Factors associated with breastfeeding in the fi rst hour of life

OBJECTIVE: To identify factors associated with breastfeeding in the fi rst hour of life (Step 4 of the Baby-Friendly Hospital Initiative). METHODS: A cross-sectional study was conducted with a representative sample of mothers who gave birth in maternity wards in the city of Rio de Janeiro, Southeastern Brazil, between 1999 and 2001. Newborns or mothers with restriction to breastfeeding were excluded, resulting in a sample of 8,397 pairs. A random effect – at maternity hospital level – Poisson model was employed in a hierarchical approach with three levels: distal, intermediate and proximal for characteristics of the mother, of the newborn, and of prenatal and hospital assistance. RESULTS: Only 16% of the mothers breastfed in the fi rst hour of life. Breastfeeding in this period was less prevalent among neonates with immediate intercurrences after birth (PR = 0.47; CI99% 0.15;0.80); among mothers who did not have contact with their newborns in the delivery room (PR = 0.62; CI99% 0.29;0.95); among mothers submitted to cesarean section delivery (PR = 0.48; CI99% 0.24;0.72); and among mothers who gave birth at private maternity hospitals (PR = 0.06; CI99% 0.01;0.19) or at maternity hospitals contracted out to National Health System (SUS) (PR = 0.16; CI99% 0.01;0.30). The context effect of maternity wards was statistically signifi cant. CONCLUSIONS: At an individual level, breastfeeding within one hour after birth was constrained by inappropriate practices in private or SUS-contracted maternity hospitals. The group effect of maternity hospitals and the absence of individual maternal-related factors that explain the outcome suggest that mothers have little or no autonomy to breastfeed their babies within the fi rst hour of life, and depend on the institutional practices that prevail at the maternity hospitals. DESCRIPTORS: Breast Feeding. Postpartum Period. Hospitals, Maternity. Maternal-Child Health Services. Cross-Sectional Studies.


INTRODUCTION
Breastfeeding in the fi rst hour of life is recommended by the World Health Organization (WHO) a and corresponds to step four of the Baby-Friendly Hospital Initiative (BFHI).This is one of the fundamental strategies to promote, protect and support breastfeeding in Brazil 21 and it is based on the newborns' (NB) capacity to interact with their mothers in the fi rst minutes of life.This contact a United Nations International Children's Emergency Fund.Breast-feeding management and promotion in a baby-friendly hospital: an 18-hour course for maternity staff.New York: 1993.
is important to establish the mother-baby bond. 4In addition, it increases the duration of breastfeeding, 2,4,13,18,25 the prevalence of breastfeeding in maternity hospitals, 25 and reduces neonatal mortality. 10However, the practice of breastfeeding in the fi rst hour of life is relatively low in Brazil (43%) (National Survey of Demography and Maternal and Child's Health, 2008).b The set of practices, structures and routines and the quality of human resources of maternity hospitals may interfere in the time that elapses until the fi rst breastfeed, 3 since the hospitals' fi nancing source (public, military, private or contracted out to National Health System (SUS) produces differences in the quality of assistance, of the practices and in morbidity. 15prior study evaluated factors associated with the time that elapses between birth and the fi rst breastfeed in the fi rst 24 hours of life. 3Then, there was the need to perform an evaluation focusing on step four of BFHI, in view of its importance in the breastfeeding policy in Brazil. 21In relation to the fi rst paper, there were changes: in the type of outcome, in time (from continuous to dichotomic), and in the connection functions of the statistical models (from survival analysis in the fi rst paper to Poisson distribution with robust variance in the current paper).The variables are similar, but some of them were excluded from the current paper, as they were associated with what happened after the fi rst hour of life.Thus, the present study aimed to identify factors associated with breastfeeding in the fi rst hour of life.

METHODS
Cross-sectional study with data extracted from a research carried out in the municipality of Rio de Janeiro, Southeastern Brazil, between 1999 and 2001 to investigate perinatal morbidity and mortality. 15A stratifi ed sample was used, proportional to the number of deliveries expected in the period in all maternity hospitals with more than 200 deliveries/year in the municipality.The 47 selected maternity hospitals were grouped into three strata, according to the fi nancing source: 1-municipal and federal (n=12; 34.8% of deliveries); 2-SUS-contracted philanthropic and private, military and state-owned (n=10; 34.4% of deliveries); 3-private (n=25; 30.8% of deliveries). 15 our study, 10,071 mothers with valid information about their deliveries were interviewed.Mothers who were incapable of and/or prevented from breastfeeding due to one or more of the following characteristics were excluded: NB with very low birth weight (inferior to 1500g); gestational age (Capurro method) inferior to 32 weeks; early fetal or neonatal death; maternal death; NB in Intensive Care Unit (ICU); puerperal woman in ICU; HIV-positive serology during prenatal care documented in the mother's medical record; and Apgar score lower than seven at the fi fth minute.Overall, 895 deliveries (8.9% of the sample) were excluded.Approximately 8.5% of the mothers (n=779) did not want to or did not know how to answer the question about breastfeeding or not in the fi rst hour of life (missings); therefore, 8,397 deliveries were analyzed in the current study.Missing and non-missing groups were similar regarding the study's main variables, like age, level of schooling and parity (99% confi dence interval -99%CI).Of the total number of deliveries, 101 referred to two or more babies.In these cases, the time that elapsed until the breastfeeding of the fi rst baby was considered.
The outcome -breastfeeding in the fi rst hour of life (yes/no) -was obtained from the interview questionnaire with the puerperal woman through a 24-hour recall.Breastfeeding in the first hour of life was considered to be the offer of the breast in up to sixty minutes after birth.
For the analysis, 26 variables of the questionnaire were selected: household characteristics (basic sanitation and people/room ratio), personal characteristics (level of schooling, marital status, parity, skin color), gestational characteristics (desire to become pregnant, physical aggression suffered by the pregnant woman during pregnancy, smoking, alcohol ingestion, maternal age, support given by the child's father), prenatal care characteristics (receiving information about breastfeeding during prenatal assistance, abortion attempt, prenatal score), hospital assistance characteristics (newborn taken to the mother right after birth, presence of a companion in the delivery room), and characteristics of the evaluation of delivery assistance and breastfeeding information (Tables 1, 2 and 3).
Other variables were obtained from the medical record: newborn intercurrences immediately after delivery, birth weight, presence of congenital anomalies, number of fetuses, newborn's sex, use of oxytocin during hospitalization and type of delivery.The fi nancing source of the maternity hospitals was established by the authors of the original research based on information available in the SUS.
Bivariate analysis (Pearson's chi-square test) was performed for each variable in order to investigate the association with the outcome.Variables with p value < 0.20 were selected for the statistical model, thus avoiding residual confounding.The variables parity, maternal age and birth weight were categorized for this stage of the analysis, but the last two were considered continuous in the other stages of the study, and parity was considered ordinal.Then, the statistical models were estimated, and hierarchized multilevel regression analysis with Poisson distribution -robust variance -was employed to evaluate the outcome.The program R c was utilized in the analysis.
Regression with Poisson distribution was used for point and interval estimates because logistic regression with binomial distribution can overestimate the associations with the outcome, as the event is not rare. 6The hierarchization of variables enabled to evaluate the effect of the groups of variables according to proximity to the outcome.The multilevel approach was useful to evaluate the group effect (cluster) of the 47 studied maternity wards.
Values between zero and one obtained from statistical models represent protective factors, while values higher than one have a risk meaning.As breastfeeding in the fi rst hour of life is the evaluated outcome, the Prevalence Ratios (PR) obtained in the models that present values higher than one should be interpreted as factors that increase the prevalence of breastfeeding in the fi rst hour.Factors between zero and one should be interpreted as factors that reduce the prevalence of breastfeeding in the fi rst hour, both in relation to a certain reference category.
The residuals were evaluated by means of the functions of the R-package.The complete model was compared by analysis of deviance to the other models without the variables with p > 0.05, under the null hypothesis that the embedded model with deviance, with less parameters, better adjusts the data to the estimated parameters. 11e standardized residuals were graphically observed (versus the linear predictors), as well as the normality of the residuals, the infl uential observations (Leverage), and the distance measure of the vector of estimates caused by setting the i-th information aside (Cook's distance). 11To evaluate the quality of the models' adjustment, the goodness-of-fi t test was used, with chi-square distribution, under the null hypothesis that the model is well adjusted (with 5% level).After estimating the effects of the third model, multilevel analysis was performed (model 4), in which the 47 maternity hospitals were inserted in order to control for group effect, with intercept variation. 12At the end of this process, the possible interactions between the individual level variables were tested through analysis of deviance (99%CI).
As models 3 and 4 were not nested, the comparison between them was performed by means of the Akaike Information Criteria -AIC.
The models were composed of: six variables of the distal level (model 1); nine of the intermediate level (model 2) and ten variables of the proximal level (model 3) (Figure 1).
The study was submitted to the Research Ethics Committee of ENSP/FIOCRUZ (protocol no.16 of 2007).

RESULTS
The study showed that 16.1% of the NB were breastfed during the fi rst hour of life.Large disparities were observed in breastfeeding initiation between the strata of fi nance of the maternity wards: more than one third of the babies born in municipal and federal maternity hospitals were breastfed in the fi rst hour, compared to less than 2% of the babies born in private maternity hospitals (Table 1).This proportion also varied between babies born by Cesarean section (5.8%) and those delivered vaginally (26.4%).
Approximately half of the puerperal women underwent Cesarean section.The majority received the newborn right after birth, one third could have a companion in the delivery room and one tenth of the newborns presented some type of intercurrence after birth, like hypoxia, bradycardia, among others (Table 1).
Around one fourth of the women did not receive any type of information on breastfeeding during the prenatal assistance (Table 1).Among maternal and NB characteristics (Table 2), only number of fetuses, presence of congenital anomalies and birth weight were not associated with the outcome.One third had concluded secondary or higher education (Table 3).
The prevalence of breastfeeding in the fi rst hour of life was approximately 50% lower among newborns with intercurrences right after birth and newborns that were not taken to their mothers while in the delivery room, and also among mothers submitted to Cesarean section.This prevalence was approximately 90% lower at private, military, state-owned or SUS-contracted maternity hospitals when compared to the babies born at municipal or federal maternity wards.
The distal and intermediate variables lost statistical signifi cance when the more proximal variables and the context effect of the maternity hospitals were added (Table 4).
The inclusion of the maternity hospitals as group effect (model 4) was statistically signifi cant, which indicates that the maternity hospital determines the prevalence of breastfeeding in the fi rst hour of life.After this, the effects of the variables included in model 3 maintained its estimates similar to those of the model without this group effect (Table 4).However, the increase in the length of the confi dence intervals made the variable "use of oxytocin during delivery" loses statistical signifi cance.
Models 1, 2 and 3 were statistically different among each other (deviance test, p<0.01), and had a good adjustment of residuals.Model 4 could not be compared to the others by means of the deviance test due to the inclusion of the group variable (maternity hospitals).
All the possible interactions between the correlated variables of model 4 were tested and, although three interactions presented statistical relevance, they did not add much to the model in terms of signifi cance (analysis of deviance, 99%CI).Therefore, it was decided not to consider the interactions in order to maintain parsimony.

DISCUSSION
Although it is a practice recommended by the WHO, breastfeeding in the NB's fi rst hour of life was not very frequent in the studied population and there was great variation in this outcome across the strata of maternity hospitals.
No maternal factors explained the outcome, which may indicate that mothers have little or no autonomy to breastfeed their babies in the first hour of life.Therefore, they depend on the institutional practices that prevail in the maternity hospitals and on the professionals involved in the birth of their babies.Not always are the mothers' feelings and will respected in the moment of birth, and at this time of fragility, the professional conduct may be a determinant of breastfeeding in the delivery room.
Thus, the health professional plays a governing and regulating role over breastfeeding based on a constructed scientifi c knowledge, 19 and s/he must act ethically, respecting the mother-baby binomial to strengthen their bonds.
The prevalence of breastfeeding in the fi rst hour of life that was observed in the maternity hospitals of the city of Rio de Janeiro is below the prevalence found in the Southeast region in three national surveys carried out in: 1996 (38.7%;National Survey on Demography and Health-PNDS d ), 2006 (37.7%;PNDS b ), and 2008 (63.5%;Research on Prevalence of Breastfeeding in the Brazilian Capitals and Federal District, 2009 e ).However, it was similar to another study conducted in two maternity wards of the city in the same period (19.5% in a public maternity hospital and 15.8% in a private hospital). 8These differences may derive from a memory bias resulting from the methodology of the PNDS, d whose target population was formed by mothers of children aged up to 59 months.The discrepancies observed in the initiation of breastfeeding between public and private hospitals remained even after adjusting the model by individual and group factors.These differences were also observed in a study conducted in the city of São Paulo 23 and in Rio de Janeiro, 8 both in Southeastern Brazil, and they may be explained by the investment in the implementation of the BFHI in public maternity hospitals and by the strong growth of this initiative in the last decades in Brazil. 14 the time the study was carried out, the federal maternity hospitals were managed by the municipal sphere of Rio de Janeiro, except for Hospital Servidores do Estado, and were grouped in the same stratum.The municipal maternity hospitals traditionally receive training and follow the breastfeeding policies of the municipality; 5 many of them were undergoing the BFHI accreditation process and had a Human Milk Bank.
The fi rst paper of the original study 15 grouped the maternity hospitals in three strata, according to the proportion of low birth weight, and the second stratum concentrated the most heterogeneous group of hospitals.In this stratum, the contracted maternity hospitals did not follow the public breastfeeding policies, although they were fi nanced by the SUS.The other maternities of the same stratum, except for a university hospital, were not implementing the BFHI at the time.
The private hospitals (third stratum) tended not to adopt any policy for the promotion of breastfeeding at the time the study was carried out.
These differences between the breastfeeding policies adopted by the maternity hospitals were refl ected on the differences in the prevalence of breastfeeding within the fi rst hour of life.
About the statistical modeling process, the main advantage of maintaining the variables of a previous level that lost statistical relevance is to be able to observe of the process of intermediation of effects when a group of variables from a more proximal level is added.
In the univariate analysis, it was observed that the married mothers, with lower people/room ratio, with access to basic sanitation, who wanted to get pregnant, who did not suffer physical aggression during pregnancy, did not smoke and did not drink alcoholic beverages during pregnancy, who received support during pregnancy, received information on breastfeeding during pregnancy, who did not try to abort and had a better follow-up during prenatal assistance (distal and intermediate variables), had lower probability of breastfeeding in the fi rst hour, while the contrary was expected.When these variables are analyzed together in a statistical model they cease to be signifi cant, because maternal variables like age and parity can explain, for example, the access to the maternity hospitals and the type of delivery to which these women are submitted. 1e mothers' socioeconomic indicators may correlate with the type of hospital in which the babies are born. 15he decision to have a Cesarean section may also be determined during the pregnant woman's contact with the healthcare services. 9In the current study, the effect of these variables was mediated by factors related to delivery care (like the type of delivery, satisfaction with the provided assistance, use of oxytocin, and fi nancing source of the maternity hospital) and by the maternity hospital as context effect.
In model 4, this group or context effect was so important that the factors "parity" and "use of oxytocin in the delivery room" ceased to be statistically signifi cant in the fi nal model.
The Cesarean section was responsible for reducing by half the prevalence of breastfeeding in the fi rst hour in the maternity hospitals, considering the fi nal model (model 4).The effect of the Cesarean delivery on the delay of the fi rst breastfeed is shown by several studies, 1,3 and it may be related to the anesthetic and to the surgical procedures performed in the postpartum period.The study by D'Orsi et al, also conducted in Rio de Janeiro, found higher prevalence of breastfeeding at birth in the normal deliveries (33% in a public maternity hospital and 23.7% in a private maternity hospital) than in the Cesarean deliveries (6.9% in the public and 8% in the private maternity hospital). 8eastfeeding in the fi rst hour of birth was less prevalent among NB with immediate intercurrences after birth.Although this effect was expected, this study's design removed a priori from the analysis the NB that could have some diffi culty in feeding from their mother's breast (such as babies with very low birth weight and Apgar score lower than seven).As this specifi c question was obtained by means of an interview with the mothers, and as the discrimination of the types of intercurrences varied considerably (it referred to an open question in the structured questionnaire), it was not possible to categorize this variable.
The mother not having contact with the newborn in the delivery room reduced the prevalence of breastfeeding in the fi rst hour of life.Many times, this contact is delayed or minimized in favor of the several routines, care procedures and practices determined by the team of professionals and by the institutions, and many of them are possibly ineffi cient. 16qualitative study found out that, although the mothers felt awkward to breastfeed in the delivery room, and although it was strange for them to see the newborn dirty with blood and amniotic fl uid, breastfeeding right after the birth was accepted by the mothers and understood as a component of maternity. 17en in a hospital that institutionalizes humanization routines there may be differences in the assistance procedures, and this can reduce the empowerment potential contained in the decision to breastfeed if women's sociocultural differences are not considered. 24other study observed that, despite the importance given to the initial contact with the NB immediately after birth, they were separated from each other, which generated, in the mothers, feelings of fear and concern, but they also valued this separation, as they believed in the importance of these procedures for the maintenance of the physical integrity of their children. 7e maternity hospitals produce a group effect on breastfeeding in the fi rst hour of life, that is, the prevalence of this practice is very similar among newborns at one single maternity hospital, but it varies across maternity hospitals independently of the individual factors that may favor or not breastfeeding in the fi rst hour.The norms, routines, number of beds (and their form of occupation), investment in the professionals' qualifi cation, the institutional tradition, the context of the neighborhood, accessibility and other possible factors that were not measured generate a unique reality in each maternity hospital regarding breastfeeding in the fi rst hour of life, which could be verifi ed by means of the multilevel approach.
The main limitation of the study was the fact that it was not originally designed to evaluate breastfeeding in the fi rst hour of life.Thus, important pieces of information were not collected, such as: if the fi rst breastfeed occurred in the delivery room, which professional took the baby to the mother's breast, aspects related to latch-on and breastfeeding position, among others that could enrich the analysis and discussion of the theme.
One of the ways of modifying the current reality is professional qualification.It is necessary that the professionals develop competencies and skills in breastfeeding to perform adequate interventions and overcome the possible barriers to breastfeeding, especially in the delivery room.To achieve this, state and municipal health departments should offer institutional support through BFHI training provided for professionals from the public, SUS-contracted and private networks (WHO/PAHO/UNICEF a ).
In addition, it is fundamental that all the institutions, mainly the private, SUS-contracted, state-run, philanthropic and military ones, invest in the implementation of institutional and public policies, like the BFHI, that promote the practice of breastfeeding in the fi rst hour already during prenatal assistance.The approach to women concerning breastfeeding should be continuous: it should be stimulated during prenatal care, its practice should be initiated early in the maternity hospital and supported during the mother-baby follow-up.All of these should be performed by professionals who are qualifi ed in the handling of lactation and in hearing mothers' experiences and doubts.The interaction between the BFHI and the Breastfeeding-Friendly Primary Care Unit Initiative should be intensifi ed, so that the breastfeeding promotion, protection and support policies succeed in extending the duration of exclusive breastfeeding, a challenge today in Brazil. 20thers must be empowered to breastfeed in the delivery room, respecting their particularities and sociocultural diversities.The women should be the subject in the act of breastfeeding in the fi rst hour of life; this should not be one more procedure to which they are submitted in behalf of humanization ideas.This empowerment should begin during prenatal assistance, through a dialog between the health team and the woman about all the potential benefi ts of breastfeeding in the fi rst hour of life, so that she is able to evaluate and construct her choices.

Figure .
Figure.Breastfeeding in the fi rst hour of life according to a hierarchized, random-effect model at maternity hospital level.Municipality of Rio de Janeiro, Southeastern Brazil, 1999-2001.

Table 1 .
Proportion of children who were breastfed in the fi rst hour of life, according to prenatal care and hospital assistance characteristics in the maternity hospitals.Municipality of Rio de Janeiro, Southeastern Brazil, 1999-2001.

Table 2 .
Proportion of children who were breastfed in the fi rst hour of life, according to maternal and newborns' characteristics.Municipality of Rio de Janeiro, SoutheasternBrazil, 1999-2001.

Table 3 .
Proportion of children who were breastfed in the fi rst hour of life, according to maternal and newborns' characteristics.Municipality of Rio de Janeiro, SoutheasternBrazil, 1999-2001.

Table 4 .
Factors associated with breastfeeding in the fi rst hour of life in maternity hospitals.Municipality of Rio de Janeiro, SoutheasternBrazil, 1999-2001.