Clinical Trial Details
— Status: Completed
Administrative data
| NCT number |
NCT03209999 |
| Other study ID # |
UWashington |
| Secondary ID |
|
| Status |
Completed |
| Phase |
N/A
|
| First received |
July 3, 2017 |
| Last updated |
July 5, 2017 |
| Start date |
January 1, 2017 |
| Est. completion date |
July 1, 2017 |
Study information
| Verified date |
July 2017 |
| Source |
University of Washington |
| Contact |
n/a |
| Is FDA regulated |
No |
| Health authority |
|
| Study type |
Observational
|
Clinical Trial Summary
Using cross-sectional samples from over 40 Demographic and Health Surveys, the investigators
studied the association between maternal employment and 3 indicators of Infant and Young
Child Feeding (IYCF): exclusive breastfeeding (EBF) among children less than 6 months old
(N=39,791) and minimum diet diversity (MDD) and minimum meal frequency (MMF) (N=137,208)
among children 6 to 23 months old. Mothers were categorized as formally employed, informally
employed, or non-employed. The investigators first used adjusted logistic regression models
to assess the associations within each country. The investigators then used meta-analysis to
pool associations across all countries and by region.
Description:
Data Source and Population
The investigators analyzed cross-sectional data from the Demographic and Health Surveys
(DHS). DHS surveys employ standardized questionnaires and nationally representative,
multi-stage cluster sampling to allow for cross-country comparisons (34).
Our study included DHS datasets that were administered between 2010 and April 2017 and that
contained data on women's employment status and indicators of IYCF. Analyses included women
who had at least one child between ages 0-24 months. If mothers had more than one child, the
investigators included the youngest child in the household. Children residing outside the
household were excluded.
In models that evaluated MDD and MMF as the dependent variables, the final analytic sample
included 137,208 children aged 6 to 23 months from 50 countries. Small cell sizes were
prohibitive to exploring the employment-EBF associations in 10 countries that met the
aforementioned inclusion criteria; therefore, when modeling EBF as the dependent variable,
the final analytic sample included 39,791 children aged 0 to 5 months in 40 countries.
Primary Dependent Variables
EBF, MDD, and MMF served as the primary dependent variables of interest as prior literature
suggests changes in women's income and time, stemming from women's employment, may affect
these indicators of IYCF. These indicators were created using data from the 24-hour recall of
foods/food groups available in DHS. EBF, a binary variable, was defined as the proportion of
infants 0 to 5 months of age who were fed exclusively with breast milk. MDD, a binary
variable, was defined as proportion of children 6 to 23 months of age who received foods from
4 or more of the following 7 food groups: grains, roots and tubers; legumes and nuts; dairy
products (milk, yogurt, cheese); flesh foods (meat, fish, poultry and liver/organ meats);
eggs; vitamin-A rich fruits and vegetables; and other fruits and vegetables. MMF, a binary
variable, was defined as the proportion of breastfed and non-breastfed children 6 to 23
months of age who receive solid, semi-solid, or soft foods the minimum number of times or
more. For breastfed children, the minimum number of times varies with age (2 times if 6 to 8
months and 3 times if 9 to 23 months). Non-breastfed children ages 6 to 23 months must be fed
4 or more times per day to meet the MMF indicator.
Primary Independent Variable
The investigators modeled maternal employment as a 3-category variable: formally employed,
informally employed, and non-employed based on prior research which suggests: 1) a large
proportion of women in LMIC are engaged in less formalized employment and 2) wages earned are
more than 60% lower in the informal sector. Women are described as non-employed because this
term includes persons who choose to not seek employment whereas unemployed describes persons
without jobs who are actively seeking employment.
Employment type was defined based on 4 indicators: 1) employment status in the last 12 months
(employed, non-employed); 2) aggregate occupation category (skilled, unskilled); 3) type of
earnings (cash only, cash and in-kind, in-kind only, unpaid); and 4) seasonality of
employment (all year, seasonal/occasional employment). Formal employment included the
following 3 combinations: 1) employed, skilled occupation, cash only earnings, employed all
year; 2) employed, skilled occupation, cash only earnings, seasonal/occasional employment;
and 3) employed, unskilled occupation, cash only earnings, employed all year. Other employed
women were categorized as informally employed.
Confounders and Effect Measure Modifiers
The investigators identified confounding factors a priori using a directed acyclic graph,
which is a causal diagram used to characterize the relationship among variables that
influence the primary independent and the dependent variables based on both theorized and
documented relationships. In all models, confounders included maternal education (< primary
school complete, [Symbol] primary school complete), maternal age (years), marital status
(married or living together versus single, widowed, divorced), parity, morbidity (presence of
diarrhea or fever in the last two weeks), child age (months), and urban/rural status. The
investigators aimed to specify covariates in a way that allowed for the same specification in
each country and for each outcome. For variables that were in their continuous form in the
DHS (maternal age, child age, parity), the investigators assessed their linearity with the
outcomes by specifying disjoint indicator variables. Because variables were approximately
linearly associated with the outcomes in most countries, they were retained in their
continuous form to minimize the number of observations dropped from the model due to small
cell sizes. Marital status and maternal education were dichotomized because there were very
few people in some categories (e.g. divorced).
The investigators hypothesized that the employment-diet association would vary by countries'
stage in the nutrition transition. Therefore, the investigators explored differences in the
country-level associations by log-gross domestic product (GDP) per capita, adjusted for
purchasing power parity, a theorized driver of the nutrition transition (41). Data were
obtained from the World Development Indicators database, corresponding to the survey year
(42). GDP-per capita was log-transformed to reflect the expected influence of a percent
increase (e.g. 5%), rather than an absolute dollar increase (e.g. $5).
Statistical Analysis
Within-country analyses. It is expected that trends in employment (i.e. the percent of women
in formal versus informal employment) as well as diet to differ depending on countries' stage
of the nutrition transition. Therefore, the investigators aimed to keep samples comparable by
selecting countries with a recent DHS (2010-2017) and we allowed for different relationships
in each country by starting with disaggregate, country-specific estimates. The investigators
first employed separate multivariable logistic regression models for each country to test the
association between maternal employment and IYCF indicators (EBF, MDD, MMF). In these
country-specific models, the investigators utilized sampling weights to account for
differential probability of selection and response and Taylor series linearized standard
errors accounted for DHS' clustered design.
Between-country and region analyses. After obtaining disaggregated estimates for each
country, coefficients for the employment-EBF, employment-MDD, and employment-MMF associations
were entered into a random effects meta-analysis to obtain odds ratios pooled across all
countries and by world region (East Asia and Pacific, Europe and Central Asia, Latin American
and Caribbean, Middle East and North Africa, South Asia, and sub-Saharan Africa). Random
effects meta-analysis, used to generate pooled odds ratios (POR), is the statistical
combination of the estimates from separate countries and assumes that the associations
between employment and IYCF may differ by country and/or region.
Country-specific beta coefficients were also entered into a random-effects meta-regression to
assess whether the associations between employment and IYCF varied by country-level log-GDP.
Sensitivity Analyses. Sensitivity analyses included modeling several alternative outcomes
including: 1) continued breastfeeding at 1 year, 2) diet diversity score, and 3) minimum
acceptable diet. Continued breastfeeding at 1 year, a binary variable, was defined as the
proportion of children 12 to 15 months of age who are fed breast milk (37). Diet diversity
score, a continuous variable estimated among children aged 6 to 23 months, was based on the
aforementioned 7 foods groups used to calculate MDD (37). For each of the 7 food groups,
children received 1 point if any food in the group was consumed (i.e. minimum DDS =0, maximum
DDS =7). Minimum acceptable diet was modeled as a binary variable and was defined as the
proportion of children 6 to 23 months of age who received a minimum acceptable diet as
determined by minimum food frequency, diet diversity, and breastfeeding status (37). Because
in some cases the within-country sample size for the associations between employment and EBF
were somewhat smaller, we also employed a single logistic regression model with all 40
countries to test the association between employment and EBF. In this specification, each
country was included as a fixed-effect, and therefore controlled for baseline country-level
differences; but this model assumes the association between employment and EBF is homogeneous
across countries. Alpha was set to 0.05. Analyses were performed using Stata 14.1 (StataCorp
LP, College Station, Texas). No institutional review board review was obtained given that all
analyses used secondary data.
