Pan American Journal Original research of Public Health

Social inequalities in maternal mortality among the of

Antonio Sanhueza,1 Jakeline Calle Roldán,2 Paulina Ríos-Quituizaca,3 Maria Cecilia Acuña,1 and Isabel Espinosa1

Suggested citation Sanhueza A, Calle Roldán J, Ríos-Quituizaca P, Acuña MC, Espinosa I. Social inequalities in maternal mortality among the . Rev Panam Salud Publica. 2017;41:e97.

ABSTRACT Objective. This study set out to describe the association between the maternal mortality ratio (MMR) estimates and a set of socioeconomic indicators and compute the MMR inequali- ties among the provinces of Ecuador. Methods. A cross-sectional ecological study was conducted, using data for 2014 from the country’s 24 provinces. The MMR estimate was calculated for each , as well as the association and its strength between MMR and specific socioeconomic indicators. For the indi- cators that were found to be significantly associated with MMR, inequality measurements were computed. Results. Despite a relatively low MMR for Ecuador overall, ratios differed substantially among the provinces. Five socioeconomic indicators proved to be statistically significantly associated with MMR: total fertility rate, the percentage of indigenous population, the percent- age of households with children who do not attend school, gross domestic product, and the percentage of houses with electrical service. Of these five, only three had MMR inequalities that were significant: total fertility rate, gross domestic product, and the percentage of households with electricity. Conclusions. This study supports research arguing that national averages can be mislead- ing, as they often hide differences among subgroups at the local level. The findings also suggest that MMR is significantly associated with some socioeconomic indicators, including ones linked with significant health outcome inequalities. In order to reduce health inequities, it is crucial that countries look beyond national averages and identify the subgroups being left behind, explore the particular social determinants that generate these health inequalities, and examine the specific barriers and other factors affecting the subgroups most vulnerable to maternal health inequalities.

Keywords Maternal mortality; health inequalities; social determinants of health; health equity; Ecuador; Latin America.

The year 2015 marked the end of the health and education for millions of peo- advances at the national level, insufficient Millennium Development Goal (MDG) ple. In terms of maternal mortality, Ecua- progress was made at the subnational era. For Ecuador, that period had been one dor was one of the countries of Latin level, with thousands of people still liv- of economic growth (1), improvements in America and the Caribbean that experi- ing in poverty and thousands of women water quality and in sanitation, and an enced the steepest reductions in the ma- losing their lives due to preventable expansion of social services, including ternal mortality ratio. For the country as a pregnancy-related causes (3). whole, the ratio decreased from 185 ma- Today, Ecuador and other countries 1 Pan American Health Organization, Washington, D.C., United States of America. Send correspon- ternal deaths per 100 000 live births in face a new global health agenda that dence to Antonio Sanhueza, [email protected] 1990 to 64 deaths per 100 000 live births in prioritizes universal health and equity, 2 Ministry of Public Health of Ecuador, , Ecuador. 3 Facultad de Ciencias Médicas, Universidad 2015, a reduction of 65.4% (2). Despite including through the Sustainable De- Central del Ecuador, Quito, Ecuador. these impressive economic and social velopment Goals (SDGs) and the new

Rev Panam Salud Publica 41, 2017 1 Original research Sanhueza et al. • Social inequalities in maternal mortality in Ecuador

Global Strategy for Women’s, Children’s was considered a proxy for poverty. On socioeconomic indicator, and the fifth and Adolescents’ Health (4). In order for the other hand, the relationship between quintile (Group 5) represents the most these nations to take on these ambitious fertility and social factors helps to put advantaged provinces for the indicator. targets, they have to start looking be- the fertility rate into a larger, overall The AD and the RR indicate the gap be- yond national averages and identify the ­context, including identifying the most tween the most disadvantaged and the subgroups that are being left behind. It vulnerable sectors. most advantaged groups. The AD and is also important that the countries start The analyses included the MMR esti- the RR were computed by, respectively, exploring the specific barriers and other mate for each province, which was calcu- subtracting and dividing the MMR in conditions affecting these subgroups, lated by dividing the number of maternal each of these two groups. Higher values and identify the mechanisms generating deaths by the total number of live births of AD and of RR indicate greater inequal- current health inequalities. reported in the year 2014, expressed as ity in maternal mortality. To help Ecuador prepare for these the number of maternal deaths per The MSII and PRSII were computed tasks, this study had two objectives: (1) 100 000 live births. A 95% confidence in- through regression model fitting. In or- to describe the association between ma- terval (CI) for the MMR in each province der to fit the models, the complete data ternal mortality and a set of socioeco- was also computed. set of the provinces was first ordered by nomic indicators and, (2) based on those First, an exploratory data analysis was socioeconomic indicator status, from the socioeconomic indicators, to compute carried out in order to ascertain the MMR most disadvantaged to the most the inequalities in maternal mortality and the selected socioeconomic indica- advantaged. among the provinces of Ecuador, using tors in Ecuador. Then, a study of MMR The MSII is based on the estimated data for the year 2014. inequality was conducted for each of the slope parameter by fitting a linear regres- selected socioeconomic indicators by sion model that considers the MMR as MATERIALS AND METHODS taking two steps: (1) study the associa- the dependent variable and the Ridit (the tion and its strength between MMR and cumulative relative position of each For this research, a cross-sectional eco- each of the socioeconomic indicators and province with respect to the socioeco- logical study was conducted in 2014, us- (2) for the socioeconomic indicators that nomic indicator, which ranges between ing data from the 24 provinces of were found to be significantly associated 0 and 1) as the independent variable. The Ecuador. with MMR, compute the MMR inequal- weighted least squares method was used The variables considered in this study ity measurements. in this case. were one heath indicator, nine socioeco- A weighted least squares regression The MSII was computed as MSII = b nomic indicators, and one demographic model was used to study the association (RiditMin - RiditMax), where b is the esti- indicator. between MMR and the socioeconomic mated slope parameter computed by fit- The one heath indicator was the mater- indicators. The weights were the number ting the linear regression model. nal mortality ratio (MMR). of live births in each province. The The RiditMin and the RiditMax are, respec- The nine socioeconomic indicators strength of the association was assessed tively, the observed minimum and maxi- were: (1) total fertility rate; (2) percent- with the Pearson correlation coefficient. mum Ridit values. The theoretical age of indigenous population; (3) per- The bootstrap method was used by con- minimum and maximum Ridit values are centage of households with children who sidering 2 000 resamples in order to esti- 0 and 1, so that MSII = -b. do not attend school; (4) gross domestic mate the Pearson correlation and its The 95% CI for the MSII was computed product (GDP) per capita; (5) average bias-corrected 95% CI. as [bU (RiditMin - RiditMax), bL (RiditMin - Rid- household income; (6) percentage of Simple and complex measures of in- itMax)], where bL and bU are, respectively, poverty with unsatisfied basic needs; (7) equality were computed to explore the the lower and upper limits computed percentage of households with inade- magnitude of MMR inequality for each from the 95% CI for the slope parameter. quate services (e.g., no connection to a socioeconomic indicator. The simple In order to compute the PRSII, the piped water supply or to the sewer sys- measures included computing the abso- Poisson regression model was fitted by tem or a septic tank); (8) percentage of lute difference (AD) and the relative ratio considering the number of maternal households with electricity; and (9) aver- (RR) (8). The complex measures included deaths as the response variable and the age number of persons per bedroom. the modified slope index of inequality Ridit as the independent variable. The one demographic indicator was (MSII) proposed by Bacallao in 2007 (9) The formula for the PRSII was PRSII = live births. (which is an absolute measure of in- exp{b (RiditMin - RiditMax)}, where exp is the The indicators used in this study were equality) and the Poisson relative slope exponential function and b is the estimated derived from three key information index of inequality (PRSII) (which is a slope parameter computed by fitting sources (5–7) on socioeconomic condi- relative measure proposed for this work the Poisson regression model. RiditMin tions in Ecuador and its provinces. that utilizes the Poisson regression and RiditMax are, respectively, the observed Some of the associations between the model) (10). We computed 95% CIs for minimum and maximum Ridit values. socioeconomic indicators and the MMR all those inequality measurements. The theoretical minimum and maxi- are intuitive, but others may not be. For In order to compute the AD and the mum Ridit values are 0 and 1, so that the instance, the average number of people RR, quintiles of provinces for each of the PRSII = exp{-b}. per bedroom is an indicator of the qual- socioeconomic indicators were created. The formula for the 95% CI for the ity of life and of living space. It especially For the purpose of this study, the first PRSII was [exp{bU (RiditMin - RiditMax)}, reflects overcrowding and thus social quintile (Group 1) represents the most ­exp{bL (RiditMin - RiditMax)}], where bL and deprivation. Therefore, this indicator disadvantaged provinces in terms of the bU are, respectively, the lower and upper

2 Rev Panam Salud Publica 41, 2017 Sanhueza et al. • Social inequalities in maternal mortality in Ecuador Original research limits of the 95% CI for the slope param- MMR estimate was Zamora Chinchipe, Bolívar province having the lowest value eter computed by fitting the Poisson re- with 142.2 deaths per 100 000 live births (US$ 582.0) and having gression model. (Figure 1). the highest value (US$ 1 859.0). In terms Data entries were made in Microsoft The average total fertility rate in Ecua- of living in poverty with unsatisfied ba- Excel software, and the statistical analy- dor is 2.9 children per woman of child- sic needs, that is true for 39.0% of the ses were carried out using SAS version bearing age, ranging between 2.1 in population overall, with values ranging 9.2 software. Graphs were generated us- Pichincha and 4.1 in Morona Santiago from 16.6% in to ing Tableau version 9.3 software. (Table 1). 62.9% in (Table 1). Ethical review was not required for Only 20% of the population overall is In the case of housing conditions in Ec- this study since it does not contain any considered to be indigenous, with the uador, the percentage of households human research data (we utilized data percentage ranging from 0.1% in Manabí with inadequate services ranges from that are publicly available online). to 70% in Napo (Table 1). With respect to 6.4% in to 52.8% in Ore­ education, Ecuador is relatively well off. llana province, with a national median of RESULTS Overall, 0.9% of the households still have 22.2%. Further, although the overall per- children who do not attend school; centage of households with electricity is As of 2014, the estimated national among the provinces, that ranges from relatively high in the country (94.0%), the MMR in Ecuador was 49.3 deaths per 0.1% to 3.5% (Table 1). percentage still ranges from 73.8% to 100 000 live births. Despite this relatively The median GDP per capita in Ecua- 99.2% (Table 1). low national average, 14 provinces had dor is US$ 3 920.5, with values ranging In terms of the association between the MMR estimates higher than that. The from US$ 2 545 in Morona Santiago prov- MMR and studied socioeconomic indica- province with the lowest MMR was ince to US$ 9 021 in Pichincha province. tors in Ecuador, MMR was only statisti- Azuay, with 19.1 deaths per 100 000 live The median of the average household cally significantly associated with five births. The province with the highest income per person is US$ 786.2, with socioeconomic indicators: total fertility

FIGURE 1. The maternal mortality ratio (MMR) (number of maternal deaths per 100 000 live births) in the provinces of Ecuador, in study of social inequalities in maternal mortality, 2014a

150

125

100

75 MMR

National average 50

25

0

Loja Azuay Cañar Napo Guayas Carchi EI Oro Manabí Bolívar Los Ríos Cotopaxi Pastaza Orellana Galápagos Esmeralda Pichincha Imbabura Tungurahua Sucumbíos ChimborazoSanta Elena Santo Domingo Morona Santiago Zamora Chinchipe a The maternal mortality ratio for Galápagos was 0 in 2014.

TABLE 1. Exploratory analysis of socioeconomic indicators in study of social inequalities in maternal mortality in Ecuador, 2014

Indicator Minimum Maximum Mean Standard deviation Median Gross domestic product (US$ per capita) 2 544.5 9 020.9 4 296.4 1 588.6 3 920.5 Household average monthly income (US$ per capita) 582.2 1 858.5 816.3 251.9 786.2 Poverty with unsatisfied basic needs (%) 16.6 62.9 39.0 14.1 37.8 Households with inadequate services (%) 6.4 52.8 25.6 15.5 22.2 Households with children who do not attend school (%) 0.1 3.5 1.0 0.7 0.9 Average number of persons per bedroom 1.5 2.2 1.8 0.2 1.8 Households with electricity (%) 73.8 99.2 94.0 0.1 96.3 Indigenous population (%) 0.1 70.0 20.0 20.0 10.0 Total fertility rate (births per women) 2.1 4.1 2.9 0.5 2.9

Rev Panam Salud Publica 41, 2017 3 Original research Sanhueza et al. • Social inequalities in maternal mortality in Ecuador rate, percentage of indigenous popula- also have a higher MMR. The strength of domestic product (P value = 0.001). tion, percentage of households with the association, which is provided by the Given that the strength of this associa- children who do not attend school, per- correlation between these two variables, tion is -0.68 (95% CI [-0.84, -0.45]), one centage of houses with electrical ser- is 0.57 (95% CI [0.25, 0.79]), which is the can conclude that the association is nega- vices, and GDP (Table 2). While three of second strongest association (Table 2). tive, which means that the provinces them are positively associated (meaning Regarding the association between with a higher gross domestic product the higher the value of the socioeco- MMR and percentage of indigenous have a lower MMR. Out of the five socio- nomic indicator, the higher the MMR), population, it is evident that they are economic indicators that are statistically two of them are negatively associated positively associated, with a P value of significantly associated with MMR, gross (meaning the higher the value of the so- 0.0256. The strength of the association is domestic product has the strongest asso- cioeconomic indicator, the lower the 0.38 (95% CI [0.06, 0.65]), which proves to ciation (Table 2). MMR). The socioeconomic indicators be the weakest association out of the Another negative association can be with the strongest associations with five socioeconomic indicators (Table 2). found between MMR and the percentage MMR are gross domestic product and to- Analyzing the association between of households with electricity (P value = tal fertility rate. Figure 2 shows the dis- MMR and the percentage of households 0.0039). The strength of the association is tribution of MMR across the quintiles of with children who do not attend school, -0.44 (95% CI [-0.68, -0.10]) (Table 2). provinces for each of these five socioeco- it is possible to conclude that they are Of the five socioeconomic indicators nomic indicators. positively associated (P value = 0.0189), that were statistically associated with In terms of the association between with a 0.39 strength of association (95% MMR, only three proved to have statisti- MMR and total fertility rate, results indi- CI [0.03, 0.65]) (Table 2). cally significant MMR inequality mea- cate that they are positively associated Another socioeconomic indicator that sures when using the MSII and PRSII: (P value = 0.0031), which means that proves to be statistically significantly as- total fertility rate, GDP, and the percentage provinces with a higher total fertility rate sociated with MMR in Ecuador is gross of households with electricity (Table 3).

TABLE 2. Association between the maternal mortality ratio and socioeconomic indicators in study of social inequalities in maternal mortality in Ecuador, 2014

Indicator Coefficient estimate P value Pearson correlation Lower limit Upper limit Gross domestic product −0.004 0.0138 −0.68 −0.84 −0.45 Average household income −0.034 0.1205 0.24 −0.57 0.15 Poverty with unsatisfied basic needs 0.470 0.0633 0.34 −0.01 0.64 Percentage of households with inadequate services 0.462 0.0652 0.32 −0.02 0.63 Percentage of households with children who do not attend school 16.881 0.0189 0.39 0.03 0.65 Average persons per bedroom 21.698 0.3947 0.35 −0.03 0.64 Percentage of households with electricity −232.380 0.0039 −0.44 −0.68 −0.10 Percentage of indigenous population 48.751 0.0256 0.38 0.06 0.65 Total fertility rate 23.842 0.0031 0.57 0.25 0.79

FIGURE 2. Disaggregated maternal mortality ratio (MMR) (number of maternal deaths per 100 000 live births) across the quintiles of socioeconomic indicators, in study of social inequalities in maternal mortality in Ecuador, 2014a

Gross domestic product

Percentage of households with children who do not attend school

Percentage of households with electricity

Percentage of indigenous population

Total fertility rate

30 40 50 60 70 80 90 100 110 MMR Groups by socioeconomic dimension G1 G2 G3 G4 G5 a The first quintile, G1, represents the most disadvantaged provinces in terms of the socioeconomic indicator, and the fifth quintile, G5, represents the most advantaged provinces.

4 Rev Panam Salud Publica 41, 2017 Sanhueza et al. • Social inequalities in maternal mortality in Ecuador Original research

The AD between the group with the Regarding the MMR inequality that the group with the highest percentage of highest total fertility rate (Group 1) and GDP produces, the AD value of 60 (Table households with electricity (Group 5). that with the lowest total fertility rate 4) indicates that there are 60 additional The RR value of 1.9 (Table 4) indicates (Group 5) was 70 (Table 4). This means maternal deaths per 100 000 live births that women from Group 1 have almost that there are 70 additional maternal occurring in the group with the lowest twice the risk of dying from maternal deaths per 100 000 live births occurring GDP (Group 1), compared to the group causes as do women from Group 5. in the group of provinces with the high- with the highest GDP (Group 5). The RR Considering all the values of each est total fertility rate than in the prov- value of 2.4 (Table 4) indicates that province, the MSII value of 22.6 (Table 3) inces with the lowest. Further, the RR of women from Group 1 have more than indicates that there are an additional 23 MMR between Group 1 and Group 5 twice the risk of dying from maternal women per 100 000 live births who die was 2.8 (Table 4), indicating that women causes as do women from Group 5. from maternal causes in the province who live in the group of provinces Considering the values for each prov- considered most disadvantaged in terms that are the least advantaged in terms of ince in Ecuador, the MSII value of 28.8 of the percentage of households with this socioeconomic indicator are almost (Table 3) indicates that there are approxi- electricity as compared to the women in 3 times as likely to die from maternal mately 32 additional maternal deaths per the most advantaged province. Lastly, in causes as are women who are born in the 100 000 live births occurring in the prov- terms of PRSII, the value of 1.6 (Table 3) group of provinces that are the most ince with the lowest GDP than in the indicates that women who are from the advantaged. province with the highest GDP. Further, most disadvantaged province in terms of With respect to the total fertility rate the PRSII value of 1.8 (Table 3) indicates this socioeconomic indicator have a risk for each province, the MSII value of 26.1 that women from the most disadvan- of dying from maternal causes that is 1.6 (Table 3) indicates that there are 26.1 taged province are almost twice as times as high as that of women from the more maternal deaths per 100 000 live likely to die from maternal causes as most advantaged province. births in the province with the highest are women from the most advantaged total fertility rate (the most disadvan- province. DISCUSSION taged). The PRSII value of 1.7 (Table 3) In terms of housing conditions, the AD indicates that women who live in the value of 42 (Table 4) indicates that there In recent years, Ecuador has made most disadvantaged province in terms are approximately 42 additional mater- significant achievements in terms of of total fertility rate have almost twice nal deaths per 100 000 live births hap- overall economic growth, poverty re- the risk of dying from maternal causes pening amongst women from the group duction, and social service expansion. as do women from the most advantaged with the lowest percentage of house- However, the country is still struggling province. holds with electricity (Group 1) than in to ensure that economic progress is

TABLE 3. Complex measurements of inequality, with confidence intervals, for socioeconomic indicators, in study of social inequalities in maternal mortality in Ecuador, 2014

Complex measurements (based on regression models) Socioeconomic indicator Linear regression Poisson regression MSIIa Lower limit Upper limit PRSIIb Lower limit Upper limit Gross domestic product 28.8 7.7 50.0 1.8 1.1 2.9 Percentage of households with children who do not attend school 17.5 −7.1 42.2 1.4 0.8 2.4 Percentage of households with electricity 22.6 0.2 45.1 1.6 1.1 2.6 Percentage of indigenous population 5.0 −19.8 29.8 1.1 0.7 1.8 Total fertility rate 26.1 4.4 47.8 1.7 1.1 2.8 a MSII = modified slope index of inequality. b PRSII = Poisson relative slope index of inequality.

TABLE 4. Simple measurements of inequality, with confidence intervals, for socioeconomic indicators, in study of social inequalities in maternal mortality in Ecuador, 2014

Simple measurements (based on quintiles) Socioeconomic indicator Absolute risk Relative risk ADa Lower limit Upper limit RRb Lower limit Upper limit Gross domestic product 60 14 106 2.4 1.5 4.0 Percentage of households with children who do not attend school 43 2 85 2.0 1.1 3.4 Percentage of households with electricity 42 −2 85 1.9 1.1 3.4 Percentage of indigenous population 22 −10 53 1.4 0.9 2.3 Total fertility rate 70 19 120 2.8 1.6 4.9 a AD = absolute difference. b RR = relative risk.

Rev Panam Salud Publica 41, 2017 5 Original research Sanhueza et al. • Social inequalities in maternal mortality in Ecuador made at all levels of society, and that insight into the specific mechanisms (so- One of the limitations of the statistical improvements in effective social ser- cial determinants) that generate these analysis is that maternal mortality data vices coverage and health outcomes are health inequalities. Furthermore, the ev- produced at the second subnational level shared by everyone, including the most idence from this study can lead to addi- (the cantons) are not taken into account. vulnerable. Similarly, Ecuador also dis- tional research that examines the specific Thus, the results of the analysis in this plays significant differences in terms of barriers and other factors affecting the work have to be considered as a first de- health outcomes, such as maternal mor- subgroups most vulnerable to maternal scription of the maternal mortality in- tality. For example, the province in Ec- health inequalities. equalities, and further studies are uador with the lowest MMR (Azuay, One of the strengths of this paper is needed. In a future study we will con- with 19.1 deaths per 100 000 live births) the use of the Poisson regression model sider the use of multilevel analysis, by has an MMR that is similar to that of for computing a novel relative mea- utilizing data from the cantons and Trinidad and Tobago, which has one sure. On the other hand, there are some provinces. of the highest human development in- weaknesses in this study, such as the dex (HDI) values in the Americas (8). use of the cross-sectional ecological de- Conclusions In contrast, the province with the sign that produces an ecological fallacy. highest MMR (Zamora Chinchipe, with This means that apparent associations This study is one of the few studies 142.2 deaths per 100 000 live births) has between different provinces in Ecuador analyzing maternal mortality inequali- an MMR similar to that of the Solomon may not accurately reflect the true as- ties in Ecuador. By carrying out an anal- Islands (2), which is considered to sociation between individuals within ysis combining descriptive measures, have one of the lowest HDI values in those provinces. However, inequality association measures, and inequality Oceania (11). measurements using data for provinces measures, we hope to provide decision- The results presented in our study (the first subnational level in Ecuador) makers with the type of information support previous research that argues provide relevant and more accurate in- needed for priority-setting. Further, the that national averages often hide differ- formation for the development of local mixed analysis used in this article is in- ences at the local level, and that these health policies because they make it novative in the sense that it expands on disparities can be strongly associated possible to identify the provinces that the methodology traditionally used in with different socioeconomic indicators require equity-based interventions. studies that measure health inequali- (11). For example, while our statistical as- Limitations on the actual distribution ties, and it offers additional information sociation study indicated that five of the of the maternal mortality data in the that may enrich the understanding of nine socioeconomic indicators assessed provinces by socioeconomics indicators maternal mortality inequalities within a were statistically significantly associated may have also been a weakness in our country. with MMR, the analysis using health in- analysis. As for the maternal mortality equality measures indicated that three data, they come from the Epidemiologi- Acknowledgments. We acknowledge of those five socioeconomic indicators cal Surveillance System of the Ministry the entire group working in the Gerencia (total fertility rate, GDP, and the percent- of Public Health (Ministerio de Salud de Mortalidad Materna at the Ministry of age of households with electricity) were Pública (MSP)). In dealing with a sus- Public , and Dr. Gina statistically significant in terms of the pected case of maternal death, a multi- Tambini and Dr. Adrian Diaz from the MMR inequalities. disciplinary team carries out a thorough Pan American Health Organization In reviewing previous research, we investigation to either confirm or rule (PAHO) Country Office in Ecuador. We have found that there are some studies out that initial analysis. Additionally, the would also like to thank Ramon Marti- on the association between socioeco- vital records system (INEC in Spanish) nez, of the Department of Noncommuni- nomic indicators and maternal mortality and the MSP perform a semiannual pro- cable Diseases and Mental Health at in Latin American and Caribbean coun- cess of active search for maternal deaths, PAHO, for his support of this work. tries. However, few of these studies ana- through the review of death registers lyze the maternal mortality inequalities and local research processes. Conflicts of interest. None declared. related to those socioeconomic factors. The data on the socioeconomic indi- Our Ecuador study is important because cators for 2014 were obtained from Disclaimer. The authors hold sole re- it can provide decisionmakers with the ­various sources, including the informa- sponsibility for the views expressed, type of information that is needed to de- tion system of the Central Bank of which may not necessarily reflect the termine which subgroups to target in ­Ecuador and the INEC survey on ur- opinion or policy of the RPSP/PAJPH or order to reduce current health inequali- ban employment unemployment, and the Pan American Health Organization ties. The information can also provide underemployment. (PAHO).

6 Rev Panam Salud Publica 41, 2017 Sanhueza et al. • Social inequalities in maternal mortality in Ecuador Original research

REFERENCES

1. Cord LJ, Lucchetti L, Rodriguez-Castelan 5. Banco Central del Ecuador. Cuentas 9. Bacallao J. Indicadores basados en la no- C. Shifting gears to accelerate shared Regionales. Aplicativo Cuentas Provinciales ción de entropía para la medición de las prosperity in Latin America and 2014. Available from: https://www.bce.fin. desigualdades sociales en salud. Rev Cub Caribbean. Washington D.C.; World Bank;​ ec/index.php/component/k2/item/293- Salud Publica. 2007;33(4). 2013. cuentas-provinciales/ Accessed 15 April 10. Tang W, He H, Tu XM. Applied categorical 2. World Health Organization. Trends in ma- 2016. and count data analysis. Boca Raton: CRC ternal mortality: 1990 to 2015: estimates by 6. Instituto Nacional de Estadística y Censos. Press; 2012. WHO, UNICEF, UNFPA, World Bank Población y Demografía. Available from: 11. Jahan S. Human development report: Group and the United Nations Population http://www.ecuadorencifras.gob.ec/cen- work for human development. New York; Division. Geneva: WHO; 2015. so-de-poblacion-y-vivienda/ Accessed 7 United Nations Development Programme; 3. Fukuda-Parr S. From the Millennium April 2016. 2015. Development Goals to the Sustainable 7. Ministerio Coordinador de Desarrollo Development Goals: shifts in purpose, Social. Sistema de Indicadores Sociales del concept, and politics of global goal setting Ecuador. Available from: http://www.siise. for development. Gend Dev. 2016;24(1):​ gob.ec/siiseweb/ Accessed 10 April 2016. 43-52. 8. Mackenbach JP, Kunst AE. Measuring the 4. World Health Organization. The Global magnitude of socio-economic inequalities in Strategy for Women’s, Children’s and health: an overview of available measures Manuscript received on 13 June 2016. Revised Adolescents’ Health (2016-2030). Geneva; illustrated with two examples from Europe. version accepted for publication on 29 November WHO; 2016. Soc Sci Med. 1997 Mar 1;44(6):757-71. 2016.

RESUMEN Objetivo. El propósito de este estudio fue describir la asociación entre la razón de mortalidad materna y un conjunto de indicadores socioeconómicos, y calcular las desigualdades en la razón de mortalidad maternal entre las distintas provincias del Las desigualdades sociales en Ecuador. Métodos. Se consideró un estudio ecológico transversal utilizando datos prove- cuanto a la mortalidad materna nientes de las 24 provincias de Ecuador en el 2014, calculándose la razón de mortal- entre las provincias del Ecuador idad materna para cada provincia, así como estudiando la asociación y su fuerza entre la razón de mortalidad materna y el conjunto de los indicadores socioeconómicos. Se obtuvieron las medidas de la desigualdades para aquellos indicadores socioeconómi- cos que mostraron una asociación estadísticamente significativa con la mortalidad materna. Resultados. A pesar de que la razón de mortalidad materna en Ecuador es relativa- mente baja a nivel mundial, las razones de la mortalidad materna difieren mucho entre las provincias. Hubo cinco indicadores socioeconómicos que resultaron estar asociados siginificativamente con la razón de mortalidad materna: la tasa total de fecundidad, el porcentaje de población indígena, el porcentaje de hogares con niños que no asisten a la escuela, el producto interno bruto y el porcentaje de hogares con servicio eléctrico. De estos cinco, solo tres mostraron desigualdades estadísticamente significativas en la mortalidad materna: la tasa total de fecundidad, el producto interno bruto y el porcentaje de hogares con electricidad. Conclusiones. Este estudio respalda las investigaciones que sostienen que los prome- dios nacionales pueden ser engañosos, pues a menudo ocultan diferencias entre sub- grupos a nivel local. Los resultados también indican que la razón de mortalidad materna esta asociada significativamente con algunos indicadores socioeconómicos, incluyendo algunos que resultaron en desigualdades significativas en salud materna. Para reducir las inequidades en materia de salud, es crucial que los países adopten un enfoque que trascienda a los promedios nacionales y detecten los subgrupos que van quedando rezagados, analicen los determinantes sociales particulares que generan esas desigualdades en materia de salud y examinen los obstáculos específicos y otros factores que afectan a los subgrupos más vulnerables a las desigualdades en salud materna.

Palabras clave Mortalidad materna; desigualdades en la salud; determinantes sociales de la salud; equidad en salud; Ecuador; América Latina.

Rev Panam Salud Publica 41, 2017 7