Outcome variable

Barriers to accessing health care were the outcome variables of interest. In the MDHS 2015–16, it was defined the percentage of currently married women age 15–49 who reported that they have experienced serious barriers to accessing health care for themselves when they are sick. Responses were categorized by type of barrier: (1) Getting permission to go to the doctor? (2) Getting money needed for advice or treatment? (3) The distance to the health facility? (4) Not wanting to go alone? [21]. All four indicators were pooled together as a single entity and recoded into three groups (0 = no barriers at all, 1 = had faced one barrier, 2 = had faced more than one barrier).

Other relevant socio-demographic factors included in the MDHS data were considered as independent factors in the analysis. The variable of husband’s occupation was recorded as either white-collar (professional/technical/managerial/clerical) or blue-collar (agricultural/manual). DHS sample weights were used in all analyses to make the sample data representative of the entire population [19]. Before doing the multinomial logistic regression analyses with confounders adjusted for the survey sampling design, the multicollinearity between variables was checked, and univariate analysis was done. Only variables with significant associations (p-value < 1) in the univariate analyses were included in the multinomial logistic regression. Reference categories were set according to the MDHS (2015–16) report [13] and a specific coding system [20]. All analyses used the svyset command in STATA 14.