Predictive Models for Characterizing Disparities in Exclusive Breastfeeding Performance in a Multi-ethnic Population in the US

Abstract

Objectives

Maternal lactation performance varies across populations, yet the relative impact of maternal sociodemographics, perinatal factors, and birth outcomes on disparities in exclusive breastfeeding (XBR) outcomes is not well known. We aimed to develop predictive models and compare the relative contribution of predictors for XBR initiation and XBR ≥ 6 months.

Methods

Infant feeding data were obtained from women with children aged 0–6 years (n = 1471) in a multi-ethnic cross-sectional study in the US (2011–2012). We compared discriminant ability of predictors for ever XBR and XBR ≥ 6 months using discriminant function analysis, respectively. We also calculated adjusted ORs for factors associated with XBR outcomes and breast-bottle feeding (BrBot) subgroups.

Results

Maternal sociodemographics (education level, marital status, nativity, and age at childbirth) had greater discriminating abilities in predicting ever XBR and XBR ≥ 6 months than birth outcomes and perinatal factors. Foreign-born women were two-fold more likely to initiate XBR but not necessarily continue to 6 months compared to their US-born counterparts. Factors associated with BrBot subgroups differed from those associated with XBR outcomes, whereas maternal age was the only predictor consistently associated with ever XBR, XBR ≥ 6 months, and BrBot subgroups. The areas under the receiver operating characteristic curves for models predicting ever XBR and XBR ≥ 6 months were 0.88 (95 % CI 0.85, 0.91) and 0.90 (95 % CI 0.88, 0.93), respectively.

Conclusions

Findings underscore the importance of educational, clinical, and social support to promote XBR in mothers with sociodemographic factors predictive of none or poor XBR outcomes.

Introduction

Breastfeeding has both short-term and long-term health benefits for mothers and infants [1, 2]. Both the WHO and the American Academy of Pediatrics (AAP) recommend exclusive breastfeeding (XBR) for 6 months with introduction of complementary foods and continued breastfeeding thereafter [3, 4]. Despite the rising rates of breastfeeding in the US [5], the latest reported rates of XBR for 6 months (18.8 %) among 2011 births [6] are lower than the Healthy People 2020 Target (25.5 %) [7]. However, it is worth noting that the reported overall rates obscure sociodemographic including racial/ethnic variation in breastfeeding practices [8-10]. Therefore, it is important for healthcare and public health professionals to understand these variations in relation to women’s XBR initiation and duration.

In addition, the trend of mothers returning to work shortly after childbirth is on the rise in the past decades [11]. Cumulatively, 44.2 % of new mothers returned to work by 3 months after childbirth in 2005–2007, compared to 32.9 % in 1981–1984 [12]. Consequently, some mothers who initiate XBR may have to supplement breast milk with formula or completely switch to formula feeding. Studies investigating factors associated with breast and bottle feeding (BrBot) are warranted to help health practitioners identify women at risk of early cessation of XBR.

Despite a rich body of literature on the factors associated with breastfeeding initiation and duration [13-17], determinants of XBR and BrBot outcomes have been less studied. Moreover, comparisons across previous studies are limited by inconsistent or unclear definitions of XBR, residual confounding, and varied sociocultural context. For instance, among three studies assessing factors associated with XBR using nationally representative survey data in the US [18-20], one failed to apply the WHO recommendation of XBR for 6 months [18]; all three collected breastfeeding data from survey respondents rather than exclusively from biological mothers; and none adjusted for maternal reproductive factors [i.e., prepregnancy body mass index (BMI), gestational weight gain, and pregnancy complications], which are associated with XBR [16, 21]. Additionally, despite factors identified to be associated with XBR, there is a paucity of data on the relative extent to which these factors contribute to XBR initiation and XBR ≥ 6 months.

Therefore, the objectives of this study were to (1) develop predictive models and compare the relative discriminant ability of predictors for XBR since birth and XBR ≥ 6 months; and (2) explore factors associated with BrBot subgroups classified by the relative frequency of each feeding in a multi-ethnic population of mother–child dyads in the US.

Methods

Study Design and Population

Data were collected during the National Children’s Study Formative Research in Anthropometry, a cross-sectional study across eight study sites in the US (2011–2012), whose names and locations can be found in the acknowledgements. A convenience sample of mother-offspring dyads of various ethnicities (non-Hispanic White, Hispanic, non-Hispanic Black, and other including Chinese, South East Indian, Filipino, Pacific Islander, etc.) were recruited at daycare centers, churches, clinics, and community centers by study coordinators, word of mouth, and referral as described previously [22, 23]. Eligibility criteria included: mothers aged 18–49 years and non-institutionalized; and offspring who were aged 0–5.9 years, healthy, of the same ethnicity as and living with the mother, and had not suffered from any illness associated with weight loss within the past week. Among 1634 eligible dyads, 1538 had complete infant feeding data. If more than one infant/child of the same mother was recruited, the youngest one was included in this analysis to reduce the cluster effect due to the same mother (n = 1471). The study was approved by the Institutional Review Boards at the Eunice Kennedy Shriver National Institute of Child Health and Human Development and each study center. Informed consent was obtained from the mothers.

Infant Feeding Practices

Infant feeding practices were reported by mothers using an interviewer-administered questionnaire with modified questions based on the Nurses’ Health Study (NHS) Mother’s questionnaire, a retrospective cohort study of the mothers of the NHS participants. Specifically, according to questions about (1) whether the child was ever breastfed/formula-fed; (2) if ever breastfed, the age when the child was first fed something other than breast milk on a daily basis; and (3) if ever fed breast- and formula-fed, whether the child was fed “formula only, formula and breast milk with equal frequency, formula more frequently than breast milk, or breast milk more frequently than formula”, infant feeding practices were classified as: (1) XBR (by the WHO definition) [24]: if the mother reported she had only fed the child breast milk without anything else except for drops and syrups containing vitamins, minerals, and medicines; (2) BrBot: if the mother reported breastfeeding and formula-feeding at the same time; or (3) exclusive bottle-feeding (XBot): if the mother reported she had fed her child formula (or any other non-human milk liquid) from a bottle but never fed breast milk. The BrBot group was further stratified into three subgroups based on the question about the relative frequency of breast milk versus formula fed to the child: breast milk >, = , or < formula. Duration of XBR was defined as the age before introduction of anything else other than breast milk if the child was exclusively breastfed since birth.

Covariates

A priori selected potential determinants of XBR outcomes were obtained via interviewer-administered questionnaire and included: (1) maternal characteristics: ethnicity, education level, marital status, nativity, employment status at the time of the interview, age at the index childbirth, prepregnancy BMI [self-reported prepregnancy weight (kg)/height(m)²], gestational weight gain, smoking during pregnancy, and pregnancy complications (gestational diabetes, gestational hypertension, pre-eclampsia/eclampsia, and only proteinuria); (2) child characteristics: age, sex, birth year, birthweight, gestational age, and birth order; and (3) household number.

Statistical Analysis

Data preprocessing approaches were reported elsewhere [25]. Descriptive statistics of continuous and categorical variables were presented as the mean (SD) or frequency, respectively. Bivariate associations between infant feeding practices (XBR, BrBot, and XBot) and maternal and child factors were assessed by ANOVA for continuous variables or the χ² test for categorical variables.

Discriminant analysis for ordinal responses was used to develop predictive models discriminating mothers who initiated XBR since birth or maintained XBR for 6 months versus those initiating XBot since birth, respectively. Factors associated with infant feeding practices in the bivariate analyses meeting the selection criteria (P < 0.25) [26] were included in the initial multivariate discriminant model for XBR since birth, along with child’s birth year (range 2005–2012) based on the secular trend of increasing XBR rates over time [5]. Stepwise selection using entry (P = 0.10) and removal (P = 0.05) criteria were used to identify variables that contributed significantly to the final discriminant function for XBR since birth. These factors were also used for model development for XBR ≥ 6 months. Of note, compared to logistic regression, discriminant analysis provides a unique approach to assess the relative discriminant ability of predictors by comparing the size of standardized canonical discriminant function coefficientis [27]. In addition, these standardized coefficients and their 95 % CIs were derived based on 1000-fold bootstrap samples (random sampling with replacement for 1000 times), which can improve model robustness against sensitivity to the assumption of multivariate normality [28].

To assess the independent associations of predictors included in the final discriminant functions with XBR since birth and XBR duration C 6 months, mixed-effects bivariate logistic regression with study center as a random effect was used to estimate the unadjusted and adjusted odds ratio (OR) and 95 % confidence interval (95 % CI) of each endpoint. Likewise, mixed-effects multivariate logistic regression estimated the unadjusted and adjusted OR (95 % CI) for BrBot subgroups (i.e., breast milk > , = , < formula) with “breast milk < formula” as the comparison group. Only variables reaching P < 0.25 [26] in univariate analysis were included in the multivariate logistic regression model. Because maternal reproductive factors and birth outcomes at birth are often correlated, variance inflation factors (VIFs) were calculated to assess the multicollinearity of covariates. Results indicated no concerns about multicollinearity with VIFs all below the recommended cut-off point of 2.5 [29]. The area under the curve (AUC) statistic of the receiver operating characteristic curve was calculated to quantify the ability of the multivariate logistic models to discriminate mothers who exclusively breastfed the child since birth or for C6 months from those who exclusively bottle-fed since birth. All analyses were conducted with IBM SPSS 21 (IBM Corp, Armonk, New York). Statistical significance was set at a 2-tailed P < 0.05.

Results

Among the 1471 children included in this analysis, 256 (17.4 %) were exclusively breastfed since birth, 871 (59.2 %) were mixed fed with both breast milk and formula, and the remainder (23.4 %) were exclusively bottle-fed (Table 1). Of the 1127 (76.6 % of 1471) mothers who ever breastfed, the rates of XBR ≥ 6 months were 15.3 %. The variable reduction procedure via stepwise selection resulted in a final predictive model of ten variables for XBR since birth. These variables were ordered by the size of standardized canonical discriminant coefficients (Table 2). Maternal education had the greatest discriminating ability while prepregnancy BMI had the least impact on predicting mothers in the XBR group versus the XBot group. Similarly, the same variables except birth year remained statistically significant in the final model discriminating mothers who had XBR ≥ 6 months from those who initiated XBot since birth.

Table 1.

Characteristics of the mother-offspring dyads by infant feeding practices

	All (n = 1471)	XBR (n = 256)	BrBot (n = 871)	XBot (n = 344)	P valuea
Maternal characteristics
Ethnicity (%)					<0.001
Non-Hispanic White	20.1	40.6	18.0	9.9
Hispanic	45.1	32.4	49.7	43.0
Non-Hispanic Black	26.0	11.0	23.3	44.2
Other	8.8	16.0	9.0	2.9
Education (%)					<0.001
<High school	19.6	11.7	18.4	28.9
High school graduate or equivalent	19.2	8.6	17.9	30.3
>High school	61.0	79.7	63.7	40.8
Employed (%)	47.5	43.9	50.5	43.1	0.036
Marital status (single mother) (%)	29.6	11.3	27.2	50.0	<0.001
Nativity (US born) (%)	74.2	70.8	70.8	87.5	<0.001
Age at childbirth (year), mean (SD)	28.4 (6.2)	29.9 (5.3)	28.6 (6.2)	26.6 (6.4)	<0.001
Prepregnancy BMI (kg/m2) (%)					<0.001
Underweight (<18.5)	4.4	6.7	3.6	4.8
Normal (18.5–24.9)	47.0	56.3	48.6	39.8
Overweight (25.0–29.9)	20.9	18.7	22.0	21.4
Obese (≥30.0)	25.9	18.3	25.8	34.0
Weight gain during pregnancy (%)					<0.001
<14 lbs	15.6	9.5	15.2	21.9
15–19 lbs	10.5	9.1	11.4	9.8
20–29 lbs	28.0	27.7	29.2	26.9
30–40 lbs	28.9	38.7	27.8	26.0
>40 lbs	15.7	15.0	16.4	15.4
Pregnancy complications (≥1) (%)	24.9	20.7	23.7	31.8	0.003
Smoking during pregnancy (%)	5.8	1.6	4.7	12.0	<0.001
Primiparous (%)	37.7	39.5	39.0	35.1	0.27
Mother–child dyad characteristics
Household number, mean (SD)	4.5 (1.6)	4.5 (1.6)	4.4 (1.5)	4.6 (1.8)	0.41
Child characteristics
Age (year), mean (SD)	2.0 (1.7)	1.8 (1.7)	2.0 (1.7)	2.1 (1.7)	0.055
Males (%)	51.8	52.0	51.2	53.2	0.76
Birthweight (kg), mean (SD)	3.2 (0.6)	3.4 (0.6)	3.2 (0.6)	3.1 (0.6)	<0.001
Preterm birth (<37 week) (%)	10.9	9.0	10.3	14.0	0.059

XBR exclusive breast feeding, BrBot breast-bottle feeding, XBot exclusive bottle-feeding

^aANOVA for continuous variables and χ² tests for categorical variables

Table 2.

Standardized canonical discriminant function coefficients of factors predicting ever XBR since birth and XBR duration ≥6 months

	Ever XBR (n = 256)a Coefficient (95 % CI)	XBR ≥ 6 months (n = 172)a Coefficient (95 % CI)
Maternal education (>high school)	0.50 (0.35, 0.63)	0.49 (0.35, 0.62)
Marital status (single mother)	0.42 (0.29, 0.55)	0.38 (0.24, 0.50)
Maternal nativity (foreign born)	0.28 (0.11, 0.42)	0.24 (0.05, 0.42)
Maternal age at childbirth (year)	0.27 (0.13, 0.41)	0.30 (0.16, 0.44)
Child’s birthweight (g)	0.25 (0.10, 0.39)	0.29 (0.16, 0.42)
Child’s birth year (year)	0.24 (0.10, 0.38)	0.004 (−0.13, 0.16)
Pregnancy complications (none)	0.21 (0.05, 0.35)	0.19 (0.03, 0.35)
Maternal ethnicity (non-Hispanic)	0.18 (0.03, 0.33)	0.26 (0.11, 0.40)
Smoking during pregnancy (no)	0.17 (0.04, 0.29)	0.14 (0.01, 0.26)
Maternal prepregnancy BMI (kg/m2)	−0.16 (−0.32, −0.01)	−0.26 (−0.41, −0.11)

XBR exclusive breast feeding

^aThe comparison group was exclusively bottle-feeding (n = 344)

Logistic regression analyses revealed associations of individual variables with breastfeeding outcomes (i.e., ever XBR since birth, XBR duration C 6 months, and BrBot subgroups). Non-Hispanic Black (NHB) mothers were least likely to initiate XBR since birth or prolong it to 6 or more months in either unadjusted or adjusted analyses compared to their NHW counterparts (Table 3). In unadjusted analyses, the likelihoods of initiating XBR since birth were higher among mothers who received education beyond high school, who were married or living with a partner, born in foreign countries, older at childbirth, and whose index child was born more recently (range 2005–2012); whereas the likelihoods were lower among those who were obese before pregnancy, had more than one pregnancy complication, smoked in pregnancy, and had an index child born <2500 g. These factors were also independently associated with ever XBR in the adjusted model, except for maternal prepregnancy obesity and pregnancy complications. Combining overweight and obese women into one group did not appreciably alter the results.

Table 3.

Unadjusted and adjusted ORs (95 % CIs) of factors associated with ever XBR since birth and XBR duration^a

	Ever XBR (n = 256)a		XBR ≥ 6 months (n = 172)a
	Unadjusted OR (95 % CI)	Adjusted OR (95 % CI)b	Unadjusted OR (95 % CI)	Adjusted OR (95 % CI)b
Maternal ethnicity
Non-Hispanic White	1.00	1.00	1.00	1.00
Hispanic	0.19 (0.10, 0.34)	0.27 (0.13, 0.56)	0.14 (0.07, 0.26)	0.28 (0.12, 0.63)
Non-Hispanic Black	0.07 (0.04, 0.13)	0.12 (0.06, 0.25)	0.05 (0.02, 0.10)	0.11 (0.05, 0.25)
Other	1.10 (0.44, 2.79)	0.77 (0.24, 2.45)	0.88 (0.32, 2.37)	0.85 (0.25, 2.88)
Maternal education
<High school	1.00	1.00	1.00	1.00
=High school	0.75 (0.39, 1.42)	0.89 (0.42, 1.87)	0.61 (0.24, 1.48)	0.63 (0.22, 1.77)
>High school	4.37 (2.64, 7.24)	2.99 (1.54, 5.80)	5.76 (3.02, 11.00)	3.41 (1.48, 7.83)
Married or living with a partner	6.09 (3.82, 9.72)	2.38 (1.36, 4.15)	6.17 (3.54, 10.74)	2.22 (1.14, 4.33)
Foreign born	2.54 (1.61, 4.03)	2.44 (1.29, 4.59)	2.15 (1.25, 3.69)	2.01 (0.91, 4.44)
Maternal age at childbirth (year)	1.07 (1.04, 1.11)	1.05 (1.01, 1.10)	1.08 (1.04, 1.12)	1.07 (1.02, 1.12)
Maternal prepregnancy BMI (kg/m2)
Underweight (<18.5)	0.94 (0.43, 2.04)	1.15 (0.42, 3.20)	0.88 (0.36, 2.12)	1.02 (0.32, 3.230)
Normal (18.5–24.9)	1.00	1.00	1.00	1.00
Overweight (25.0–29.9)	0.66 (0.41, 1.05)	0.59 (0.33, 1.07)	0.70 (0.41, 1.18)	0.59 (0.31, 1.14)
Obese (≥30.0)	0.44 (0.28, 0.69)	0.66 (0.37, 1.19)	0.33 (0.19, 0.58)	0.50 (0.24, 1.04)
Pregnancy complications (≥1)	0.65 (0.43, 0.97)	0.63 (0.37, 1.08)	0.56 (0.34, 0.90)	0.64 (0.34, 1.21)
Smoking during pregnancy	0.13 (0.04, 0.38)	0.16 (0.04, 0.55)	0.14 (0.04, 0.49)	0.23 (0.06, 0.89)
Child’s birthweight (g)
Low (<2500)	0.30 (0.15, 0.60)	0.44 (0.20, 0.98)	0.25 (0.11, 0.60)	0.30 (0.11, 0.83)
Normal (2500–3999)	1.00	1.00	1.00	1.00
High (≥4000)	1.49 (0.79, 2.80)	1.28 (0.57, 2.84)	1.70 (0.85, 3.42)	1.49 (0.60, 3.75)
Child’s birth yearc	1.19 (1.06, 1.32)	1.33 (1.16, 1.53)	0.94 (0.83, 1.06)	1.10 (0.93, 1.30)

XBR exclusive breastfeeding

^aThe comparison group was exclusively bottle-feeding (n = 344)

^bAdjusted for all other variables in the table

^cRange: 2005–2012

Overall, the ORs of factors associated with XBR ≥ 6 months were of similar magnitude and statistical significance as those for ever XBR, whereas the significant associations of maternal nativity, prepregnancy BMI, and pregnancy complications with XBR ≥ 6 months in the crude analysis did not persist in the adjusted model. However, after combining overweight and obese groups, the adjusted OR of XBR ≥ 6 months was 0.55 (95 % CI 0.31, 0.97; data not shown) for overweight/obese women compared to their normal-weight counterparts. The AUC statistics of the receiver operating characteristic curves for initiating XBR since birth and XBR ≥ 6 months based on the multivariate logistic regression models were 0.88 (95 % CI 0.85, 0.91) and 0.90 (95 % CI 0.88, 0.93), respectively (Fig. 1).

Fig. 1.

Receiver operating characteristic curves of the multivariate mixed-effects logistic regression models for a ever exclusive breastfeeding since birth and b exclusive breastfeeding ≥6 months. AUC area under the curve

Among 871 breast-bottle fed children, 60.4 % was fed breast milk less frequently than formula, while 191 (21.9 %) were fed breast milk more frequently than formula and 154 (17.7 %) were fed the two milks with equal frequency. Factors associated with “breast milk > formula” differed from those associated with “breast milk = formula” (Table 4). Mothers who were married or living with a partner were more likely to feed the child breast milk more frequently than formula rather than feeding the two milks with equal frequency. Maternal age at childbirth was positively associated with increased likelihoods of both “breast milk > formula” and “breast milk = formula”. Although mothers who were obese before pregnancy were 51 % less likely to feed the child breast milk more frequently than formula compared to their normal weight counterparts, the adjusted ORs of “breast milk = formula” did not vary by prepregnancy BMI. For children born more recently (range 2005–2012), their mothers were 27 % more likely to feed the two milks with equal frequency versus feeding formula more frequently. Of note, although maternal education, smoking during pregnancy, and child’s birthweight were significantly associated with ever XBR and XBR ≥ 6 months, these factors were not significant determinants of BrBot.

Table 4.

Unadjusted and adjusted ORs (95 % CIs) of factors associated with breast-bottle feeding subgroups^a

	Breast milk > formula (n = 191)a		Breast milk = formula (n = 154)a
	Unadjusted OR (95 % CI)	Adjusted OR (95 % CI)b	Unadjusted OR (95 % CI)	Adjusted OR (95 % CI)b
Maternal ethnicity
Non-Hispanic White	1.00	1.00	1.00	1.00
Hispanic	0.80 (0.51, 1.28)	1.00 (0.62, 1.61)	1.37 (0.78, 2.44)	1.47 (0.79, 2.74)
Non-Hispanic Black	0.48 (0.28, 0.83)	0.69 (0.39, 1.22)	1.16 (0.62, 2.18)	1.56 (0.80, 3.05)
Other	0.94 (0.47, 1.89)	1.04 (0.50, 2.18)	2.10 (0.96, 4.61)	1.94 (0.80, 4.70)
Married or living with a partner	2.31 (1.52, 3.51)	1.81 (1.15, 2.85)	1.81 (1.17, 2.79)	1.52 (0.95, 2.43)
Foreign born	1.04 (0.71, 1.52)	0.83 (0.53, 1.29)	1.55 (1.05, 2.28)	1.13 (0.71, 1.79)
Maternal age at childbirth (year)	1.06 (1.03, 1.09)	1.06 (1.03, 1.09)	1.06 (1.03, 1.10)	1.06 (1.03, 1.10)
Maternal prepregnancy BMI (kg/m2)
Underweight (<18.5)	1.21 (0.50, 2.96)	1.23 (0.49, 3.01)	2.08 (0.84, 5.17)	1.75 (0.67, 4.53)
Normal (18.5–24.9)	1.00	1.00	1.00	1.00
Overweight (25.0–29.9)	0.97 (0.63, 1.49)	0.89 (0.57, 1.40)	1.63 (1.04, 2.58)	1.49 (0.92, 2.41)
Obese (≥30.0)	0.56 (0.36, 0.87)	0.49 (0.31, 0.79)	1.00 (0.62, 1.56)	0.78 (0.47, 1.30)
Child’s birth yearc	1.10 (0.99, 1.23)	1.10 (1.00, 1.23)	1.25 (1.11, 1.42)	1.27 (1.12, 1.45)

^aThe comparison group was “breast milk < formula” (n = 526)

^bAdjusted for all other variables in the table

^cRange: 2005–2012

Discussion

The predictive models for ever XBR and XBR ≥ 6 months included maternal sociodemographic factors (education level, marital status, nativity, and age at childbirth) with greater discriminating abilities than birth outcomes (birthweight) and perinatal factors (complications and smoking during pregnancy and prepregnancy BMI). Amongst the predictors, women who are obese prior to pregnancy and pregnant smokers are those who can be targeted for strategies to promote breastfeeding from their first prenatal visit, whereas those who experience complications in pregnancy can be targeted upon diagnosis. Additionally, although most sociodemographic factors are difficult to modify, identification of these predictors provide potential targets for effective interventions. For instance, breastfeeding education and counseling provided to mothers with low educational attainments, clinical and familial support offered to single mothers, or a combination of both strategies might be used to promote XBR. Thus, the predictive models are empirically driven by variables with strong potential for public health interventions.

There could be several reasons why less educated, young, and single mothers were less likely to initiate or prolong XBR to 6 months. Previous data show 66.3 % of mothers with a bachelor’s degree or more enjoyed paid leave compared to 18.5 % of mothers with less than a high school education level [12]. Educated and older mothers were more likely to be aware of the WHO/AAP recommendation of XBR for 6 months and had greater intention to meet the recommendation [30]. Also, given that paternal knowledge of breastfeeding recommendations [31] and ability to help mothers prevent and manage lactation difficulties [32] are positively associated with breastfeeding initiation and duration, lack of such support may compromise lactation performance among single mothers. Indeed, most of these maternal sociodemographic characteristics are difficult to ameliorate; however, these factors might be indicators of the underlying psychosocial mechanisms of poor lactation performance. Future research investigating determinants of these related psychosocial factors is important to better our understanding in structuring effective interventions.

Our results revealed substantial ethnic disparities in infant feeding practices. In this study, NHB mothers were least likely to initiate XBR or prolong it to 6 months among all ethnicities. These findings are consistent with previous results of nationally representative surveys in the US [18, 19, 33]. In contrast with previous observations, Hispanic mothers in this study were less likely to ever exclusively breastfeed or extend XBR to 6 months compared to their NHW counterparts. This could be partly due to the heterogeneity in race and national origin of the Hispanic population. In particular, 86.4 % of the Hispanic sample in this study were recruited from one study center (San Antonio, TX), who were more likely to be homogeneous compared to the nationally representative samples. Thus, future breastfeeding promotion programs incorporating information on specific ethnic background are warranted to optimize intervention strategies.

As reported previously, high prepregnancy BMI was associated with poor lactation performance [34-39]. Likewise, our results showed prepregnancy obesity was associated with reduced likelihoods of ever XBR and XBR ≥ 6 months in unadjusted analyses. However, the association with ever XBR did not remain significant after adjusting for other covariates. Similarly, pregnancy complication was associated with XBR outcomes in the unadjusted but not adjusted model. Besides the potential for insufficient power, these differences between the crude and adjusted models could also be partly due to the intercorrelation between the prepregnancy BMI and pregnancy complications. Indeed, obese women have a higher risk of pregnancy complications such as gestational diabetes, gestational hypertension, and pre-eclampsia [40-42]. However, few studies have included pregnancy complications as potential confounders [16, 43, 44]. Kitsantas and Pawloski [44] reported an independent effect of prepregnancy overweight/obesity on breastfeeding initiation only among mothers with pregnancy complications but not among healthy ones. Collectively, these findings suggest pregnancy complication is an important covariate to consider when assessing the association between maternal prepregnancy BMI and breastfeeding/XBR outcomes.

In agreement with previous observations [18, 20, 45], our results also indicated that maternal smoking during pregnancy and child’s low birthweight (<2500 g) decreased the likelihood of ever XBR and XBR for 6 months, respectively. A review of maternal smoking and breastfeeding suggested that psychosocial factors rather than physiological ones are largely responsible for poor lactation performance in smoking women [46]. Women who perceive smoking as a barrier to breastfeeding [47] are less likely to seek help for breastfeeding problems [48]. Thus, efforts to address these challenges are needed in XBR promotion initiatives. As for child’s low birthweight, one major contributor is preterm birth [49]. The separation between mothers and their hospitalized infants in the neonatal intensive care unit could contribute to the difficulty in initiating XBR. Given the controlled environment of the neonatal intensive care unit, special clinical support is of upmost importance to promote XBR in this particular population.

Our study also revealed immigrant women were more likely to exclusively breastfeed the child since birth but not necessarily for 6 months compared to US-born women. Similarly, the 2007 National Survey of Children’s Health reported increased likelihood of ever breastfeeding but not 6-month XBR in foreign-born women relative to US-born women [20]. We speculate that whilst cultural traditions and beliefs inherent to the mother’s origin could influence the mother’s decision to initiate XBR, social support or the lack thereof from families and health professionals would influence the XBR duration. Future studies to identify barriers to prolonging XBR among women who initiate XBR in a culturally sensitive approach are warranted.

In addition, findings illustrate that factors associated with BrBot subgroups differed from those with XBR outcomes, whereas older maternal age at childbirth was the only factor consistently associated with ever XBR, XBR ≥ 6 months, and BrBot subgroups. Given that pediatricians in 2004 versus 1995 were five times more likely to recommend against breastfeeding for mothers who were perceived as “too young or immature” according to the AAP Annual Survey of Fellows selected from a random sample of non-retired AAP members in the US [50], it is of particular importance to provide appropriate clinical support to young mothers.

There are a number of strengths of our study. To our knowledge, it is one of the few first studies exploring factors associated with BrBot subgroups, which contribute to better understanding of mothers who feed both breast milk and formula to different extents. Also, we applied the WHO definition of XBR and recommendation of XBR for 6 months based on breastfeeding data obtained from biological mothers. In discriminant function analysis, a comparison of standardized discriminant coefficients revealed the relative discriminant ability of each determinant, which is of great significance in formulating effective public health initiatives to address disparities in lactation performance.

Certain limitations of the study need to be noted. First, infant feeding practices were based on maternal recalls. Thus, random errors due to retrospective recall may exist, which may partly contribute to the wide confidence intervals for effect estimates. However, such self-reported data have been used successfully previously [51, 52] and have shown to provide valid and reliable estimates of breastfeeding duration with recall lengths varying from 3 years after the practice to 20 years after delivery [53, 54]. Second, some factors which may influence mother’s choice of infant feeding practices are not available in this study, e.g. maternal intention to initiate breastfeeding, prior lactation experience, psychosocial factors (e.g., knowledge of breastfeeding recommendations and benefits, mental health, social support) [38], and maternity leave. In addition, since gestational weight gain was self-reported in categories rather than on a continuous scale based on questions used previously [55], women cannot be precisely classified as below/within/above the Institute of Medicine guidelines by prepregnancy BMI. Moreover, besides breastfeeding directly from the breast, recent data predominantly from the US Caucasian population revealed that infants were also frequently fed expressed breast milk from the bottle [56, 57]. Future research to investigate the determinants of not only the type and duration of milk fed to the child but also the mode of milk delivery in a multiethnic population is warranted. Finally, the study population was not sampled to be nationally representative which would limit the generalizability.

In conclusion, the present study developed predictive models for characterizing disparities in XBR performance (i.e., ever XBR and XBR ≥ 6 months). Discriminant function analysis provided a unique approach to compare the relative discriminant abilities of determinants of ever XBR and XBR ≥ 6 months. Findings revealed that maternal sociodemographic factors, more than birth outcomes and maternal perinatal factors, had greater discriminating abilities to predict ever XBR and XBR ≥ 6 months. Foreign-born women were two-fold more likely to initiate XBR but not necessarily continue to 6 months compared to their US-born counterparts. Moreover, maternal age at childbirth was the only factor consistently associated with ever XBR, XBR ≥ 6 months, and BrBot subgroups, whereas maternal education, smoking during pregnancy, and child’s birthweight were associated with XBR outcomes but not BrBot. Our findings highlight the importance of public health efforts to promote XBR initiation and XBR for 6 months, especially among young women, and suggest the need to encourage longer duration of XBR among immigrant women despite their intention to initiate XBR.

Significance.

Maternal lactation performance varies across populations, yet the relative impact of sociodemographics, perinatal factors, and birth outcomes on exclusive breastfeeding (XBR) outcomes is not well known. The predictive models developed in this study suggest that sociodemographic factors, more than perinatal factors or birth outcomes, had greater discriminating abilities to predict ever XBR and XBR ≥ 6 months. Foreign-born women were two-fold more likely to initiate XBR but not necessarily to continue for 6 months compared to their US-born counterparts. Moreover, maternal age at childbirth was the only factor consistently associated with ever XBR, XBR ≥ 6 months, and breast-bottle feeding subgroups.