Case study on the use of electronic data collection in a modular household survey as part of the IDEAS project. Presented by Keith Tomlin at LSHTM's 'Enhancing data capture in health research' RDM event on November 20th, 2015.
Scanning the Internet for External Cloud Exposures via SSL Certs
Electronic data collection for a modular household survey in Ethiopia
1. Electronic data collection for
a modular household survey
in Ethiopia
Keith Tomlin
IDEAS Data Manager/Epidemiologist
Improving health worldwide
www.lshtm.ac.uk
2. The Project
• IDEAS (Informed DEcisions for ActionS)
• Aims to improve the health and survival of mothers and newborns
by generating evidence to inform policy and practice.
• Working in Ethiopia, North-Eastern Nigeria and the Indian States of
Uttar Pradesh and West Bengal.
3. The Survey
• Ethiopia follow-up survey
• Aim to gather data on pregnancies, deliveries and newborn care
from women in households, from frontline workers and from
health facilities
• Household survey comprised in-depth interviews with women of
child-bearing age
• 80 randomly sampled clusters
• 75 households per clusters (6,000 in total)
• 12 survey teams, each with 6 interviewers and one supervisor
• Fieldwork carried out during April 2015
6. Clusters (80) Cluster Name
04 MADO TEMKET
Households (75)
04/01
04/02
04/03
04/04
04/05
04/06
04/07
04/08
etc…
Household survey structure
75 households within each
cluster
7. Household survey structure
Clusters (80) Cluster Name
04 MADO TEMKET
Households (75)
04/08
Residents (household roster)
ID number Name Gender Date of Birth Eligible
04/08/01 Teklay Male 08/05/1970 No
04/08/02 Azmera Female 08/12/1982 Yes
04/08/03 Birhdan Female 24/02/1987 Yes
04/08/04 Semere Male 15/10/1990 No
04/08/05 Meaza Female 30/08/1993 Yes
04/08/06 Mebrhit Female 05/06/2002 No
A roster is taken of all residents in
each household
8. Household survey structure
Clusters (80) Cluster Name
04 MADO TEMKET
Households (75)
04/08
Residents (household roster)
ID number Name Gender Date of Birth Eligible
04/08/01 Teklay Male 08/05/1970 No
04/08/02 Azmera Female 08/12/1982 Yes
04/08/03 Birhdan Female 24/02/1987 Yes
04/08/04 Semere Male 15/10/1990 No
04/08/05 Meaza Female 30/08/1993 Yes
04/08/06 Mebrhit Female 05/06/2002 No
The programme identifies women of
child-bearing age (15-49)
9. Household survey structure
Clusters (80) Cluster Name
04 MADO TEMKET
Households (75)
04/08
Women aged 15 - 49
ID number Name Gender Date of Birth Eligible
04/08/02 Azmera Female 08/12/1982 Yes
04/08/03 Birhdan Female 24/02/1987 Yes
04/08/05 Meaza Female 30/8/1993 Yes
• Request to interview each woman if present
• Return to the household up to three times if a
woman was temporarily absent
10. Household survey structure
Women aged 15 - 49
ID number Name Gender Date of Birth Eligible
04/08/03 Birhdan Female 24/02/1987 Yes
Pregnancy roster
Baby ID Name Gender Date of Birth
04/08/03/01 Azmera Female 12/09/2007
04/08/03/02 Semere Male 03/07/2009
04/08/03/03 Mebrhit Female 12/09/2013
A roster of births for
each woman was taken
11. Household survey structure
Women aged 15 - 49
ID number Name Gender Date of Birth Eligible
04/08/03 Birhdan Female 24/02/1987 Yes
Pregnancy roster
Baby ID Name Gender Date of Birth
04/08/03/01 Azmera Female 12/09/2007
04/08/03/02 Semere Male 03/07/2009
04/08/03/03 Mebrhit Female 12/09/2013
If the most recent birth
had taken place in the
two years prior to the
interview, additional
questions were asked
13. Software considerations
• Affordable (and preferably free)
– and runs on affordable hardware
• Accommodate modular structure, and automatically generate &
carry forward ID numbers between clusters, households, adults
and pregnancies, with no manual entering of numbers
• Accepts Unicode to allow for Amharic and Tigrinya languages
• No dependence on the internet, because there mostly isn’t any
14. Software comparison
(as at June 2014)
Provider No cost to
develop/deploy
Runs on
Android OS
Accommodates
modular structure
Recognises
Unicode
CSPro
Android
EpiCollect
ODK
Pendragon Forms
Surveybe
15. Hardware considerations
• Run Android
• Inexpensive (84 units + cases + extra batteries + import duties)
• Readily available in bulk
• Robust
• Excellent battery life
• Small enough to be held in one hand, but with a large enough
screen so that long option lists could be displayed without scrolling
17. In-field data reporting
In each survey team, data were
transferred at the end of each day
from the tablets to the
supervisor’s laptop, using USB
Cables.
18. In-field data reporting
Data were
automatically
imported into a
Microsoft Access
database to create
a daily monitoring
and query report
22. ስለተዋወቅን…..Some non-Latin characters (such as those used in Amharic) can take up more computer memory
than Latin characters. This has the potential to slow down data collection programmes. Many
thanks to the CSPro developers for rapidly responding to this problem and releasing a new
version of the Android software in time for our survey.