ORIGINAL ARTICLES Ecuadorian infant mortality linked to socioeconomic factors during the last 30 years Natalia Romero-Sandoval 1 https://orcid.org/0000-0001-6881-6581 Diego Del Alcázar 2 https://orcid.org/0000-0002-8157-3622 Jacob Pastor 3 https://orcid.org/0000-0002-4415-7793 Miguel Martín 4 https://orcid.org/0000-0001-6156-0739 1 Facultad de Ciencias Médicas de la Salud y de la Vida. Universidad Internacional del Ecuador. Av. Jorge Fernandez and Simon Bolivar. Quito 170113. Ecuador. E-mail: nromero@uide.edu.ec 2 Facultad de Ciencias Sociales y Comunicación. Universidad Internacional del Ecuador. Quito, Pichincha, Ecuador. 3 Instituto Geografico Militar. Quito, Pichincha, Ecuador. 4 Unidad de Bioestadística. Facultad de Medicina. Universidad Autonoma de Barcelona. Bellaterra, Catalunya, Spain. Abstract Objectives: to analyze the difference among geographical units and the evolution of infant mortality rate (IMR) based on Ecuadorian censuses (1990-2001-2010). Methods: artificial Neural Network analyzed the impact of sociodemographic factors over the variability of IMR. Poisson regression analyzed the variation of the standardized IMR (sIMR). Results: the decrease in the national IMR was 63.8%; however, 42.8% provinces showed an increase in 2001-2010. The variability was explained mainly by illiteracy decrease. The adjusted RR between provincial sIMR with illiteracy and poverty revealed a trend towards the unit. Conclusions: the variation of IMR reflects a complex interaction of the sociodemographic factors. Key words Infant mortality, Ecuador This article is published in Open Access under the Creative Commons Attribution http://dx.doi.org/10.1590/1806-93042019000200003 license, which allows use, distribution, and reproduction in any medium, without restrictions, as long as the original work is correctly cited. Rev. Bras. Saúde Mater. Infant., Recife, 19 (2): 295-301 abr-jun., 2019 295 Romero-Sandoval N et al. Introduction 2001 and 2010, by implementing an Artificial Neural Network model (ANN). Infant mortality rate (IMR) is defined as the number Additionally, in order to compare the risk of of infant deaths in the first year of life for every 1000 infant mortality among those with illiteracy and live births. According to the United Nations Inter- poverty we implemented an ecological analysis agency Group for Child Mortality Estimation, the using a standardized IMR (SMR) at three temporal IMR world-wide has decreased since 1990 from an cut-off points (1990, 2001 and 2010). estimated rate of 63 deaths per 1000 live births to 32 deaths per 1000 live births in 2015.1 Meanwhile in Methods Ecuador, a developing country, the IMR has dropped down from an estimated of 30.24 deaths per 1000 We implemented an ecological study using data from live births in 1990 to 10.95 per 1000 live births in three population censuses published by INEC, in 2010, as reported by the National Institute of 1990, 2001 and 2010. The data is made available in Statistics and Censuses of Ecuador (INEC, by its two structured digital formats (official sources): (1) Spanish acronym).2 Datasets of live births and general deaths grouped by IMR is related to multiple genetic, environ- sex. 14 (2) Census Database provide data of social mental, and socioeconomic factors.3-5 For example, indicators (illiteracy and poverty). Ethnicity was in developing countries an increase in education included only in 2001 and 2010.15 levels leads to decreases in marriage at a young age, Data were available for the 21 provinces (spatial first sexual experience, first birth and a better ecological unit) confirming Ecuador for the years economic environment for the woman, all which 1990, 2001 and 2010. Currently (2017), Ecuador is lead to decrease of infant mortality.6 divided into 24 provinces; however, this study only Although a strong correlation has not been found considers the 21 provinces defined in the year 1990. between poverty reduction and lower IMR, for We also have excluded from the analysis all the areas middle- and lower-income countries a 10% increase belonging to unlimited jurisdictions. The unlimited of the gross domestic product seems to be correlated areas are three and refer to the most recent with a reduction of 10% in IMR. Similarly, an geographic boundaries published by INEC, and not increase in inequality is correlated with an increase to the 1990 administrative boundaries as this data is of the IMR.5-7 not available in geographic information system Lower socioeconomic outcomes and higher IMR format. We argue that this discrepancy is negligible had been reported as disparities among ethnic since the un-delimited areas hold less than 1% of the minorities, like afro-descendants or indigenous total population (0.72% in 1990). For this analysis, social groups.8,9 In Latin America and in Ecuador - the Galápagos Islands were excluded due to socio- without exception, indigenous groups exhibit higher demographic and economic differences with the rates of poverty, lower education outcomes, higher continental Ecuadorian territory, and the lack of fertility rates and limited access to basic health sufficient observations to include it in the analysis. services compared to their non-indigenous counter- The spatial ecological unit was the province, parts.10 which is the traditional geographical unit for polit- In Ecuador, 7% of the population self-identify as ical and administrative purposes and were 21. indigenous, of them, almost 75% were poor based on Illiterate: Person older than 15 years who cannot unsatisfied basic needs (UBN) at 2010. Percentage read and write.15 of illiteracy in the indigenous population is 20.4% Poverty by UBN: A person is considered poor by compared to 5.1% of the non-indigenous population, UBN if they live in a household with the following and an IMR 1.5 to 3 times higher than the national features: a household head with two or less years of rate; when considering 1997 figures.11 education and one employed person for every four Policies of inclusion to foster equity and poverty or more people living at home; living in a house reductions have been implemented in Ecuador over made of irregular material such as cane walls; having the last ten years. 12 These policies have allowed a a child between 6 to 12 years old who is not reduction of around 4.9 points in the Gini index attending school; more than three people living per between 2005 and 2010.13 room; and, if the house does not have a connection In this context, we decide to analyze the quanti- to the water supply network or uses latrines or toilet tative contribution of illiteracy, poverty by unsatis- facilities without a sanitation system.16 fied basic need (UBN) and indigenous ethnicity Indigenous ethnicity: People living in a house- proportion to the variability of IMR in years 1990, hold whose head self-describes as indigenous.15 296 Rev. Bras. Saúde Mater. Infant., Recife, 19 (2): 295-301 abr-jun., 2019 Ecuadorian infant mortality during the last 30 years Illiteracy, poverty and indigenous ethnicity factors explained in 45.2% by Illiteracy, in 39.1% by are presented in percentages. poverty and in 15.7% by the proportion of the The Artificial Neural Network model aims to indigenous ethnic group. Meanwhile, 2010 data weight each the factors that explain to the IMR vari- showed that IMR variability for illiteracy and ability. ANN are mathematical non-linear models poverty remained the same as 2001. However, it that analyze the variability of the dependent variable, increased by 5.1% in the proportion of the indige- based on the connection of the variables to be nous ethnic group. studied. 17,18 In this analysis the independent vari- The variability of SMR by provinces respect to ables are illiteracy, poverty and share of indigenous the country in the periods 1990-2001 and 2001-2010 population, and the dependent variable was provin- is shown in Figure 1. In the period 1990-2001, of 21 cial infant mortality crude rates. The variability is provinces, 14 presented a decrease of the SMR with expressed between individual provinces for each a variation between 7% and 45%, while 7 provinces census. had an increase of the SMR with a variation between We performed an indirect standardization of the 2% and 21%. In the period 2001-2010, 9 of 21 IMRs (SMR) at every province, taking as reference provinces showed a decrease of the SMR between the population of the whole country. We show the 2% and 35%, whereas in 12 provinces increased the change of each provincial SMR between the census SMR between 1% and 83%. Nine (42.8%) provinces 2001 - 1990 and 2010 - 2001. showed a decrease and then an increase and vice First, we analyze the co linearity between illit- versa for 4 (19.0%). eracy and poverty by Pearson coefficient. Second, Previously, a correlation analysis between illi- we did an adjustment of the SMR infant mortality of teracy and poverty was made (r=0.46, p<0.05). Table 21 provinces, applying to Poisson regression model, 2 also presents the results of fitting Poisson model given that the number of events occurring in disjoint for the SMR of infant mortality, expressing the intervals of space-time are independent,19 in order to results as relative risks and their CI95%, for each examine the adjusted relative risk (ARR) of SMR percentile point of change in the factors stated. We related to illiteracy and poverty percentages, by observe a significant increase in the SMR associated census years. The results were expressed as ARR and to higher percentages of illiteracy. In contrast, the its 95% confidence interval. As indicated, ethnicity poverty presents significant inverse associations data was not available until the census of 2001, with the rate studies. Nevertheless, both indicators therefore it was not included in this analysis. We showed a convergence to the unit in the relative risks used the IBM SPSS Statistics 20 (IBM Corp., during the studied period. Armonk, NY, USA), to run the models. Discussion Results One of the Millennium Development Goals estab- Ecuador's IMRs for years 1990, 2001 and 2010, both lished by the United Nations, in 2000, was to for the whole country and by gender, are shown in decrease the under-age of 5 mortality rate by two Table 1. The Table shows a decreasing IMR with a thirds between 1990 and 2015. While the reduction total decrease of 63.8% (19.29 deaths per 1000 live of infant mortality in most countries has been acce- births) nationally, from 1990 to 2010. lerated, most developing countries will need many Illiteracy, poverty, and indigenous ethnicity years after 2015 to achieve that goal.20 This objec- percentages are presented in Table 2. Between 1990 tive encompassed Latin America as well but was not and 2010, illiteracy shows a reduction of 42.7%, achieved globally, 5 apparently due to high propor- with a bigger drop among women (44.2%) than men tions of neonatal deaths which suggest that more (39.6%). Poverty declined from 1990 to 2010 in efforts should have focused at the earliest ages.21-23 24.5%, Indigenous ethnicity proportion remained In Ecuador, a country that went from being a constant at approximately 7%. lower-middle income country in the 1990’s to an Table 2 presents the proportion of the national upper-middle income country in 2010,12 the results IMR variability explained by each of three factors of this study showed a similar reduction of IMR studied in the three cut-off points. For 1990, the illi- figures with a drop from 30.24 to 10.95 per 1000 live teracy explained the 66.2% of the variation of the births (63.8%), between 1990 and 2010. Since 2006, national IMR compared to 33.8% of the variability there have been several new policies implemented explained by poverty. When the three factors are directed to improve literacy, reduce poverty and considered, IMR’s variability for the year 2000 was inclusion of marginalized groups such as indigenous Rev. Bras. Saúde Mater. Infant., Recife, 19 (2): 295-301 abr-jun., 2019 297 Romero-Sandoval N et al. Table 1 Ecuador’s Infant mortality rate (IMR) by gender in the years 1990, 2001, 2010. Year Deaths within one year, per 1000 Number of births Ecuador live births (IMR) population (millions) Female Male Total 1990 27.95 32.44 30.24 263582 10.26 2001 15.16 19.29 17.25 278160 12.55 2010 10.26 11.61 10.95 292329 14.16 Table 2 Illiteracy, poor people by unsatisfied basic needs and indigenous people rates, variability proportion of national IMR explained by them, and adjusted relative risk of standardized IMR, Ecuador census 1990-2001-2010. Census year Gender Illiteracy PP / UBN IndP VPN-IMR Adjusted VPN-IMR Adjusted VPN-IMR Adjusted (%) by Ill RR (%) by RR (%) by IndP RR (CI95%) PP/UBN (CI95%) (CI95%) 1990 Female 13.79 78.41 NA Male 9.47 80.48 NA Total 11.78 0.662 1.039 79.43 0.338 0.981 (1.035- (0.978- NA NA NA 1.043) 0.984) 2001 Female 10.29 70.38 3.50 Male 7.65 72.19 3.33 Total 9.00 0.452 1.023 0.985 (1.014- 71.27 0.391 (0.983- 6.83 0.157 NA 1.033) 0.988) 2010 Female 7.70 59.17 3.57 Male 5.74 60.82 3.45 Total 6.74 0.420 1.026 0.989 (1.011- 59.98 0.373 (0.986- 7.03 0.208 NA 1.041) 0.992) Ill= Illiteracy; PP/UBN= Poor people by unsatisfied basic needs; IndP= Indigenous people; VPN – IMR= Variability proportion of national Infant Mortality Rate; NA= Not available. 298 Rev. Bras. Saúde Mater. Infant., Recife, 19 (2): 295-301 abr-jun., 2019 Ecuadorian infant mortality during the last 30 years Figure 1 Variability of SMR by provinces respect to the country in the periods 1990-2001 and 2001-2010. 10400000 10400000 500000 600000 700000 800000 900000 1000000 1100000 500000 600000 700000 800000 900000 1000000 1100000 -700000 -600000 -500000 -400000 -700000 -600000 -500000 -400000 10050000 10300000 10050000 10300000 N N 9900000 9900000 10200000 10200000 Colombia Colombia 10100000 10100000 Pacific Ocean Pacific Ocean 10000000 10000000 9900000 9900000 9800000 9800000 9700000 9700000 Peru Peru 9600000 9600000 SMR variation 1990-2001 SMR variation 2001-2010 -0.450 to -0.300 -0.350 to -0.090 -0.300 to -0.160 -0.090 to -0.020 -0.160 to 0.080 -0.020 to 0.100 9500000 9500000 -0.080 to 0.090 0.100 to 0.200 0.090 to 0.210 0.200 to 0.830 0 100 200 km Undelimited boundaries 0 100 200 km Undelimited boundaries 9400000 9400000 SMR= Standardized Infant Mortality Rate. people,24 factors whose influence on infant mortality Intensification of governmental effort including have been described.7,9 health care, primarily health care. 25 In the period In this context, we set out to analyze the contri- studied, improvement in the health sector could bution of these factors in the variability of the IMR. produce a decrease in the magnitude of the interven- An ANN model helped us to establish the variability tion in education and the reduction of poverty. The proportion of the national IMR explained by illi- effect of the intervention, through a reform of the teracy, poverty and indigenous ethnicity percentages. health sector and the injection of funds, could be Based on the ANN analysis, we identify that among reflected in the reduction of infant mortality in less the three factors studied, the predominant contri- time, compared with the time required to observe the butor to explain the variability of IMR in Ecuador is effect of reduction of both factors, poverty and illi- illiteracy. teracy, over infant mortality. 12,25 (2) Although a These results motivated to quantify the relative declining trend was found in IMR in Ecuador weight of the study factors to explain the differences throughout the 3 periods of study, the evolution of between the provincial rates standardized by age and IMR, in indigenous population the rate seems to sex, with respect to the country's overall. This remain stable. analysis, carried out through a Poisson regression Indigenous population has been linked with model, showed a trend of adjusted relative risks to factors such as poverty and illiteracy,10 all of which the unit over time. are barriers that make cooperation and coordination Plausible explanations to these results are: (1) harder, especially when is related to health services. Rev. Bras. Saúde Mater. Infant., Recife, 19 (2): 295-301 abr-jun., 2019 299 Romero-Sandoval N et al. These factors are reflected in the lower access to three of the main factors related to the IMR which maternal-infant services among indigenous can give public policies to continue with the reduc- compared to mestizos in Ecuador. 24,26 In this tion of IMR.24,29 (2) In order to ensure the standar- analysis the proportion of indigenous ethnicity over dization of the definition of each variable studied, the variability of IMR remained constant, and so its we took only three temporal cut-off points (1990, effect. 2001 and 2010), which limits the analysis of the On the other hand, the SMR of the ecological variability proportion of the IMR. (3) In the lasts units resulted disparities between the provinces in census the quality of data acquisition and registra- both periods (2001-1990 and 2010-2001). Almost tion could have improved especially among half of the ecological units showed a decrease in the provinces historically marginalized of informatics SMR during the first period, but the second showed aspects which is reflected in an increase in the noti- an increase in the SMR, corresponding to a strong fication of the indicators studied. Government action focused on the social axis. This An in-depth analysis of the relationship between apparent contradiction could be explained by the illiteracy, poverty, and ethnicity over the time, and improvement in the registration of births and deaths. their relationship with infant mortality, allows the The increase registration of births and deaths, moti- country a better understanding of the current situa- vated by a greater monitoring of the population, as tion, in order to create strategies to improve the well as the standardization of records of births and living conditions of the population. infant deaths have an initial effect of higher mortality, due to the visualization of the problems. This effect has been described in other countries.27,28 Author’s contribuition However, this effect was not observed in all ecolo- gical units, which may be explained by the different Romero-Sandoval N and Martín M conceived, degrees of structural modification of the records, design and coordinated the study. Romero-Sandoval over time. N, Martín M and Pastor J prepared the manuscript Our study has the following limitations: (1) We and contributed to the analysis of data. Romero- analyzed the evolution of IMR considering the polit- Sandoval N, Martín M, Alcázar D and Pastor J were ical-administrative division of Ecuador at 1990 responsible for data management and statistical which could change if the actual division is consid- analysis. All the authors were involved in the inter- ered, the unavailable data regarding ethnicity in pretation of the data and drafting the manuscript. All 1990 as well as the absence of other variables of the authors read and approved the final manuscript. set of factors that influence the variability of the infant morbidity and mortality. However, we analyze References 1. You D, Hug L, Ejdemyr S, Beise J. Levels and trends in 6. Grépin KA, Bharadwaj P. Maternal education and child child mortality. Report 2015. Estimates developed by the mortality in Zimbabwe. J Health Econ. 2015; 44: 97-117. UN Inter-agency Group for Child Mortality Estimation; 7. O’Hare B, Makuta I, Chiwaula L, Bar-Zeev N. Income and 2015. child mortality in developing countries: a systematic review 2. Villacís B, Carrillo D, Martínez AG. Estadística demográ- and meta-analysis. J R Soc Med. 2013; 106: 408-14. fica en el Ecuador: diagnóstico y propuesta. Quito Inst Nac 8. Rossen LM, Khan D, Schoendorf KC. Mapping geographic Estad Censos. 2011; 1-74. variation in infant mortality and related Black–White 3. Hanmer L, Lensink R, White H. Infant and child mortality disparities in the US. Epidemiol. 2016; 27 (5): 690-6. in developing countries: Analysing the data for Robust 9. Haider SJ. Racial and ethnic infant mortality gaps and determinants. J Dev Stud. 2003; 40; 101-18. socioeconomic status. Focus. 2014; 31: 18-20. 4. Rosano A, Botto LD, Botting B, Mastroiacovo P. Infant 10. Hall G. Indigenous peoples, poverty and human develop- mortality and congenital anomalies from 1950 to 1994: an ment in Latin America. (Springer; 2005). international perspective. J Epidemiol Community Health. 11. Cujilema M, Dávila GO. La Salud de los Pueblos 2000; 54: 660-6. Indígenas. Salud En Glob. 2003; 169. 5. Ward JL, Viner RM. The impact of income inequality and 12. Malo-Serrano M, Malo-Corral N. Reforma de salud en national wealth on child and adolescent mortality in low Ecuador: nunca más el derecho a la salud como un privi- and middle-income countries. BMC Public Health. 2017; legio. Rev Peru Med Exp Salud Publica. 2014; 31: 754-61. 17: 429. 300 Rev. Bras. Saúde Mater. Infant., Recife, 19 (2): 295-301 abr-jun., 2019 Ecuadorian infant mortality during the last 30 years 13. Albuja J, Navas A, Paguay D, Moreno A, Nájera P. 22. Ferre JC. Economic Inequalities in Latin America at the Technological GINI: a study of the inequality in Ecuador. Base of Adverse Health Indicators. Int J Health Serv Plan In: Democracy & Government (ICEDEG), 2015 Second Adm Eval. 2016; 46: 501-22. International Conference on 133–137 (IEEE, 2015). 23. Målqvist M. Neonatal mortality: an invisible and margin- 14. Instituto Nacional de Estadística y Censos. Anuario de alised trauma. Glob Health Action. 2011; 4: 5724. Estadísticas de Nacimientos y Defunciones (Generales y 24. López-Cevallos DF, Chi C. Health care utilization in Fetales); 2011. Ecuador: a multilevel analysis of socio-economic determi- 15. Censos INE. (Instituto Nacional de Estadística y Censos). nants and inequality issues. Health Policy Plan. 2010; 25: Población y Demografía. Instituto Nacional de Estadística 209-18. y Censos. [Accessed: 4th March 2018]. Available at: 25. Lucio R, Villacrés N, Henríquez R. Sistema de salud de http://www.ecuadorencifras.gob.ec/censo-de-poblacion-y- Ecuador. Salud Pública México. 2011; 53: S177-S187. vivienda/ 26. Kuang-Yao Pan W, Erlien C, Bilsborrow RE. Morbidity and 16. Censos INE (Instituto Nacional de Estadística y Censos). mortality disparities among colonist and indigenous popu- Pobreza por Necesidades Básicas Insatisfechas. [Accessed: lations in the Ecuadorian Amazon. Soc Sci Med. 2010; 70 4th March 2018]. Available at: http://www.ecuadoren- (3): 401-11. cifras.gob.ec/pobreza-por-necesidades-basicas-insatisfe- 27. Alvarez G, Harlow SD, Denman C, Hofmeister MJ. Quality chas/ of cause-of-death statements and its impact on infant 17. Zhang W, Barrion A. Function approximation and docu- mortality statistics in Hermosillo, Mexico. Rev Panam mentation of sampling data using artificial neural networks. Salud Publica Pan Am. 2009; 25 (2): 120-7. Environ Monit Assess. 2006; 122: 185. 28. Wuhib T, Mc Carthy BJ, Chorba T, Sinitsina TA, Ivasiv. IV, 18. Cilimkovic M. Neural networks and back propagation algo- McNabb SJ. Underestimation of infant mortality rates in rithm. Inst. Technol. Blanchardstown Blanchardstown Road one republic of the former Soviet Union. Pediatrics. 2003; North Dublin 15; 2015. Available at: http://www.datamin- 111 (5): e596-600. ingmasters.com/uploads/studentProjects/NeuralNetworks.p 29. Anderson I, Robson B, Connolly M, Al-Yaman F, Bjertness df E, King A, et al. Indigenous and tribal peoples’ health (The 19. Hayat MJ, Higgins M. Understanding poisson regression. J Lancet-Lowitja Institute Global Collaboration): a popula- Nurs Educ. 2014; 53: 207-15. tion study. Lancet Lond Engl. 2016; 388: 131-57. 20. Lozano R, Wang H, Foreman KJ, Rajaratnam JK, Naghavi M, Marcus JR, Dwyer-Lindgren L, Lofgren KT, Phillips D, Atkinson C, Lopez AD, Murray CJ. Progress towards Millennium Development Goals 4 and 5 on maternal and child mortality: an updated systematic analysis. Lancet Lond Engl. 2011; 378: 1139-65. 21. Servan-Mori E, Torres-Pereda P, Orozco E, Sosa-Rubí SG. An explanatory analysis of economic and health inequality changes among Mexican indigenous people, 2000-2010. Int J Equity Health. 2014; 13: 21. ______________ Received on May 10, 2018 Final version presented on February 25, 2019 Approved on March 27, 2019 Rev. Bras. Saúde Mater. Infant., Recife, 19 (2): 295-301 abr-jun., 2019 301