Clinical Trial Details
— Status: Withdrawn
Administrative data
| NCT number |
NCT02015325 |
| Other study ID # |
0266-13-EP |
| Secondary ID |
|
| Status |
Withdrawn |
| Phase |
N/A
|
| First received |
|
| Last updated |
|
| Start date |
April 2013 |
| Est. completion date |
May 2, 2013 |
Study information
| Verified date |
August 2023 |
| Source |
University of Nebraska |
| Contact |
n/a |
| Is FDA regulated |
No |
| Health authority |
|
| Study type |
Interventional
|
Clinical Trial Summary
The objective of the study is to evaluate the impact of the Planet Water Foundation's
intervention to improve Water, Sanitation and Hygiene related illness for reducing
absenteeism and diarrheal episodes among school children resulting in improvement of their
learning performance and ultimately long term economic prosperity.
Description:
A multidimensional approach is needed to meet the Millennium development Goals (MDG) to
address Water and Sanitation Hygiene Knowledge, Attitudes and Practices. The MDG of reducing
child mortality by 2015 will require an increased focus on countries and regions where
children death rates are the highest, and a renewed commitment to reach the most vulnerable
children (UN, 2013). India and Nigeria account for more than one third of all deaths in
children less than five years of age (UN, 2013). Using a combination of a research based
literature review, expert opinion and the policy reports a multifactor model was developed to
determine the necessary variables that need to be addressed so that an integrated approach to
WASH related problems could be designed.
Variables were identified that impact WASH outcomes. These variables have been classified
into several categories including:
1. Socio-demographics
2. Environmental and Access
3. Knowledge, Attitudes, and Practices
4. Family Health Profile
5. School Characteristics
The proposed study is to conduct a randomized control clinical trial to meet the overall
study objectives by evaluating Planet Water Foundation's Program through the developed
multidimensional framework. The proposed clinical trial will include 12 schools (6
residential and 6 non-residential). Six schools (n=3 residential; n=3 non-residential) each
will be included into Group 1 (No Planet Water intervention) and Group 2 (With Planet Water
intervention) from two locations including districts of Pune and Thane in the State of
Maharashtra. Locations were matched based on similar water sources at school, toilet
facilities, water quality testing, water purification techniques, geographic setting, and
geographic population size.
Planet Water Foundation's Program (PWP) provides a school-based program with an aim to
improve health outcomes in children using three components: (a) access to safe water, (b)
access to hand washing facilities, and (c) access to water-health and hygiene education. As
part of the PWP, access to safe water is provided through the use of an AquaTower, which
incorporates an Ultra Filtration (UF) system. PWP's Water-Health & Hygiene education program
is a four week program including four modules: (1) importance of clean water, (2) how to wash
your hands, (3) when to wash your hands, and (4) protecting against germs through a hands-on,
activity-based program that incorporates games, drama, song, and dance.
A mixed methods approach was used with open-ended and multiple-choice questions in the form
of questionnaires. Questionnaires were translated from English into the local language
(Marathi), and back-translated to English for analysis by BAIF Development Research
Foundation in India. Students, household caregivers/parents, and teachers were given separate
questionnaires. Three different interviews were conducted: (1) a household questionnaire
conducted at the home of each student included in the study, (2) a teacher questionnaire
administered at the schools, and (3) a student questionnaire administered with each student
individually by the interviewer in a separate room away from other students. Group 1 and
Group 2 were given the same questionnaires. All questionnaires were administered in
cooperation with BAIF Development Research Foundation, a local foundation in India. The
foundation received data collection and data entry training prior to the study. Data will be
gathered on paper forms and all the data will be entered in Microsoft Excel®. The online data
repository will be stored in an encrypted manner and will be password protected. The
information will be stored behind a secure firewall provided by the local research foundation
in India. All the data gathered on the paper forms will be stored in a locked cabinet, and
information will be accessed to only those individuals who are directly involved in the
research. Consent will be obtained ensuring that (1) only approved personnel are present
during the consent process, (2) fewest number of individuals possible are aware of the
subject's participation in the research, and (3) research activities are performed in as
private of a place as possible.
The study sample will include six Group 1 (no PWP intervention) schools and six Group 2 (With
PWP intervention) schools, with 60 students per school for a total of 360 students in Group 1
schools and 360 students in Group 2 schools. The sample size among the 12 schools was
calculated assuming an intra-cluster correlation coefficient of p=0.01, this design will
provide 80% power (Donner and Klar, 1996) to detect a medium effect size (Cohen 1992) of
0.50. The sample size will be increased by 15%, to 414 for intervention and control, to
account for attrition.
Descriptive statistics will be computed for all variables to ensure data quality and to
evaluate the assumptions of statistical tests. Variable distributions will be described with
histograms, box plots, central tendencies, measures of variation and frequency distributions.
We will test the assumption of normality for the outcome measures, within the PWP and control
groups. Additionally, we will test the assumption that the PWP and control groups have equal
variances. Generalized linear mixed models (GLMMs) with unequal variances (McCulloch et al.,
2008) can be applied when the ideal conditions of normality and equal variances are not met.
Testing for the equality of variances is available with the COVTEST statement, which is based
on the ratio of residual likelihoods or pseudo-likelihoods. Residuals at each level of the
model will be examined to assess model fit. Generalized linear mixed models using SAS PROC
GLIMMIX also provides the flexibility to model different types of outcome variables (binary,
count and continuous data). All analyses will be conducted using SAS/STAT® software for
Windows version 9.2 or higher.
Prior to testing the study hypotheses, we will examine data for possible covariates.
Potential confounding variables will be controlled for by entering these variables as
covariates in the regression models. These explanatory variables will improve the detection
of treatment effects by entering these variables as covariates in the GLMMs. These variables
will be entered as a covariate if the univariate two-sided chi-square test comparing the PWP
and control groups is significant at the alpha level of 0.20. In preliminary analysis, we
will examine potential covariates using single-level however will we also examine potential
covariates using hierarchical linear models. We will use a liberal alpha of 0.20 for the
decision of whether to adjust for a covariate so that important confounding effects are not
overlooked.
Hierarchical, correlated data structure: A hierarchical linear models (HLM) approach
(Raudenbush and Bryk, 2002) will be used to account for the correlation induced by repeated
measurements over time for students nested within schools. School will be modeled as a random
effect. These HLMs will be implemented in SAS using generalized linear mixed models, which
can be applied to analyze binary, count and continuous repeated measurement data, collected
at unequal time intervals following a hierarchical structure (McCulloch and Searle, 2001;
Liang and Zeger 1986). These models address the correlated nature of multiple measurements
that come from each subject over time. We anticipate that generalized linear models will be
appropriate for the data, but will also explore using the method of generalized estimating
equations (GEE) (McCulloch and Searle, 2008). The GEE method does not require assumptions of
normality and equal variances, but uses less information than linear mixed models since only
the first two moments of the distribution are employed in GEE.
