Infant mortality in Brazil during recent periods of economic crisis  



Maria da Conceição Nascimento Costa; Eduardo Luiz Andrade Mota; Jairnilson Silva Paim; Lígia Maria Vieira da Silva; Maria da Glória Teixeira; Carlos Maurício Cardeal Mendes

Instituto de Saúde Coletiva da Universidade Federal da Bahia. Salvador, BA, Brasil





OBJECTIVE: To analyze time trends in infant mortality in Brazil during a recent period of economic crisis (1980-1998).
METHODS: Time-series study based on Ministry of Health Mortality Information System, IBGE Foundation, and Fundação Nacional de Saúde (National Health Foundation) databases. Serial parameters were described using Autoregressive Integrated Moving Average (ARIMA) models, and the association between infant mortality rates and a number of determinants was evaluated using Spearman correlation coefficients.
RESULTS: Infant mortality showed a declining trend (-59.3%) and a strong correlation with most of the indicators analyzed. However, only correlations between infant mortality rate and total fertility and birth rates presented a significant difference between the two decades.
CONCLUSIONS: Variations in fertility rate were the main cause for the persistent decline in infant mortality in the 1980’s. In the subsequent period, causes related to living conditions, especially healthcare, may have been more important.

Keywords: Infant mortality. Mortality rate. Fertility rate. Birth rate. Information systems. Socioeconomic factors. Time series.  




Until the 1960’s, infant mortality was understood as a problem related to economic underdevelopment,1 due to its almost exclusive inverse association with indicators such as Gross Internal Product (GIP),18 unemployment rates11, and Gini’s index.25 This relationship was explained by the role these indicators played in setting income, education, and sanitation profiles, and in determining the accessibility of health programs and services16, which promoted improvements in population-wide living conditions.

However, from the end of that decade onwards, persistently declining infant mortality trends in developing countries coexisted with severe economic crises, characterized by reductions in GIP and in the real value of minimum wage, and increasing unemployment rates. This apparent paradox served as a stimulus for conducting studies aimed at identifying the reasons behind such a sustained decline in mortality rates.6,20 In addition, subsequent reductions in pace, stagnation, or reversions in this trend observed throughout the 1990’s in a number of countries19 revived the interest in expanding the amount of knowledge currently available on the role played by different infant mortality determinants in developing countries. Several investigations along these lines have suggested that healthcare interventions are a major determinant of the variations observed in infant mortality levels.20

In Brazil, where a sustained decrease in infant mortality was observed in a scenario of economic recession, certain authors have indicated, as possible explanations for this trend, the adoption of public policies, such as medical-sanitary interventions, expansion of the water supply network, and increases in schooling,14 with the subjacent aid of the reduction in fertility rates also observed during this period.13,17

However, attempts to implement healthcare-related interventions in the 1980’s were not always successful.15 On the other hand, certain programs and initiatives capable of reducing infant mortality – including Programas de Atenção Integral à Saúde da Mulher (Integral Women’s Healthcare Programs – PAISM), oral rehydration therapy (ORT), Programa Nacional de Imunização (National Immunization Program – PNI), and social interventions, such as the expansion of the water supply network and reductions in illiteracy – began to be implemented in the Brazil only in the second half of the 1980’s. It is thus reasonable to assume that such policies began to yield more substantial effects on the maintenance of the declining trend in infant mortality only from the 1990’s onwards.

From this perspective, it is of fundamental importance to clarify what factors could have had such a marked variation as to contribute to the persistent declining trend in infant mortality, in a period in which economic, social, and healthcare indicators in Brazil suggested such an unfavorable scenario. With this in mind, we analyzed time trends of infant mortality in Brazil between 1980 and 1998, in order to identify potential factors which may have influenced the course of infant mortality evolution.



An ecological time series study of infant mortality in Brazil between 1980 and 1998 was carried out using mortality data relative to all state capitals nationwide. The single exception was Palmas, in the state of Tocantins, excluded due to the absence of information for the entire period.

Infant mortality rate (IMR) and its neonatal (NMR) and post-neonatal (PNMR) components, as well as proportional infant mortality (PIM) were calculated. The number of deaths among infants (age <1 year) was obtained through the Brazilian Mortality Information System (MIS).4 The entire population in this age group was used as a denominator for the above mentioned rates due to the lack of reliable estimates of the number of live births for the entire series. For the years 1980, 1991, and 1996, data from Demographic Censuses and Population Counts4 were used; for the remainder, estimates were made using the geometric growth method.21

Values for total fertility rate (TFR), birth rate (BR), unemployment rate (UR), illiteracy rate (IR), annual per capita GIP variations, Gini’s Index, proportion of households connected to main water supply network, and proportion of households connected to sanitary sewage network or with septic tank were obtained from publications and from the Fundação Instituto Brasileiro de Geografia e Estatística (Brazilian Institute for Geography and Statistics Foundation – IBGE Foundation) web page*.

Annual per capita GIP was used as a percentage rather than an absolute value. We considered the former as an easier means for the reader to understand GIP dynamics. It may also increase the chances of capturing any possible influence this actor may have had on the evolution of infant mortality.

Due to the lack of infant healthcare indicators for the whole series, a proxy was built based on the number of health facilities per 10,000 population, in spite of its lack of discriminatory power. Anti-measles immunization coverage among infants was used as a marker for children’s general access to healthcare,13 due to its ability to express improvements in health facility prevention activities, in addition to greater impact of this vaccine on infant mortality. Both indicators were obtained from the IBGE Foundation* and from the Ministry of Health’s Centro Nacional de Epidemiologia (CENEPI) of the Fundação Nacional de Saúde (FNS) internet pages.4

Infant mortality trend analysis was done through graphic representation and through a description of movements observed through visual inspection of time trend curves. IMR curve variations were smoothened using third order moving average techniques.

Time series structure description was carried out by applying Autoregressive Integrated Moving Averages (ARIMA) models to the number of deaths occurred among infants resident in the set of state capitals between 1980 and 1998, clustered by trimester. This technique allows for a description of the degree of auto-explanation between observations based on parameters “p” and “Ps”, associated to autoregression, which provide information concerning the order of structural dependence existent between adjacent observations, indicating the existence of autocorrelation. In the equation, f stands for serial correlation coefficient; fs, “d”, and “Ds” specify the number of times the series must be differentiated in order to become stationary; and “q” and “Qs” indicate the number of moving average terms. In each parameter, the first notation refers to serial trend, and the second to seasonal trend.3

Coefficients of variation were also calculated, and annual percentage variations in IMR and its components were examined both for the entire period and for the 1980-1989 and 1990-1998 periods separately.

Statistical associations between IMR and analyzed indicators were verified based on Spearman partial correlation coefficients and their respective 95% confidence intervals for the 1980-1998 period, year by year; and for the 1980 and 1990 decades separately. A test of equality between two correlations10 was applied in order to identify the occurrence of possible temporal changes in these rates. Statistical software used include Stata 7.0 (2001) and Statistica for Windows 4.3 (1993);23 a 5% significance level was adopted.



Between 1980 and 1998, IMR in state capitals declined from 67.1%0 to 27.3%0, a 59.3% decrease. The great variability in IMR during this period deserves special attention, especially the periods of stagnation, between 1983 and 1984; and of increase, in 1988, 1992, and 1993 (Table 1 and Figure 1). The 15.0% reduction observed between 1990 and 1991was ascribed mostly to the Northeastern capitals. IMR variation rate in the 1980’s (17.7%) was greater than that observed in the 1990-98 period (9.5%). PIM had a more uniform decline – approximately 70% between 1980 and 1989 – when it fell from values above 20% to approximately 7% (Table 1).




Figure 1 – Infant, neonatal and postnatal mortality rates in Brazilian capitals* according to year of occurrence. Brazil, 1980–1998.


The 29.2%0 neonatal mortality rate observed in 1980 fell to 21.4%0 in 1989, a 26.7% decline. In the 1990-98 period this variation was only 5.7% (from 19.1%0 to 18.0%0) and between 1992 and 1995 there was a slight increase, more pronounced in 1995 (+5.1%). NMR only surpassed post-neonatal infant morality in 1987. The latter suffered a 50.1% reduction in the 1980-89 period (from 37.3%0 to 18.6%0), despite an increase in 1984 and a stagnation period in 1987-88. The 1990-98 period saw a 47.1% reduction, in spite of an increase in 1992 and stagnation in the two following years (Table 1). PNMR variation rate was 23.6% in the 1980-98 period and 21.0% in the 1998-98 period. For NMR these numbers were 11.5% and 4.8% for the two decades, respectively.

ARIMA models applied to quarterly distributions of the number of infant deaths between 1980 and 1998 yielded a final model (1,1,1) (1,1,0) 4 and f and fs values of 0.58 (p=0.001) and –0.49 (p=0.0001). This means that the infant mortality curve had serial and seasonal trends with first-degree autoregression (p=1; Ps=1) and serial and seasonal correlation coefficients of f=0.58 (p=0.001) and fs=-0.49 (p<0.0001), respectively. Hence the first number of the series was positively correlated (58.0%) with the second one, which, in terms of serial trend, means that if the first observation is large, so will be the second one; if it is small, so will be the second one as well. However, in the case of seasonal trends, an inverse correlation was verified between seasonally adjacent observations, that is, a large first observation was followed by a small second one. Figure 2 clearly shows such trends; one can also see that both the amplitude of seasonal variation and the decline in the number of infant deaths decreased from the last years of the 1980’s onwards.


Figure 2 – Death for children under 1-year-old in Brazilian capitals according to year and month of occurrence. Brazil*, 1980–1998.


Tables 2 and 3 show that the illiteracy rate fell from 25.5% in 1981 to 9.0% in 1998, being strongly correlated with IMR (r=0.929; p=0.00) in the 1990-1998 period, for which a larger number of observations were available for this indicator. Both anti-measles immunization coverage among infants and the proportion of households connected to sewage networks or with septic tank were higher and had greater variation in the 1990’s; their correlation with infant mortality during this period was also higher.

Tables 2 and 3 also show that the greatest levels of unemployment occurred during the first five years of the 1980’s decade. In this period these rates had a strong positive correlation with IMR (r=0.864; p=0.01). The number of children per woman in reproductive age varied between 4.0 and 2.8 between 1980 and 1989, with a 3.4 average. This number fell to 2.5 in the following decade, ranging from 4.0 to 2.4. There was a very strong correlation between TFR and IMC values, especially between 1980 and 1989 (r=0.985’p,0.001). The greatest negative variations in per capita GIP were registered in 1981 and 1983, but this indicator was not correlated with IMR. The remaining indicators provided no further statistically significant correlations. Statistical tests for equality of correlations were performed in order to compare correlation coefficients from the 1980-89 and 1990-98 periods. Only the correlation coefficients related to the association between TFR/BR (collinear variables) and IMR yielded statistically significant differences.



The interpretation of study results based on secondary data requires caution, especially if dealing with studies based on Brazilian Information Systems, which experienced different stages of development and organization during the period studied, including variations in coverage and availability that most certainly affect the quality of historical series. Furthermore, one must consider the lack of a number of live births for the entire period, the arguable limitations of the independent variables employed in order to reflect population-wide living conditions, and the fact that the annual fertility rates used refer to estimates obtained through indirect techniques, not always based on a full set of the necessary data for such calculations. Result interpretation in the present study also faces difficulties posed by the recognition, to any degree of certainty, of the impact exerted by economic crises. This may be ascribed to the adoption of compensatory interventions, intended to counteract the effects of crisis. The lack of knowledge concerning the time period necessary for such effects to become perceptible poses further problems.

However, despite all these limitations, the declining trend observed for infant mortality in the present study can be considered perfectly valid, since more trustworthy data were used, provided by state capitals. Data obtained directly from the Mortality Information System (MIS) have greater sensitivity in order to express variations possibly produced by economic crises or demographic alterations, as long as one assumes their quality to be constant. In this sense, it is important to note that the most marked reduction, observed in 1991 in the time-trend infant mortality curve, may be a result of specific operational problems due to the beginning of the MIS decentralization process, in 1993 (there was, at that time, a two-year lag in data processing in this system).

Therefore, despite the great variability observed in infant mortality levels in Brazil between 1980 and 1998, their declining trend is evident, as demonstrated by previous studies.24 It is hence pertinent to consider the fluctuations observed in the infant mortality curve as a result of the impact of economic crises occurred in the period,13 despite the attenuation provided by the adoption of public compensatory policies.

The greater pace of the decrease in infant mortality observed during the 1980’s may be explained by the its higher levels at the beginning of the period. On the other hand, the intensive implementation of oral rehydration therapy from 1988 on may account for the substantial decline in post-neonatal deaths due to infectious intestinal disease. Reductions in the amplitude of seasonal infant mortality variation observed in the present investigation may be considered as an evidence in favor of this hypothesis.

As expected, demographic, socioeconomic, and healthcare indicators were in general strongly associated with infant mortality. The absence of a statistically significant negative association between infant mortality and “percentage of households connected to sewage network or with septic tank” in 1980-89 and “anti-measles immunization coverage” in both 1980-89 and 1990-98 may be due to the little variability observed in these two indicators during both periods. It is thus possible that reductions in illiteracy, increases in anti-measles immunization coverage, and the expansion of the water supply network registered in the 1990’s may have contributed to the decrease in the numberof avoidable deaths among live births.

Nevertheless, there was a reduction in the absolute number of infant deaths due to all causes except for congenital anomalies and external causes. In addition, it was only from the 1990’s onwards that the restructuring of the healthcare sector in Brazil began, after the implementation of the Sistema Único de Saúde (Unified Healthcare System) in 1988. In the preceding decade, the inflationary process experienced by the country – along with increases in medical/hospital/outpatient care complexity and costs – contributed to a further hampering of service accessibility and to reductions in efficiency and efficacy in the public healthcare sector.5 It can thus be argued whether the interventions implemented in order to minimize the effects of crisis – most of which were adopted from the end of the 1980’s onwards – were able, by themselves, to sustain the decline in infant mortality during that period.

In order to consider a factor as explicative of any given trend, it needs to have some degree of temporal variation. Of the variables analyzed, this was the case only with birth and total fertility rates, since these were the only variables whose correlation coefficients with infant mortality differed significantly from one decade to the next.

The rapid reduction in fertility rates was recognized as an actuality in the 1980’s, especially during the first half of the decade, when it was as high as 19.0% (in the entire previous decade these rates declined 25%; this occurred especially among poorer women and in regions where living conditions were precarious).7 In the 1990’s, a 14.8% reduction in fertility rates (from 2.7 to 2.3 children per woman) was observed between 1992 and 1999. The difference in mean number of children per woman between the Northeast and Southeast Regions fell from 1.0 to 0.5.8 In light of this evidence, it is reasonable to propose the hypothesis that fertility rate variations were the main cause for the sustained decline in infant mortality in the 1980’s.

Among the mechanisms by which fertility rates may influence the persistence of the decline in infant mortality rates in Brazil, we must mention the relative reduction, in the 1980’s, of the number of children born from mothers living in precarious conditions – hence at greater risk of dying during their first year of life. Other mechanisms include: the mediation provided by factors such as the reduction in the number of preterm and shortly spaced births and the number of high parity and older mothers, as occurred in Brazil; increases in economic efficiency due to reductions in inter-familiar dependence and in the demand for healthcare services, and to an increase in the amount of time available for mothers to care for their children.13

The hypothetical contribution of a reduction in illiteracy rates contributing towards the maintenance of the decline in infant mortality via a decrease in fertility rates seems improbable, since, according to data from the 1980 and 1991 Demographic Censuses, the proportion of illiterate women between ages 15 and 49 years in this period in Brazil was 17.5% and 29.4%, respectively.Furthermore, it has been demonstrated that, in that occasion, the reduction observed in fertility rates was due to economic causes,2 reaching especially poorer women living in regions with more precarious living conditions, of whom a large proportion used permanent contraceptive methods (sterilization).7

As to the 1990’s, it is likely that increases in service availability and other social advances be the factors that most contributed to the maintenance of the declining trend in infant mortality8,13 This is especially the case in the North and Northeast, the less economically and socially developed regions. Thus, when analyzing factors associated with infant mortality trends in underdeveloped countries from the 1990’s onwards, no single factor or small group of factors may be considered as most relevant for the reduction of infant mortality levels, since the behaviors observed in relation to some of these factors, such as fertility rate, breastfeeding, and infant nutrition were of little significance in this period. They may have therefore played a minor role in infant mortality trends.19

The decentralization of healthcare services and interventions registered in the 1990’s in Brazil, especially after the implementation of the Normas Operacionais Básicas (basic operational norms - NOB) in 1993,12 may have contributed towards the development of healthcare strategies directed towards the more vulnerable groups. Such decentralization resulted in increased immunization, antenatal care, and medical appointment coverage,9 positively affecting infant mortality rates. Although such social accumulation may constitute compensatory policies or short-term stores of protective factors, the question remains as to whether they will be sufficient to guarantee the pace of the decline in infant mortality in case economic crises persist, increasing social exclusion and intensifying inequities in access to healthcare and consumer goods that assure better living conditions.



1. Ageitos ML, Rico Velazco J, Ojeda NS, Yunes J. La mortalidad de niñez en las Américas: tendencias y diferenciais. Análisis de casos. OPAS/OMS Programa de Salud Maternoinfantil; 1991. p. 348.        

2. Berquó E. Brasil, um caso exemplar – anticoncepção e partos cirúrgicos – à espera de uma ação exemplar. Rev Estudos Feministas (Rio de Janeiro) 1993;1:366-81.        

3. Box GEP, Jenkins GM. Times-series analysis: forecasting and control. 21th ed. San Francisco: Holden-Day; 1976. p. 575.        

4. Buss PM. Saúde e desigualdade: caso do Brasil. In: Buss PM, Labra ME, organizadores. Sistemas de saúde: continuidades e mudanças. Argentina, Brasil, Chile, Espanha, Estados Unidos, México e Quebec. São Paulo: Hucitec; 1995. p. 61-101.        

5 Corteguera RR, Massabot NER, Sabatela RC. Infant mortality in Cuba. Bull Panam Health Organ 1989;23:273-83.        

6. Costa SH, Pinto CS. Os efeitos demográficos da concepção – a queda da fecundidade. In: Conselho Nacional dos Direitos da Mulher (CNDM). Quando a paciente é mulher: Relatório do Encontro Nacional Saúde da Mulher: um direito a ser conquistado. Brasília (DF): Ministério da Justiça; 1989. p. 85-90.        

7. Fundação Instituto Brasileiro de Geografia e Estatística. Estudos e pesquisas - informações demográficas e socioeconômicas nº5. Síntese de Indicadores Sociais 2000. Rio de Janeiro: IBGE; 2001. p. 21-23, 47-50. R1.        

8. Instituto de Saúde Coletiva. Centro de Análise da Situação de Saúde. Análise de situação de saúde nos Estados e municípios do Nordeste. Produção ambulatorial em atenção básica à saúde: avaliação de regiões e Estados, Brasil, 1996-1999. Salvador: UFBA; 2002. p. 51. (Número especial com informações de todos os Estados, 8).        

9. Kleinbaum DG, Kupper LL, Muller KE. Applied regression analysis and other multivariable methods. 2nd ed. Belmont, California: Duxbury Press; 1988. p. 718.        

10. Lardelli P, Blanco JL, Delgado-Rodriguez M, Bueno A, Luna JD, Galvez R. Influence of socioeconomic and health care development on infant and perinatal mortality in Spain 1975-86. J Epidemiol Commun Health 1993;47:260-4.        

11. Levcovitz E, Lima LD de, Machado CV. Política de Saúde nos anos 90: relações intergovernamentais e o papel das Normas Operacionais Básicas. Ciên Saúde Col 2001;6:269-91.        

12. Monteiro CA. Contribuição para o estudo do significado da evolução do coeficiente de mortalidade infantil no município de São Paulo, SP (Brasil) nas três últimas décadas (1950-1979). Rev Saúde Pública 1982;16:7-18.        

13. Monteiro CA, Benício MHA, Freitas ICM. Evolução da mortalidade infantil e do retardo de crescimento nos anos 90: causas e impacto sobre desigualdades regionais. In: Monteiro CA, organizador. Velhos e novos males da saúde no Brasil: a evolução do país e suas doenças. 2a ed. São Paulo: Hucitec/Nupens/USP; 2000. p. 393-420.        

14. Monteiro CA, Iunes RF, Torres AM. A evolução do país e de suas doenças: síntese, hipóteses e implicações. In: Monteiro CA, organizador. Velhos e novos males da saúde no Brasil: a evolução do país e suas doenças. São Paulo: Hucitec/Nupens//USP; 1995. p. 349-56.        

15. Noronha JC, Levcovitz E. AIS-SUDS-SUS: os caminhos do direito à saúde. In: Guimarães R, Tavares R, organizadores. Saúde e sociedade no Brasil: anos 80. Rio de Janeiro: Relume-Dumará; 1994. p. 73-111.        

16. Paim JS, Costa MCN. Decréscimo e desigualdade da mortalidade infantil: Salvador, 1980-1988. Bol OPAS 1993;114:415-28.        

17. Rosero Bixby L. Determinantes del descenso de la mortaidad infantil en Costa Rica. Bol OPAS 1985;99:510-27.        

18. Rutstein SO. Factors associated with trends in infant and child mortality in developing countries during the 1990s. Bul WHO 2000;78:1255-70.        

19. Sandiford P, Morales P, Gorter A, Coyle E, Smith GD. Why do child mortality rates fall? An analysis of the Nicaraguan experience. Am J Public Health 1991;81:30-7.        

20. Santos JLF. Principais técnicas de análise e projeção. In: Santos JLF, Levy MSF, Szmrecsányi, T, organizadores. Dinâmica da população: teoria, métodos e técnicas de análise. São Paulo: T. A. Queiroz; 1980. p. 103-55. (Biblioteca Básica de Ciências Sociais, 3).        

21. Simões CCS, Monteiro CA. Tendência secular e diferenças regionais da mortalidade infantil no Brasil. In: Monteiro CA, organizador. Velhos e novos males da saúde no Brasil: a evolução do país e suas doenças. São Paulo: HUCITEC; NUPENS/USP; 1995. p. 153-6.        

22. Victora CG, Blank N. Epidemiology of infant mortality in Rio Grande do Sul, Brazil. J Trop Med Hyg 1980;83:177-86.        



Correspondence to
Maria da Conceição Nascimento Costa
Rua Padre Feijó, 29, 4o andar, CHR Canela
40110-170 Salvador, Bahia, Brasil
E-mail: mcncosta@ufba.br

Received on 29/7/2002. Reviewed on 3/7/2003. Approved on 21/7/2003.



 * Data from the IBGE Foundation web page: http//www.ibge.gov.br [January, 2001]

Faculdade de Saúde Pública da Universidade de São Paulo São Paulo - SP - Brazil
E-mail: revsp@org.usp.br