Data source
The data used in this cross-sectional study was from the Uganda Demographic Health Survey (DHS) 2016. The DHS is a nationally representative cross-sectional survey that takes place every half a decade. A representative sample of 20,880 households was obtained using a multi-stage cluster survey design and, the information came from self-reports using a survey questionnaire[26]. The questionnaires were adapted to reflect the population and health issues relevant to Uganda. In this study, we used the data from the couples module for women of reproductive age from 15-49years and, a review of each birth history[27]. We used a nationally representative sample of 1541 married couples in 2016
Description of the variables
The sample description was adopted from Dunlop, Benova [25], Married women of reproductive age who have given birth in the 5years period preceding the survey were included in the analysis. The independent variable was the decision to seek health care by the mother which was labelled as ‘person who should have a greater say on the responders’ health care’. Descriptive analysis was performed using the categorical variable which was classified into respondent alone, respondent and wife/partner, and wife or partner alone. The observations in the category of ‘someone else’ and ‘other’ was merged with the category ‘wife or partners alone’ due to extremely small number of respondent. “Respondent alone” was used as a base category in the crude and adjusted analysis to compare the relationship of category respondent and wife/partner and wife or partner alone on health facility-based delivery.
The outcome variable was ‘place of birth for the most recent child’, which is a binary variable of place childbirth and, is categorized into ‘hospital, health facility’ or ‘other’ [28].
The other covariates included the age of the mother at birth of the most recent child, which was categorized into groups of 10years starting from 10-19years, 20-29years and 30-39years. Women in age 40years and more were merged with those in the category 30-39years of age. The age of the mothers at the time of the study was categorized into the group of 5years. The mother and father’s level of education (those who did not know their level of education were merged with those who has no education) was classified as no education, primary, secondary and higher. Due to the many categories of religion, it was reclassified into Anglican, Catholics, Muslims and Pentecostal. The other categories were merged with the closely related group basing on practices and literature. The region was also reclassified into the five cardinal geographical locations in Uganda. They included Northern, Southern, Eastern, Central and Western. The type of residence described as being located in an urban or rural area. Distance to the facility to seek medical care for the mother was categorized as a ‘big problem’ and ‘not a big problem’. Need for family planning was labelled as ever used anything or tried to delay or avoid getting pregnant and the response was ‘yes’ or ‘no. ‘Wealth index combined’ was categorized as poorest, poorer, middle, richer and richest. The wealth index described the asset index quintile that the coupled fall into, comparing their asset ownership to the Ugandan population. Finally, ‘ANC uptake’ described if the woman accessed any ANC services during the last pregnancy. ANC uptake was asked in DHS only for the latest birth in the survey recall period, that is, if the woman had more than one birth in the recall period, this variable was only available for the most recent birth. Women who could not recall the ANC were merged under the category of women who have not gone for ANC.
Statistical Analysis and model estimation
Univariable analysis.
The data were subjected to a descriptive and inferential analysis using the Statistical Package of STATA® version 16 (StataCorp LLC, Texas USA). We used the chi-squared test to compare the categorical variables using the level of significance of p-value < 0.05. Descriptive statistics including, percentages and frequencies, was used to provide an account of the study findings. Meaningful patterns were identified from the summarized data to give further insight into the multivariable analysis.
Multivariable analysis
Binary logistics regression was used to analyses the data because the outcome variable is binary taking on the value of one, if the mother delivers in the hospital/health facility or zero otherwise. Odds ratio and 95% confidence interval of health facility delivery were estimated for each independent variable using level of significant of p < 0.05. Delivering in places other than the hospital was the base category and compared with delivery in the hospital/health facility. In this study, we estimated the maximum likelihood for the odds of places of childbirth in relation with decision to seek health care while holding other variables constant.
The place of birth for the most recent child for a woman i was modelled as a function of a vector of decision to seek health care (Xd) and a vector of covariates (Xs) relating to health systems, socioeconomic and social factors influencing health facility delivery so that;
Where Yi is the dependent variable taking a value of one if a woman delivers in the hospital/health facility and value of zero otherwise and εi is the stochastic error term. The logistics regression model assumes that the logit of the probability Pi follows a linear model and defined Pi = P(Yi = 1) as the probability that a woman i delivers at the hospital/ health facility. The probability Pi depends on a vector of observed covariates so that
Where f(.) Is the logistic cumulative distribution function (i.e. ), β is a vector of regression coefficients and a vector x'i including the Xd and Xs. Estimation of , unbiased estimates of the model coefficients β and it can now be demonstrated that
So that the estimated probability of place of delivery can be estimated for each patient using and appropriate values for x'i
The model for the study was simulated using three multivariable models built through a forward stepwise approach. The First model includes main independent variable and the couples’ demographic characteristics; age at birth of first child, current age at the time of the study, level of education of the mother and father, religion, region and the residence for the mother. The second model included the household combined wealth index. The last model incorporated the independent variables associated with health facility delivery which included ANC visit, distance to seek care, family planning used.