Socioeconomic and geographical inequalities in health care coverage in Mozambique: a repeated cross-sectional study of the 2015 and 2018 national surveys

This was a repeated cross-sectional study that used the Survey of Indicators on Immunisation, Malaria and HIV/AIDS in Mozambique (IMASIDA) 2015 and the 2018 Malaria Indicator survey (MIS). The surveys consist of four questionnaires: the household, biomarkers, and a survey for men and women each. For this study, only the women’s questionnaire was used.

The Demographic and Health Survey (DHS) Programme conducted both surveys using nationally representative samples of 6,964 and 6,279 households in 2015 and 2018, respectively.

The DHS follows a two-stage probability sample design drawn from the most recent census, stratified by geographic province and by urban/rural areas within each province. The primary sampling units were the census enumeration areas, forming the survey clusters. In the second stage, a complete household listing was conducted in each of the selected clusters.

Three variables representing different dimensions of health care coverage were selected: the use of insecticide-treated nets (ITNs) for children under five, fever treatment of children under five, and Fansidar malaria prophylaxis for pregnant women. These variables were selected because of the following reasons: (i) to maintain consistency, since this work is part of a bigger research project focused on maternal and child health where similar outcomes were used; (ii) the variables are part of the WHO framework that monitors universal health coverage [13]; (iii) they were available in both surveys and (iv) they capture different aspects of health service coverage.

The ITN subsample consisted of 4,756 women with children under five in 2015 and 4,204 in 2018. The fever treatment sub-sample consisted of 1,134 and 1,144 women in 2015 and 2018, respectively. For Fansidar prophylaxis, this sub-sample consisted of 3,977 women in 2015 and 3,524 in 2018.

The ITNs are freely distributed for target groups (women and children) at antenatal care services or through communities’ health campaigns. Fever treatment and Fansidar prophylaxis are also provided free of charge at district hospitals (level II) and health posts and health centres (level I) according to need [14]. The use of ITNs was defined by asking a woman with children less than five years old if the child or children had slept under an ITN the night before the survey. Lack of ITN use was categorised as “yes” if at least one child had not slept under the bed net.

Lack of fever treatment — a standard treatment for children of a certain age (indicated for a range of common childhood diseases, e.g., malaria) — was assessed by asking women with a child under the age of five if the child had had a fever two weeks before the survey. If the answer was affirmative, the follow-up question was if she had sought counselling or fever treatment. Lack of treatment was then dichotomised into “no” and “yes”.

In Mozambique, the national antenatal care guidelines recommend that pregnant women take three doses of Fansidar to prevent malaria infection [15]. Women were asked: “During the pregnancy of your last child, did you take Fansidar to prevent malaria?” If the answer was yes, then the follow-up question was how many times. If any of the recommended doses could not be verified by the antenatal card or reported by the mother, then the participant was classified as not having received the recommended doses.

The independent variables, based on the data available in both surveys, included age, education, wealth, place of residence, and region. Age of the mother was categorised into three groups: 15 to 24, 25 to 39, and 40 to 49 years. Education was classified into three categories: no education, completed primary school, and completed secondary school, or higher than secondary. Wealth index, a proxy for household income, was calculated by the DHS program based on the following households´ assets: television; car; and dwelling characteristics such as flooring material, type of drinking water source, and toilet facilities [16]. Place of residence was dichotomised into rural or urban residence, and the 11 administrative provinces were grouped into three regions: northern (Niassa, Cabo Delgado, and Nampula), central (Zambezia, Sofala, Manica, and Tete), and southern (Maputo, Gaza, and Inhambane). Urban and the southern region were used as reference categories since they are considered as more socially advantaged areas [17] .

The population characteristics were summarised using descriptive statistics to calculate the prevalence of each of the health care outcomes in 2015 and 2018. Then absolute risk differences (ARDs) were estimated as the measure of association between the socioeconomic variables and the outcomes. In addition, the slope index of inequality (SII) was calculated to obtain the absolute social gradient in the outcomes [18]. The SII is a weighted summary measure of health inequality that represents the absolute difference in the estimated values of a health indicator between the most and the least advantaged group, while taking into consideration the population distribution across the social categories [19]. To estimate the SII, ridit scores, corresponding to the mid-point of the average cumulative proportion of the population in each category of the socio-economic variable, were first calculated [20]. Then, SII coefficients were obtained by generalised linear models with the outcome regressed on the ridit scores, separately for each socioeconomic indicator. If there is no inequality, the SII takes the value of zero, while other values indicate social inequality in health. In this study, positive values indicated higher use of the outcome in the socially advantaged subgroups, while negative values indicated higher use in the disadvantaged subgroups. Finally, an interaction term between the ridit scores of the socio-economic variables and time period was included to calculate the SII difference that quantifies and tests the statistical significance of changes in socioeconomic inequalities between 2015 and 2018 [21]. Analyses considered the two-stage probability design of the surveys (using the Stata svyset command), the weighting procedure (as recommended by the DHS Program) and were adjusted for age. All regression models were estimated using a generalized linear model with a binomial distribution and the 95% confidence intervals of the ARD, SII, and SII differences were used to express statistical inference. If the interval did not include zero, the difference was considered to be statistically significant. The analyses were conducted using Stata 15.1 statistical software [22].

For this study, IMASIDA and MIS data were obtained from the DHS website [http://​www.​measuredhs.​com]. These data are anonymous and publicly available; their usage is covered by the ethical approval secured by the DHS for the data collection. The IMASIDA and MIS informed consent forms provide details that participation is voluntary, that the respondent may refuse to answer any question or terminate participation at any time, and that the respondent’s identity and information will be kept strictly confidential. Before each interview, the informed consent statement was handed to the respondent, who could accept or decline to participate.