2 exceptional cases
 Both India and Sri Lanka having trouble with msg-identifier; could be because msg-identifier
getting truncated by the 160 char SMS constraint
 Recommendation :: put msg- identifier in subject header (but may cutoff rest due to 160 char
SMS); use the term “reference number” instead or both

CAP Short-text message over SMS, 84 responses for 4 different messages
CAP Short-text message over SMS, 76 responses for 4 different messages
Summarize message:
17%
Recommendations:
Other Delivery
www.lirneasia.net
Credibility, Persuasiveness, Validity
Counts, disease,
locations, response
32%
Message Authenticity
14% 8%
18%
28%
Call MOH
Call Issuer
Refer In-ternet
Other
No Ans
45%
Verify Authenticity?
18%
14%
8% 14%
Call MOH
Call Issuer
Refer In-ternet
Other
No Ans
72%
7%
11%
11%
Disease
locations
sender
Disease
locations
sender re-sponse
Other
No Ans
7%
7%
70%
Mention
patient de-tails
Adequate
Other
No Ans
25%
3%
14%
9%
49%
Email
Email &
Web
Voice
Other
No Ans
Expected response
Use Web/Contact localize Sender Name Email, Web, Voice

www.lirneasia.net
Some assumptions
❒ The health facilities serve, on average, 100 patients a day; hence, the
digitizing capacity of the a single data-entry assistant or single nurse is
constrained by this number
❒ Calculations for the existing programs are based on least case and the costs
for the RTBP are based on the worst case; e.g. accounted for annual training
in RTBP but not in existing system
❒ Indian Primary Health Center (PHC) is both a health facility and a health
department; Sri Lanka Hospitals/Clinics and MOH are separate
❒ Reporting is confined to PS List and Morbidity (India); H544, H399, H499,
WER (Sri Lanka)
❒ No software licensing fees included such as for the T-Cube Web Interface
but that is less than 5% of TCO
❒ Considered only total cost of ownership for the economic analysis, did not
incorporate impact related costs such as quality adjusted life years,
productivity improved, lives saved, etc

TCO macro-costs and the marginal differences
System delivery, system support, and data center costs are < 7% of overall cost; hence the focus of the economic
analysis is on the bulk: health facilities, health departments, and health workers
Explanation of marginal difference of RTBP macro cost > 20% than existing system
System delivery :: unable to get actual program design, development, and implementation cost,
most likely funded by INGO, however, the per district monthly cost is very small.
System Admin/Support :: no established budget, each department spends money for repairs as and
www.lirneasia.net
when needed. RTBP accounts for it.
Data Center :: India – DPH&PM system is one component of several managed by the National
Information Center, in comparison to decentralizing the data centers to be managed at districts
Health Facilities :: major portion of the cost is the new human resource bundled with technology for
health record digitization

Comparison of expenses in relation to the data collection, event detection, and alerting components
Comparison of expenses in relation to the health facility, health department, and health workers
Digitizing data at the point of care
removes the bulk health department
expenses of labor intensive data
aggregation and consolidation.
Worst case scenario of bundling frontline
data digitizing with new resource person
increases the health facility investment.
Health facility cost increase < health
department money saved; India: 61% <
86%, Sri Lanka: 72% < 87%
www.lirneasia.net
Total Cost of Ownership
India and Sri Lanka invest very little or no
resources on real-time event detection
and alerting, ~ 88% in data collections
RTBP can reduce TCO > 35%,
moreover, increase timeliness, and
introduce rapid detection and alerting
Existing trend analysis is for long term
planning only; dual data-entry at
departments.
Comparison of Entity Costs in India and Sri Lanka
- existing paper-based vs introduced RTBP
[ Existing (IN) = present system in India (Integrated Disease Surveillance Program); Existing (LK) = present system in Sri Lanka (Disease
Surveillance and Notification Program); RTBP (IN), RTBP (LK) = Real-Time Biosurveillance Program in India and Sri Lanka, respectively]

Incremental Cost Effectiveness Ratios (ICER)
Going from Existing system to RTBP
Data collection – cheaper, more data, more attributes, data
available same day (No further analysis required)
Event Detection – cheaper in Sri Lanka, almost same in India,
real-time/automated event detection, reports at the touch of a
button, globally accessible, no humans needed to feed data,
better visualization and analystcs (no further analysis required)
Alerting – new investment and new concept but health
workers will be better informed of the regional health status
for preparedness and early response (needs further analysis)
Expensive &
effective→
needs further
analysis
Existing RTBP Existing RTBP
www.lirneasia.net
Relative
RTBP
Cost
Relative
RTBP
benefits
Expensive &
ineffective→
resource
unacceptable
Existing
Cheaper &
ineffective→
needs further
analysis
Cheaper &
effective→ no
further
analysis

India/Sri Lanka – almost
zero resources on detection
and mitigation
www.lirneasia.net
Data center
Health facility
10,000.00
5,000.00
ICT System delivery
System Admin/Support
Health worker
Health department
0.00
Alerting (USD)
Detection (USD)
Collection (USD)
Internal Entities
Cost (USD) / District / Month
20,000.00
10,000.00
NIC IDSP System delivery
Data center
Health facility
Health worker
Health department
0.00
Alerting / situ-aware
Event Detection
Date Collection
Internal Entities
20,000.00
10,000.00
ICD/WER System delivery
Data center
Health facility
Health worker
Health department
0.00
Alerting / Situ-aware
Event Detection
Data Collection
Internal Entities
RTBP can fix the imbalance
Ideally health facilities
should be powered for data
collection, health
departments for detection
and alerting, with health
workers fully on response
RTBP
Sri Lanka – health
departments consume bulk
of the resources

Economic Efficiencies :: India (annual cost per district)
Present IDSP RTBP
www.lirneasia.net
 Report LIMITED morbidity/PS-list
disease: 77
US$ 3,650.00 / disease
 Report LIMITED morbidity/PS-List avg
cases::600
470.00 / case
 Program for District pop: 1150753
US$ 0.24 / inhabitant
 Report ALL diseases: 117
US$ 800.00 / disease
 Report ALL cases: 6175 avg per month
15.00 / case
 Program for District pop:1150753
Data Collection
Event Detection
 Monitor LIMITED PS-list disease: 25
 Monitor ALL disease: 117
Alerting
 Disseminate LIMITED PS List disease: 25
650.00 / disease
 Disseminate Communicable disease: 43

Economic Efficiencies :: Sri Lanka (annual cost per district)
Present SNS RTBP
www.lirneasia.net
 Report LIMITED Notifiable disease: 25
US$ 10,800.00 / disease
 Report LIMITED WER cases::70
321.00 / case
 Report ALL diseases: 179
 Report ALL cases: 22835 avg per month
4.50 / case
 Program for District pop:1550000
Data Collection
Event Detection
 Monitor LIMITED Notifiable disease: 25
 Monitor ALL disease: 179
US$ 54.00 / disease
Alerting
 Disseminate LIMITED PS List disease: 25
285.00 / disease
 Disseminate Communicable disease: 49

Sensitivity was applied to only parameters II – IV that affect individual components differently. Parameters I and V
are direct proportionality that affects the entire cost and not individual components.
www.lirneasia.net
One-way Sensitivity Analysis
Parameter: Typical Value
India Sri Lanka
(I) Health Districts per State 32 2
(II) Health Facilities per District 47 44
(III) Health Departments per District 48 21
(IV) Health Workers per Facility 10 21
(V) Currency exchange rate to US$ 45.00 110.00
Mobilizing a Health Worker
in the existing system with
the infrastructure: paper
forms,books, archiving
cupboards, workspace and
operational expenses: travel
costs is much greater than
using RTBP technology:
mobile phones and simple
server with affordable
network infrastructure
RTBP technology reduces
labor, duplicate data entry,
consolidation/aggregation,
and manual analyses
required by health
departments
Digitizing of all patient
records in RTBP system may
require introducing a new
staff member at the health
facility

www.lirneasia.net
Conclusions
 mhealthSurvey is a worthy candidate for patient disease/syndrome digitization;
however, must be robust to minimize the noise and delays; need a better GUI if
Medical Officers are to enter high volume real-time data opposed to a data entry
clerk
 Need a complete and comprehensive standardized disease syndrome ONTOLOGY
perhaps a hybrid of SNOMED-CT and LOINC
 T-Cube false alarm rates must be minimized through the iterative enhancement
and machine learning
 Sahana CAP Broker SMS, Email, and Web messaging has proven to be a winner
for real-time adverse health event information dissemination but need Voice as well
 Although the value is seen in T-Cube Web Interface and CAP/EDXL Messaging
The policies must be reformed to go beyond the century old paper based system to
using ICT based event detection and alerting/situational-awareness
 There should not be any institutional fears arising from the cost reductions instigated
by the introduction of ICTs (e.g. RTBP) as is will still require the same work force
 Before the cost benefits can take affect the laws and regulations must be changed
to remove the paper and the storage cupboards that are government mandates

Conclusions
RTBP costs are less, benefits are better, and efficiency gains
are higher than the existing disease surveillance and
notification systems
The laws and regulations must be changed to replace the
legal forms and registers with electronic health records.
Health record security, privacy, and unique identifiers must
be addressed before national implementation
Be ready to accept change; especially the paradigm of
comprehensive disease/syndromic active surveillance
without paper
There is a severe cost associated with false alarms (false
positives) and missed alarms (true negatives), which needs
further study and rectifying
www.lirneasia.net

Project Partners:
International Development Research Center, Canada - www.idrc.ca
Ministry of Health and Nutrition, Wayamba Privince, Sri Lanka - http://www.health.gov.lk/
Health and Family Welfare Department, Tamil Nadu, India - http://www.tnhealth.org/dphpm.htm
LIRNEasia, Sri Lanka - www.lirneasia.net
Lanka Jathika Sarovodaya Shramadana Society, Sri Lanka - www.sarvodaya.org
IIT-Madras's Rural Technology and Business Incubator, India - www.rtbi.in
Carnegie Mellon University Auton Lab, USA - www.autonlab.org
University of Alberta, Canada - www.ualberta.ca
National Center for Biological Sciences, India - www.ncbs.in
Respere Lanka (Pvt) Limited, Sri Lanka - www.respere.com
Listed in alphabetical order

www.lirneasia.net
References related to RTBP
[1] S. Prashant and N. Waidyanatha (2010). User requirements towards a biosurveillance program, Kass-
Hout, T. & Zhang, X. (Eds.). Biosurveillance: Methods and Case Studies. Boca Raton, FL: Taylor &
Francis, Chapter 13, pp .240-263.
[2] Kannan T., Sheebha R., Vincy A., and Nuwan Waidyanatha (2010). Robustness of the
mHealthSurvey midlet for Real-Time Biosurveillance, Proceedings of the 4th IEEE International
Symposium on Medical Informatics and Communication Technology (ISMICT '10), Taipei, Taiwan.
[3] M. Sabhnani, A. Moore, and A. Dubrawski (2007). Rapid processing of ad-hoc queries against large
sets of time series. Advances in Disease Surveillance, Advances in Disease Surveillance, Vol 2, 2007.
[4] S. Ray, A. Michalska, M. Sabhnani, A. Dubrawski, M. Baysek, L. Chen, J, and Ostlund (2008). T-Cube
Web Interface: A Tool for Immediate Visualization, Interactive Manipulation and Analysis of
Large Sets of Multivariate Time Series, AMIA Annual Symposium, 2008:1106, Washington, DC, 2008.
[5] A. Dubrawski, M. Sabhnani, S. Ray, J. Roure, and M. Baysek (2007). T-Cube as an Enabling
Technology in Surveillance Applications. Advances in Disease Surveillance 4:6, 2007.
[6] G. Gow and N. Waidyanatha (2010). Using Common Alerting Protocol To Support A Real-Time
Biosurveillance Program In Sri Lanka And India, Kass-Hout, T. & Zhang, X. (Eds.). Biosurveillance:
Methods and Case Studies. Boca Raton, FL: Taylor & Francis, Chapter 14, pp 268-288.
[7] G. Gow, Vincy. P.., and N. Waidyanatha (2010). Using mobile phones in a Real-Time
Biosurveillance Program: Lessons from the frontlines in Sri Lanka and India. Proceedings of the 2010
IEEE International Symposium on Technology and Society (ISTAS '10), New South Wales, Australia.
[8] Ganesan M., S. Prashant, Janakiraman N., and N. Waidayanatha (2010), Real-time Biosurviellance
Program: Field Experiences from Tamil Nadu, India. IASSH conference paper. Varanasi, Uttarpradesh,
India.
[9] A. Dubrawski (2009). Detection of Events in Multiple Streams of Surveillance Data. In Infectious
Disease Informatics: Public Health and Biodefense, Eds. C. Castillo-Chavez, H. Chen, W. Lober, M.
Thurmond, and D. Zeng. Springer-Verlag (in press).
[10] D. Neill, and G. Cooper (2009). A Multivariate Bayesian Scan Statistic for Early Event Detection
and Characterization. Machine Learning (in press).
[11] M. Wagner (2008). Methods for testing Biosurveillance systems, Handbook of Biosurveillance (eds.
Wagner, M., Moore, M., and Aryel, R.), pp 507-515, Elsevier academic press.
Go To our Blog
Search with key terms:
real-time
biosurveillance,
mhealth, outbreak
detection, common
alerting protocol,
alerting, situational
awareness,
mHealthSurvey, T-cube,
Sahana
Messaging Module

Real-Time Biosurveillance Program Pilot - India & Sri Lanka

More Related Content

Similar to Real-Time Biosurveillance Program Pilot - India & Sri Lanka (20)

More from Nuwan Waidyanatha (20)

Recently uploaded (20)

Real-Time Biosurveillance Program Pilot - India & Sri Lanka