IMD30 as a measure of area-based 2 Rayfield S, et al. BMJ Open 2015;5:e009144. doi:10.1136/bmjopen-2015-009144 Open Access on February 1, 2022 by guest. Protected by copyright. http://bmjopen.bmj.com/ BMJ Open: first published as 10.1136/bmjopen-2015-009144 on 13 November 2015. Downloaded from deprivation. Pregnancy and delivery characteristics included parity (subdividing multiparous women by previous breastfeeding experience of more or less than 6 weeks), type of delivery, admission to the NNU, length of stay in hospital, gestation at delivery, antenatal feeding intention and peer feeding behaviour (whether they had known other mothers during pregnancy and which feeding methods they had used). Statistical analysis Logistic regression was used to calculate ORs for the association between breastfeeding support and breastfeeding at 10 days and 6 weeks. Analysis of the factors associated with breastfeeding at 10 days was based on those who initiated breastfeeding. Similarly, the analysis of breastfeeding at 6 weeks was based on those who were breastfeeding at 10 days. The analysis was conducted separately for late preterm (34+0–36+6 weeks) and term (>37+0 weeks) infants and was performed in stages due to large numbers of variables: first the sociodemographic variables (model A), then the antenatal and delivery characteristics (model B), and finally the breastfeeding support variables (model C). At each stage, variables that were not statistically significant (p>0.05) were removed from the model. The remaining statistically significant variables within models A and B were combined with model C with further dropping of variables that were not statistically significant as necessary. Antenatal feeding intention and the peer feeding variables were included in the model as a priori confounders, regardless of their p value as they are recognised as being highly predictive of subsequent feeding behaviour.21 31– 33 The final multivariable model for breastfeeding at 10 days included any remaining statistically significant variables from each of models A, B and C in addition to these two variables. These final regression models for the late preterm and term infants at 10 days were then used as the respective models for breastfeeding at 6 weeks. Outcomes are presented as adjusted ORs (aOR), with 95% CIs. STATA V.13 was used to conduct the analysis with ‘survey commands’ to take account of the weighted sample. All percentages and ORs are presented as weighted values, whereas frequencies are unweighted values. RESULTS There were 15 104 singletons born at more than 34 weeks gestation, of which 14 525 (95.9%) were full term and 579 (4.1%) were late preterm. Overall, mothers of late preterm infants had a younger age distribution, were more likely to be from a non-white ethnic group and were more deprived than mothers of fullterm infants (table 1). In addition, late preterm infants were less likely to have been born by normal vaginal delivery (NVD), were more likely to be admitted to the Least deprived *Per cent rounded to 1 decimal point (weighted). Frequencies (n) are unweighted values. IMD, Index of Multiple Deprivation. Rayfield S, et al. BMJ Open 2015;5:e009144. doi:10.1136/bmjopen-2015-009144 5 Open Access on February 1, 2022 by guest. Protected by copyright. http://bmjopen.bmj.com/ BMJ Open: first published as 10.1136/bmjopen-2015-009144 on 13 November 2015. Downloaded from enough help with feeding in hospital (aOR 0.23, 95% CI 0.09 to 0.60, p=0.003) compared to mothers who did feel they had enough help (table 5) and more likely to be breastfeeding at 10 days if their mother reported that she was given contact details for community support groups (aOR 3.14, 95% CI 1.40 to 7.04, p=0.006) compared to mothers who reported that they were not given these details. The late preterm infants who reported no feeding problems at home were significantly less likely to be breastfeeding at 10 days compared to those who had experienced problems with feeding at home and had received help for them (aOR 0.08, 95% CI 0.02 to 0.33). In contrast to term infants, no types of breastfeeding support were associated with breastfeeding late preterm infants at 6 weeks. DISCUSSION Our study found a statistically significantly lower prevalence of breastfeeding at 6 weeks among late preterm infants compared to term infants. When mothers reported they had received enough help in hospital and were given contact details for support groups in the community, this was associated with a higher likelihood Table 3 Descriptive analysis* of antenatal and delivery characteristics Variable Breastfeeding at 10 days (%) Breastfeeding at 6 weeks (%) Gestation 34–36+6 37+ 34–36+6 37+ N (sample size) 365 (437) 9230 (11 292) 281 (365) 7547 (9230) Parity *Per cent rounded to 1 decimal point (weighted). Frequencies (n) are unweighted values. NVD, normal vaginal delivery. 6 Rayfield S, et al. BMJ Open 2015;5:e009144. doi:10.1136/bmjopen-2015-009144 Open Access on February 1, 2022 by guest. Protected by copyright. http://bmjopen.bmj.com/ BMJ Open: first published as 10.1136/bmjopen-2015-009144 on 13 November 2015. Downloaded from of breastfeeding in late preterm and term infants at 10 days and also term infants at 6 weeks.