Transcript Gadgets for Good How Computer Innovation Can Help

Gadgets for Good
How Computer Innovation
Can Help Save Lives
in Low-Income Countries.
Neal Lesh
Harvard School of Public Health
~40 million HIV infected people
~3 million died in 2004
~9 years on ave. to live w/o treatment ~6 million need treatment
>75% unaware of status
~700,000 receiving treatment
ARVs
Samuel Morin, 2001, Haiti
One year later,
after treatment
Clinical staff using Partners in
Health’s EMR in Belladere, Haiti.
Roadmap
 me
 the world
 international public health
 computers
 me
My Background
• Computer science experience:
– 1991-1997: PhD in A.I. at U. Washington
– 1997-1998: postdoc at U. Rochester
– 1998-2004: research at MERL
• Areas of work:
– planning
– story sharing
– collaboration
– optimization
– data mining
– engagement
– indoor
navigation
– inference
intention
– probabilistic
reasoning
– information
visualization
– intelligent
tutoring
– data
exploration
Currently
• Full-time student, masters of public health (MPH)
– taking classing, field trip to India, starting some
research projects.
• Goals for this talk:
– Give you flavor of the field
– Generate excitement
– Get invited back in a couple years
Warning!
• Many approximations:
“There is a tendency for all knowledge, like all
ignorance, to deviate from the truth in an
opportunistic direction.”—Gunnar Myrdal.
• Neglecting lots, e.g.
– disadvantaged people in rich countries
• Glossing over a lot of complexity
• Assuming you know about what I did 1 year ago
How are we doing?
~six billion people
World Population Growth
Population and year
Time to add a billion
1 billion in 1804
2 billion in 1927
3 billion in 1960
4 billion in 1974
1,001,804 years
123 years
33 years
14 years
5 billion in 1987
6 billion in 1999
7 billion in 2012
13 years
12 years
13 years
8 billion in 2026
8.9 billion in 2050
14 years
26+ years
How are we doing?
How are we doing?
one billion people
in rich countries
five billion people in
middle- or low-income countries
Poverty as a Risk Factor
for surviving the Titanic.
70
% survived
60
50
40
30
20
10
0
1st
2nd
class of service
3rd
Poverty as a Risk Factor
for dying young.
Malawi
U.S.
Life expectancy
at birth
38 yrs.
77 yrs.
Prob. of dying before 5
years old.
18.3%
.8%
Prob. of dying before
40 year old.
49.8%
13% die before
60 yr.
HIV rate among 18-49
year olds (2001)
15%
.6%
GDP per capita
$585
$35,991
Roadmap
me
the world
 international public health
 computers
 me
Unit of measurement
• Need to quantify population health
– measure success
– allocate resources
• Measure health by counting deaths?
Canada
Deaths per 1000
per year (2003 est.)
7.61
Mexico
4.97
(answer: Mexicans are younger than Canadians)
Life Years Lost
• Select a target/ideal length of life.
– e.g., 80 years for men, 82 for women
• For each death, calculate life years (LY) lost
relative to target length.
– E.g. death of a 40 year old woman =
82 – 40 = 42 LY lost
How many Life Years lost?
• Tsunami:
deaths X LY per death (my guess) =
300,000 x 65 =
19,500,000 LY lost
• Malawi:
population X death rate X LY per death =
12,000,000 X .024 X 42 =
12,096,000 LY lost
• Sub-Saharan Africa:
650,000,000 X .018 X 34 =
397,800,000 LY lost =
20 tsunami’s worth of LY lost per year
DALYs: Disability Adjusted LY
• Assign weights to
health states:
– E.g. “Give 1000 people
a year of healthy life or
2000 people a year of
paralyzed life?”
• Assign weights years
– E.g. 25th year worth
more than 5th or 65th
• Discount future years
– E.g. 3% per year
cause of lost DALYs
1
Lower respiratory
infection
6.4%
2
Perinatal conditions
6.2%
3
HIV/AIDS
6.1%
4
Unipolar depression 4.4%
5
Diarrhoea
4.2%
6
Ischaemic heart
3.8%
7
Cerebrovascular
3.1%
8
Road traffic
2.8%
9
Malaria
2.7%
10 Tuberculosis
2.4%
What can we do?
Phys/Human
Capital
Income
E.g. being pushed
into poverty by
medical expenses
Demography
and Health
E.g. hard to learn
when ill, or if
working because
parent is ill.
Reducing Child Mortality
HIV Prevention
Roadmap
me
the world
international public health
 computers
 me
Information & Communication
• Had another revolution in the
last 10-15 years:
– ease of communication
– availability of information
– tracking of objects
• Many opportunities to
address fatal information
deficits in healthcare.
But...
Information Kiosk
Less than $5 on her
healthcare, annually
Information Deficits
for Medication
• Tele-medicine
• Electronic medical
records (EMR)
• Decision support
• Intelligent tutoring
• Sensor networks
• Data mining and
visualization
• Connectivity for
low-income regions
• What’s in stock, expirations
• Healthcare workers
– medical expertise
– patient’s medical history
• Population/policy
– Needs assessment
– What’s working
• Individual
– When to seek care
Tele-health
• Addresses information deficits due to
– unfortunate distribution of medical expertise
– burden of travel
• Many options
–
–
–
–
doctor to patient, never meet
doctor to patient, meet occasionally
doctor to doctor
doctor to data repository (HealthNet)
• Technical challenges
– sensors for health data
– max. use of bandwidth
– user interface
Electronic Medical Records
• Info. management in med. care:
–
–
–
–
patient history at point-of-service
drug inventory, and prediction
decision support
monitoring and evaluation
• Challenges for computerization:
– expense
– electricity & connectivity
– expertise
Nurses in India, using
EMR by Dimagi and AIIMS.
Ca:sh
(Community Access to Sustainable Health)
• Handhelds for nurses
• Targets antenatal care,
immunization, disease
management
• 80,000 records since
February 2002
• 25¢ per patient per year
• Now using desktops &
car batteries in clinics.
• By Dimagi, AIIMS
• Encode standard protocols
to guide health workers
• Working on HIV protocols
• First target: filter out easy
“no change needed” cases
• Information periodically
uploaded
• Led by Marc Mitchell,
Hilarie Cranmer
Symptoms
fever 
□
RR > 40/50 or
chest indrawing □
diarrhea □
abd. pain □
rash □
next 
Need Research?
"The task before us is very urgent, so we
must slow down.”
Analogy: 10/90 gap in medical research
Behavior Change
• Information deficits in
caretakers of children:
– keep children away from smoke
– don’t withhold food from children
w/ diarrhea
– don’t rub dirt into umbilical cord
• Possible tools:
– interactive tutoring/testing
– games, animation
– virtual reality
Cost Effectiveness
• Behavior change system
– laptops, PDA, phones,
projectors, VR goggles, etc.
– operated by one person
• Cost
– $1000 per year for equipment
– $4000 per year operational
• Reach
– present to 10 people per day
– 200 presentations saves a
child’s life
• Impact
lower
child
mortality
– $333
per
life
– ~$10 per DALY
fertility
–reduced
World Bank
says
$150 per DALY is
bettercost
health
& wealth
effective
Passive Surveillance
Crisis Mapping
• Field personnel register location of
– physical resources (e.g., medicine)
– activities (NGO’s)
– situations (people, disease)
• Upload to GIS system to improve
– coordination of responders
– cooperation between NGO’s
“It's such an obvious idea that
no one has done it. Go figure.”
Related Challenges
• Predicting path of
fleeing refugees
• Population counting
for refugee camps
Connectivity
• Vehicle-mounted hubs (Pentland)
• Boosting 802.11b (Brewer,
Pentland)
– many hardware/power issues
– unconventional networking
– specialized protocols
• DVDs by Postal service (Wang)
Parting thoughts
the
answer
• Easy pickings for exciting ideas
• Must work with people in field
• Funding etc. a challenge
• My next years: visit many sites
and field-test variety of ideas.
Inspiration
• “We’re going to be a millionaire of a different sort.
We’re going to try to affect the lives of a million
people.” - Vikram Kumar, CEO of Dimagi.
• The new abolitionist: someone working to
eliminate extreme poverty this century.
Thanks!
To keep in touch, email me at
[email protected]
Age distribution
Surveillance
• Def: ongoing & standardized data collection
• Crucial for:
– Resource allocation
– Evaluation
– Outbreak detection
• Currently inadequate:
– Often rely on studies & models
– Push for “evidence-based medicine”
Road traffic safety
Road traffic injuries
expected to move to 3rd
leading cause of DALY’s
by 2020.
Medical Records
& Decision Support
• Many of the world’s poor:
– never see physician
– not reached by standard treatment
protocols, e.g., case management
for diarrhea or measles
– have no continuity of care
• Computerization improves:
– patient info. at point-of-service
– decision support, latest protocols
– collection of data
Life-threatening shortages of...
• Human expertise
– 2 physicians per 100000 Malawians
• Information
– recently ‘found’ 250,000,000 cases of malaria
• Efficiency
– many drugs expire in rural clinics
• Coordination/communication
– tremendous overlap of activity in humanitarian efforts
How and when
to introduce technologies?
Shortage
Human
Expertise
Example
2 physicians per
100,000 Malawians
Tools
 telemedicine
 HealthNet
 decision support
 intelligent training
systems
Information
recently ‘found’ 250
 passive sensing
million cases of malaria  standardized records
 pattern detection in
health data
Efficiency
drugs expiring in clinics  drug inventory in EMR
 path prediction of
fleeing refugees
Comm. &
Coord.
importance of email
 better connectivity
Phys/Human
Capital
Income
Demography
and Health
Applied computer science to make
new tools for healthcare efforts
Leading causes of DALYS
Cause
%Total DALYS (2002)
1 Lower respiratory infections
2 Perinatal conditions
3 HIV/AIDS
4 Unipolar depressive disorders
5 Diarrhoeal diseases
6 Ischaemic heart disease
7 Cerebrovascular disease
8 Road traffic accidents
9 Malaria
10 Tuberculosis
6.4%
6.2%
6.1%
4.4%
4.2%
3.8%
3.1%
2.8%
2.7%
2.4%