Data source
This study used self-weighted data from the national representative Family Health Survey (FHS) conducted by the General Authority for Statistics (GaStat) in 2018 in collaboration with several actors in the KSA, including the Ministry of Health (MOH), Saudi Health Council, as well as the private and academic sectors [21]. This nationally representative survey covers the 13 administrative regions of the country and collects information relating to geography, basic characteristics of household members, family income and expenditure, marriage and family planning, fertility and mortality, and health status of individuals, including whether or not they suffer from any chronic diseases, among other topics. The FHS collected a total sample of 15,265 responses randomly selected across all 13 regions of the KSA. For this study, the analysis was limited to respondents who were aged 18 years or older, and provided complete information on all variables of interest, resulting in a sample of 10,785 respondents.
Variables
The survey asked respondents to report their average household monthly OOP expenditure on health and their average household monthly income (in Saudi Riyal [SR]). The outcome variable of interest in this study is the relative OOP health expenditure, estimated as the percentage of income spent on healthcare. OOP health expenditure relates to direct medical costs paid by individuals to access healthcare services, including medicines, consultations, diagnosis, tests, laboratory, radiology, and admission, along with non-medical costs such as transport for both inpatients and outpatients, among others. The Saudi FHS survey expresses OOP expenditure as the total costs paid for healthcare services and does not discriminate whether any or all of the amount has been reimbursed by private health insurance; hence, caution must be exercised in interpreting the outcome, which represents the general OOP health expenditure rather than the net OOP health expenditure that could also include unobservable reimbursements.
The exploratory variables (covariates) are informed by Andersen’s Health Behaviour model, distinguishing the following three categories of factors affecting access to healthcare services [22]: (i) enabling factors, including gender, age group, and marital status; (ii) predisposing factors, including educational level, economic status, employment status, and health insurance coverage status; and (ii) need factors, including periodic check-ups and self-rated health. In addition, these socioeconomic factors are based on evidence from previous studies that demonstrated their relative influence on OOP health expenditure [23,24,25].
Statistical analysis
Difference in OOP expenditure by background characteristics
We calculated the absolute mean OOP health expenditure for each group (with and without chronic illness) separately. We then calculated the relative OOP health expenditure, estimated as the percentage of income spent on healthcare, across the two groups. We further analysed the difference in absolute and relative OOP health expenditure across the socioeconomic covariates.
Inequality in relative OOP health expenditure
We used the concentration curve and index to gauge the extent of inequality in mean relative OOP between individuals with and without chronic illness. The generalized concentration index was used to gauge the extent of inequality in relative OOP health expenditure between the two groups [26]. The concentration index has been frequently used to measure relative inequality in one health variable over the distribution of a socioeconomic variable [27]. To create a variable that ranks households from the poorest to the richest, we used the income to classify households into wealth quintiles.
The concentration index was then derived as follows:
$$\begin{array}{c}\mathrm{CI}= \frac{2}{{\upmu }_{\mathrm{h}}}\sum_{\mathrm{i}=1}^{\mathrm{n}}\left({\mathrm{h}}_{\mathrm{i}}-{\upmu }_{\mathrm{h}}\right)\left({\mathcal{R}}_{\mathrm{i}}- \frac{1}{2}\right)\\ =\frac{2}{{\upmu }_{\mathrm{h}}}\mathrm{cov}\left(\mathrm{h}, \mathcal{R}\right)\end{array}$$
(1)
where n denotes the number of observations, hi is the health variable, μ is the mean of h, and \({\mathrm{R}}_{i}-\frac{1}{2}\) is the fractional socioeconomic rank, ranging from the poorest to the richest [28].
Blinder-Oaxaca decomposition
Given the difference in relative OOP health expenditure across individuals with and without chronic illness, we further ascertained the factors that drive this difference. Toward this end, we applied the Blinder-Oaxaca (B-O) decomposition approach, which has been extensively used to assess differences in health outcomes between two groups [28]. First, we measured the difference in the mean relative OOP between the chronically and non-chronically ill. Second, we applied B-O decomposition models [29] to examine how much of the observed gap can be explained by the differences in the characteristics of the groups by decomposing the differences into two components: one that is explained by the effect of the covariates (covariate effects/explained component) and another that is explained by the difference in the effect of the covariates (coefficient effects/unexplained component) [30].
This decomposition is represented as follows:
$$\mathcal{R}=\mathrm{\rm E}\left({X}_{A}\right) -\mathrm{\rm E}\left({X}_{B}\right)$$
(2)
where E(X) denotes the expected relative OOP accounted for by the group differences, A refers to one group (chronically ill), and B refers to the other group (non-chronically ill). The two-fold decomposition is divided into two components as follows:
$$R=Explained(Q) + Unexplained (U)$$
(3)
$$R={\lfloor E\left({X}_{A}\right)-E\left({X}_{B}\right)\rfloor}^{\mathrm{^{\prime}}} {\beta }^{*}+ \left[E{\left({X}_{A}\right)}^{\mathrm{^{\prime}}}\left({\beta }_{A}-\beta \right)+E{\left({X}_{B}\right)}^{\mathrm{^{\prime}}}\left({\beta }^{*}-{\beta }_{B}\right)\right]$$
(4)
where the first component is attributed to the group differences in the predictors (endowment/explained effect), as follows:
$$Q=\lfloor\mathrm{\rm E}\left({X}_{A}\right)-\mathrm{\rm E}\left({X}_{B}\right)\rfloor\mathrm{^{\prime}} {\beta }^{*}$$
(5)
The second component of Eq. (4) is the contribution of the difference in the coefficients (coefficient/unexplained effect), as follows:
$$U=\mathrm{\rm E}\left({X}_{A}\right)\mathrm{^{\prime}} ({\beta }_{A}-\beta )+E({X}_{B})\mathrm{^{\prime}} ({\beta }^{*}-{\beta }_{B})$$
(6)
Ethical clearance
This study was based on the use of secondary data from the FHS, which was conducted, commissioned, funded, and managed in 2018 by GaStat that was in charge of all ethical procedures. All procedures performed in this study involving human participants complied with the institutional and/or national research committee ethical standards, and with the 1964 Helsinki Declaration and subsequent amendments or equivalent ethical standards. Informed consent was obtained from all participants. All personal identifiers were removed from the dataset by GaStat to allow for secondary data use. GaStat granted permission to use the data and thus no further clearance was necessary as this was performed at the data collection phase.