Women's intentions to breastfeed: a population-based cohort study

Authors

  • O Lutsiv,

    1. Department of Epidemiology, Biostatistics and Occupational Health, McGill University, Montreal, QC, Canada
    Search for more papers by this author
  • E Pullenayegum,

    1. Department of Clinical Epidemiology & Biostatistics, McMaster University, Hamilton, ON, Canada
    2. Centre for Evaluation of Medicines, St Joseph's Healthcare, Hamilton, ON, Canada
    Search for more papers by this author
  • G Foster,

    1. Department of Clinical Epidemiology & Biostatistics, McMaster University, Hamilton, ON, Canada
    2. Biostatistics Unit, Father Sean O'Sullivan Research Centre, St Joseph's Healthcare, Hamilton, ON, Canada
    Search for more papers by this author
  • C Vera,

    1. Departmento de Obstetricia y Ginecologia, Escuela de Medicina, Pontificia Universidad Catolica de Chile, Santiago, Chile
    Search for more papers by this author
  • L Giglia,

    1. Division of General Pediatrics, Department of Pediatrics, McMaster University, Hamilton, ON, Canada
    Search for more papers by this author
  • B Chapman,

    1. Better Outcomes Registry and Network (BORN) Ontario (formerly OPSS), Ottawa, ON, Canada
    Search for more papers by this author
  • C Fusch,

    1. Division of Neonatology, Department of Pediatrics, McMaster University, Hamilton, ON, Canada
    Search for more papers by this author
  • SD McDonald

    Corresponding author
    1. Department of Clinical Epidemiology & Biostatistics, McMaster University, Hamilton, ON, Canada
    2. Division of Maternal–Fetal Medicine, Departments of Obstetrics & Gynecology and Diagnostic Imaging, McMaster University, Hamilton, ON, Canada
    • Correspondence: Dr SD McDonald, Division of Maternal–Fetal Medicine, Departments of Obstetrics & Gynecology, Radiology, and Clinical Epidemiology & Biostatistics, McMaster University, 1280 Main St W., Hamilton, ON, Canada L8S 4K1. Email mcdonals@mcmaster.ca

    Search for more papers by this author

Abstract

Objective

Given that intention to breastfeed is a strong predictor of breastfeeding initiation and duration, the objectives of this study were to estimate the population-based prevalence and the factors associated with the intention to breastfeed.

Design

Retrospective population-based cohort study.

Setting

All hospitals in Ontario, Canada (1 April 2009–31 March 2010).

Population

Women who gave birth to live, term, singletons/twins.

Methods

Patient, healthcare provider, and hospital factors that may be associated with intention to breastfeed were analysed using univariable and multivariable regression.

Main outcome measures

Population-based prevalence of intention to breastfeed and its associated factors.

Results

The study included 92 364 women, of whom 78 806 (85.3%) intended to breastfeed. The odds of intending to breastfeed were higher amongst older women with no health problems and women who were cared for exclusively by midwives (adjusted OR 3.64, 95% CI 3.13–4.23). Being pregnant with twins (adjusted OR 0.73, 95% CI 0.57–0.94), not attending antenatal classes (adjusted OR 0.58, 95% CI 0.54–0.62), having previous term or preterm births (adjusted OR 0.79, 95% CI 0.78–0.81, and adjusted OR 0.87, 95% CI 0.82–0.93, respectively), and delivering in a level–1 hospital (adjusted OR 0.85, 95% CI 0.77–0.93) were associated with a lower intention to breastfeed.

Conclusions

In this population-based study ~85% of women intended to breastfeed their babies. Key factors that are associated with the intention to breastfeed were identified, which can now be targeted for intervention programmes aimed at increasing the prevalence of breastfeeding and improving overall child and maternal health.

Introduction

Breastfeeding has been recognised by many authorities as the optimal method of infant feeding,[1, 2] because of its well-established role in promoting infant growth, immunity, and cognitive development.[3-5] The child health benefits of breastfeeding extend even further to include lower incidences of: infections, sudden infant death syndrome, obesity, diabetes, lymphoma, and asthma.[6] In addition to child benefits, breastfeeding is also associated with lower risks of breast cancer, ovarian cancer, and diabetes in mothers.[6-9]

Despite the recognised benefits of breastfeeding, its prevalence has been on the decline for several decades, with more mothers opting to formula feed.[10, 11] Although there has been a recent resurgence in breastfeeding,[12] the rates still remain disappointingly low.[13, 14]

As maternal intention to breastfeed has been shown to be strongly correlated with the initiation and longer duration of breastfeeding,[15-19] determining the factors that are associated with it could help in an effort to increase and sustain breastfeeding. Although some factors have already been identified, the current knowledge about the true prevalence of intention to breastfeed and its associated variables is inadequate because of the methodological limitations of existing studies (mainly self-administered surveys), including volunteer bias, selection bias, and recall bias, which are addressed by the use of population-based research. Consequently, there have been calls for additional, methodologically sound research to be undertaken.[20]

It is important to identify the factors associated with a woman's intention to breastfeed, so that effective counselling practices and promotion policies can be implemented to improve breastfeeding rates. As patient factors alone can only partially predict a woman's decision to breastfeed, a thorough exploration of other variables, including healthcare provider and hospital factors, is necessary. Our aims were to determine population-based rates of intention to breastfeed, and to understand the patient, healthcare provider, and hospital factors associated with a woman's decision to breastfeed.

Methods

Study design

A retrospective population-based cohort study was conducted according to the STROBE (Strengthening the Reporting of Observational Studies in Epidemiology) guidelines.[21] This is part of a series of studies on breastfeeding: the details of the first study, which focused on exclusive breastfeeding at hospital discharge, have been reported elsewhere.[14]

Data source

The population-based Better Outcomes Registry & Network (BORN) database was used as the source of information on all hospital births in the province of Ontario, the most populous province in Canada.[22] BORN is a prescribed registry under Ontario's Personal Health Information Protection Act,[23] which maintains high data quality by the formal training of new users and logic-checking mechanisms that are built into the system to minimise missing or erroneous data. Quality reports are conducted regularly and sent to each hospital in order to edit the data and correct any errors. A recent quality audit of BORN data found that for the variables that were included in this study, the percentage agreement with chart data ranged from the mid to high 90s.[24]

Inclusion and exclusion criteria

Women were eligible for study inclusion if they gave birth to a live singleton or twin, at term (from 37 weeks and 0 days of gestation to 41 weeks and 6 days of gestation), between 1 April 2009 and 31 March 2010, and if they had information on breastfeeding at maternal–infant discharge. Women were excluded if they had any of the following: triplets and higher order pregnancies; stillborn infants; neonatal deaths; infants with congenital anomalies; infants discharged to a location other than home (to another hospital), or discharged from a neonatal intensive care unit or special care unit (because their information is not included in BORN); and planned home births (which are not included in BORN).

Outcome

Our primary outcome was antenatal intention to breastfeed. This was usually assessed antenatally in preregistration clinics during the third trimester, or by a nurse when the woman was admitted to the labour floor. The outcome was recorded as (‘Yes’, ‘No’, or ‘Unknown’) in the patients’ BORN charts. We treated ‘unknown’ and blank responses as missing information.

Variables

The variables that were explored to determine their associations with intention to breastfeed were grouped into maternal, antenatal, and hospital factors. The following maternal variables were included: age; number of pervious term and preterm babies; physical or mental health problems; illicit drug use; and smoking. Although individual patient income and employment data were not available, information about the patients' median neighbourhood income (entered in $CDN20 000 increments in the regression model), and the median proportion employed and in the labour force in their neighbourhood, was provided by BORN from linkage with Statistics Canada (these data were defined according to Statistics Canada census data,[25] based on the postal code, which is a six-character code defined and maintained by the national postal system for the purpose of sorting and delivering mail).

The antenatal variables that were analysed include: obstetrical/pregnancy complications [with one or more of the following: gestational diabetes, hypertension, suspected intrauterine growth restriction (IUGR)/small for gestational age (SGA), large for gestational age (LGA), periodontal infection, placenta praevia, placental abruption, pre-eclampsia, premature rupture of membranes, preterm labour, preterm premature rupture of membranes, and urinary tract infection (UTI)]; twin pregnancy; use of assisted reproductive technology; type of antenatal care provider (exclusive care by: midwife, family doctor, nurse practitioner, or obstetrician, or no exclusive care provider); and attendance at prenatal classes.

The hospital variable analysed was hospital level, categorised into four groups based on the most acute level of care available at the hospital: level 1, care of mothers with healthy pregnancies, and their infants, with births at ≥36 weeks of gestation; level 2, care of mothers with low-to-moderate risk pregnancies, and their infants, with births at ≥32 weeks of gestation; level 2+, care of mothers with moderate risk pregnancies, and their infants, with births at 30–32 weeks of gestation; and level 3, subspecialty care for high-risk pregnancies, including mothers with severe medical complications, unwell/unstable newborns, multiple pregnancies of triplets and higher order multiples, and births at <32 weeks of gestation.[26]

Statistical analysis

Baseline descriptive statistics were performed to characterise women who were intending to breastfeed compared with those who were not. Each variable that is potentially associated with intention to breastfeed (maternal, antenatal, and hospital variables) was first examined in a univariable fashion using logistic regression. In a multivariable logistic regression model, we then adjusted for all maternal, antenatal, and hospital variables, to yield adjusted odds ratios (ORs) for intention to breastfeed. To decrease the probability of any given woman having more than one pregnancy included in the data, only a single year of data was analysed. Missing variables were imputed using multiple imputation,[27, 28] and the results with and without multiple imputation were compared in a sensitivity analysis. The area under the curve (AUC) was also calculated as a goodness-of-fit measure for the multivariable logistic regression model.

Sample size

The a priori sample size calculation was performed for the initial study in this series, which focused on exclusive breastfeeding,[14] and determined an 80% power to detect relative risks (RRs) as small as 1.3 (= 1.0/0.77) that are associated with rare occurrences (e.g. twins, ~2% incidence). With intention to breastfeed one would expect a slightly lower power to detect the same RR, as intention was more common than exclusive breastfeeding.

Analyses were performed using sas 9.2 (SAS Institute Inc., Cary, NC, USA). The study was approved by the Faculty of Health Sciences/McMaster University Research Ethics Board.

Results

A total of 142 373 women gave birth between 1 April 2009 and 31 March 2010 in Ontario, Canada. We did not have any information about planned home births (n = 2677), as they are not included in BORN. The remaining 139 696 mothers delivered in hospitals in Ontario, Canada (Figure 1). From these we excluded women who gave birth to premature or postdates infants (<37 weeks or ≥42 weeks of gestation, n = 12 362). Furthermore, we excluded women with triplets or higher order multiple gestations, neonatal deaths, babies with congenital anomalies (as these could impact the ability to be breastfed), and infants discharged to another hospital or from a neonatal intensive care unit or special care unit, as they do not have information in BORN (n = 12 113 total). This yielded 115 221 women who gave birth to live singletons or twin infants at term. Of these, there were 22 316 (15.7% of 142 373) women with missing information on infant feeding at discharge. Baseline characteristics of the women with and without information on feeding at discharge were similar on all characteristics compared (maternal ages 29.8 versus 30.6 years, respectively; neighbourhood median family incomes $CDN72 794 versus $CDN73 650; mean number of previous pregnancies 0.89 versus 0.89; and median gestational ages at delivery 39 versus 39 weeks).

Figure 1.

Flow chart of subject selection; N = number.

Thus there were 92 364 women who met our inclusion criteria (delivering at term from 37 weeks and 0 days of gestation to 41 weeks and 6 days of gestation) with information about breastfeeding at maternal–infant discharge. Of these, 78 806 (85.3%) women were antenatally intending to breastfeed.

Descriptive statistics

Maternal variables

The baseline characteristics of the study sample, categorised according to the women's antenatal breastfeeding intentions, are shown in Table 1. Women who were intending to breastfeed were on average older (29.9 years old versus 28.1 years old, respectively), and had a higher median neighbourhood family income ($CDN72 983 versus $CDN66 206, respectively), than women who were not intending to breastfeed. Women intending to breastfeed were also less likely to have had a previous term or preterm baby compared with women not intending to breastfeed (53.0 versus 67.4% and 4.2 versus 5.6%, respectively), have any physical health, mental health, or drug problems (18.4 versus 21.0%, 6.5 versus 10.0%, and 3.6 versus 6.0%, respectively), and be smokers (10.8 versus 31.0%).

Table 1. Descriptive variables by intention to breastfeed
Maternal factorsIntending to breastfeed n (%)aNot intending to breastfeed, n (%)a
Numberb 78 806 (85.3%)8747 (9.5%)
Maternal age, mean (SD)29.90 (5.44)28.14 (5.93)
Median family income in the neighbourhoodc in $CDN
Median (25 percentile,75 percentile)$72 983 ($55 359, $90 469)$66 206 ($50 193, $82 799)
Median proportion of labour force employed in the neighbourhoodc median (25 percentile,75 percentile)94.0% (91.0, 96.0)94.0% (91.0, 96.0)
Median proportion in labour force in the neighbourhoodc median (25 percentile,75 percentile)70.1% (62.8, 76.3)67.6% (60.0, 74.8)
Previous term baby
Yes41 725 (53.0%)5885 (67.4%)
No37 014 (47.0%)2849 (32.6%)
Previous preterm baby
Yes3321 (4.2%)490 (5.6%)
No75 270 (95.8%)8213 (94.4%)
Maternal physical health problem(s)
Yes14 129 (18.4%)1792 (21.0%)
No62 721 (81.6%)6741 (79.0%)
Maternal mental health problem(s)
Yes4987 (6.5%)853 (10.0%)
No71 863 (93.5%)7680 (90.0%)
Maternal drug problem(s)
Yes2746 (3.6%)509 (6.0%)
No74 104 (96.4%)8024 (94.0%)
Smoker   
Yes8232 (10.8%)2595 (31.0%)
No67 867 (89.2%)5770 (69.0%)
Antenatal factorsIntending to breastfeed, n (%)aNot intending to breastfeed, n (%)a
Obstetrical/pregnancy complication(s) d
Yes19 998 (25.4%)2278 (26.0%)
No58 808 (74.6%)6469 (74.0%)
Twin gestation
Yes525 (0.7%)70 (0.8%)
No78 281 (99.3%)8677 (99.2%)
Reproductive assistance
Yes1243 (1.8%)70 (0.9%)
No65 951 (98.2%)7607 (99.1%)
Antenatal care provider
Exclusive care by midwife6092 (7.9%)186 (2.2%)
Exclusive care by family physician9861 (12.7%)1285 (15.1%)
Exclusive care by nurse practitioner98 (0.1%)14 (0.2%)
No exclusive care provider (care provided by more than one type of care provider)17 285 (22.3%)2111 (24.7%)
Exclusive care by obstetrician44 038 (56.9%)4939 (57.9%)
Prenatal classes
Yes17 590 (25.5%)967 (12.4%)
No51 376 (74.5%)6842 (87.6%)
Hospital factorsIntending to breastfeed, n (%)aNot intending to breastfeed, n (%)a
  1. Numbers for variables do not add to the final total because of missing values.

  2. a

    Number (n) and % are shown, unless otherwise stated.

  3. b

    Numbers do not add up to 100% because 5.2% of the sample had no information on intention to breastfeed.

  4. c

    Median data for the neighbourhood are presented based on Statistics Canada data based on the postal code.

  5. d

    Yes/no to one or more of the following: gestational diabetes, hypertension, intrauterine growth restriction/small for gestational age, large for gestational age, periodontal infection, placental abruption, pre-eclampsia, premature rupture of membranes, preterm labour, preterm premature rupture of membranes, urinary tract infection.

  6. e

    Level 1, care of mothers with healthy pregnancies, and their infants, with births at ≥36 weeks of gestation; level 2, care of mothers with low-to-moderate risk pregnancies, and their infants, with births at ≥32 weeks of gestation; level 2+, care of mothers with moderate-risk pregnancies, and their infants, with births at 30–32 weeks of gestation; and level 3, subspecialty care for high-risk pregnancies, including mothers with severe medical complications, unwell/unstable newborns, multiple pregnancies of triplets and higher order multiples, and births at <32 weeks of gestation.

Hospital level e
111 719 (14.9%)1804 (20.6%)
235 952 (45.6%)3512 (40.2%)
2+24 076 (30.6%)2668 (30.5%)
37059 (9.0%)763 (8.7%)

Antenatal variables

Women who were intending to breastfeed were similar to women who were not intending to breastfeed in terms of the likelihood of having a twin gestation and receiving exclusive antenatal care from a nurse practitioner. Women intending to breastfeed were less likely to have pregnancy complications (25.4 versus 26.0%), and have antenatal care provided exclusively by a family physician or receive antenatal care by more than one type of care provider (12.7 versus 15.1%, and 22.3 versus 24.7%, respectively). Conversely, women intending to breastfeed were more likely to have received reproductive assistance (1.8 versus 0.9%), have attended prenatal classes (25.5 versus 12.4%), and have antenatal care provided exclusively by a midwife (7.9 versus 2.2%).

Hospital variables

The level of the hospital where women delivered their baby was also associated with a woman's intention to breastfeed. Women intending to breastfeed were more likely to give birth in a level–2 hospital than women not intending to breastfeed (45.6 versus 40.2%); however, they were less likely to give birth in a level–1 hospital (14.9 versus 20.6%).

Multivariable logistic regression

Maternal variables

The odds ratios for women's intentions to breastfeed, adjusted for all maternal, antenatal, and hospital factors, are shown in Table 2. Older women were more likely to intend to breastfeed (adjusted OR 1.54, 95% CI 1.46–1.61, for 10–year increments). Median family income in the neighbourhood was not associated with a woman's intention to breastfeed (adjusted OR 0.98, 95% CI 0.96–1.01, for $CDN20 000 increments), although living in a neighbourhood with a higher median employment rate was associated with a woman's intention to breastfeed (adjusted OR 1.01, 95% CI 1.00–1.01). The odds of intending to breastfeed were decreased with each additional previous term or preterm baby (adjusted OR 0.79, 95% CI 0.78–0.81, and adjusted OR 0.87, 95% CI 0.82–0.93, respectively). Women were more likely to intend to breastfeed their babies if they did not have physical health problems (adjusted OR 1.13, 95% CI 1.07–1.20), mental health problems (adjusted OR 1.19, 95% CI 1.09–1.29), or drug problems (adjusted OR 1.14, 95% CI 1.03–1.27). Non-smokers also were more than twice as likely to intend to breastfeed their babies as smokers (adjusted OR 2.31, 95% CI 2.18–2.44).

Table 2. Multivariable logistic regression of intention to breastfeed
Maternal factorsOdds ratio(95% CI) P Adjusted odds ratio (95% CI)a P
Maternal age (10–year increments)1.76 (1.69–1.83)<0.00011.54 (1.46–1.61)<0.0001
Median family income in the neighbourhoodb ($CDN20 000 increments)1.22 (1.20–1.24)<0.00010.98 (0.96–1.01)0.2043
Median proportion of labour force employed in the neighbourhood b 1.02 (1.02–1.03)<0.00011.01 (1.00–1.01)0.0093
Median proportion in labour force in the neighbourhood b 1.02 (1.02–1.02)<0.00011.01 (1.00–1.01)<0.0001
Previous term baby 0.79 (0.78–0.81)<0.00010.79 (0.78–0.81)<0.0001
Previous preterm baby 0.81 (0.77–0.87)<0.00010.87 (0.82–0.93)<0.0001
No maternal physical health problem(s) 1.19 (1.12–1.25)<0.00011.13 (1.07–1.20)<0.0001
No maternal mental health problem(s) 1.59 (1.47–1.71)<0.00011.19 (1.09–1.29)<0.0001
No maternal drug problem(s) 1.74 (1.58–1.91)<0.00011.14 (1.03–1.27)0.0118
Non-smoker 3.47 (3.30–3.65)<0.00012.31 (2.18–2.44)<0.0001
Antenatal factorsOdds ratio (95% CI)PAdjusted odds ratio (95% CI)a P
No obstetrical/pregnancy complication(s) c 1.05 (1.00–1.11)0.04331.05 (1.00–1.11)0.0558
Twin gestation 0.80 (0.63–1.01)0.06230.73 (0.57–0.94)0.0160
No reproductive assistance 0.52 (0.41–0.66)<0.00010.81 (0.63–1.03)0.0856
Antenatal care provider
Exclusive care by midwife3.54 (3.06–4.11)<0.00013.64 (3.13–4.23)<0.0001
Exclusive care by family doctor0.86 (0.81–0.92)<0.00011.19 (1.11–1.27)<0.0001
Exclusive care by nurse practitioner0.76 (0.55–1.05)0.09831.10 (0.77–1.57)0.5987
No exclusive care provider (care provided by more than one type of care provider)0.92 (0.87–0.97)0.00211.10 (1.04–1.16)0.0015
Exclusive care by obstetricianReferenceReference
No prenatal classes 0.47 (0.44–0.50)<0.00010.58 (0.54–0.62)<0.0001
Hospital factorsOdds ratio (95% CI)PAdjusted odds ratio (95% CI)a P
  1. a

    Adjusted for all variables in Tables 1 and 2 (maternal, antenatal, and hospital factors).

  2. b

    Median employment rate as a percentage of those in the labour force for the neighbourhood are presented based on Statistics Canada data based on the postal code.

  3. c

    Yes/no to one or more of the following: gestational diabetes, hypertension, intrauterine growth restriction (IUGR)/small for gestational age (SGA), large for gestational age (LGA), periodontal infection, placental abruption, pre-eclampsia, premature rupture of membranes, preterm labour, preterm premature rupture of membranes, and urinary tract infection (UTI).

  4. d

    Level 1, care of mothers with healthy pregnancies, and their infants, with births at ≥36 weeks of gestation; level 2, care of mothers with low-to-moderate risk pregnancies, and their infants, with births at ≥32 weeks of gestation; level 2+, care of mothers with moderate-risk pregnancies, and their infants, with births at 30–32 weeks of gestation; and level 3, subspecialty care for high-risk pregnancies, including mothers with severe medical complications, unwell/unstable newborns, multiple pregnancies of triplets and higher order multiples, and births at <32 weeks of gestation.

Hospital level d
10.74 (0.67–0.80)<0.00010.85 (0.77–0.93)0.0004
21.13 (1.04–1.22)0.00321.01 (0.93–1.10)0.8577
2+1.00 (0.92–1.09)0.97400.94 (0.86–1.03)0.1730
3ReferenceReference

Antenatal variables

Women who were pregnant with twins were less likely to intend to breastfeed (adjusted OR 0.73, 95% CI 0.57–0.94), as were women who did not attend prenatal classes (adjusted OR 0.58, 95% CI 0.54–0.62). Women receiving antenatal care exclusively by midwives had approximately three and a half times the odds of intending to breastfeed their babies than women receiving care exclusively by obstetricians (adjusted OR 3.64, 95% CI 3.13–4.23). Women whose antenatal care was provided exclusively by family doctors and those who were not receiving care exclusively by one type of care provider were also more likely to intend to breastfeed, compared with women cared for exclusively by obstetricians (adjusted OR 1.19, 95% CI 1.11–1.27, and adjusted OR 1.10, 95% CI 1.04–1.16, respectively). Having any obstetrical/pregnancy complications or receiving reproductive assistance were not significantly associated with intention to breastfeed.

Hospital variables

Women delivering in a level–1 hospital were less likely to intend to breastfeed compared with women delivering in a level–3 hospital (adjusted OR 0.85, 95% CI 0.77–0.93). No significant difference in intention to breastfeed was found between women who delivered in a level–2 or -2+hospital, compared with those who delivered in a level–3 hospital.

The sensitivity analysis comparing results with and without multiple imputation revealed no significant differences (results not shown). The AUC was calculated to be 0.71 (95% CI 0.70–0.71), suggesting a fair goodness of fit for the multivariable logistic regression model.

Discussion

In this population-based study, we found that ~85% of pregnant women intended to breastfeed their babies. This intention was shown to be associated not only with various maternal factors, such as age, smoking, and various health problems, but also with antenatal, healthcare provider, and hospital factors.

Previous research on intention to breastfeed is mainly limited to cross-sectional studies (i.e. surveys) and qualitative studies, which are likely to be affected by selection, volunteer, and recall bias. Although a recent Canadian survey of 6421 new mothers found that 90% of them reported that they had intended to breastfeed their babies prior to giving birth, the survey was conducted at 5–14 months postpartum, and hence was subject to both recall and volunteer bias.[29] Differences in antenatal intentions to breastfeed also exist between other countries, with large non-population based studies from the UK, Scotland, and Hong Kong reporting considerably lower rates (55, 76, and 54%, respectively).[30-32]

Existing literature on intention to breastfeed has identified a number of associated factors, such as women's knowledge about the benefits of breastfeeding and their attitude about breastfeeding in public.[33] Our study complements the existing knowledge on intention to breastfeed by confirming some of the other established associations, such as older age,[31, 32, 34, 35] primiparity,[31, 32, 34, 36] primigravidity,[37] and not smoking,[31, 37] and more importantly by exploring a number of novel variables, for example, maternal physical health problems, mental health problems, and drug problems, pregnancy complications, use of reproductive assistance, antenatal care provider type, attendance at prenatal classes, and hospital level. Detection of these factors antenatally could prompt healthcare providers to identify women who may benefit from more breastfeeding education and support.

We found that a higher median employment rate in the neighbourhood, as well as a higher median proportion in the labour force in the neighbourhood, were significantly associated with intention to breastfeed, even when adjusted for all other factors. Conversely, median family income in the neighbourhood was not significantly associated with intention to breastfeed. Various other studies have examined the role of family income; however, whether it is significantly associated with intention to breastfeed,[29, 31] or not,[32, 37, 38] remains inconclusive. Although we did not have the data to analyse these variables on an individual level, we were able to analyse them on a neighbourhood level. Interestingly enough, studies have shown that social factors may have a larger impact on health outcomes than factors on an individual level.[39, 40]

Our exploration of non-patient factors influencing intention to breastfeed highlights important variables that had not previously been reported. Compared with women who chose to have care provided solely by obstetricians, women who chose to have care provided by any other health professional were more likely to intend to breastfeed, with the most pronounced difference between obstetric and midwifery patients (midwifery patients had three and a half times the odds of intending to breastfeed). Although this difference may in part be explained by the characteristics of the patients who seek midwifery care in Canada, where it is a relatively newly recognised field, unlike in the UK, breastfeeding promotion and support by the healthcare provider is also likely to play an important role. The findings of an Iowa study support this idea, with 64% of nurse midwives, 13% of family practitioners, and 7% of obstetricians stating that they were strong advocates of breastfeeding.[41] The important role of the healthcare provider has been highlighted by other studies, which found that counselling from care providers about the benefits of breastfeeding can influence women to change their decision from formula feeding to breastfeeding.[42, 43]

Hospital variables were also assessed because women would have been exposed to the hospital and staff prenatally when coming in to preregister for their delivery. Delivering in a level–1 hospital, which is designed to care for healthy infants and mothers, ironically was associated with a lower intention to breastfeed. Although some of these hospitals are smaller and may have fewer resources to support breastfeeding, assessing and improving current hospital practices may help to promote and strengthen the decision to breastfeed.

The strengths of this study include the population-based design, with data from Ontario, the most populous province in Canada, with more than one-third of the country's births. In addition, the strengths of the database include real-time data entry and information on a broad range of variables. Because of the large sample size we were able to examine rare factors, such as being pregnant with twins, subcategories of factors, such as non-exclusive care by healthcare providers, and factors not previously reported, including healthcare provider and hospital level.

The limitations of this study include the fact that we did not have information about certain factors, such as family support, marital status, or ethnicity, which might influence a woman's intention to breastfeed. Other variables, such as body mass index, had incomplete information, and were not included in the analysis. Indeed, the measure of goodness of fit for the multivariable logistic regression model was fair (AUC 0.71, 95% CI 0.70–0.71), suggesting that other unmeasured variables account for part of the intention to breastfeed. Also, we are unable to generalise our results to certain subpopulations (i.e. planned home births), as they were excluded from the study. Furthermore, we do not have validation data on our primary outcome (intention to breastfeed) for the BORN database; however, if non-differential misclassification of the outcome did occur, we would expect our reported results to be attenuated and closer to the null than the actual ORs. Although 5% of women did not have information about their antenatal intentions to breastfeed, a further comparison of baseline characteristics revealed that women with this data were similar to those with missing data. We also lacked information on the degree of influence of the baby formula industry, which generates a $CDN5–6 billion profit annually.[44] Early supplementation with baby formula and the distribution of formula samples have been found to significantly reduce the duration of breastfeeding among women, even if they had prenatal intentions to breastfeed.[45]

Our findings suggest that along with the women themselves, healthcare providers and hospitals play an important role in a woman's decision to breastfeed, and can therefore be targeted in an effort to increase the prevalence of positive intention to breastfeed. Further research is also required to develop new tools and strategies that promote breastfeeding at all levels, and ensure increased hospital participation in the Baby Friendly Hospital Initiative,[46] along with more rigorous enforcement of the WHO International Code for Marketing of Breast-Milk Substitutes,[47] which prohibits the promotion and distribution of formula in healthcare facilities.

Conclusion

In summary, the results of this population-based study reveal that 85% of women intended to breastfeed their babies, which was associated with various maternal, antenatal, healthcare provider, and hospital factors. Given that intention to breastfeed is strongly associated with the initiation of actual breastfeeding, the identified factors can be used in the development of breastfeeding promotion strategies.

Disclosure of interests

The authors do not have any personal, financial, or other conflicts of interest to declare.

Contribution to authorship

OL, CV, LG, BC, and CF were involved in the conception and design of the study, drafting of the article, critical revision of the article for important intellectual content, and final approval of the version to be published. SDM, EP, and GF were involved in the statistical analysis and interpretation of data, as well as the conception and design of the study, drafting of the article, critical revision of the article for important intellectual content, and final approval of the version to be published.

Details of ethics approval

The study was approved by the Faculty of Health Sciences/McMaster University Research Ethics Board on 3 September 2009 (approval ref. no. 08045C).

Funding

This work was supported by a Canadian Institute of Health Research (CIHR) operating grant (#MCH97589), and Sarah D McDonald is supported by a CIHR New Investigator Award (#CNl95357). The authors do not have financial relationships with any organisations that might have an interest in the submitted work in the previous 3 years. CIHR had no role in the study design, data analysis, interpretation of the data, writing of the report, or decision to submit the article. All researchers are independent of CIHR.

Commentary on ‘Women's intentions to breastfeed: a population-based cohort study’

Breastfeeding is a major predictor of health outcomes for both mothers and infants; however, not all women decide to breastfeed.

In this issue, Olha Lutsiv and colleagues report a population-based study of breastfeeding intention in Ontario, Canada. Women were asked their feeding plans in the third trimester or upon admission to the labour ward. The authors found a variety of factors – including neighborhood employment, birth in a higher-acuity hospital, and antenatal care provider – were associated with breastfeeding intention.

Exclusive care by a midwife was the strongest predictor of breastfeeding. Importantly, most women decide whether to breastfeed prior to pregnancy, so it difficult to determine to what extent midwifery care reflects the pre-existing preferences of women planning to breastfeed. Nevertheless, obstetricians underestimate the impact of their advice on infant feeding (Taveras EM et al. Pediatrics 2004;113:e405–e411), and these results suggest that obstetricians might glean effective strategies for breastfeeding support from their midwifery colleagues.

Several health factors were associated with lower odds of breastfeeding intention, including drug problems, tobacco use, and maternal mental health problems, probably reflecting both maternal desire and provider advice. Although active substance abuse is a contraindication to breastfeeding, for mothers who smoke, observational data suggest improved respiratory outcomes among breastfed versus bottle-fed infants (Woodward A et al. Journal of Epidemiology & Community Health 1990;44:224–230). These data suggest that providers should encourage mothers to breastfeed, even if they are unable to stop smoking. Likewise, women with mental health problems need accurate information to balance the risks of breastfeeding while taking medication versus the risks of formula feeding.

Importantly, pregnancy intention was not included in the current analysis. Unplanned or mistimed pregnancy is associated with smoking, maternal depression, and markedly lower rates of breastfeeding initiation (Cheng D et al. Contraception 2009;79:194–198). Thus mental health problems and substance use may in fact be proxies for unplanned pregnancy.

Consistent with other studies, multiparous women were less likely to intend to breastfeed. Parous women who have struggled with breastfeeding in the past may decide to formula feed. Better support – and compassionate, respectful care for women who are unable to achieve their breastfeeding goals – may enable parous women who have encountered obstacles with prior births to consider breastfeeding again in the future.

Lutsiv's analysis suggests important avenues for future investigation. In particular, their model and the underlying data allow them to identify hospitals, providers, and neighborhoods with unexpected breastfeeding success. Focusing on such ‘positive deviants’ may identify novel local strategies that enable more women to initiate breastfeeding (Bradley EH et al. Implement Sci 2009;4:25).

Future questions could include: how do midwives in Ontario provide breastfeeding guidance, and how might such practices be incorporated into other models of care? What healthcare provider practices and policies foster breastfeeding among women who smoke, or who have mental health problems? What strategies increase the proportion of pregnancies that are planned? What type of support has enabled women with prior breastfeeding difficulties to consider trying again in a subsequent pregnancy?

Identifying and implementing strategies that work in the context of local cultural norms will ultimately enable more women to initiate breastfeeding, thereby improving health outcomes for mother and child.

  • AM Stuebe

  • University of North Carolina School of Medicine, Chapel Hill, NC, USA

Ancillary