Skip to main content
Maternal & Child Nutrition logoLink to Maternal & Child Nutrition
. 2010 Sep 16;6(4):318–327. doi: 10.1111/j.1740-8709.2009.00206.x

A multilevel model for the study of breastfeeding determinants in Brazil

Daniela Wenzel 1,, Ricardo Ocaña‐Riola 2, Gracia Maroto‐Navarro 2, Sônia Buongermino De Souza 1
PMCID: PMC6860883  PMID: 21050386

Abstract

The benefits of breastfeeding for the children's health have been highlighted in many studies. The innovative aspect of the present study lies in its use of a multilevel model, a technique that has rarely been applied to studies on breastfeeding. The data reported were collected from a larger study, the Family Budget Survey‐Pesquisa de Orçamentos Familiares, carried out between 2002 and 2003 in Brazil that involved a sample of 48 470 households. A representative national sample of 1477 infants aged 0–6 months was used. The statistical analysis was performed using a multilevel model, with two levels grouped by region. In Brazil, breastfeeding prevalence was 58%. The factors that bore a negative influence on breastfeeding were over four residents living in the same household [odds ratio (OR) = 0.68, 90% confidence interval (CI) = 0.51–0.89] and mothers aged 30 years or more (OR = 0.68, 90% CI = 0.53–0.89). The factors that positively influenced breastfeeding were the following: higher socio‐economic levels (OR = 1.37, 90% CI = 1.01–1.88), families with over two infants under 5 years (OR = 1.25, 90% CI = 1.00–1.58) and being a resident in rural areas (OR = 1.25, 90% CI = 1.00–1.58). Although majority of the mothers was aware of the value of maternal milk and breastfed their babies, the prevalence of breastfeeding remains lower than the rate advised by the World Health Organization, and the number of residents living in the same household along with mothers aged 30 years or older were both factors associated with early cessation of infant breastfeeding before 6 months.

Keywords: breastfeeding, prevalence, determinants, hierarchical (multilevel) analysis

Introduction

Monitoring child nutrition is fundamental in order to assess the state of child health in a given country (Monteiro et al. 1993). Specialists worldwide recommend the practice of breastfeeding because it provides a superior source of nutrients for infants, an effective form of protection against illness and the expression of affective closeness between mother and child (Ergenekon‐Ozelci et al. 2006). Breastfeeding is defined as giving breast milk to children (whether directly from the breast or pumped and ingested later), independently of the ingestion of solid foods, semi‐solid foods or liquids (including artificial milk) (Ministério da Saúde 2002). The World Health Organization recommends breast milk for infants, particularly up to the age of 6 months (WHO 1991), thus explaining the rationale for the research carried out specifically within this age group.

Since the 1980s, several strategies have been established to increase breastfeeding prevalence in Brazil. According to data from national studies, breastfeeding has increased considerably in all regions of the country. In 1975, 33% of infants aged 0–6 months were breastfed. By 1989, this rate had increased to 49% (Venâncio & Monteiro 1998).

A number of direct and indirect variables are associated with breastfeeding. For instance, an important factor to consider is the age of the mother: older mothers are more likely to breastfeed their children than younger mothers (González et al. 2002). The mother's ethnic group is also linked to breastfeeding: in Brazil, white mothers breastfeed their children less than black or mixed‐race mothers (Rea 1994). Social factors such as the level of the mother's education and income are also related to breastfeeding rates: mothers with a lower level of education and income breastfeed their children more than mothers with a higher level of education and income (Bentley et al. 2003; Haas et al. 2006; Romero et al. 2006; Yeoh et al. 2007). However, the trend is currently changing: the number of better‐educated women with higher incomes and easier access to information who breastfeed their children is on the increase. The trend in Brazil for less well‐educated mothers to breastfeed is not mirrored in the UK or Western Europe, where breastfeeding is associated with higher levels of education. The area of residence, rural or urban, is also correlated to breastfeeding (Marques et al. 2001; Batista Filho & Rissin 2003; Kronborg & Vaeth 2004). According to a survey carried out in Latin America and the Caribbean, women residing in rural areas breastfeed more than those residing in urban areas (Pérez‐Escamilla 2003).

There is a high level of heterogeneity in the literature as regards variables associated to breastfeeding, disfavouring or stimulating it. Levels depend on the methodologies applied, e.g. the size of the sample, the location of the study and data collection. In Brazil, this heterogeneity is observed because of the above factors, largely owing to the cultural differences among geographical regions. However, few national studies have analysed this issue.

Knowledge regarding regional breastfeeding trends and determinants is fundamental in order to assess breastfeeding promotion and support strategies and introduce proposed changes and adjustments. To date, only global studies concerning breastfeeding determinants have been carried out, which have not focused on patterns of dependence or correlation among individuals from a particular region. Dependence between observations means that the use of traditional regression models for the analysis of breastfeeding determinants is unsuitable for studying the different regions. Recent research has suggested new methodological approaches, such as the use of multilevel models for the epidemiological designs of this type (Leyland & Goldstein 2007).

Therefore, this study aimed to estimate the current breastfeeding prevalence in Brazil and its regions, to identify the determinants for breastfeeding, and to study their individual and contextual factors on the prevalence of breastfeeding in children aged up to 6 months.

Key messages

  • • 

    The prevalence of breastfeeding in Brazil was 58%. The northern region of the country has the highest prevalence with 63%, while the lowest was found in the southeastern region with 51%.

  • • 

    The factors unfavourable to breastfeeding in this sample were mothers aged over 30 years and over four persons in the same household. The factors favourable to breastfeeding were having more than two infants less than 5 years in the same household, higher income and dwelling in rural areas of the country.

  • • 

    This study demonstrated low variability of the determinants of breastfeeding among Brazil's large regions. Further studies are needed to examine why the practice of breastfeeding, although frequent throughout the country, remains lower than levels advised by the World Health Organization, despite the pro‐breastfeeding policies adopted in Brazil.

Subjects and methods

Data and study populations

Brazil has a large national representative household‐based study with a complex sample selected in multiple stages involving stratification and clustering, such as the Family Budget Survey [Pesquisa de Orçamentos Familiares (POF)], carried out in 2002–2003 by the Brazilian Institute of Geography and Statistics [Instituto Brasileiro de Geografia e Estatística (IBGE)].

The POF involved interviews conducted in a sample of 48 470 households. The geographical stratification ensured the geographical representativeness of the sample by guaranteeing the participation of all parts of the Brazilian territory. The sampling of POF was structured to produce estimates representative of the country as a whole, its major regions (north, north‐east, south‐east, south and central west) and within both urban and rural settings. A complete overview and further details are described elsewhere (IBGE 2004).

In the questionnaire, sex was registered under four categories: male, female, pregnant and breastfeeding. On this basis, our sample was selected based on all households with children under 6 months, and numbered 1477 children, and assigned the condition of breastfeeding to all children under 6 months whose mothers had reported breastfeeding at the time.

The independent variables in the study were the following: mother's age (<20, 20–25, 26–30 and >30 years), mother's level of education (schooling of up to 4, 5–8 and 9 years or more), the number of residents in the household (1–3, 4–5 and over 6), the number of children in the household under the age of 5 years (1, 2, 3 or more), the ethnic group [white, black, mixed race (i.e. black and white)], the region (north, north‐east, south‐east, south and central west), the household per capita income (according to income quartiles), day nursery attendance, and the type of area (rural or urban). In this paper, the cut‐off points chosen were those commonly used in scientific publications on breastfeeding, allowing results obtained to be compared with those of other studies.

Statistical analysis

Breastfeeding prevalence was estimated for the country as a whole and for the individual regions, calculated by dividing breastfed infants aged 0–6 months by the total number of infants aged 0–6 months, with a 95% confidence interval.

The distribution of breastfeeding determinants in the various regions was analysed by using contingency tables and the chi‐square test with a 5% significance level.

A two‐level hierarchical model was used to perform the analysis, given the structure of the data that were grouped by region. The two levels were subjects (level 1) and regions (level 2) (Bryk & Raudenbush 1992; Goldstein 1995). The effectiveness of this model is higher than that of other models that do not take into account the hierarchical structure of the data, which is why it has also been used by public‐health researchers (Gatsonis et al. 1995; Leyland & Boddy 1997; Sixma et al. 1998).

The hierarchical model used in the analysis establishes that the dependent variable Yij follows the binomial distribution YijBinomial(1,πij) with the conditional variance var(yij|πij) = πij(1 − πij) in which yij is the observed value of the dependent variable and πij is the probability of not breastfeeding for subject i from region j. Therefore, the binomial model establishes:

graphic file with name MCN-6-318-e001.jpg

in which z 0ij is the binomial variance, defined by πij(1 − πij), and e 0ij is the level 1 random error. The variance of this random error is equal to that of the unit in the event of a perfect fit to the binomial distribution or may be estimated based on the data to compare the overdispersion or extra‐binomial variation of the model. If the variance of e 0ij is greater than 1, then there will be an overdispersion in the model, and therefore, an incorrect fit in the data.

The dependent variable may be explained by a series of independent variables represented by X 1, . . . , Xp. With the above conditions, the multilevel logistic regression model may be expressed in the form

graphic file with name MCN-6-318-e002.jpg

in which u 0j is the random effect corresponding to the second level, with normal distribution of zero average and variance Inline graphic (Goldstein 1995).

As the equation of the logit model represents the logarithm of the odds presenting the characteristic of interest, the exponential of the coefficients of the regression model may be interpreted in terms of odds ratios (Goldstein 1995; Leyland & Goldstein 2007).

To identify the statistically significant predictor variables, the Wald test is used, with a level of significance of 10% (Greenland 1989; Mickey & Greenland 1989; Leyland & Goldstein 2007).

To estimate the share of variability in breastfeeding by the independent variables, the expression proposed by Snijders & Bosker (2003) was used, which states:

graphic file with name MCN-6-318-e004.jpg

The share explained by the variables is Inline graphic (variance of the predictor values of the model), and the unexplained share is Inline graphic. In this unexplained variation, σu refers to level two and Inline graphic refers to level one, fixed at 3.29 (Snijders & Bosker 2003).

The proportion of variability unexplained by the model, 1 − R 2, may be broken down into two parts. On the one hand, Inline graphic represents the unex plained variance of level 2. On the other hand, Inline graphicrepresents the unexplained variance of level 1.

The SPSS statistical package, version 15 (SPSS Inc., Chicago, IL, USA), was used for the descriptive analysis, while R software, version 2.4.1 (R Foundation for Statistical Computing, Vienna, Austria), was used for the multilevel data analysis. Generalized linear mixed models via penalized quasi‐likelihood (glmmPQL) were fitted using the glmmPQL function from the MASS library in R. The study was approved by the Research Ethics Committee of the Faculty of Public Health of the University of São Paulo.

Results

In Brazil, 58% of infants aged 0–6 months were breastfed (Table 1).

Table 1.

Breastfeeding prevalence and their respective confidence intervals (CI 95%) in Brazil's macro‐regions, 2002–2003

Regions of Brazil Breastfeeding
Infants Prevalence (%) 95% CI
North 290 63 57–68
North‐east 660 59 55–63
South‐east 200 51 44–57
South 118 61 52–70
Central west 209 56 49–62
Brazil 1477 58 55–60

A total of 75% of the Brazilian population lived in urban areas. The mean age of mothers was 26 years, with an average of 6 years' education. The average income was $485.00. Fifty‐six per cent of the population was classified as mixed race. Forty‐four per cent of all households consisted of four or five people. Table 2 depicts the population spread among Brazil's regions.

Table 2.

Characteristics of the Brazilian population in the different Brazilian regions, 2002–2003

Variables Macro‐regions of Brazil
Brazil North North‐east South‐east South Central West P *
n (%) n (%) n (%) n (%) n (%) n (%)
Area <0.001
 Rural 373 (25.3) 96 (33.1) 177 (26.8) 45 (22.5) 18 (15.3) 37 (17.7)
 Urban 1104 (74.7) 194 (66.9) 483 (73.2) 318 (77.5) 100 (84.7) 172 (82.3)
Mother's education (years) <0.001
 ≤4 481 (34.4) 101 (37.0) 257 (41.3) 58 (30.2) 27 (23.7) 38 (19.5)
 5–8 471 (33.7) 93 (34.1) 189 (30.3) 65 (33.9) 43 (37.7) 81 (41.5)
 ≥9 445 (31.9) 79 (28.9) 177 (28.4) 69 (35.9) 44 (38.6) 76 (39.0)
Children <5 years <0.001
 1 804 (54.4) 132 (45.5) 336 (50.9) 129 (64.5) 93 (78.8) 114 (54.5)
 2 496 (33.6) 113 (39.0) 232 (35.2) 57 (28.5) 20 (16.9) 74 (35.5)
 ≥3 177 (12.0) 45 (15.5) 82 (13.9) 14 (7.0) 5 (4.3) 21 (10.0)
Nursery <0.001
 No 560 (37.9) 119 (41.0) 288 (43.6) 71 (35.5) 35 (29.7) 47 (22.5)
 Yes 917 (62.1) 171 (59.0) 372 (56.4) 129 (64.5) 83 (70.3) 162 (77.5)
Number of residents <0.001
 1–3 342 (23.2) 49 (16.9) 155 (23.5) 55 (27.5) 34 (28.8) 49 (23.4)
 4–5 659 (44.6) 136 (46.9) 259 (39.2) 95 (47.5) 60 (50.8) 109 (52.2)
 ≥6 476 (32.2) 105 (36.2) 246 (37.3) 50 (25.0) 24 (20.4) 51 (24.4)
Mother's age (years) <0.001
 <20 359 (26.1) 78 (28.7) 166 (26.9) 41 (21.9) 21 (18.8) 53 (27.6)
 20–25 330 (23.9) 65 (24.0) 158 (25.6) 37 (19.8) 21 (18.8) 49 (25.5)
 26–30 324 (23.5) 69 (25.5) 137 (22.2) 51 (27.3) 24 (21.4) 43 (22.4)
 >30 365 (26.5) 59 (21.8) 155 (25.3) 58 (31.0) 46 (41.0) 47 (24.5)
Income <0.001
 1st quartile 378 (25.6) 64 (22.1) 245 (37.1) 34 (17.0) 6 (5.1) 29 (13.9)
 2nd quartile 406 (27.5) 100 (34.5) 186 (28.2) 43 (21.5) 21 (17.8) 56 (26.8)
 3rd quartile 326 (22.1) 69 (23.7) 111 (16.8) 54 (27.0) 40 (33.9) 52 (24.9)
 4th quartile 367 (24.8) 57 (19.7) 118 (17.9) 69 (34.5) 51 (43.2) 72 (34.4)
Ethnic group <0.001
 White 563 (38.4) 78 (26.8) 193 (30.3) 105 (52.1) 102 (82.5) 85 (41.5)
 Black 104 (5.5) 12 (2.9) 42 (5.9) 20 (7.8) 14 (5.3) 16 (5.7)
 Mixed race 810 (56.1) 200 (70.3) 399 (63.8) 80 (40.1) 20 (12.2) 111 (52.8)
*

P‐values by chi‐square test.

Table 3 compares the relationship between demographic, social and economic factors, and breastfeeding rates for infants aged 0–6 months to the remaining age groups.

Table 3.

Relation between breastfeeding and demographics, social and economic factors in a multilevel logistic model. Brazil, 2002–2003

Breastfeeding
No (%) Yes (%) OR 90% CI
Children <5 years
 1 42.4 57.2 1
 2 43.7 58.3 1.25 1.00–1.58
 3 40.1 59.9 1.26 0.90–1.74
Nursery
 No 40.2 59.8 1
 Yes 43.0 57.0 0.87 0.71–1.06
Mother's education (years)
 ≤4 41.0 59.0 1
 5–8 39.3 60.7 1.00 0.80–1.26
 ≥9 35.3 64.7 1.17 0.89–1.55
Number of residents
 1–3 34.2 65.8 1
 4–5 42.3 57.6 0.68 0.51–0.89
 ≥6 46.8 53.2 0.63 0.45–0.86
Income
 1st quartile 42.8 57.2 1
 2nd quartile 43.6 56.4 1.05 0.81–1.37
 3rd quartile 40.5 59.5 1.28 0.95–1.72
 4th quartile 40.3 59.7 1.37 1.01–1.88
Ethnic group
 White 34.8 65.2 1
 Black 36.3 63.7 1.02 0.66–1.56
 Mixed race 35.2 64.8 1.03 0.84–1.25
Mother's age (years)
 <20 35.4 64.2 1
 20–25 38.2 61.8 0.85 0.65–1.11
 26–30 35.5 64.5 0.95 0.74–1.25
 >30 44.6 55.4 0.68 0.53–0.89
Area
 Urban 42.4 57.6 1
 Rural 40.5 59.5 1.25 1.00–1.58
Levels Variance Overdispersion
 1 0.998791
 2 0.1227

OR, odds ratio; CI, confidence interval.

The rate for the maintenance of breastfeeding in infants aged 0–6 months is lower in the case of older mothers and also if there are over four persons in the same household. Children born to mothers aged over 30 years have a 32% higher likelihood of not being breastfed than those born to younger mothers. If there are four or five residents in the same household, then children have a 32% higher likelihood of not being breastfed and a 37% higher likelihood if there are over six persons (Table 3).

The variables income, number of children in the household under the age of 5 years and type of area all significantly influenced breastfeeding. Infants from families with higher incomes are 37% more likely to be breastfed than those belonging to lower income quartiles. Families with over two infants under 5 years were 25% less likely to not breastfeed than families with only one infant. Twenty‐five per cent fewer infants living in rural areas of Brazil are not breastfed compared with infants in urban areas (Table 3).

The variance of breastfeeding rates explained by the model's variables was 12%. The remaining 88% is the result of the variables that were not considered in this study. The breakdown of the unexplained variance between the levels of analysis demonstrated a variance of 2% in the second level and 98% in the first level, indicating that there is negligible variability among the regions of Brazil. The overdispersion parameters in the analysis showed no evidence of inappropriate data adjustment in this binomial model. Residuals fitted a normal distribution, and there were no outliers or influent points in the data.

Discussion

This study was carefully planned and carried out so as to avoid sampling and information bias. The selected sample was of an appropriate size and representative of the Brazilian population. Data were collected properly. Therefore, the results obtained and presented in this paper accurately reflect breastfeeding practices in Brazil for the studied period. It is crucial to prevent information bias in order to obtain credible results in research. In this study, biases were neutralized or minimized by thoroughly defining the studied variables, thus guaranteeing the internal validity of the study.

Multilevel analysis is an extension of traditional multivariate regression models. In this sense, all the reported potential confounders were controlled. Maybe there are other unknown key confounders. However, we used those that are most relevant in the scientific literature. As some authors showed, the number of confounders that need to be adjusted for will be limited in order to reduce the risk of overfitting of a multilevel multivariable model (Groenwold et al. 2008).

In the last few decades, according to the data from national surveys, breastfeeding prevalence in Brazil has risen. In 1975, 33% of 6‐month‐old infants were breastfed, compared with 49% in 1989 (Venâncio & Monteiro 1998), 51% in 1996 (BEMFAM/DHS/IBGE/MS/UNICEF 1997) and 58% currently, according to the results of this study. Compared with other countries, breastfeeding prevalence in Brazil is high: prevalence is 33% in the USA (Ryan et al. 2002), 36% in Hong Kong (Lee et al. 2007), 24% in Greece (Antoniou et al. 2005; Bakoula et al. 2007), 58% in Germany (Peters et al. 2006), 47% in Russia (Grjibovski et al. 2005) and 76% in China (Xu et al. 2006).

Breastfeeding prevalence in Brazil is also high when compared with countries in Latin and Central America. Studies carried out in the 1970s and 1980s (Pérez‐Escamilla 2003) demonstrated a slight increase in breastfeeding rates in Ecuador, Peru, and Trinidad and Tobago, no change in Mexico, and a decline in Columbia and the Dominican Republic. In the case of Brazil, the highest prevalence was in the northern region at 63%.

The innovative aspect of the present study lies in its use of a multilevel model, a technique that has rarely been applied to studies on breastfeeding. This relatively recent approach is most suited to research involving hierarchical data and for explaining the individual and contextual variables related to the families' and infants' biological and socio‐economic backgrounds as breastfeeding determinants.

Four people or more in a single household proved to be an unfavourable factor as regards breastfeeding. This situation could be linked to excessive work burden for the mothers and a consequent decrease in the amount of attention given to the child. These data agree with the results presented in the National Survey of Children's Health which indicates that, where a family is composed of members other than the biological parents, there is a risk of children not being breastfed (Singh et al. 2007).

The link between living conditions and breastfeeding practices during the first few months of an infant's life is clear. The results of this study proved that infants aged 0–6 months whose families' income was higher were more likely to be breastfed. It may be assumed that their more privileged economic situation provides these families with easier access to information concerning the benefits of breast milk, which would explain why the more privileged classes value breastfeeding more highly than the lower classes. Factors such as pressure on mothers to return to work, aggressive formula milk marketing campaigns and cultural backgrounds that do not value breastfeeding also increase the likelihood of infants in poorer families being introduced to supplementary baby food at an early age (Victora et al. 1992; Rea et al. 1997; Vasconcelos et al. 2006; Oliveira et al. 2007).

Another important factor is maternal age. Older mothers breastfeed their children less. A plausible explanation for this would be their more stable professional situation. Women who are committed to their career might not have the time needed to take care of their babies or lack the dedication breastfeeding requires. A further reason is the mothers' personal decision not to breastfeed. These results agree with those presented in studies carried out in Hong Kong and the USA, where the intention not to breastfeed was more common in older mothers than younger ones (Ryan et al. 2002; Leung et al. 2003). However, some studies have reported the opposite trend, where older mothers have increased the length of their breastfeeding period (González et al. 2002; Hwang et al. 2006).

Even amid the current trend towards urbanization, mothers in rural areas are more likely to breastfeed their children in the Latin American and Caribbean region (Trussel et al. 1992; Pérez‐Escamilla 2003). This may indicate a certain degree of reversal in the anticipated trend of adopting dominant cultural values by the underprivileged strata of the population. However, according to a study carried out by Ludvigsson (2003) in Bolivia, the opposite situation is true; mothers living in rural areas were four times less likely to breastfeed their children than urban mothers.

Our results indicate that families with over two infants under 5 years are more likely to breastfeed their children. One possible explanation for this finding is the fact that multiparous mothers have had previous experience with breastfeeding, or these mothers possessed the habit and perception necessary for the practice of breastfeeding. Another hypothesis is that even if the mother has not breastfed a child, the fact of being more experienced in relation to maternity could encourage the practice of breastfeeding at other times. In a study representative of the national population in Greece involving 2860 parent–child pairs followed for 8–12 months after childbirth, previous experience of breastfeeding by the mother was shown to be significantly associated with a longer period of breastfeeding (Bakoula et al. 2007). According to Kronborg & Vaeth (2004), in a cohort study conducted in 471 Danish mothers who were followed for up to 4 months after delivery, previous experience of breastfeeding was also positively related to breastfeeding rates.

One of the public‐health authorities' main objectives is to increase the rate of breastfeeding, especially among groups that are less likely to breastfeed at all and those who are likely to stop prematurely. Counselling programmes about breastfeeding, information campaigns (Venâncio & Monteiro 2006), the development of programmes such as the Baby Friendly Hospital Initiative (Bartington et al. 2006), goals such as those set by Healthy People 2010 (CDC – Division of Nutrition, Physical Activity 2007), and the Brazilian trading standard for infant products have all had a positive, significant effect on breastfeeding rates. Successful breastfeeding requires the care of the mother's family, friends and health professionals (Cattaneo et al. 2001). Information is crucial to encourage changes in mothers' beliefs and attitudes (Ergenekon‐Ozelci et al. 2006), and continue the increasing trend in breastfeeding seen over the last decade.

The choice of a multilevel model in this study contributed towards a better understanding of breastfeeding determinants by highlighting the hierarchical relationships among different Brazilian regions as well as the determining factors provided in the clustering levels. A complex system of interrelationships was defined. The results obtained can be valuable for planning effective campaigns in the fields of nutrition and children's health, as well as in more successful monitoring of risk factors linked to breastfeeding.

Some of the variability in breastfeeding not explained by this model could be explained by other individual variants that were not considered in this study. Further research is needed to study individual characteristics. As variability among the regions proved low, it can be assumed that breastfeeding determinants are similar in both regional and national studies.

In any case, the estimation of variability revealed by the independent variables shows that the issue remains under theoretical investigation wherever a logistical regression model is involved. Alternatives suggested by other authors may yield different results but are not yet available in current statistical programmes (Browne et al. 2005).

Overall, there is a need for further analysis of the reasons why the return to breastfeeding, although rapidly increasing throughout the world, is not progressing more swiftly in view of the pro‐breastfeeding policies adopted by governments worldwide. These findings may be interpreted as indicating that mothers can be helped to fully develop their breastfeeding abilities through improvements in the living conditions of children and their families, and the provision of broader access to essential goods and services.

Source of funding

This study was financed by the Federal Government through the Ministry of Planning, Budgeting and Management and the Ministry of Health. It was prepared by the Brazilian Institute of Geography and Statistics (IBGE).

Conflicts of interest

The authors declare that they have no conflicts of interest.

Acknowledgements

The authors thank Wolney L. Conde, from Faculdade de Saúde Pública – São Paulo University, Brazil, for his assistance and excellent comments.

References

  1. Antoniou E., Daglas M., Iatrakis G., Kourounis G. & Greatsas G. (2005) Factors associated with initiation and duration of breastfeeding in Greece. Clinical and Experimental Obstetrics & Gynecology 32, 37–40. [PubMed] [Google Scholar]
  2. Bakoula C., Veltsista A., Prezerakou A., Moustaki M., Fretzayas A. & Nicolaidou P. (2007) Working mothers breastfeed babies more than housewives. Acta Paediatrica 96, 510–515. [DOI] [PubMed] [Google Scholar]
  3. Bartington S., Griffiths L.J., Tate A.R. & Dezateux C. (2006) Millennium Cohort Study Health Group. Are breastfeeding rates higher among mothers delivering in Baby Friendly accredited maternity units in the UK? International Journal of Epidemiology 35, 1178–1186. [DOI] [PubMed] [Google Scholar]
  4. Batista Filho M. & Rissin A. (2003) A transição nutricional no Brasil: tendências regionais e temporais. Caderno de Saúde Pública 19 (Suppl. 1), S181–S191. [DOI] [PubMed] [Google Scholar]
  5. BEMFAM (Sociedade Civil Bem‐Estar Familiar no Brasil)/DHS (Demographic and Health Survey)/IBGE (Fundação Instituto Brasileiro de Geografia e Estatística)/MS (Ministério da Saúde)/UNICEF (Fundo das Nações Unidas para a Infância) (1997) Pesquisa Nacional sobre Demografia e Saúde. BEMFAM/DHS/IBGE/MS/UNICEF: Rio de Janeiro, Brazil. PNDS. [Google Scholar]
  6. Bentley M.E., Dee D.L. & Jensen J.L. (2003) Breastfeeding among low income, African‐American women: power, beliefs and decision making. The Journal of Nutrition 133 (1), 305S–309S. [DOI] [PubMed] [Google Scholar]
  7. Browne W.J., Subramanian S.V., Jones K. & Goldstein H. (2005) Variance partitioning in multilevel logistic models that exhibit overdispersion. Journal of the Royal Statistical Society. Series A 168 (Pt 3), 599–613. [Google Scholar]
  8. Bryk A.S. & Raudenbush S.W. (1992) Hierarchical Linear Models. SAGE Publications: Newbury Park, CA. [Google Scholar]
  9. Cattaneo A., Borgnolo G. & Simon G. (2001) Breastfeeding by objectives. European Journal of Public Health 11, 397–401. [DOI] [PubMed] [Google Scholar]
  10. CDC – Division of Nutrition, Physical Activity (2007) Breastfeeding trends and updated national health objectives for exclusive breastfeeding – United States, birth years 2000–2004. MMWR Morbidity and Mortality Weekly Report 56, Article 30. [PubMed] [Google Scholar]
  11. Ergenekon‐Ozelci P., Elmaci N., Ertem M. & Saka G. (2006) Breastfeeding beliefs and practices among migrant mothers in slums of Diyarbakir, Turkey 2001. European Journal of Public Health 16, 143–148. [DOI] [PubMed] [Google Scholar]
  12. Gatsonis C.A., Epstein A.M., Newhouse J.P., Normand S.L. & McNeil B.J. (1995) Variations in the utilization of coronary angiography for elderly patients with an acute myocardial infarction: an analysis using hierarchical logistic regression. Medical Care 33, 625–642. [DOI] [PubMed] [Google Scholar]
  13. Goldstein H. (1995) Multilevel Statistical Models. Wiley: London. [Google Scholar]
  14. González M.E., Cebrián D.M., Santana R.M., Villanueva E.G. & Santana P.S. (2002) Fatores relacionados com o abandono da lactancia matena. Anales Españoles de Pediatría 56, 144–150. [PubMed] [Google Scholar]
  15. Greenland S. (1989) Modeling and variable selection in epidemiology analysis. American Journal of Public Health 79, 340–349. [DOI] [PMC free article] [PubMed] [Google Scholar]
  16. Grjibovski A.M., Yngve A., Bygren L.O. & Sjöström M. (2005) Socio‐demographic determinants of initiation and duration of breastfeeding in northwest Russia. Acta Paediatrica 94, 588–594. [DOI] [PubMed] [Google Scholar]
  17. Groenwold R.H.H., Van Deursen A.M.M., Hoes A.W. & Hak E. (2008) Poor quality of reporting confounding bias in observational intervention studies: a systematic review. Annals of Epidemiology 18, 746–751. [DOI] [PubMed] [Google Scholar]
  18. Haas D.M., Howard C.S., Christopher M., Rowan K., Broga M.C. & Corey T. (2006) Assessment of breastfeeding practices and reasons for success in a military community hospital. Journal of Human Lactation 3, 439–445. [DOI] [PubMed] [Google Scholar]
  19. Hwang W.J., Chung W.J., Kang D.R. & Suh M.H. (2006) Factors affecting breastfeeding rate and duration. Journal of Preventive Medicine and Public Health 39, 74–80. [PubMed] [Google Scholar]
  20. Instituto Brasileiro de Geografia e Estatística (IBGE) (2004) Pesquisa de orçamentos familiares, 2002–2003. Aquisição alimentar domiciliar per capita, Brasil e grandes regiões. IBGE: Rio de Janeiro, Brazil. [Google Scholar]
  21. Kronborg H. & Vaeth M. (2004) The influence of psychosocial factors on the duration of breastfeeding. Scandinavian Journal of Public Health 32, 210–216. [DOI] [PubMed] [Google Scholar]
  22. Lee W.T., Wong E., Lui S.S., Chan V. & Lau J. (2007) Decision to breastfeed and early cessation of breastfeeding in infants below 6 months old – a population‐based study of 3204 infants in Hong Kong. Asia Pacific Journal of Clinical Nutrition 16, 163–171. [PubMed] [Google Scholar]
  23. Leung T.F., Tam W.H., Hung E.C., Fok T.F. & Wong G.W. (2003) Sociodemographic and atopic factors affecting breastfeeding intention in Chinese mothers. Journal of Paediatrics and Child Health 39, 460–464. [DOI] [PubMed] [Google Scholar]
  24. Leyland A.H. & Boddy F.A. (1997) Measuring performance in hospital care: length of stay in gynaecology. European Journal of Public Health 7, 136–143. [Google Scholar]
  25. Leyland A.H. & Goldstein H. (2007) Multilevel Modelling of Health Statistics. Wiley: New York. [Google Scholar]
  26. Ludvigsson J.F. (2003) Breastfeeding intentions, patterns, and determinants in infants visiting hospitals in La Paz, Bolivia. BMC Pediatrics 3, 5. [DOI] [PMC free article] [PubMed] [Google Scholar]
  27. Marques N.M., Lira P.I.C., Lima M.C., Da Silva N.L., Filho M.B., Huttly S.R.A. et al. (2001) Breastfeeding and early weaning practices in northeast Brazil: a longitudinal study. Pediatrics 108, e66. [DOI] [PubMed] [Google Scholar]
  28. Mickey R.M. & Greenland S. (1989) The impact of confounder selection criteria on effect estimation. American Journal of Epidemiology 129, 125–137. [DOI] [PubMed] [Google Scholar]
  29. Ministério da Saúde. Secretaria de Política de Saúde. Organização Pan Americana de Saúde (2002) Guia Alimentar para crianças menores de dois anos. Ministério da Saúde: Brasília, Distrito Federal, Brazil. [Google Scholar]
  30. Monteiro C.A., Benicio M.H.D., Iunes R., Gouveia N.C. et al. (1993) ENDEF e PNSN: Para Onde Caminha o Crescimento Físico da Criança Brasileira? Caderno de Saúde Pública 9 (Suppl. 1), 85–95. [PubMed] [Google Scholar]
  31. Oliveira L.P.M., Barreto M.L.B., Assis A.M.O., Braga‐Junior A.C.R. et al. (2007) Preditores do retardo de crescimento linear em pré‐escolares: uma abordagem multinível. Caderno de Saúde Pública 23, 601–613. [DOI] [PubMed] [Google Scholar]
  32. Peters E., Wehkamp K.H., Felberbaum R.E., Krüger D. & Linder R. (2006) Breastfeeding duration is determined by only a few factors. European Journal of Public Health 16, 162–167. [DOI] [PubMed] [Google Scholar]
  33. Pérez‐Escamilla R. (2003) Breastfeeding and the nutritional transition in the Latin American and Caribbean Region: a success story? Caderno de Saúde Pública 19 (Suppl. 1), S119–S127. [DOI] [PubMed] [Google Scholar]
  34. Rea M.F. (1994) Avaliação das práticas diferenciais de amamentação: a questão da etnia. Revista de Saúde Pública 28, 365–372. [DOI] [PubMed] [Google Scholar]
  35. Rea M.F., Venâncio S.I., Batista L.E., Santos R.G. & Greiner T. (1997) Possibilidades e limitações da amamentação entre mulheres trabalhadoras formais. Revista de Saúde Pública 31, 149–156. [DOI] [PubMed] [Google Scholar]
  36. Romero S.Q., Bernal R., Barbiero C., Passamonte R. & Cattaneo A. (2006) A rapid ethnographic study of breastfeeding in the North and South of Italy. International Breastfeeding Journal 1, 1–8. [DOI] [PMC free article] [PubMed] [Google Scholar]
  37. Ryan A.S., Wenjun Z. & Acosta A. (2002) Breastfeeding continues to increase into the new millennium. Pediatrics 110, 1103–1109. [DOI] [PubMed] [Google Scholar]
  38. Singh G.K.S., Kogan M.D. & Dee D.L. (2007) Nativity/immigrant status, race/ethnicity, and socioeconomic determinants of breastfeeding initiation and duration in the United States 2003. Pediatrics 119 (Suppl. 1), S38–S46. [DOI] [PubMed] [Google Scholar]
  39. Sixma H.J., Spreeuwenberg P.M. & Van Der Pasch M.A. (1998) Patient satisfaction with the general practitioner: a two‐level analysis. Medical Care 36, 212–229. [DOI] [PubMed] [Google Scholar]
  40. Snijders T.A.B. & Bosker R.J. (2003) Multilevel Analysis. An Introduction to Basic and Advanced Multilevel Modelling. SAGE Publications: London. [Google Scholar]
  41. Trussel J., Grummer‐Strawn L., Rodriguez G. & Vanlandingham M. (1992) Trends and differentials in breastfeeding behaviour: evidence from WFS and DHS. Population Studies 46, 285–307. [Google Scholar]
  42. Vasconcelos M.G.L., Lira P.I.C. & Lima M.C. (2006) Duração e fatores associados ao aleitamento materno em crianças menores de 24 meses de idade no estado de Pernambuco. Revista Brasileira de Saúde Materno Infantil 6, 99–105. [Google Scholar]
  43. Venâncio S.I. & Monteiro C.A. (1998) A tendência da prática da amamentação no Brasil nas décadas de 70 e 80. Revista Brasileira de Epidemiologia 1, 40–49. [Google Scholar]
  44. Venâncio S.I. & Monteiro C.A. (2006) Individual and contextual determinants of exclusive breast‐feeding in São Paulo, Brazil: a multilevel analysis. Public Health Nutrition 9, 40–46. [DOI] [PubMed] [Google Scholar]
  45. Victora C.G., Huttly S.R.A., Barros F.C., Lombardi C. & Vaughan J.P. (1992) Maternal education in relation to early and late child health outcomes: findings from a Brazilian cohort study. Social Science & Medicine 34, 899–905. [DOI] [PubMed] [Google Scholar]
  46. WHO (1991) Indicadores para evaluar las prácticas de lactancia materna. WHO: Geneva. [Google Scholar]
  47. Xu F., Liu X., Binns C.W., Xiao C., Wu J. & Lee A.H. (2006) The decade of change in breastfeeding in China's far north‐west. International Breastfeeding Journal 1, 1–7. [DOI] [PMC free article] [PubMed] [Google Scholar]
  48. Yeoh B.H., Eastwood J., Phung H. & Woolfenden S. (2007) Factors influencing breastfeeding rates in south‐western Sydney. Journal of Paediatrics and Child Health 43, 249–255. [DOI] [PubMed] [Google Scholar]

Articles from Maternal & Child Nutrition are provided here courtesy of Wiley

RESOURCES