Decomposing socioeconomic inequalities in contraceptive use among Kurdish women: a cross-sectional analysis of the ravansar cohort study | BMC Public Health

Study population

This study utilized data from 4,823 Kurdish women enrolled in the Ravansar Non-Communicable Disease (RaNCD) cohort, a prospective epidemiological study based in Ravansar County, Kermanshah Province, Iran. The RaNCD cohort is a component of the larger Prospective Epidemiological Research Studies in Iran (PERSIAN) cohort, which aims to investigate the risk factors associated with non-communicable diseases across various Iranian subpopulations. The RaNCD cohort includes individuals aged 35 to 65 years, and this analysis focused specifically on women within this age range.

Participant recruitment was carried out using a structured census approach. In urban areas, trained research assistants conducted door-to-door visits to register eligible residents. In rural regions, household information was obtained from local health units that maintain regularly updated demographic records. Recruitment commenced in March 2015, and by February 2017, a total of 10,065 individuals had been enrolled, reflecting a high participation rate of 93.2%. All eligible participants received comprehensive information about the study’s objectives and procedures prior to providing informed consent [22].

For this analysis, we included women who had been married at least once—encompassing those currently married, widowed, and divorced—in order to provide a comprehensive evaluation of CU among individuals with marital experience. Accordingly, younger married women outside the cohort’s age range, as well as single individuals, were excluded, as the study focused on family planning within the context of formal marital relationships. The data were derived from the recruitment phase of the RaNCD cohort conducted in 2017.

In the initial stage of analysis, we employed the Pearson Chi-square (χ2) test to examine associations between categorical variables, with a particular focus on evaluating differences in CU across socioeconomic quintiles.

Conceptual model

Our study examines how socioeconomic and demographic determinants influence CU among Kurdish women in western Iran. The framework is structured as follows:

  1. 1.

    Independent Variables (Determinants)

    • • Demographic Factors: Age, marital status, age at first pregnancy, number of live births.

    • • Socioeconomic Factors: Socioeconomic quintile, place of residence (urban/rural).

  2. 2.

    Mediating Factors

  3. 3.

    Outcome Variables

Measures of inequality

To assess socioeconomic-related inequality in CU, we employed the normalized concentration index (NC), a widely accepted metric that quantifies the degree of socioeconomic-related inequality in a health variable. The NC ranges from −1 to + 1, with zero indicating no inequality, negative values suggesting a higher concentration of CU among poorer individuals, and positive values indicating a greater prevalence among wealthier individuals. The NC was calculated using the Eq. 1 [23]:

$$NC= frac{2}{nmu (1-mu )} sumnolimits_{i=1}^{n}{y}_{i}{r}_{i}-1$$

(1)

where ({y}_{i}) represents the history of contraceptive use or tubectomy for the (i)-th participant, ({r}_{i}) denotes the fractional rank of the (i)-th participant based on SES, and (mu) is the mean of the outcome variable (history of contraceptive use/tubectomy).

Decomposition of inequality

To identify the primary factors contributing to socioeconomic inequality in CU, we applied the Wagstaff decomposition method. This analytical approach allows for the quantification of each explanatory variable’s contribution to overall inequality. The variables considered in the decomposition included age, marital status, age at first pregnancy, number of live births, place of residence (urban vs. rural), and SES.

SES was operationalized through a composite index constructed using principal components analysis (PCA). Using this method, we synthesized a broad array of indicators reflecting household wealth and living standards. These included housing characteristics (per capita living space, per capita number of rooms), housing tenure (ownership vs. rental), and ownership or access to household assets such as freezers, washing machines, dishwashers, computers, internet, automobiles, motorcycles, color televisions, bathrooms, and vacuum cleaners. Additionally, the index incorporated cultural and educational dimensions, including the number of books read in the past year, participation in pilgrimage trips, the frequency of national and international travel, and the highest level of educational attainment. PCA was employed to reduce the dimensionality of these correlated variables and generate a single SES index that captures the underlying variance in socioeconomic position. Based on their SES scores, women were subsequently classified into five quintiles. This SES index served as a key independent variable in analyzing inequalities in contraceptive us.

We selected the Wagstaff decomposition method because it offers a comprehensive breakdown of socioeconomic inequality while accounting for the full distribution of SES. This method is widely recognized and aligns with best practices in health inequality research [24,25,26]. Given that the outcome variables in this study—history of contraceptive use and tubectomy—are binary, we employed the extended decomposition technique proposed by Wagstaff (2005), which allows the normalized concentration index (NC) for binary outcomes to be decomposed accordingly. This decomposition is represented in Eq. 2 [27]:

$$NC=frac{{sum }_{k}(frac{{beta }_{k}{overline{X} }_{k}}{mu }){C}_{k}}{1-mu }+frac{frac{{C}_{e}}{mu }}{1-upmu }= {C}_{widehat{y}}+varepsilon$$

(2)

In this equation, the normalized concentration index (NC) is decomposed into two components: the explained component (({C}_{widehat{y}})), which captures the inequality attributable to the observed explanatory variables included in the study, and the unexplained residual component ((varepsilon)). The term ({beta }_{k}) represents the marginal effects of the explanatory variables, derived from a logistic regression model and Ck is concentration index for explanatory variables. Utilizing marginal effects enhances the interpretability of logistic regression outcomes by expressing the impact of predictor variables in terms of changes in predicted probabilities, which is particularly valuable in decomposition analyses of health inequalities [28,29,30].

The elasticity of each explanatory factor, expressed as (left(frac{{beta }_{k}{overline{X} }_{k}}{upmu }right)), measures how responsive the outcome variable (i.e., history of contraceptive use or tubectomy) to changes in that particular variable. Here, ({overline{X} }_{k}) is the mean or proportion of the k-th explanatory variable, and μ is the mean of the outcome variable.

To quantify the absolute contribution of each explanatory variable to total inequality, we multiplied its elasticity by its concentration index (({eta }_{k}{C}_{k})). This allowed us to quantify the specific contribution of each factor to the overall socioeconomic inequality in CU. Furthermore, we computed the percentage contribution of each explanatory factor by dividing its absolute contribution by the total value of the normalized concentration index. This enabled a clear assessment of the relative importance of each determinant in explaining the observed inequality.

The direction of each variable’s contribution to inequality in CU indicates its directional effect. A positive contribution means the variable is associated with lower CU among poorer individuals, whereas a negative contribution suggests lower CU among wealthier individuals.

Given that our analysis involved two distinct outcome variables—(1) a history of contraceptive use and (2) tubectomy—we constructed separate analytical models for each to better capture the nature of socioeconomic disparities. Tubectomy, or female sterilization, is a permanent method of contraception that involves surgically obstructing the fallopian tubes to prevent pregnancy. It is regarded as one of the most reliable long-term family planning strategies and is generally selected by women who do not wish to have more children or who seek a permanent solution to contraception [31].

In this study, we defined CU based on lifetime experience. Specifically, a participant was classified as having a history of contraceptive use if she had ever used any temporary method of contraception—excluding withdrawal—at any point in her life, irrespective of current use status. This broader definition enables a more comprehensive assessment of contraceptive behavior over time, rather than limiting analysis to current or recent practices.

Data extraction was conducted using Microsoft Excel 2016, while statistical analyses were carried out using Stata version 17 (StataCorp, College Station, TX, USA).

Continue Reading