Design and recruitment

This was a longitudinal analysis of the Delta Healthy Sprouts participants’ breastfeeding intent, knowledge, and beliefs measured at baseline [enrollment; gestational month (GM) 4 visit] and the last gestational (GM 9) visit as well as breastfeeding behaviors (initiation and duration) in the postnatal period [postnatal month (PM) 1 through PM 12 visits]. A comprehensive description of the Delta Healthy Sprouts Project has been published elsewhere [9]. Briefly, 82 pregnant women were enrolled in this project spanning three Lower Mississippi Delta counties. Recruitment activities included publicizing the study in the local media via the distribution of flyers and brochures and active recruitment by study staff at local health clinics and medical facilities serving pregnant women and at local health fairs. Women also were referred to the study by health clinic/department staff, WIC nutritionists, social service agencies, and through word of mouth by currently enrolled participants. Inclusion criteria included: female gender; at least 18 years of age; less than 19 weeks pregnant with first, second or third child; singleton pregnancy; and resident of Washington, Bolivar, or Humphreys County in Mississippi. Participant enrollment occurred on a rolling basis; hence baseline data were collected between March 2013 and December 2014.

The target enrollment was 75 women in each of the two arms (control and experimental) of the project. The sample size of 150 women was based on the following assumptions: 20% attrition rate, 37% of control participants with gestational weight gain with the Institute of Medicine recommendations, and a 22% difference between treatment arms for gestational weight gain within recommendations. Additional power and sample size calculations for the postnatal primary outcomes –postpartum weight loss and child obesity at 1 year of age – were performed [9]. However, recruitment was stopped by the study’s Principal Investigator prior to reaching these numbers due to unexpected difficulties recruiting pregnant women meeting study criteria. Recruitment was extended as long as possible, but fiscal issues eventually necessitated the closing of this period. Data collection was completed in May 2016. Figure 1 illustrates the CONSORT diagram.

Delta Healthy Sprouts was designed to evaluate the impact of the Parents as Teachers® (PAT) curriculum compared with a nutrition and physical activity enhanced PAT curriculum (PATE) on maternal gestational weight gain and postpartum weight control and childhood obesity prevention. Parents as Teachers is a nationally recognized, evidence based, home visiting program that seeks to increase parental knowledge of child development, improve parenting practices, provide early detection of developmental delays, prevent child abuse, and increase school readiness [13]. Participants were randomly assigned to one of two treatment arms [PAT control (N = 43) or PATE experimental (N = 39)]. Participants were followed for 18 months, starting at approximately 4 months gestation through 12 months postnatal. At the baseline (GM 4) visit, demographic data and anthropometric measures were collected, 24-h dietary recalls were conducted, and physical activity and other questionnaires were administered.

Intervention

The control arm of the intervention was based on the PAT curriculum that included one-on-one home visits, optional monthly group meetings, developmental screenings, and a resource network for families. Using the PAT model, Parent Educators provided parents with evidence based information and activities during home visitation. Materials were responsive to parental information requests and were tailored to the age of the child (or gestational age of the fetus).

The experimental arm of the intervention built upon the PAT curriculum by adding culturally tailored maternal weight management and early childhood obesity prevention components. The PATE curriculum was guided by the theoretical underpinnings of the social cognitive theory [14] and the transtheoretical model of behavior change [15]. Additionally, the PATE curriculum included foundational elements from the Diabetes Prevention Program and the Infant Feeding Activity and Nutrition Trial. Elements based upon the Diabetes Prevention Program principles included a flexible, culturally sensitive, individualized educational curriculum taught on a one-to-one basis [16]. Elements taken from the Infant Feeding Activity and Nutrition Trial included anticipatory guidance and parenting support principles [17]. Anticipatory guidance involves providing practical, developmentally appropriate, child health information to parents in anticipation of significant physical, emotional, and psychological milestones [18]. Parenting support emphasizes children’s psychological and behavioral goals, logical and natural consequences, mutual respect, and encouragement techniques [19].

Intervention components of the PATE arm included appropriate weight gain during pregnancy and weight management after pregnancy, nutrition and physical activity in the gestational (mother) and postnatal (mother and infant) periods, breastfeeding, appropriate introduction of solid foods, and parental modeling of healthful nutrition and physical activity behaviors. Lessons included weight gain (gestational) and loss (postnatal) charts, hands-on activities, instructional DVDs, and goal setting and barrier reduction for both diet and exercise.

Parent Educators provided both PAT and PATE participants (regardless of breastfeeding intent) with the Parents as Teachers® handouts titled “Why Breastfeed” and “Formula Feeding” as well as the monthly newsletters that featured local breastfeeding classes (location, dates, and times). Additionally, Parent Educators offered to set up a meeting with a lactation specialist for those participants who expressed interest in or ambivalence towards breastfeeding. Hence all participants, regardless of treatment arm, were encouraged to breastfeed. However, during the GM 8 visit, PATE participants watched the Breastfeeding with Bravado DVD (Bravado! Designs, 2008, 25 min in length) which featured discussions with mothers who have breastfed, a mother breastfeeding, and advice from experts. During the GM 9 visit, Parent Educators again discussed the benefits of breastfeeding for both the mother and her infant with PATE participants. During this visit, discussions also included other feeding options (mixed breast and formula, and formula only), infant feeding cues, and maternal postnatal nutrition.

Both arms of the intervention were delivered in the home to women beginning early in their second trimester of pregnancy by trained, community based Parent Educators. Parent Educators were African American, college educated women residing in the target communities. They were trained to deliver the nutrition and physical activity lessons and to collect data from participants, including dietary intake, by senior research staff members who were certified master trainers in the Nutrition Data System for Research (NDSR) software. Home visits occurred monthly and were approximately 60-90 min in length for the PAT lessons, and approximately 90-120 min for the PATE lessons. Both PAT and PATE participants received incentives at every visit. Gift cards were provided at the baseline and first and last postnatal visits. Other incentives included items such as diapers and baby bottles, books, and toys. Additional details regarding Parent Educator training, study methodology, and lesson plan outlines have been published elsewhere [9].

Measures

Anthropometric measures obtained on the participants at the baseline visit included height which was measured in duplicate using a portable stadiometer (model seca 217, seca, Birmingham, UK) and weight which was measured using a digital scale (model SR241, SR Instruments, Tonawanda, NY). Both measures were performed without shoes or heavy clothing. Pre-pregnancy body weight was self-reported. Body mass index was calculated as weight (kg) divided by height (m) squared where height was averaged if the two measurements differed. Weight also was measured at each of the 17 subsequent (5 gestational and 12 postnatal) visits.

Breastfeeding intent, knowledge, and beliefs were measured at the baseline and GM 9 visits. Breastfeeding intent was captured with the survey item “When my baby is born, I intend to” with 3 exclusive responses – breastfeed only, bottle feed formula only, and combine breastfeed with formula feed. Breastfeeding knowledge (7 items) and beliefs (12 items) were assessed using true/false statements such as “breastfeeding is healthier for babies than formula feeding” and “breastfeeding is embarrassing.” These 19 items were taken from a study conducted with pregnant WIC recipients in 18 county health departments in Mississippi [20] and the national Loving Support Makes Breastfeeding Work campaign, a mail survey of low income postpartum women that was conducted in 2000 in Mississippi [21]. For each item, one point was given if the response was in the desired way (i.e., reflected the current state of knowledge about breastfeeding) and 0 points if the response was otherwise. While these items were taken from validated instruments, internal consistency or reliability of the two scales, knowledge and beliefs, was assessed in this study. Breastfeeding initiation was assessed at PM 1 visit by recording the participant’s response to the question “Are you currently breastfeeding?” If the response was “yes” or “no – stopped” then breastfeeding was considered as initiated. If the response was “no – never started” then breastfeeding was considered as non-initiated. Breastfeeding duration was assessed at PM 2 through PM 12 visits using the same question. Duration was measured conservatively as the last month at which the participant indicated she was currently breastfeeding.

Participants also provided information regarding demographic characteristics (e.g., age, marital status, household size, education, employment, household income, insurance, prenatal care), WIC participation, health history, current health conditions, and psychosocial constructs of diet and physical activity (expectations, social support, self-efficacy, and barriers) [22–25]. Details regarding other measures and questionnaire data that were collected, but are not relevant to the present paper, have been published elsewhere [9]. All measures and questionnaires were collected or administered by trained research staff (Parent Educators) using laptop computers loaded with relevant software (i.e., Snap Surveys).

Statistical analyses

Because there were two distinct periods in this study with associated measures of interest, analyses were run using a gestational cohort (all participants enrolled in the study; n = 82) and a postnatal cohort (participants who completed the gestational period and had at least one visit in the postnatal period; n = 54). Five participants who completed the gestational period but dropped out of the study prior to the PM 1 visit were excluded from the postnatal cohort. Further, preliminary analyses indicated that differences in breastfeeding outcomes between the two treatment arms were not significant. Hence, participants were classified into two groups, those that initiated breastfeeding (BF) and those that did not initiate breastfeeding (NBF) to determine if and what differences existed between these two groups.

Statistical analyses were performed using SAS® software, version 9.4 (SAS Institute Inc., Cary, NC). Descriptive statistics, including means, standard deviations, frequencies, and percentages, were used to summarize participants’ demographic characteristics, anthropometric measures, and breastfeeding outcomes – intent, knowledge, beliefs, initiation, and duration.

Chi square tests of association or Fisher’s exact tests (categorical measures) and two sample t tests (continuous measures) were used to assess differences between BF and NBF participants’ baseline characteristics, between gestational period study completers’ and non-completers’ baseline characteristics, and between postnatal period study completers’ and non-completers’ baseline characteristics. Gestational period study completers were defined as participants who had their GM 9 visit or those who had at least two visits in the gestational period and their PM 1 visit. The second definition was used because a substantial proportion of PAT and PATE participants (36 and 42%) who had their PM 1 visit, missed their GM 9 visit due to the early birth of their infant. Postnatal period study completers were defined as participants who had their PM 12 visit.

To determine the structure or relationships within the sets of breastfeeding knowledge and beliefs items, multiple correspondence analysis was used. Correspondence analysis is conceptually similar to principal component analysis, but applicable to categorical rather than continuous data. It provides a means of graphically representing the structure of cross tabulations to help in elucidating underlying mechanisms [26]. Item profiles that fall in approximately the same direction away from the origin and are located in approximately the same region of space are associated with each other. Further, the overall spatial variation (inertia) in each set of data points is quantified and assists in the interpretation of the plot.

Generalized linear mixed models, using maximum likelihood estimation, were used to test for significant group, time, and group by time (interaction) effects on breastfeeding outcomes. Maximum likelihood estimation is an approach for handling missing data in repeated measures. Group (BF vs. NBF) was modeled as a fixed effect for all outcomes. Breastfeeding knowledge and belief outcomes were modeled using a Gaussian distribution with an identity link function and time (GM 4 and GM 9 visits) was modeled as a repeated measure using a variance covariance structure. Least squares means with 95% confidence limits were computed using these models. Due to small cell sizes, the original three categories of breastfeeding intent – exclusive breastfeeding, mixed breast and formula feeding, and formula feeding only – were collapsed into two categories – exclusive breastfeeding plus mixed breast and formula feeding vs. formula feeding only. Hence breastfeeding intent was modeled as a binomial distribution with a logit link function and time (GM 4 and GM 9 visits) was modeled as a repeated measure using an exchangeable covariance matrix structure. A sensitivity analysis that excluded the 19 (8 BF and 11 NBF) participants who completed the breastfeeding questionnaire during the PM 1 visit due to a missed GM 9 visit also was conducted. This was due to the concern that these participants may have answered the breastfeeding intent question according to their postnatal infant feeding behavior and not their intent in the last month of pregnancy.

Exploratory univariate analyses (Fisher’s exact tests and two sample t tests) indicated that relationships existed between breastfeeding initiation and breastfeeding intent at GM 4 and GM 9 and beliefs at GM 9. Hence, logistic regression was used to determine if breastfeeding initiation was predicted by breastfeeding intent (exclusive breastfeeding/mixed feeding vs. formula feeding only) and beliefs while controlling for baseline characteristics that differed between breastfeeding groups. Odds ratios with 95% Wald confidence limits (CL) were computed. The significance level of the tests was set at 0.05.

Breastfeeding outcomes in the gestational period

Based on the correspondence plots, correct responses for the knowledge items appeared to cluster in the same region of the graph, while incorrect responses clustered together in another region. Likewise, generally more positive belief responses appeared to cluster in the same region of the graph, while more negative belief responses clustered in another region. The total inertia (variability) accounted for by the first two principal axes was 99.9% for knowledge items and 76.7% for beliefs, indicating that the two-dimensional approximation provided a good representation of the data points. Based on these analyses, it was concluded that both the knowledge items and the belief items likely represented one underlying construct, respectively, in the data. Hence knowledge and belief scales were created by summing the responses in each set of items.

At baseline, 98% (80 out of 82) of participants indicated that breastfeeding is healthier for babies than formula feeding. Correct responses for the other six knowledge items ranged from 33% (breastfeeding will make it less likely that I will get breast cancer) to 76% (breastfeeding will help me to lose weight after pregnancy). Breastfeeding knowledge significantly increased from baseline to GM 9 (4.6 to 5.6 out of 7 points, p < 0.001). Also at baseline, 95% (78 out of 82) of participants indicated that they did not believe that breastfeeding would make people think they did not have any money. Positive responses for the other 11 belief items ranged from 33% (breastfeeding hurts; false response) to 93% (breastfeeding is nasty; false response). The increase in breastfeeding beliefs from baseline to GM 9 was not significant (8.3 to 8.6 out of 12 points, p = 0.291). Likewise, participants were equally as likely to indicate their intent was to exclusively breastfeed or mixed breast and formula feed their infant vs. formula feed only at baseline (47 vs. 53%) as compared to GM 9 (55 vs. 45%, p = 0.900).

Breastfeeding outcomes in the postnatal period

Of the 54 participants, 21 (39%) indicated they had initiated breastfeeding, although only 10 (48%) of these 21 participants were currently breastfeeding at the PM 1 visit. Only 2 participants continued breastfeeding their infants until at least PM 3 and only 1 participant exclusively breastfed her infant and did so for 7 months. All other participants who breastfed also fed their infants formula.

Breastfeeding intent at GM 4, beliefs at GM 9, and pre-pregnancy weight class significantly predicted breastfeeding initiation (84% concordant). Participants who indicated their intent was to exclusively breastfeed or mix breast with formula feed their infant at GM 4 were 6.0 times (95% Wald CL = 1.38 to 27.25) as likely to initiate breastfeeding than participants who indicated their intent was to formula feed their infant at GM 4. Additionally, for every 1 point increase in the mean breastfeeding beliefs score at GM 9 (mean score = 8.5), the odds of initiating breastfeeding increased by 1.6 (95% Wald CL = 1.02 to 2.42). Further, participants classified as underweight or at a healthy weight prior to becoming pregnant were 5.6 times (95% Wald CL = 1.14 to 27.12) as likely to initiate breastfeeding as compared to their overweight or obese counterparts. Participant age and parity were not significant model covariates for breastfeeding initiation.