Wealth stratified inequalities in service utilisation of breast cancer screening across the geographical regions: A pooled decomposition analysis

Background: Breast cancer is the most commonly occurring cancer among women in low-resourced countries. Reduction of its impacts is achievable with regular screening and early detection. The main aim of the study was to examine the role of wealth stratified inequality in the utilisation breast cancer screening (BCS) services and identified potential factors contribute to the observed inequalities. Methods: A population-based cross-sectional multi-country analysis was used to study the utilisation of BCS services. Regression-based decomposition analyses were applied to examine the magnitude of the impact of inequalities on the utilisation of BCS services and to identify potential factors contributing to these outcomes. Observations from 140,974 women aged greater than or equal to 40 years were used in the analysis from 14 low-resource countries from the latest available national-level Demographic and Health Surveys (2008-09 to 2016). Results: The population-weighted mean utilisation of BCS services was low at 15.41% (95% CI: 15.22, 15.60), varying from 80.82% in European countries to 25.26% in South American countries, 16.95% in North American countries, 15.06% in Asia and 13.84% in African countries. Women with higher socioeconomic status (SES) had higher utilisation of BCS services (15%) than those with lower SES (9%). A high degree of inequality in accessing and the use of BCS services existed in all study countries across geographical areas. Older women, access to limited mass media communication, being insured, rurality and low wealth score were found to be significantly associated with lower utilisation of BCS services. Together they explained approximately 60% in the total inequality in utilisation of BCS services. Conclusions: The level of wealth relates to the inequality in accessing BCS amongst reproductive women in these 14 low-resource countries. The findings may assist policymakers to develop risk-pooling financial mechanisms and design strategies to increase community awareness of BCS services. These strategies may contribute to reducing inequalities associated with achieving higher rates of the utilisation of BCS services. The study was population-based and multi-country using a cross-sectional design using available standard Demographic and Health Survey (DHS) data. The DHSs are nationally-representative household surveys that provide data for a wide range of monitoring and impact evaluation indicators in the areas of population, health, and nutrition. Approval was received from DHS to use data. Data were extracted from the most recent DHS, covering 14 low-resource countries for the period from 2008 to 2016 [24–37]. The DHS surveys have been part of a long-standing worldwide program that includes individual and household-level, socio-demographic, health indicators and health care data in the context of low-resource countries. These national-level surveys, generally conducted every 3 years, capture information related to maternal and child health, mortality, fertility, family planning, and nutrition-related parameters. A two-stage stratified cluster sampling was used. In the first stage, samples were selected from the main DHS sampling frame developed from enumeration areas. In the second stage, systematic random sampling was employed. The detailed information regarding survey sampling, quality control, management, and survey instruments are reported elsewhere [24–37]. Trained interviewers collected data using face-to-face interviews. Written consent was collected from the respondents before conducting the survey. The survey response rate varied between 85 and 95%. A sample was drawn from the DHS database for analysis, which resulted in a total of sample 140,974 women living in 14 low- resource countries. India had the highest proportion of participants (43,502 women, 31% of the total sample), followed by Egypt (18,254 women, 13% of the sample). The average (standard deviation) age of the participants was 49.54 (± 2.32) years. Of the 90 countries where the DHS surveys have been implemented, BCS related questions in 18 countries (20.0%) [24, 38]. The common themes identified were disease knowledge, screening knowledge, screening practice, and screening outcomes. In this study, data on the utilisation of BCS services was used from the 14 low-resource countries during the period of 2008–2016 namely: Albania (2008–09), Burkina Faso (2010), Colombia (2015), Cote d’Ivoire (2011–12), Dominican Republic (2013), Egypt (2015), Honduras (2011–12), India (2015–16), Jordan (2012), Kenya (2015), Lesotho (2014), Namibia (2013), Philippines (2013), and Tajikistan (2012) [24–37]. However, Equatorial Guinea and Peru were excluded from the analysis because their data were not publicly accessible. Armenia was excluded because it lacked sufficient information related to the study variables. Brazil was excluded due to obsolete data in 1986. Countries were grouped across geographical regions according to the continent such as Africa (i.e., Kenya, Burkina Faso, Egypt, Lesotho, Namibia, Cote d’Ivoire), Asia (e.g., India, Philippines, Jordan, Tajikistan), Europe (e.g., Albania), North America (e.g., Honduras, Dominican Republic) and South America (e.g., Colombia). The study participants were restricted to women aged 40 years or more at risk of developing breast cancer [39–44]. Several types of studies also used a similar inclusion criterion for epidemiological, observations and clinical studies [39–43]. This is because breast cancer diagnosis among younger women is more complex because their breast tissue is usually more dense compared to their older counterpart [40–42]. Participants were asked questions related to their utilisation of BCS services [38]. For example, or ‘have you ever had a mammogram?’ or ‘have you had a clinical breast cancer examination?’ Participants self-reported as their responses in the form of a dichotomous (‘yes’ or ‘no’) and this information was used as the outcome variable in the analytical exploration. Explanatory variables were selected based on different criteria, including epidemiology and published studies on the utilisation of BCS and these data were examined for potential confounders [3, 7, 38, 44]. Explanatory variables were selected based on the available in the DHS data sets. The participants’ characteristics, including age, education, sex of household head and age at the time of respondent’s first birth, were selected as potential predisposing factors in the analyses. Age was grouped as follows: 40–44 years or ≥ 45 years. Participant’s educational background was categorised as: no education, primary education, secondary education or higher education. The head of the participant’s household was defined as ‘male’ if the participants lived in the male-dominated household, or ‘female’ if otherwise. The number of live births was classified as < 4 births, 4–5 births, or > 5 births. Participant’s mass media exposure was assessed by means of access to radio and television in the household. Health insurance coverage, body mass index, and wealth status were considered enabling factors. Health insurance coverage in households was dichotomous (‘yes’ if insured of the participants household or ‘no’ if uninsured). The height and body weight of the participants were measured by trained field research staff. Weight was measured once, with light clothing on and without shoes, by digital weighing scales placed on a flat surface. Height was measured once using a standard clinical height measuring scale with the participant standing without shoes. Body mass index (BMI) was calculated as the ratio of weight in kilograms (kg) to height in meters (m) squared (kg/m2). SES was based on the ownership of durable assets [45]. This method has been used in previous studies using DHS data from developing countries [38, 46, 47]. Each household’s characteristics (assets) were dichotomised (‘yes’ if present and ‘no’ if not). Country-specific principal components analysis (PCA) was performed using this ownership of durable assets [37]. Weights were estimated by factor scores derived from the first principal component in the PCA. The constructed wealth index values were then assigned to individuals based on accessible variables. The wealth index was divided into five groups: poorest (lowest poor 20%), poorer, middle, richer, and richest (top 20%). Furthermore, the wealth index recorded participants into three groups: 40% bottom (poor), middle 40% (middle) or top 20% (rich). Another control variable, the location of residence, was dichotomised as either urban or rural. For the inequality analysis, utilisation of BCS services was performed across wealth quintiles. The standard measures of concentration index (CI) were employed to examine the magnitude of household wealth-related inequality and the trends in utilisation of BCS services across 14 developing countries. The CI was estimated as the covariance of the utilisation of BCS services and the proportional rank in wealth score distribution [47] as follows: where CI is the concentration index, y¯ is the mean utilisation of BCS services, ri is the cumulative proportion that each individual represents over the total population once the latter has been ranked by the distribution of wealth score. The values of CI are bounded between y¯−1 and 1−y¯; y¯−1≤CI≤1−y¯ when y is dichotomous [48, 49]. CI acquires a negative value when the curve lies above the line of equality, which indicates a disproportionately lower prevalence of BCS service utilisation among the poor (i.e., pro-poor). A positive value of CI signifies a higher concentration of health indicators among the rich (i.e., pro-rich). There is no socioeconomic inequality in the distribution of utilisation of BCS services (y) when the value of CI is zero and the concentration curve (CC) coincides with the 45° line. The dichotomous character of the utilisation of BCS services may result in unstable bounds in response to varying means; therefore, the normalised standard index was estimated to check the robustness of the estimation [50, 51]. In addition, when the outcome variable is dichotomous, the CI has to be corrected in order to allow comparisons between groups of individuals from different time periods that may show different levels of use of health services [52]. In the context of a dichotomous outcome variable, the Erreygers’s CI is the CI multiplied by four times the mean health or outcome of interest [53]. Erreygers’ suggested corrected CI can be expressed as: where ymax and ymin are the boundary of y (utilisation of BCS services). When the Erreygers’ corrected index is used, the decomposition of inequality is generally expressed as: This estimate produces an index that satisfies various attractive axiomatic properties for an inequality index, including the sign condition, scale invariance and mirror properties [53, 54]. The adjusted CI method allows for an examination of the causes of (and their corresponding contributions to) and levels of changes in inequalities in terms of the utilisation of BCS services [54]. In addition, multiple logistic regression was applied to measure the likelihood of utilisation of BCS services. Adjusted odds ratios (AORs) with a 95% confidence interval (CI) were estimated for identifying influencing factors on utilisation of BCS services at a 5% or lower level of significance. All statistical analyses were performed with Stata/SE-13 software (StataCorp, College Station, TX, USA).