e-IMCI: Improving Pediatric Health Care
in Low-Income Countries
Brian DeRenzi
Quals Talk
November 19, 2007
University of Washington
e-IMCI

Project
PDA-based decision support for clinicians at the point of
care
 Increase quality of care delivered


Result


Significantly increased adherence to medical protocol
without substantially increasing patient visit time
Contribution
Adapted code base to implement the protocol for
pediatric health care
 Ran two-month field study in rural Tanzania to pilot the

Outline









Motivation
Introduction
Background on Project
Integrated Management of Childhood Illness (IMCI)
e-IMCI
Field Study
Results
Future work
Acknowledgements
Motivation




This year almost 10 million children will die before
reaching the age of 5
Most live in low-income countries
10% of infants die during
their first year, compared to
0.5% in wealthy countries
Almost 2/3 could be saved
by the correct application
of affordable interventions
Motivation

Every 6 seconds a child dies unnecessarily
Introduction

UNICEF, WHO and others develop medical protocols



e.g. Integrated Management of Childhood Illness (IMCI)
Clinical guidelines for busy facilities
Easy to use for lowly-trained health workers
Introduction - IMCI




Originally developed in
1992
Adopted by over 80
countries worldwide
Children 0-5 years old
Common illness
Cough
 Diarrhea
 Fever
 Ear Pain
 Malnutrition

IMCI
IMCI Barriers



Expense of training ($1150 -$1450)
Not sufficient supervision
Chart booklet
Takes a long time to use
 Natural tendency to be less rigorous
 Social pressure


Result - not often followed in health clinics
Related Work

Automating procedural tasks


Harvard University Program on AIDs (HUPA) Project





Designing medical protocol in South Africa
Decision support in India
TRACNet, OpenMRS, IHRDC study
Gary Marsden
Computable protocols


Using mobile devices can help under high workloads
GLIF
Artificial Intelligence
e-IMCI



Put IMCI protocol on PDA
Guide health workers step-by-step
Potential benefits
Better adherence to protocol
 Easier and faster than book
 Data collection is a by-product of care
 Can handle more complex protocols
 Interface with other devices and EMR
 Reduce training time and cost
 Strong supervision

Background
How the project started and how I got involved.
D-Tree International





Medical algorithms on mobile devices
Help over-burdened health workers
Gather data from the field
Work with governments to implement sustainable
programs
HUPA project
HUPA Project






Started in Cape Town
HIV screening algorithm
Counselors can quickly
determine if patient needs
to see doctor
Huge shortage of doctors
29.1% national HIV
prevalence1
Less than 1% in US
1 http://www.avert.org/safricastats.htm
South Africa

Worked with Right To Care





Non-profit at Helen Joseph
Hospital
Second site for HUPA
project
Gained experience with the
HUPA code
Delivered PDAs,
established workflow
Introduced to health
facilities and field work
South Africa
Tanzania


Worked with IHRDC
Met with the Tanzanian government and other NGOs
IMCI
Integrated Management of Childhood Illness.
IMCI Example
IMCI Example
IMCI Example
IMCI Example
IMCI Example
e-IMCI
Electronic delivery of IMCI.
e-IMCI Interface
e-IMCI



Implemented subset of IMCI
protocol for pilot study
Contains cough, diarrhea,
fever and ear pain
questions and treatment
First visit, ages 2 weeks to
5 years
Field Study
Real clinicians. Real patients. Real world.
Mtwara, Tanzania



Worked with IHRDC in Mtwara, Tanzania
Southern Tanzania
Rural
Subsistence farming
 Fishing


Piloted e-IMCI at a
dispensary
Study Design



Started with five clinicians
Four clinicians completed study
Goals:
Discover usability
issues
 Discover if e-IMCI
increases adherence
 Determine how
e-IMCI affects
patient visit

IMCI Protocol Use

Ideal case


“Current practice”


Follow paper chart booklet for every patient between
0-5 years of age
Treat patients from memory, occasionally referencing the
chart booklet
e-IMCI trials

Treat patients using the e-IMCI software system
Study Design





Started with some pretrials to fix major bugs
Semi-structured
interview of all
clinicians
Observed 24 “current
practice” IMCI sessions
27 e-IMCI sessions
Exit interview for each
clinician
Study Design




Real Patients, not actors
Used same data
collection forms for
current practice and eIMCI
Pairwise design
Basic pilot, no
randomization
Trials per Clinician
Clinician
1
2
3
4
5
Number of “current practice” trials 5
5
5
5
4
Number of e-IMCI trials
-
6
4
4
13
Results
Numbers, reactions and lessons.
Adherence


Measured adherence
using 23 items IMCI
asks the practitioner
to perform
e-IMCI significantly
improved adherence
to the IMCI protocol
p < 0.01
p < 0.01
Adherence: The Numbers
Investigation
Current
Practice
Adherence
e-IMCI
Adherence
p-value
Vomiting
66.7% (n=24) 85.7% (n=28)
-
Chest
Indrawing
75% (n=20)
94.4% (n=18)
-
Blood in Stool 71.4% (n=7)
100% (n=3)
-
Measles in Last 55.6% (n=9)
3 Months
95.2% (n=21)
<0.05
Tender Ear
0% (n=1)
100% (n=5)
-
All
61% (n=299) 84.7% (n=359) < 0.01
Adherence: Advice Numbers
Clinical
Officer
Current Practice e-IMCI Advice
Adherence
Adherence
p-value
1
20% (n=15)
76.9% (n=39)
< 0.01
3
26.7% (n=15)
66.7% (n=18)
< 0.05
4
80% (n=15)
100% (n=12)
-
5
100% (n=12)
73.3% (n=21)
-
All
56.9% (n=72)
77.4% (n=84)
< 0.01
Timing

No substantial increase in patient visit time
Clinical
Officer
Average Length of
Current Practice Patient
Visit (minutes)
Average Length of eIMCI Patient Visit
(minutes)
95% Confidence
Interval of e-IMCI
Minus Current
Practice
1
16 (n=5)
13 (n=13)
-2.1 to 7.9 †
3
6 (n=5)
8 (n=6)
-5.5 to 1.0 †
4
7 (n=5)
9 (n=4)
-5.7 to 4.7 †
5
19 (n=4)
14 (n=4)
-2.1 to 13.1 †
Total
10 (n=24)
11 (n=27)
-2.4 to 2.4 ‡
† unpaired t-test, ‡ paired t-test of 18 trials
Clinician Reaction

Unanimously cited e-IMCI as easier to use and
faster than following the chart booklet
Clinician Reaction





Wanted to use the system for Care Treatment Clinic
Liked being able to review answers to questions
Asked to be in future studies
“Sometimes since I have experience [with IMCI] I will
skip things, but with the PDA I can’t skip.”
Would “use a combination” of current practice and
the e-IMCI software and would never need to refer
to the book
Lessons Learned

Limitations
Question Grouping
 Threshold Problem


Requirements


Incorrect IMCI


Flexibility
otitis externa
Local Preference
Antibiotic
 Lab use

Conclusion




e-IMCI significantly improves adherence to IMCI
protocol
Does not substantially lengthen the patient visit time
Positive reaction from clinicians, but room for
improvement
Large number of interesting enhancements for the
future
Future Work
Where we’re going.
e-IMCI for Training





Current training lasts
11-16 days
Costs $1150 - $1450
per person
Using e-IMCI to train,
could reduce time and
cost
No need to train the
protocol as in-depth
Tutored mode
User-Driven Model




“Expert” mode
Allow users to decide
what investigations to
perform
Flexibility will encourage
long-term use
Merge with current
system-driven approach
to ensure correct care
Deploying Protocols



Interfaces for tutor,
guided and expert
modes
Automatically
generate interfaces for
different platforms
Maintain consistent
look and feel
Community Outreach


Take e-IMCI outside of the health facility
Travel village-to-village to collect health census
information and deliver care
Acknowledgments

Neal Lesh, Marc Mitchell, Gaetano Borriello, Tapan
Parikh, Clayton Sims, Werner Maokola, Mwajuma
Chemba, Yuna Hamisi, David Schellenberg, Kate
Wolf, Victoria DeMenil, D-Tree International, Dimagi
Inc., the Ifakara Health Research & Development
Centre, the Ministry of Health in Tanzania and the
clinicians in Mtwara for their support and
contribution to this work.
Questions
Extra Slides
Just in case.
Introduction
The vision.
Related Work
What others have done.
IMCI in Tanzania



Adapted and adopted by Tanzania in 1996
National policy
Main component is a medical protocol followed by
health workers at the point of care
Pre-Grad School


Volunteered with American Red Cross after
Hurricane Katrina
Volunteered with International Service Learning to
deliver medical supplies in rural Tanzania
Introduction

Digitize protocol to make it easier to use