Educational inequalities and self-reported health among men and women aged 18-49 years in Yangon Region, Myanmar: analysis of a population-based, cross-sectional study

Health status in Myanmar

Myanmar, classified as a low-income and middle-income country, experiences a significant burden of disease compared with high-income countries.1 According to the Global Burden of Disease study in 2017, Myanmar’s healthy life expectancy (HALE) at birth was 59.9 years (57.3–62.3), in contrast to the higher HALE in high-income countries such as the UK and Norway, with figures of 69.3 years (66–72.1) and 70.2 years (66.8–73.2), respectively.1 Moreover, Myanmar has undergone an epidemiological transition, shifting from a substantial decrease in the burden of communicable diseases to a rapid increase in non-communicable diseases (NCDs), resulting in a complex double burden of disease.2 In 2007, the top 5 causes of disability-adjusted life-years in Myanmar for both sexes and all ages were tuberculosis, neonatal conditions, stroke, lower respiratory infections and diarrhoeal diseases, ranked accordingly.3 However, by 2017, the order shifted to neonatal conditions, stroke, tuberculosis, lower respiratory infections and ischaemic heart disease.3 Multiple factors pose challenges to addressing Myanmar’s considerable disease burden. These include a fragile healthcare system due to decades of underinvestment in healthcare, disproportionate allocation of healthcare resources (including finance, infrastructure and human resources), challenges within the primary healthcare system stemming from its traditional focus on communicable diseases, high user fees and direct out-of-pocket payments due to the absence of a reliable health insurance system and cost-sharing policies, and language barriers arising from the nation’s diversity with 135 ethnicities speaking over 100 languages and dialects.2 4

Self-reported health

One of the goals of the 2030 Sustainable Development Agenda, embraced by every member of the United Nations in 2015, is to ensure healthy lives and promote well-being for all at all ages, as outlined in Goal 3.5 Achieving this goal begins with understanding the current situation, and self-reported health (SRH) data can provide that essential information. Because SRH is a widely used indicator reflecting an individual’s physical and mental health,6 7 as well as a strong predictor of morbidity and mortality.8–10 Though the specific wording and response choices differ in self-rated health questions, a widely employed approach is to ask individuals to assess their health status using a single question on a 4-point or 5-point scale, ranging from very good to poor, or very good to very bad.11 12 These responses are influenced by health behaviour, physical ability or performance, and current illnesses.13 Despite its subjectivity, SRH has proven to be reliable14 and valid.15 16 Additionally, owing to the ease, speed and cost-effectiveness of gathering SRH data, these measures can be valuable in national surveys of low-income countries to rapidly identify high-risk groups for disease prevention and early intervention.17

While population-based studies have estimated the prevalence of SRH in many countries, research on SRH among adult citizens in Myanmar is limited. A study conducted in 2002, encompassing 69 countries, reported that 2.9% of individuals in Myanmar reported bad or very bad health.18 But, these data are outdated and insufficient to assess the current health status of Myanmar’s adults. Since then, several population-based studies have been carried out regarding SRH in Myanmar. However, these studies have primarily focused on specific aspects, such as self-rated oral health status,19 or targeted specific populations, including the elderly20 and migrants.21 There have been no recent population-based studies in Myanmar examining the prevalence of SRH among adult citizens.

Social determinants of health

When assessing the prevalence of SRH, considering it from the perspective of social determinants of health can provide valuable insights due to their significant influence on individuals’ health outcomes at every stage of life.22 These social determinants of health encompass factors like education, economic stability, food, neighbourhood and physical environment, community and social context, and the healthcare system.23 Among these determinants, socioeconomic status is a crucial and commonly used indicator to investigate health inequalities, referring to an individual’s position within a structured social hierarchy.24 Education, one of the socioeconomic factors, emerges as a key integrated marker as it influences both occupation and income.25

A positive relationship between education and health has been reported by numerous studies.26–28 This association is explained through several mechanisms. Ross and Wu proposed three potential mechanisms, which are frequently mentioned as major mediators: work and economic conditions, healthy lifestyle and social–psychological resources.25 For example, a higher level of education is related to a higher employment rate29 and increased income.30 It is also linked to healthier behaviours, such as reduced smoking, lower alcohol consumption and increased physical activity.31 Additionally, higher education levels contribute to an increased sense of personal control32 and higher levels of social support.33 All of these factors collectively contribute to improved health outcomes.34–38 Understanding the intricate interplay of education and its impacts on health remains a complex and multifaceted endeavour.

Gender gap in educational inequalities on health

Gender differences in health are firmly established. While men face higher mortality rates39 and are at greater risk of life-threatening illnesses compared with women,1 they tend to rate their health more positively.40–42 Conversely, women are more prone to non-fatal illnesses,1 as well as minor chronic conditions and acute conditions,43 and they typically report poorer health than men.40–42 In Myanmar, mortality trends align with these gender differences,44 yet recent information on morbidity and SRH by gender is lacking.

Though the precise reasons for women’s lower SRH are not fully understood, suggested factors contributing to the gender gap in SRH from previous studies include biological distinctions between the sexes, differences in male and female psychology (such as risk perception, illness definition and coping strategies), variations in how genders tend to report their health (under-reporting or over-reporting) and societal gender inequalities.40 45

Additionally, education is widely acknowledged as a factor that can lead to improved health outcomes for both women and men. However, the extent of gender disparities in the relationship between education and health may differ by gender. Recent review studies indicate that women may experience fewer or no mortality benefits from education compared with men.46 47 While research concerning morbidity is limited, a study from the USA indicates that as educational levels rise, women tend to experience a lower burden of physical impairment compared with men.48 Moreover, the consistency of gender differences in the benefits of education on SRH varies across different studies.49–51 In Myanmar, as far as we are aware, there is one study that has researched this issue, indicating a significant negative association between years of schooling and mental distress among women and men aged more than 30 years, but not among the younger men.52 For comparison of gender differences in the benefits of education on morbidity and SRH in Myanmar, more studies are warranted.

Aims

In this study, we aimed to estimate the prevalence of good SRH in subpopulations based on the social determinants of health (sex, area of residence, family type, education, occupation and income) and to investigate the association between education (measured in years of schooling) and good SRH among men and women in the Yangon Region of Myanmar.

Study design and setting

The present manuscript reports an analysis of data from a population-based, cross-sectional study among men and women aged 18–49 years in the northern and southern districts of the Yangon Region of Myanmar, conducted in October and November 2016.52–54 Data were collected in the Yangon Region, the most developed and densely populated region in Myanmar, and included all ethnic groups residing there. However, it excluded nuns, monks, soldiers, institutionalised people and those deemed too ill physically and/or mentally to participate. Domestic violence was the motivation for conducting the original study.53 Therefore, the sample size was determined based on the overall aims of the previous study, which was to estimate the prevalence of domestic violence among married women (21%), as reported by the Myanmar Demographic and Health Survey (DHS, 2015–2016).55 A sample size of 2400 persons was estimated, based on a prevalence of 21%, a power of 80%, a significance level of 0.05 and allowing for a 20% non-response rate, as well as enabling sub-group analyses by sex.53

A multistage sampling procedure was used. Two out of the four districts in the Yangon Region, namely the north and south districts, were selected based on the 2014 Myanmar Population and Housing Census.56 These districts were chosen because they comprise both urban and rural areas, whereas the east and west districts consist solely of urban areas. Subsequently, urban wards and rural villages were listed from these two selected districts. In the northern district, there were 125 wards (urban subunits of a township) and 235 villages (rural subunits of a township), while the southern district had 110 wards and 375 villages. Eight wards and eight villages were then randomly chosen from each of the two districts, resulting in a total of 16 wards and 16 villages being selected. From these, 2,731 households were sampled proportionally to the total population sizes in the north and south districts of the Yangon Region and in proportion to the urban and rural population sizes within these two districts. Lists of households were obtained from local authorities.56 One person in each household was asked to make a list of family members aged 18–49 years. Based on this list, one of the family members was randomly asked to participate in the study through a sealed envelope method.57 From every second household, a woman or a man was randomly selected. Out of the 2,731 sampled households, we couldn’t get in touch with 89 households. Among the remaining 2,642 households, a total of 130 invitees declined to participate. Additionally, six individuals had missing data on one of the variables used in the regression analysis. Consequently, the final sample used for analysis consisted of 2,506 individuals, comprising 1,135 men and 1,371 women, resulting in a response rate of 91.8%. Given the high response rate and the representativeness of the respondents for men and women aged 18–49 years in the northern and southern districts of the Yangon Region, the data were deemed suitable for investigating the objectives of this study.

Data collection and measurement

Data were collected through structured interviews, using standard questions adapted from the Myanmar DHS (2015–2016).55 These questions had previously been translated into Burmese, the most commonly spoken language in Myanmar, and then back-translated into English by a professional translator. A pilot survey was conducted on 1 and 2 October 2016 in 54 households within a single ward (Dagon Seikkan township, which was not included in the main survey area). The purpose of this pilot survey was to assess the clarity, cultural acceptability and understanding of the questions. The interviews were conducted face-to-face using questionnaires (online supplemental appendix 1, pp1–3). (Three different questionnaires were used in the project: married women, unmarried women and men questionnaires. All three questionnaires included the same questions, which were used in the present analyses. In the appendix, we have included the English-translated original questionnaire for married women to present the relevant questions and variables.) Before the interviews, all participants were informed about the study’s purpose, potential risks and benefits of participation. After obtaining informed written consent from the interviewee, the interview proceeded. We ensured privacy, anonymity and confidentiality for all participants. Participants had the right to refuse to answer any questions or decline participation, and they could withdraw from the study at any time without facing any consequences.

Directed acyclic graph (DAG): conceptual framework

We drew a DAG58 to identify confounders, mediators and colliders as part of our strategy for the Poisson regression analyses examining the association between years of schooling and good SRH (figure 1). Drawing from prior evidence, the present research framework postulates that sex, age and area of residence are confounders because each of these factors influences both education and SRH.45 59–62 In Myanmar, according to 2016 UNESCO statistics, girls show a slightly higher completion rate in primary, lower and upper secondary school compared with boys.59 Additionally, urban areas exhibit significantly higher completion rates across these educational levels based on the same statistics.59 Due to the limited data available for Myanmar, when examining results from other countries, it’s observed that men and urban residents tend to report better health outcomes compared with women and rural area residents.45 60–62 Furthermore, we considered three categories of variables that mediate a connection between education and SRH: work and economic factors (spousal education, occupation, income, household debt)29 30 34 35; social–psychological factors (family type, marital status, number of children, mental health)32 33 38 and physical health.7 26–28 After controlling for confounding variables, we assumed a weakened association.

Study variables

Statistical methods

Data management and statistical analyses were conducted using STATA/IC V.16.0. For descriptive analyses, we estimated the prevalence of good SRH among men and women in the Yangon Region of Myanmar by characteristics such as sex, age, area of residence, family type, education, occupation and income, along with 95% CIs. We computed the 95% CI to present the uncertainty of these estimates. Analyses were stratified by sex.

Since our aim was to estimate the total effect of education on good SRH, we adjusted for confounders but not mediators. We opted for calculating prevalence ratios (PRs) with 95% CIs using Poisson regression instead of ORs with logistic regression. This choice was made because when the outcome is common (ie, prevalence>10%), ORs can potentially overstate the association.65 66

We examined interactions between the exposure and confounders and identified an interaction between years of schooling and sex. However, we did not observe any other significant interactions in other variables. Consequently, we conducted separate analyses only by sex.

For each sex, we analysed two models. Model 1 was the crude model, while Model 2 was adjusted for confounders (age and area of residence). We assessed the assumptions of the Poisson model, including linearity and homoscedasticity (constant error variance), by plotting residuals versus predicted values. We confirmed that the observations were independent of each other, as only one person from each household was included in the sample.

Patient and public involvement

None.

Descriptive Statistics

The sample comprised 2,506 participants, including 1,135 men and 1,371 women. Women accounted for a higher proportion (54.7%) than men (45.3%). The mean age of participants was 33.1 years (SD=9.1), while the mean years of schooling were 7.7 years (SD=4.0). Men had more years of education (8.0 years) than women (7.6 years).

Table 1 displays the prevalence of the four categories of SRH. It shows that the prevalence of ‘very good’ SRH was similar among men (12.4%) and women (11.2%). However, a higher prevalence of ‘poor’ SRH was observed in women compared with men (35.4%, 95% CI 32.9 to 37.9 vs 20.7%, 95% CI 18.4 to 23.2).

As shown in table 2, when dichotomising SRH as ‘good’ (very good and good) and ‘poor’ (not so good and poor), the overall prevalence of good SRH was 61.2% (95% CI 59.3 to 63.1). Notably, a higher proportion of men (72.0%, 95% CI 69.3 to 74.5) reported good SRH compared with women (52.3%, 95% CI 49.6 to 54.9). Variations in the prevalence of good SRH were observed across different age groups, decreasing from 69.2% in the youngest age group (18–29 years) to 49.0% in the oldest age group (40–49 years). This disparity was consistently evident among men but not among women. Residents in urban areas reported a higher prevalence of good SRH compared with rural areas: 63.6% (95% CI 60.8 to 66.3) versus 59.2% (95% CI 56.5 to 61.8). This difference was consistently observed among women but not among men. Regarding the family type, individuals living with extended families reported good SRH more frequently (66.6%, 95% CI 63.7 to 69.4) than those living with nuclear families (57.5%, 95% CI 54.9 to 60.0). However, this difference was found among men but not among women. Participants with less than 6 years of education had a lower prevalence of good SRH (56.3%, 95% CI 53.0 to 59.4) than those with higher educational levels (63.5%, 95% CI 60.7 to 66.2 and 66.0%, 95% CI 61.3 to 70.5). This difference was consistently found among men but not among women. Lastly, there was no difference in the prevalence of good SRH among different daily household income levels in the general population and for both sexes.

Association between years of schooling and good SRH

In separate Poisson regression analyses among men and women, adjusted for confounders (age and area of residence), the association (PR) between years of schooling and good SRH was statistically significant among men (PR=1.01, 95% CI 1.01 to 1.02, p=0.002), but not among women (PR=1.01, 95% CI 0.99 to 1.02, p=0.415) (table 3). The findings indicate that a 1-year increase in education is associated with a 1% increase in the prevalence of good SRH among men.

Patient consent for publication

Not applicable.

Ethics approval

This study involves human participants and was approved by the Norwegian Regional Committees for Medical and Health Research Ethics (2016/1195) and the University of Public Health, Yangon, Ministry of Health and Sports in Myanmar. The Regional Health Department of Yangon also granted permission and provided supporting documents to the research team to conduct the present study. Participants gave informed consent to participate in the study before taking part.