Article Text
Abstract
Objectives: To test the hypothesis that poverty is associated with infant mortality in Eastern Mediterranean countries and to measure the strength of the association.
Methods: A bibliographic search was conducted. The studies including data regarding deaths during the first year of life, socioeconomic status of the household and/or maternal literacy were selected. Nine studies, conducted in the Eastern Mediterranean region, fulfilled the inclusion criteria. These included seven cross-sectional surveys and two case–control studies. Maternal illiteracy and low socioeconomic status were used to show the level of poverty in each household. Risk estimates for low socioeconomic status and maternal illiteracy were extracted from each study. Meta-analysis was performed for the association between exposure groups of low socioeconomic status and maternal illiteracy and the outcome of death within the first year of life.
Main results: Poverty was associated with an increased risk of infant death (pooled OR 1.52, 95% CI 1.38 to 1.67), significant at p<0.0001. There was a significantly increased risk of infant death among illiterate mothers (OR 1.72, 95% CI 1.42 to 2.08) compared with literate mothers. The meta-analysis OR for an association between low socioeconomic status subgroup and infant death was 1.37 (95% CI 1.25 to 1.49), significant at p<0.0001.
Conclusions: This meta-analysis indicates that there is a significantly increased mortality risk in infants born in poor households. The results suggest that policies aimed at poverty alleviation and female literacy will substantially contribute to a decrease in infant mortality.
Statistics from Altmetric.com
Human development is a multifactorial process in which health, economic and social factors are involved. Health and economic development are associated with each other. Improvements in either result in betterment of the other. Health is a critical problem for poor countries, but it is especially critical for the poorest people within these countries. Evidence worldwide shows that those with lower socioeconomic status (SES) have higher morbidity and mortality rates.1 Extreme poverty is labelled as the world’s biggest killer and the greatest cause of ill health and suffering across the globe.2 More than 10 million children die each year, most from preventable causes and almost all in poor countries.3 The majority of these children belong to the poorest of the poor group.
Causes of higher mortality among the poor are multiple and inter-related. Researchers have tried to explore these causes and the relationship between poverty and mortality-related measures such as infant and child mortality.4–7
The risk of death among infants is associated with the SES of their family and maternal education. The infant mortality rate (IMR), defined as the risk for a live-born child to die before its first birthday, is known to be one of the most sensitive and commonly used indicators of the social and economic status of a population.8 The relation between SES and infant mortality has been studied in both developed and developing countries.9–23 However, all studies do not give coherent results, and the association between infant death and economic status has become a focus of an ongoing debate in the public health literature.
In addition to economic status, maternal education is widely recognized as an important determinant of infant mortality. The relationship between maternal education and infant mortality is complex but most studies have demonstrated reduced rates of infant and child mortality in association with increased levels of maternal education5 15–18 20 24–29 but some studies have found only weak effects of maternal education on infant and early childhood mortality.30
Maternal education can also be considered as a proxy indicator of poverty. Evidence suggests that children from poorer households are generally likely to receive less education. In a patriarchal society, if resources are scarce, sons are preferentially sent to schools rather than daughters. Girls in poor households are more likely to miss out on schooling for several reasons: the direct and indirect costs; investment in girls’ education is not considered worthwhile as daughters are expected to leave the household on marriage.31 The link between maternal education and child health, and the tendency for poorer women to be less educated, is one of the explanations of higher mortality in poorer children.
Understanding the role that poverty plays in infant mortality and identifying its strength of association is important when setting priorities and allocating resources.
A review of studies of infant mortality, from countries that share similar culture, traditions and geographic location, can give a great deal of information regarding the role of poverty and the strength of its association in these countries. Countries of the Eastern Mediterranean region share many cultural practices, such as a patriarchal society, marriage at a young age, multiparity and consanguinity,32–36 all of which affect infant health and mortality. There are few studies from this region that explore the association of infant mortality and poverty. Some of these studies show statistically non-significant associations for this important determinant of infant mortality. Combining the data from these studies, using objective and statistically valid methods, called meta-analysis, can provide an insight into the significance of the association of infant mortality with poverty, and will also determine the strength of this association.
The aim of this study was to test the hypothesis that poverty is associated with infant mortality in Eastern Mediterranean countries and to measure the strength of the association.
To the best of our knowledge, this is the first formal meta-analysis, addressing the issue of poverty and infant mortality in the region.
METHODOLOGY
Data sources
The papers and reports for review were selected from an electronic search and manual search at accessible local and national libraries. Four electronic research databases were searched: Pubmed, Academic Search Premier, the Educational Resource Information Center (ERIC) and Library Information Science and Technology Abstract (LISTA). In addition, Saudi Medbase and the websites of indexed medical journals from the Eastern Mediterranean region were also searched for relevant articles. The studies were retrieved using the following MeSH headings and keywords: “infant mortality”, “neonatal mortality”, “child mortality”, “socioeconomic”, “female literacy”, “maternal education”, “poverty”, “determinants of infant mortality”, “risk factors for infant mortality”, “infant death” and “neonatal death.” The bibliographic databases were searched for relevant papers from their inception to February 2007.
The search was narrowed down by retrieving those studies that had been conducted in Eastern Mediterranean countries, which yielded 657 studies. On reviewing the titles, 62 studies were found to be relevant. The abstracts of these studies were reviewed: many of them were non-contributory as they were review papers, antenatal health policy or clinical studies. The full text of four studies, which seemed relevant, was not accessible. The authors of these studies were contacted. Two of them responded and provided their full text articles, but two did not respond in spite of repeated requests. We did not attempt to locate unpublished studies.
Selection criteria
We included studies in this meta-analysis if they fulfilled the following criteria:
published in the English language;
included data regarding deaths during first year of life, that is infant, neonatal and/or post neonatal mortality data;
included data regarding SES of the household and/or maternal literacy;
analysed infant mortality by SES and/or maternal education, or had data that could be used for calculating the required measures of association.
A total of nine studies fulfilled the eligibility criteria and are included in this meta-analysis.
VARIABLE DEFINITION
Infant death
All deaths that occurred during the first year of life were included, including neonatal, post-neonatal and infant deaths. Data regarding deaths after the first year of life, reported in some selected studies,30 37 38 were excluded from the analysis.
Maternal literacy
The lowest educational attainment level was labelled illiterate and compared with the literate group. For this purpose, data from all literate categories were merged for studies,37 39 40 where more than one category was mentioned in educational status.
Socioeconomic status
The lowest socioeconomic/income group was compared with the remaining better off categories. For this purpose, data of all better off categories were merged, for studies,30 37 40–42 where more than one category was mentioned in SES. For studies,39 43 in which only rural/urban stratification was available, rural was considered as low SES. In one study,44where ownership of land was reported, not owning a land was categorized as low SES.
In this meta-analysis, combination of variables; maternal illiteracy and low SES are used to signify the state of poverty of the household.
Statistical analysis
For studies where risk estimates were not presented, the reported data were used to calculate these estimates, together with their 95% CIs.
The number of infant deaths and sample sizes by maternal education and/or SES of the household were extracted. Two by two and/or row by column tables were constructed and the measure of association calculated. Microsoft Excel and Open Epi45 were used for constructing the tables and computing OR and their 95% CIs.
If the measure of association and 95% CI was reported in the study, it was included in the meta-analysis. For studies41 42 where more than one category was mentioned with their ORs and 95% CIs for individual categories, the reported ORs were merged using Comprehensive Meta-analysis Version 246 to get one single measure of association.
If a measure of association was reported to have a protective effect, it was re-calculated as a measure of risk.
The process of data extraction and calculation was repeated three times to check for and eliminate errors.
Meta-analysis
Meta-analysis was performed for the association between exposure groups of low SES and maternal illiteracy and the outcome of death within the first year of life. Meta-analysis was conducted using the Comprehensive Meta-analysis Version 2. The ORs are presented as a forest plot.47 The forest plot shows study-specific ORs (and their 95% CIs) and the relative weighted contribution of each study, as well as the OR estimate, was pooled across all studies.
RESULTS
The characteristics of the nine studies selected for meta-analysis are outlined in table 1. The studies were conducted in Egypt, Iran, Pakistan, Saudi Arabia and the United Arab Emirates between the years 1980 and 2000. All studies included more than 100 deaths. Two studies,39 41 mentioned infant mortality rates instead of the number of infant deaths. All studies, except two, were cross-sectional surveys. Two case–control studies38 42 were included which investigated the influence of important risk factors on mortality for children less than 5 years. Infant mortality data was extracted and analysed, excluding the deaths after 1 year of age.
These studies reported neonatal, post-neonatal and/or infant deaths. All studies used the same definitions of infant mortality except one30, which had mentioned early infant mortality and included the age group 1–7 months in this category. Grant et al37 discussed neonatal deaths (1–28 days), post-neonatal deaths (1–11 months) and infant deaths (1–12 months) as separate entities.
The maternal educational levels and SES were defined differently in all studies. Although a variety of categories were used to assess educational status, the lowest category in seven of them was illiterate, none or no formal education.
A standard test for heterogeneity found that the pooled data were not homogeneous (Q value 152.68, p value <0.0001). This finding implies that the random effects models are more appropriate than the fixed effects models. Thus, random effects models are presented in this study.
Individual study ORs as well as meta-analysis ORs for the association between maternal illiteracy, household SES and infant mortality are presented in figure 1. The studies that presented infant mortality by groups are listed more than once, for example the studies by Casterline et al30 (neonatal and early infant mortality) and Al Hosaini et al38(Neonatal mortality and infant mortality) are listed twice, whereas the Grant et al study37 (neonatal, post neonatal and infant mortality) is listed three times. In all studies, the risk of infant mortality was higher for infants of illiterate mothers and from low SES households. However, the pattern is different and variable. Two studies,30 44 found statistically non-significant associations between SES and infant mortality as well as maternal illiteracy and infant mortality. However one study38 found the association between infant mortality and low SES statistically significant but the association between infant death and maternal illiteracy was not significant.
When the nine studies were entered into the meta-analysis, the pooled OR was 1.52 (95% CI 1.38 to 1.67) and significant at p<0.0001. Thus, there was a significantly increased mortality risk in infants born in poor household compared with their better-off counterparts.
The meta-analysis ORs were computed for risk of infant deaths in illiterate mothers compared with literate mothers. There was a significantly increased risk of infant death among illiterate mothers (OR 1.72, 95% CI 1.42 to 2.08) compared with literate mothers. The meta-analysis OR for an association between low SES subgroup and infant death was 1.37 (95% CI 1.25 to 1.49), significant at p<0.0001 (table 2).
To examine the findings further, another analysis was performed for examining the risk of infant death in the lowest SES households and illiterate mothers compared with the highest SES households and mothers with the highest educational status. The purpose was to compare the risk of infant death in the poorest group with the most privileged group in these studies. The overall meta-analysis OR was 1.81 (95% CI 1.47 to 2.23), significant at p<0.0001 (fig 2). The subgroups were also analysed individually. The meta-analysis OR was 2.23 (95% CI 1.35 to 3.70) for group comparison between illiterate mothers and highest educational status mothers. The meta-analysis OR for comparison between the lowest SES subgroup and the highest SES and infant death was 1.60 (95% CI 1.30 to 1.96) (table 2).
Three studies30 37 38 reported neonatal deaths. The data were analysed after excluding neonatal mortality. For both low SES and maternal illiteracy, the OR was 1.53 (95% CI 1.37 to 1.71), for maternal illiteracy alone, the OR was 1.74 (95% CI 1.38 to 2.20), and for low SES it was 1.39 (95% CI 1.25 to 1.54).
Further analysis was carried out for the association of neonatal deaths with low SES and maternal illiteracy. For both SES and maternal illiteracy the OR was 1.48 (95% CI 1.17 to 1.87), for maternal illiteracy alone it was 1.66 (95% CI 1.11 to 2.48) and for low SES 1.26 (95% CI 1.12 to 1.42).
DISCUSSION
During the past two decades, meta-analysis has come to play a key role in providing empirical evidence, which is being used to set new policies or make amendments in the existing ones.
This study presents the findings from a meta-analysis of available data on infant mortality in the Eastern Mediterranean region, according to the maternal education and SES of the household. Infant mortality was selected as the outcome indicator as it is regarded as the measure of ill health that is most sensitive to poverty.48 SES is an obvious indicator of poverty but, it is well known that measuring SES in research studies is problematic because, among other reasons, some people might have incentives to over-report or under-report their real income, because of insufficient recall, and the problem on non-cash income.49 In order to overcome this limitation, maternal education, which is more objective and reliable, was merged with SES to better estimate the state of household poverty.
The meta-analysis of nine studies on infant mortality reveal a significant association of poverty and infant mortality. It showed that infants who are born in a poorer household have a 52% greater risk of dying than those in an advantaged household. This finding is in accordance with other studies that have also shown that infants from low socioeconomic families are at greater risk from illness and death than those from high SES families.12 13 15–20
Results of this study support claims that both low SES and maternal illiteracy increase the probability of infant death. The meta-analysis shows that maternal illiteracy has a stronger association with infant deaths than lower SES. This association has been observed in developing countries and in developed countries.15–20
The results from our study show that infants of illiterate mothers are 72% more likely to die during their first year of life than infants of literate mothers. However, when the mothers with highest educational status were compared with the lowest, this association became even stronger. Infants of illiterate mothers are at a more than twofold risk of dying than infants of the mothers with the highest educational attainment. This is in agreement with the earlier findings that the mother’s education is a more decisive determinant of child survival.25–29 50–52 Female literacy strengthens women’s ability to create healthy and better-off households, to make proper use of health services for child bearing and child rearing and to carry out childcare activities to bring up a healthy child.
Neonatal deaths were also included in this study because some of the exogenous factors that are associated with post-neonatal mortality are also important for neonatal deaths and are associated with low SES and maternal illiteracy, for example low birth weight. Poor maternal nutrition, before and during pregnancy, has an adverse impact on pregnancy outcome and infant survival. However, it is known that the neonatal death rate is determined more by endogenous factors, whereas post-neonatal mortality is determined more by exogenous factors, that is environmental and social factors.53 This phenomenon can be observed in the present study also. The effect of poverty on infants became even stronger when the data were analysed after the exclusion of neonatal deaths. Infants from poor households were 53% more at risk of death than their better-off counterparts, and the neonates had an increased risk of 48%. Similarly, the ill effects of illiteracy were also more pronounced: infants of illiterate mothers were 74% more at risk of dying than the infants of literate mothers, and the increased risk was 66% for neonates. Low SES infants were 39% more at risk of death and neonates were 26% more at risk.
There are important caveats to keep in mind in this review. No studies were available from some of the poorest countries, for example Afghanistan, Yemen and Somalia. Different studies define SES and maternal educational level differently. Also, the possibility of publication bias cannot be ruled out. Despite these limitations, the study provides empirical evidence of the role of poverty in infant mortality.
In conclusion, this study offers evidence of a significant statistical association of poverty and infant mortality. The results suggest that policies aimed at poverty alleviation will contribute to a decrease in infant mortality. The study also provides evidence that maternal illiteracy is a stronger predictor of infant mortality than family SES or poverty in general. Therefore, policies aimed at improving female literacy in the Eastern Mediterranean region will result in decreased infant mortality. However, these policy decisions require a multifaceted approach and long-term commitment to human development by policy-makers.
What this study adds
The study reveals a significant association of poverty and infant mortality in the Eastern Mediterranean region.
Infants born in poorer households have a 52% greater risk of dying than those born in an advantaged household.
Maternal illiteracy has a stronger association with infant deaths than lower economic status.
Infants of illiterate mothers are at more than twofold risk of dying than infants of mothers with the highest educational attainment.
Policy implications
Policies aimed at poverty alleviation and female education will contribute to a decrease in infant mortality.
Acknowledgments
The author would like to thank Dr Mohammad Athar Khalil, JMicron Technology Corporation, California, for his valuable contribution in data extraction and calculation of measures of association, Dr Michael Borenstein, Biostat Company, New Jersey, for his expert advice regarding technical issues in meta-analysis and Dr Amjad Ali Omar, King Fahad Specialist Hospital, Buraidah, for his helpful comments and suggestions. I also wish to thank Dr Abdullah Mohammed Al-Saigul, Assistant Director General for Primary Health Care Administration, Qassim, Dr John B. Casterline, Lazarus Professor in Population Studies, Ohio State University and Dr Ahmad Reza Hosseinpoor, Epidemiologist, World Health Organization for providing relevant literature for this meta-analysis.
REFERENCES
Footnotes
Competing interests: None.
Linked Articles
- In this issue