Article Text
Abstract
Objectives To compare breastfeeding rates at discharge for very preterm infants between European regions and neonatal units, and to identify characteristics associated with breast feeding using multilevel models.
Methods Population-based cohort of 3006 very preterm births (22–31 weeks of gestation) discharged home from neonatal units in eight European regions in 2003.
Results Breastfeeding rates varied from 19% in Burgundy to 70% in Lazio, and were correlated with national rates in the entire newborn population. Women were more likely to breast feed if they were older, primiparous and European; more premature, smaller and multiple babies or those with bronchopulmonary dysplasia were breast fed less. Variations across regions and neonatal units remained statistically significant after adjusting for maternal, infant and unit characteristics.
Conclusion It is possible to achieve high breastfeeding rates for very preterm infants, but rates varied widely across regions and neonatal units throughout Europe.
Statistics from Altmetric.com
Introduction
Breast milk is beneficial for high risk infants as for term infants, and may confer additional protection against necrotising enterocolitis and contribute to neurological development.1 2 Furthermore, breast feeding represents an opportunity to involve mothers in infant care during hospitalisation and encourage mother–infant bonding.1 Breast feeding can be difficult for mothers because of the medical condition and prolonged hospitalisation of their preterm babies2 and studies have found lower breastfeeding rates for preterm than term babies.3 Little is known about factors contributing to successful breast feeding after preterm birth.4 5
The aim of this study was to compare breastfeeding rates at discharge home from hospital for very preterm infants between European regions and to identify maternal, infant and neonatal unit characteristics associated with breast feeding.
Methods
This study includes all very preterm infants (n=3006; from 22+0 to 31+6 weeks) discharged home in eight European regions participating in the MOSAIC study,6 a population-based study which aimed to compare medical practices and organisation of care across Europe. Data on maternal and infant characteristics were collected from medical records using a common protocol.
Seven of the 10 MOSAIC regions with comparable data on breast feeding at discharge were included, together with the French Burgundy region where the MOSAIC protocol was implemented in a second phase. Data were from 2003 in the original MOSAIC regions and from 2002 to 2004 in Burgundy.
Feeding at discharge was recorded as breast milk, formula or mixed. We considered that infants were breast fed if they were fed exclusively (only breast milk) or partly breast milk. The type of feeding at discharge home was known for over 95% of infants, except in the Netherlands (74%).
Exclusive and mixed breastfeeding rates for very preterm infants were estimated by region and rates were compared with national published breastfeeding rates7 8 using a Spearman's rank correlation.
To assess determinants, we used a multi-level random intercept model with infants (level 1) nested within neonatal units (level 2) to take into account the hierarchical structure of data and the non-independence of observations within units.9 Infants were attributed to the unit where they spent their first consecutive 48 h of life, as breastfeeding success depends on establishing lactation in the period directly after birth.2
Maternal variables included age, parity and country of birth. Infant variables included gestational age, sex, multiple birth, small for gestational age, Apgar score, congenital anomaly, surgery, neurological morbidity, bronchopulmonary dysplasia, inborn and transfer at any time after admission to a neonatal unit. Neonatal unit variables included level of care, number of annual very preterm admissions and region. Variables associated with breast feeding (p<0.1 in univariable analyses) were included in multilevel models. Only the maternal and infant variables significantly associated with breast feeding in multivariable analysis were retained for the final model. We estimated variance in breast feeding between neonatal units using a model that had no predictor variables (τ00(1)), a model with individual-level variables (τ00(2)) and a model with individual- and unit-level variables (τ00(3)). The proportion of variability between neonatal units was assessed using the proportional change in the variance (PCV=(τ00(1)−τ00(2))/(τ00(1))).
Thirty-two units had only one admission over the study period and were excluded because the interpretation of random effects is not clear in this situation, and three units (15 infants) were excluded because of unknown level of care. Our final sample for the multilevel analysis included 2959 infants admitted to 98 primary units. STATA 9 software (StataCorp, College Station, Texas, USA) and Hierarchical Linear and Nonlinear Modeling (HLM v 6) software (Scientific Software International,Lincolnwood, Illinois, USA) were used.
Results
Overall breastfeeding rates varied from 19% in Burgundy to 70% in Lazio (figure 1). Exclusive breastfeeding rates varied from 29% in Trent to 6% in Poland, and mixed rates varied from 52% in Italy to 5% in Burgundy. Breastfeeding rates among very preterm infants were higher in countries with high overall breastfeeding rates (r=0.8, p=0.02).
Women were more likely to breast feed if they were at least 25 years old, primiparous and European (table 1). In contrast, babies were less likely to be breast fed when they were more premature, smaller, from a multiple pregnancy or had bronchopulmonary dysplasia.
While breastfeeding rates were lower in less specialised and smaller units, these unit characteristics were not significant after adjustment for individual characteristics and region. Region of birth remained strongly associated with breast feeding after adjustment for all characteristics. Variations across neonatal units were statistically significant, with a baseline variance of τ00(1)=0.91 (p<0.0001). Inclusion of individual-level variables accounted for about one-quarter of the variance between neonatal units (τ00(2)=0.68 (p<0.0001); PCV=25%). All individual- and unit-level variables accounted for 59% of the variation; however, residual unit-level variance remained highly significant (τ00(3)=0.37 (p<0.0001); PCV=59%).
Discussion
We found that maternal and infant characteristics were important determinants of breast feeding at discharge, but variability remained marked between regions and units after adjustment for these characteristics.
Our study is one of the few5 that provides population-based data on very preterm breastfeeding rates in Europe, and the first to provide estimates collected with a common protocol from regions in seven countries. One limitation was the absence of information about mothers' social status; these data are not systematically recorded in medical records in Europe. While maternal social status may be associated with breastfeeding duration in very preterm infants,5 it is unlikely to explain differences between regions. For instance, 59% of women who had a baby in Lazio in 200310 had a post-secondary level of education versus 62% in Ile-de-France.8
Our results are consistent with the literature which shows that babies with health problems and younger mothers are less likely to be breastfed.3 4 In contrast, low parity was more likely to be associated with breast feeding and European mothers were more likely to breast feed. This result contrasts with research on term infants in Europe which finds that migrant women are more likely to breast feed.11 12 Migrant women may face specific barriers to breast feeding in the NICU due to socioeconomic factors (access to pumps, transport problems) or communication (complexity of procedures).
Breastfeeding rates for very preterm babies were correlated with overall national breastfeeding rates, suggesting that a breastfeeding-friendly environment, reflected by a high national rate, also favours this practice among very preterm babies. Differences in exclusive and mixed rates between regions may reflect distinct approaches to breast feeding in maternity and/or neonatal units in each region. It seems that some regions, such as Trent, have focused breastfeeding promotion practices in NICU to encourage exclusive breast feeding.
We observed significant variations in breastfeeding rates between units, as described elsewhere,4 but not according to the specialisation or activity of the neonatal unit. Variations between units may be explained by breastfeeding promotion practices,13 which have been developed in some hospitals independently of level of care or unit size, such as prevention of mother–infant separation, access to breast pumps, support for milk expression, skin-to-skin contact, provision of lactation consultants, and implementation of ‘kangaroo mother care’ or developmental care programmes.
Conclusion
It is possible to achieve high breastfeeding rates for very preterm babies, but rates varied widely across Europe. Our results suggest that promoting breast feeding in the general population and in neonatal units can help to overcome difficulties related to breast feeding these babies. A better understanding of practices in units with high breastfeeding rates could provide a basis for European-level recommendations for this high-risk population.
Acknowledgments
We thank E Draper and L Gortner for their comments on the manuscript. We acknowledge the assistance of the personnel in the maternity and neonatal units in the regions that participated in the MOSAIC project.
Footnotes
-
The MOSAIC Research Group Belgium, Flanders: E Martens, G Martens, P Van Reempts; Denmark, Eastern Denmark: K Boerch, T Weber, B Peitersen; France, Ile-de-France: G Bréart, JL Chabernaud, D Delmas, PH Jarreau, E Papiernik; France, Burgundy: H Charreire, F Michaud, C Ferdynus, E Combier, JB Gouyon; Germany, Hesse: L Gortner, W Künzel, RF Maier, B Misselwitz, S Schmidt; Italy, Lazio: R Agostino, D Di Lallo, R Paesano; Netherlands, Eastern and Central: L den Ouden, L Kollée, G Visser, J Gerrits, R de Heus; Poland, Wielkopolska and Lubuskie: G Breborowicz, J Gadzinowski, J Mazela; Portugal, Northern Region: H Barros, I Campos, M Carrapato; UK, Trent Region: E Draper, D Field, J Konje; UK, Northern Region: A Fenton, D Milligan, S Sturgiss; INSERM U953, Paris: G Bréart, B Blondel, H Pilkington, J Zeitlin; external contributors: M Cuttini, S Petrou.
-
Funding The MOSAIC project was partially funded by a grant from the European Commission Research Directorate (QLG4-CT-2001-01907) and coordinated by Assistance Publique-Hôpitaux de Paris. The Burgundy cohort was funded with a grant from IRESP (Institut de Recherche en Santé Publique/Pubic Health Research Institute). MB was supported by a research grant from the French Ministry for Higher Education and Research.
-
Competing interests None.
-
Ethics approval All regions obtained ethics approval in accordance with national guidelines for research.
-
Provenance and peer review Not commissioned; externally peer reviewed.