The relationship of women’s status and empowerment with skilled birth attendant use in Senegal and Tanzania

Study settings

This study investigated the use of SBAs in Senegal (SN) and Tanzania (TZ). These two
countries are similar with respect to maternal and child heath indicators, yet are
culturally and economically different from one another. Infant and maternal mortality
are similar across the two countries (50 per 1000 births in SN and TZ, and 320 per
100,000 live births in SN and 410 in TZ, respectively) 1], as are Total Fertility Rates (5.0 in SN and 5.4 in TZ) 36], 37].

At the same time, there are differences across the two settings with respect to income,
health service use and availability, as well as sociocultural contexts. The national
Gross Domestic Product per capita (GDP in USD) in Senegal (1032.7) is twice that of
Tanzania (516.2) 38]. In Tanzania, half of the recent births in the last five years occurred at health
facilities (50.2 %), compared to almost three quarters (72.8 %) in Senegal 36], 37]. Tanzanian women’s traditional roles and activities in the household are undergoing
change, with increases in women’s status and power that are likely to promote reproductive
health behaviors and service use 39]–41]. In general Islamic traditions are believed to negatively influence women’s status,
and women’s low social status is negatively related to maternal health services in
Senegal 42], 43]. Yet Senegalese women have been renowned in their socioeconomic and political participation
(e.g., local women’s organizations, governmental efforts including gender sensitive
programs and decentralization) 44], 45]. These advantages and freedom of mobility may represent women’s higher empowerment
status and can positively influence delivery care use.

Data

This study used data from the 2010–11 Senegal and 2010 Tanzania Demographic and Health
Surveys (DHS), nationally representative household surveys that collected data on
population, health, and nutrition issues. The study sample consisted of all births
reported by currently married women that occurred in the five years preceding each
survey. The final female study sample included 7033 women and 10,668 births in Senegal,
and 4445 women and 6748 births in Tanzania (weighted).

In the survey, the total number of women (both currently married and unmarried) who
gave birth during this period was 8148 in Senegal and 5349 in Tanzania. Questions
on household decision-making participation were asked to currently married women only,
thus unmarried women were dropped from the analysis. Furthermore, a few women were
dropped for missing data on the decision-making questions (n?=?11 in TZ) and the perceptions of gender norms questions (n?=?119 in SN and 82 in TZ). Among the births to the study female sample, some births
were excluded due to missing data on delivery assistance (n?=?4 in SN, 24 in TZ). Observations were weighted using individual and household weights
to adjust for differences in the probability of selection and interview among cases
in the sample. Given that this study is a secondary data analysis of public available
data, the study was considered exempt from IRB approval by the UCLA Institutional
Review Board.

Dependent variable

SBA use at childbirth was operationalized as the use of an SBA at childbirth(s) in the five
years preceding the survey. The variable was recoded as binary, in accordance with
the WHO definition of SBAs 2]. The SBAs included doctor or assistant medical officer, clinical officer, nurse or
midwife; non-SBAs included MCH aide, village health worker, Traditional Birth Attendant,
relative or friend, other, or no-one at the delivery.

Independent variables

Women’s education served as a proxy measure of women’s status in this analysis. The survey asked women
to report on the highest level of school that she had attended. The variable was recoded
as: no formal education; primary attended; and secondary or higher attended.

Women’s Empowerment is operationalized through four dimensions, as determined by exploratory and confirmatory
factor analysis (see below): Household decision-making power, perceptions of gender
norms against violence, perceptions of sex negotiation, and age at first marriage.

A.
Household decision-making power was examined as a summative variable. The survey asked women about their participation
in decisions regarding household matters (e.g., own health care, major household purchases,
and visits to family or relatives). The variables were first recoded into binary to
indicate whether the respondent participated in the decision, either alone or jointly
with their husband, or not. A summative variable captured the number of decisions
in which women participated (scored 0–3).

B. Two sets of questions in the DHS focused on perceived gender norms. The first domain,
perceptions of gender norms against violence, asked about women’s acceptance of wife-beating by her husband under five situations
– if she goes out without telling him, neglects the children, argues with him, refuses
to have sex with him, or burns the food. Each of the variables was first recoded as
binary (i.e., yes or no) then summed to create a scale capturing the number of situations
in which women do NOT accept the violence (scored 0–5), with higher numbers indicating
lower acceptance of gender violence and more progressive gender norms.

C. The second domain, perceptions of gender norms for sex negotiation asked about women’s perceived ability to negotiate sexual relations – if the respondent
can refuse having sex or can ask her partner to use a condom. The variables were recoded
to determine if the respondent can refuse/ask, or not (i.e., cannot refuse/ask, don’t
know, not sure, or depends). A summative variable captured the number of situations
in which women think that they can negotiate with their husband (scored 0–2).

D.
Age at first marriage was also included based on the theoretical and empirical importance of this construct
as a strategic life event and reflection of women’s empowerment 9], 17]. A continuous variable was created by MEASURE, based on calculation using the date
of the first marriage or union (“living with a man as if married”) and the date of
birth of the respondent.

Control variables

Sociodemographic characteristics of women and households included women’s age, parity, employment for payment, household wealth, marital and
household relationship, the gender composition of children, and the place of residence.
Women’s age at the time of delivery was included as a continuous variable based on
preliminary analysis indicating a linear relationship with SBA use. Parity (i.e.,
the birth order of the children) was a categorical variable (e.g., first birth; second
or third birth; fourth birth or more). Employment for payment was a binary measure
defined as a woman who had been employed for cash or in-kind in the last 12 months,
or not. Household wealth was examined using household asset data, such as ownership
of consumer items and home attributes. Principal component analysis was conducted
by MEASURE DHS to develop a ranking of household wealth according to the scores, and
households were then divided into quintiles 37]. Marital relationship was assessed as categorical – monogamous union, polygamous
as a first wife, or polygamous as a second wife or lower – to examine the potential
differences by the type of marital relationship and wife order. Household relationship
was assessed as binary – if the respondent was a household head or not. The gender
composition of children was examined as a binary variable – if the respondent had
at least one living son or not at the time of the delivery. This variable was included
based on evidence of son preference in Africa, specifically that having at least one
son has been valued for continuing the family lineage and kinship ties, as well as
transfer of property due to inheritance laws 46]. Place of residence indicated if the respondent lived in an urban or rural area.
These control variables were available in both countries. Other important variables
(e.g., religion and ethnicity) were examined in separate models, but are not presented
in the final models as they were not available in the Tanzania dataset.

Perceived difficulty in accessing health care was also included as a control variable, which assessed if the respondent perceives
difficulty when seeking health care. The survey asked: “When you are sick and want
to get medical advice or treatment, is each of the following a big problem or not?”
The answers were collected for each aspect: getting permission to go; getting money
needed for advice/treatment; the distance to the healthy facility; or not wanting
to go alone. The variables were first recoded into binary variables to show if the
respondent perceived a big problem or not (i.e., not a big problem or not a problem
at all) for the four aspects separately, then a summative scale was created (scored
0–4), with higher scores indicating higher perceived difficulties.

Analytic strategies

Data analysis was conducted in three main steps. First, descriptive analyses were
conducted using SAS 9.3. Second, exploratory and confirmatory factor analyses were
conducted using Mplus 7.3 to identify and confirm the underlying structure of the
indicators of empowerment 47]. Third, sequential regression analyses were conducted in SAS. The simple (unadjusted)
logistic regression was conducted first to examine the bivariate associations between
SBA use and each of the explanatory variables. Next, the multivariate logistic regression
was conducted that included all of the control variables found to be significant in
the bivariate models. Last, the final multivariate logistic regression models added
the measures of women’s empowerment, followed by the addition of interaction terms
between each of the empowerment domains and education. The variance inflation factor
assessed multicollinearity of variables in the model and was shown to be below cut-off
point of 10.

All of the analyses were conducted accounting for individual weights, clusters (i.e.,
Primary Sampling Unit), and sample strata using the survey analysis commands. Given
that the study examined births occurring to women nested in households, this analysis
corrected the standard errors for clustering by woman and household using the Taylor
Series linearization method 48]. Model fit was assessed though Likelihood Ratio (LR) chi-square test and Wald chi-square
test.