Cross-sectional surveys were conducted in a 12 month period in 2013–14 in Scotland, the USA and three Nordic countries. Samples represented specific socio-geographical areas rather than countries as a whole.
Recruitment of participants
In central Scotland, three schools in urban areas with a broad socio-economic mix were invited to participate with prior permission from the Local Education Authority. In Sweden, two schools in the Mälardalen region (mixed SES metropolitan area around Stockholm) were contacted and agreed to participate. Parents’ permission was sought on an ‘opt-out’ basis for pupils. A hard copy letter invited parents to respond by returning a slip to school to exclude their child; no opt-out requests were received in either country. With consent, paper-based surveys were completed and placed in sealed envelopes by participants, collected and returned to researchers unopened by teachers (Scotland, Sweden).
Identical on-line surveys [39], were distributed by teachers in schools in Norway and the USA. In Norway, 3 schools around Oslo, (mixed SES) were recruited by personal contact. Parents first received hard-copy opt-out letters as above. Pupils later completed on-line surveys in school settings, having consented to take part. In the USA, permission was requested from school district administrators in Texas. A survey link was then forwarded to parents to first seek consent and then pass on to the participant.
In Finland, a ‘snowballing’ method and a popular internet site for young people (http://irc-galleria.net) recruited school aged participants in the urban Helsinki area.
For Sweden, Norway and Finland, the questionnaire was translated and back-translated into the first language using a translation service, and checked for face validity. In the USA, some terminology was changed (e.g. ‘nappies’ (UK) to ‘diapers’ (USA)).
Most (96%) of US participant responses came from 14 urban counties in Texas, (77% breastfeeding initiation, approximating the US national average [16]). Others (4%) came from Ohio, Michigan, and Arkansas.
Participants were school pupils aged 12–18. Sample size was estimated using a priori power analysis with G*Power [40]. Using linear regression with 13 predictors, medium ES, α = .01, 90% power, we required 192 participants. We recruited additional participants to allow for missing data, and to enable sub-group analysis.
Ethical approval
The study was given ethical approval by the host institution’s University Psychology Ethics of Research Committee, including approval to collect data in Sweden, Finland and Norway subject to local permissions and with local school approval from Scottish education authorities. Authorities deemed this sufficient to negate the need for further institutional ethical review in Sweden and Norway, where additional school-based permissions were then obtained, and in Finland, where additional permission was given by a national social networking site adminstrator. The Internal Review Board of Texas A&M University (Central) gave USA approval.
All participants obtained parental consent to participate, using slightly different methods (see above). Immediately before completion the voluntary nature of the questionnaire was emphasised to potential participants who were given the opportunity not to participate. The on-line version included mandatory check boxes indicating prior parental consent and participant consent, before the questionnaire would open. It was incentivised for individual participants (young people) offering an on-line voucher for every 25th response. Inclusion in this draw, (via their email) was voluntary.
Demographics
Age was coded into 3 categories: 12–13, 14–15, 16–18. Mothers’ highest level of education provided a proxy for SES [12, 41]. In Scotland we recorded four levels; ‘standard grades’ (national qualification, age 15–16), ‘highers’ (national qualification, age 16–18), vocational qualifications and university degree. Equivalents were determined for USA and Nordic countries.
Study variables
The main study outcomes was infant feeding intentions. Predictors were infant feeding beliefs, measured in a TPB framework, measuring beliefs/attitudes, norms, perceived behavioural control and parenting beliefs, including attitudes, gender norms and confidence. Scales’ reliability was established using Cronbach’s alpha throughout (α).
Infant feeding
For consistency, we defined infant feeding at the start of questionnaires: Breast-feeding – baby gets milk from the mother’s breast; Formula-feeding – baby gets formula/powdered milk from a bottle. Combined feeding – baby gets both breast milk from the mother and formula milk from a bottle.
Own feeding experience
Participants were asked how they had been fed themselves as a newborn; (options: ‘breast-fed’, ‘formula-fed’, ‘combined-fed’, ‘don’t know’). We combined the ‘breast-fed’ and ‘combined-fed’ responses, and excluded ‘don’t know’ to create a dichotomous variable representing ‘any breast-feeding’ vs. ‘no breast-feeding’.
Future feeding intentions
We asked how they wished to feed their baby if they became a parent, with fixed choice options as above, creating a dichotomous ‘any breast-feeding’ variable for analysis.
Infant feeding behavioural beliefs
Items previously used with new parents and adolescents in Scotland [34, 36] measured behavioural beliefs about breastfeeding and formula feeding, avoiding social desirability bias towards breastfeeding as a desired health behaviour. Initially, thirteen items were scored on a Likert scale: 1 ‘strongly disagree’ to 7 ‘strongly agree’. PCA with varimax rotation identified a forced 2 factor solution with four variables eliminated (using cut-off 0.4, [42]) since they did not load onto either factor. Factors represented positive (4 items, 19.6% of variance) and negative beliefs about breastfeeding (5 items, 27.8% of variance). Positive beliefs included: ‘breastfeeding leads to a close bond between mother and baby’, ‘…helps get the mother’s figure back to normal’, ‘…is a natural way of feeding babies’, and ‘formula feeding is expensive’ (α = .59). Negative beliefs included: ‘formula feeding makes it easier for mother to go back to work’, ‘breastfeeding is embarrassing for the mother’, ‘breastfeeding spoils the shape of the mother’s breasts’, ‘breastfeeding limits the mother’s social life’, and ‘breastfeeding can be uncomfortable for the mother’ (α = .67). Reliability of factor solutions remained consistent in each cultural group.
Social norms
We assessed descriptive (what other people do) and injunctive (subjective) norms (perceived social pressure) [43].
Descriptive norm
‘Exposure’ to breastfeeding and formula feeding was calculated, from 4 social referent groups at variable social distance [36] (family member, friend, someone you don’t know, someone on television), e.g.: ‘how often have you seen a close family member breastfeeding a baby?’, scored 1–4; ‘never’, ‘1–2 times’, ‘3–10 times’, ‘more than 10 times’, summed to represent total exposure to breastfeeding and formula feeding (range 4–16).
Injunctive norms
Subjective norms (SN) were measured with 2 single items using a 7 point Likert scale with anchors as above: ‘People important to me would want me to breastfeed/formula feed my baby’.
Infant feeding confidence
Two summed items measured perceived control for each parent: ‘For the mother, breastfeeding a new baby would be..’; ‘For the father, helping to feed a new baby would be..’ scored on a 7 point Likert scale from ‘very difficult’ to ‘very easy’, (α = .52).
Parenting
Parenting intentions
We asked if participants intended becoming a parent in the future (yes =3, not sure = 2, no = 1), combining the last two groups, creating a binary variable.
Parenting attitudes
Two summed items measured attitudes to being a parent of a newborn; ‘overall being the parent of a new baby would be..’ rated on a 7 point scale from ‘very unpleasant’ to ‘very pleasant’, and ‘very bad’ to ‘very good’, with mid-points at ‘neither’. High scores represented more positive attitudes, (α = .85).
Parenting confidence
Two items measured confidence; ‘overall being a parent of a new baby would be…’, rated on a seven point scale from ‘very difficult’ to ‘very easy’, and ‘how confident would you be about being the parent of a new baby?’, from ‘very unconfident’ to ‘very confident’. Higher scores represented more confidence (α = .55).
Attitudes to shared parenting
Eight items measured attitudes, adapting and augmenting a 3 item ‘Equality in feeding’ factor from a Finnish measure of parents’ breastfeeding attitudes [35], making it applicable to non-parents. Items included; ‘baby receives breast milk’, ‘keep up with your friends’, ‘decide about feeding method’, ‘both can feed the baby’, ‘spend the same amount of time with the baby’, ‘share baby’s care equally’, ‘have time to yourself’, and ‘have time alone with the baby’. Items were scored on a 5 point Likert scale, anchors from ‘not at all important’ to ‘extremely important’. High scores reflected more positive shared parenting attitudes. Principal Components Analysis (PCA) with varimax rotation led us to discard 2 items: ‘baby receives breastmilk’ and ‘have time alone with the baby’. We identified 2 factors with eigenvalues over 1: ‘Equal parenting’ (4 items, α = .72, 38% of variance) and ‘Parental independence’ (2 items, α = .67, 23% of variance), providing better reliability than the total (α = .63). Reliability of factor solution and variance predicted was similar in each cultural group.
Gender role norms
We assessed gender role norms for parents’ participation in common tasks for newborns asking; ‘if both parents are around, how much do you think the mother or father should do these things for their newborn baby’? Eight tasks included: ‘getting up in the night to feed the baby’, ‘changing diapers’, ‘feeding the baby’, ‘playing with baby’, ‘taking baby out’, ‘taking baby to nursery’, ‘soothing baby when it cries’, and ‘babysitting’, scored from ‘mother does all of the time’ (score − 2) to ‘father does all of the time’ (score + 2), mid-point ‘both parents do equally’ (score 0). Higher scores reflected more paternal involvement. A total ‘gender role’ score summed these variables (α = .75).
Analysis
Data from Nordic participants were combined, having checked for significant differences between countries. There were no differences in feeding intentions and the overall pattern was homogeneous. Variables were checked for normality and patterns of missing data with no adjustments required. Totals reported in tables and text vary slightly due to some missing data, however missing data percentages were calculated for study variables and were generally low (between 0.5 and 6.0%). Proportions in categorical variables were examined using χ2 tests, and relationships using Pearson correlations. National group and sex differences were estimated using t-tests, one-way or factorial ANOVA. Age and sex were covariates. Planned hierarchical multiple regression predicted intended infant feeding outcomes from national group, demographic confounders, TPB infant feeding and parenting variables. Standardised beta values (95% CI for B) assessed predictors. Effect size (ES) was estimated using r, Cohen’s D, or partial η2 as appropriate.