Water quality information, WATSAN-agriculture
hygiene messages and water testing with school
students: Experimental evidence for behavioral
changes in Bangladesh
Authors[1]
Mohammad Abdul Malek (BRAC Research and
Evaluation Division, Bangladesh and University of Bonn
Center for Development Research, Germany).
Correspondence: malekr25@gmail.com.
Tahsina Naz Khan, Ratnajit Saha, Priyanka Chowdhury,
and Ikhtiar Mohammad (BRAC Research and Evaluation
Division)
[1]The
authors acknowledge intellectual inputs/support from Professor Joachim Von Braun, Dr. Nilcolas Gerber, Dr.
Evita HP, Dr. Samantha Antonio, and PhD students with WATSAN-Agriculture study project and comments/feedback
from numerous researchers/practitioners at different meetings at ZEF, Bonn and BRAC, Dhaka. Thanks goes to BRAC
RED and BRAC WASH program team for their support to implement the treatment and survey at different stages of
the project. Financial assistance from Gates Foundation is duly acknowledged.
Malek MA
Motivation
•
•
•
•
•
•
•
Bangladesh made enormous health advances in south Asia, but still challenges remain
with water supply and sanitation (Lancet 213).
Careful look on related economic literature vs public health literature.
How can the capacity of households and communities in monitoring their own water
and sanitation environment be developed? How can water, sanitation and hygiene
(WASH) behaviour be improved?
Information can increase demand for environmental quality, improve water sanitation
and hygiene behaviour, and thus improve health outcomes at the household level.
Earlier studies tested the effectiveness of different WATSAN interventions either at a
single treatment arm or multiple treatment arms.
Employing school students as agents of change and as a channel of conveying messages
for the households, farm fields and community are also limited.
Thus, we test the effectiveness of packages of WASH interventions with the school
students.
Malek MA
Study Hypotheses
Relative to the control group, it is hypothesized that individuals/
households in the treatment group will:
•Have greater changes of WASH behavior and knowledge (at home
and at farm field)
•Have changes nature and amount of investment on WASH related
activities.
•Have improved microbial quality of drinking water.
•Have less diarrhea prevalence and cost of illness
•Have improved anthropometrics of under 5 years of children
•Have less days of work/school absenteeism
Malek MA
Study methods and materials
•
RCT with BRAC developed secondary school students network (student brigades) in 6
WATSAN hotspots (sub-districts) of Bangladesh, conducted in fourth phases.
First, conducted a water quality census and drew a sample of 648 faecal-contaminated
households following a multi stage cluster random sampling.
Second, a baseline survey (i.e. pre-intervention), to establish the similarity of the
treatment and control groups. The salient WATSAN and agricultural hygiene issues.
Third, Implemented the treatments, consisting of three actions:
1) informing the households about the initial water testing results obtained from the
earlier census,
2) delivering hygiene messages through a poster related with farm field and households
use, and
3) equipping the student brigades with water testing toolkits and let them test water at
different points and communicate the results back to their households (March 10-12
March/PoU, 11-13 April/PoS, May 11-13 2015/PoU).
Finally, conducted the end-line survey and form the panel dataset that will be analysed
using standard analytical techniques.
Malek MA
Area of Interest (hotspots):
Selection Criteria
 Map-Groundwater level status
 Map-Agro-ecological zones
 Map-Flood and drought prone area
 Map-Diarrhea prevalence site
 Info-Level of development
 List-BRAC WASH intervention area
Sl
No.
Subdistricts
District
1
Atrai
Naogaon
2
Kalihati
Tangail
3
Mirzapur
Tangail
4
Bakshigonj
Jamalpur
5
Melandaha
Jamalpur
6
Bauphal
Patuakhali
Level of
Development
Peri-urban/
Advanced
rural areas
Marginal
areas
www.brac.net
Malek MA
66 schools from
hotspot areas
Treatment (28 schools
with BRAC WASH SBs)
Control (24 schools
with BRAC WASH SBs)
Water quality census: 1560 SBs’
HHs (24 students per school)
Pure Control (14 schools
from non-BRAC WASH
areas)
Eligible study population (FC
presence): 1,094 SBs’ HHs
Sample selected for treatment: 250 SBs’ HHs
Sample selected for survey: 654 (540 required at 3% CI and 95% CL) SBs’/Students’ HHs
(Treatment, control & pure control)
Fig : Sampling frame at a glance
Malek MA
Results from Water quality census
•
About 72% households drinking water faecal contaminated at POU-faecal contaminated households had more
diarrhoea prevalence rate.
Table Prevalence of diarrhea in the hotspot areas in Bangladesh (1 if yes) : Results from
logit regression
Diarrhea
Coef.
Std. Err.
P-values
Income
0.000
0.000
0.958
Agri_work (1 if yes)
ChildU5 (1 if yes)
Latrine_imprvd (1 if yes)
Soap/ash use for
handwash (1 if yes)
Advanced areas (1 if yes)
Impvd_toilet (1 if yes)
Sc_status (1 if control)
Sc_status (1 if treatment)
Inc2
inc3
inc4
_cons
0.460
-0.263
0.007
-0.663
0.211
0.202
0.211
0.206
0.029
0.193
0.972
0.001
-0.030
-0.424
0.600
-0.004
-0.533
-0.167
-0.121
-0.031
0.190
0.389
0.251
0.255
0.256
0.298
0.464
0.814
Number of obs
LR chi2(12)
Prob > chi2
0.874
0.276
0.017
0.986
0.038
0.574
0.795
0.970
600
43.33
0
The households
associated with
farming are
more likely to
diarrhoea
prevalent- this
also
necessitates
intervention
designed for
both farm
fields and
households
Malek MA
Baseline characteristics
Bonferroni multiple comparison t test
•Treatment, control and pure control households’ distance from various important
places, household characteristics were almost similar across the groups.
•Most of the community level WASH characteristics
•Student brigades quality characteristics
• Household sanitation and hygiene profile
• Household types of water sources
•Household water treatment behaviour
•None of the outcome variables (household WASH index, WATSAN expenditure,
microbial water quality, child anthropometries, etc., household adult/students
/children diarrhea prevalence, etc.) are statistically different across the groups.
Malek MA
Water Quality (FC/Feacal Coliform) Bacteria Screening by Student Brigades at Treatment Phases
Fecal Coliform Screening Result at treatment areas
100%
75%
50%
25%
Ash
PoU (1st Phase)
PoS (2nd Phase)
Bauphal
Bakshiganj
Kalihati
Atrai
Bauphal
Bakshiganj
Kalihati
Atrai
Bauphal
Bakshiganj
Black
Kalihati
0%
Atrai
Treatment SBs’ households
BlackHigher
possibility
of FC
presence
Ash-Lesser
possibility
of FC
presence
YellowAbsence of
FC
PoU (3rd Phase)
Yellow
•Except Bauphal, presence of FC at PoU reduced in third phase compared to the first phase.
•FC was also found at PoS though it is less prevalent than that of at PoU (except Bauphal)- it
may be due to use of STWs, lesser distance from the latrines/garbage place, etc.
Malek MA
Water Quality (Feacal Coliform) Testing Comparison of Lab Test (Membrane Filter
Technique) vs. Screening of Hydrogen Sulfide Producing Bacteria (presence or absence
of FC Bacteria)
HH_Feacal Coliform Presence
100
% of HH
75
Correlation
Coefficient, r=0.56
50
25
0
Endline
WQ
Atrai
Lab
Test
Endline
WQ
Lab
Test
Kalihati
Endline
WQ
Lab
Test
Bakshganj
Endline
WQ
Lab
Test
Bauphal
• Results between lab test and screening test as a whole are consistent.
Malek MA
End line Water Quality Test (screening) Result: July 2015
100
Endline Water Quality Test Result in
Color
75
Household in %
•At the baseline
(August 2014): all
samples had FC
presence
(Black/ash)
• At the end
line, absence of FC
reduced in all
locations.
•Being Mirzapur
and Melandoh as
pure control
area, absence of FC
was found among a
good no. of
samples compared
to treatment area
(Kalihati) – Need to
investigate?
50
25
0
Black
Ash
Yellow
Seasonality- Not be an issue (Ambient
condition prevailed during both baseline and
end-line water quality testing).
Malek MA
Analytical method for quantitative impact analyses
We estimate treatment effect using following DiD regression model
Where
= Outcome of interest
= dummy variable taking the value of 1 if the observation is in the treatment group and 0 otherwise.
This variable captures aggregate factors that would cause changes in the outcome even in the absence
of intervention.
= a dummy variable taking the value of 1 if the observation is from endline and 0 otherwise. This
variable captures possible differences between the treatment and control groups prior to the
intervention.
= The coefficient of interest, which is the same as a dummy variable equal to one
for those observations in the treatment group in the second period.
•Robustness check will be done by controlling the baseline (school and village) characteristics
including GPS estimates.
• The study sample can be classified into several sub-groups- thus we might also have different adoption
rate (for letter water quality info, poster message, and repeated water quality testing) –these might be
linked with the outcome measures.
• Differential impacts for girls and boys SBs’ households may also be estimated.
Malek MA
Conclusions so far..
• Experiences found the justification of the intervention packages that we
suggested for this study.
• Baseline results confirm similarities for most of the outcome indicators among
treatment and control groups.
• Impression from the implemented treatments indicates that the suggested
intervention packages can improve microbial quality of drinking water.
• If we would have impact on other outcome variables, then our intervention
could be an effective strategy to motivate households and
communities, particularly using school students as agents of change and as a
channel of conveying messages for the health and developmental/productivity
outcomes of farm households.
• Scaling up of the interventions given by this project needs only a simple
institutional/administrative arrangement with less costly investment on
training, poster and water quality testing kits.
• To have concrete results, still we need to wait for several weeks.
Malek MA