Background

The UK has one of the lowest breastfeeding rates worldwide [1] and ranked the second lowest among 32 countries in the WHO European Region with a breastfeeding rate at 6 months of 21% in 2000 [2]. Only a modest increase was observed by the 2005 Infant Feeding Survey which recorded a breastfeeding rate at 6 months of 25% [3]. Conversely, some other developed European countries, such as the Scandinavia and Germany have high breastfeeding initiation and continuation rates [4].

In recent years the UK Government has made breastfeeding promotion one of its priorities [5–7] and has amended the breastfeeding and weaning recommendations to recommend that infants should be exclusively breastfed for the first 6 months of life [8] following the revision of the WHO guidelines [9, 10]. Most recently the government proposed the Single Equality Bill under which women would get the legal right to breastfeed in public with the aim of boosting breastfeeding prevalence [11]. Some have observed that the UNICEF UK Baby Friendly Initiative is likely to increase breastfeeding initiation but not duration and suggested that other strategies which involve the provision of support for breastfeeding mothers in the early weeks after birth are required to encourage UK mothers to breastfeed for the recommended duration [12].

In 2004 Little Angels, a social enterprise based in Darwen (in Eastern Lancashire), started a collaborative project based on the WHO Ten Steps for Successful Breastfeeding [13] with the Royal Blackburn Hospital, a Baby Friendly Hospital in Lancashire, to support nursing mothers to breastfeed longer. From 2004 to 2006 they helped increase the breastfeeding rates at 6-8 weeks and 7-9 months from 20% to 56% and 4% to 13% respectively [14]. The initiative received the "Award for practices to promote the involvement of women in providing local maternity services" by the All Party Parliamentary Group (APPG) on Maternity [15]. In the Department of Health's 2007 review of progress against the National Service Framework for Children, Young People and Maternity Services, Little Angels were cited as a case study of good practice [16].

The effectiveness of peer support programmes in improving breastfeeding outcomes has been established [4, 17–19]. However breastfeeding duration depends on a number of determinants some of which may be modifiable. Knowledge about the type and importance of these determinants in peer support programmes is essential for building effective breastfeeding promotion programmes [20]. Maternal socio-demographic and obstetric factors such as ethnicity [21–23], parity [24–28], marital status [24–28], mode of delivery [24, 25] and socio-economic status [3, 4, 24, 29], and infant feeding practices [13, 29–33] are important predictors of breastfeeding cessation. The aim of this study is to identify maternal socio-demographic and obstetric factors, and in-hospital infant feeding practices that are associated with breastfeeding cessation among the mothers supported by Little Angels.

Methods

The data

Little Angels run a peer support programme which relies on trained peer support workers. They provide 24-hour telephone support to the nursing mothers in breastfeeding difficulties in addition to regular home visits. Information on feeding methods is collected at 6 weeks, 17 weeks, 6 months, and 9 months. Date of stopping breastfeeding and reasons for stopping are some of the additional information collected.

This study used data from mothers on Little Angels' database and covered all mothers resident in Blackburn with Darwen (BwD) and Hyndburn in East Lancashire who gave birth at the Royal Blackburn Hospital and initiated breastfeeding while in hospital. The database was anonymised and held 2,801 records at the time of the study. Records of mothers with babies aged at least 6 months on 30th May 2007 and with valid postcodes were extracted for analyses. In all 2,107 records were identified meeting these criteria. The main bulk of the data retrieved were for mothers who gave birth between January 2005 and December 2006. Thirteen records were for mothers who gave birth in 2004. Ethical approval was sought from the Cumbria and Lancashire Committee but it was considered not required for this study.

The socio-economic status of the mothers were obtained by linking the data to the 2004 Index of Multiple Deprivation (IMD) [34]. The resultant dataset was analysed to describe infant feeding patterns and the association between breastfeeding cessation, and maternal socio-demographic factors, maternal obstetric factors, and in-hospital infant feeding practices (see table 1 for the list of the factors).

Table 1 Predictor variables considered for the study
Full size table

The durations of breastfeeding, in weeks, were calculated from babies' dates of birth and dates of breastfeeding cessation stated by the mothers. Where the dates of breastfeeding cessation were not known the feeding methods at 6 weeks, 17 weeks and 6 months were used to estimate the durations. 1,129 (53.6%) of the records did not have dates of breastfeeding cessation of which 856 (75.8%) had entries indicating that the mothers were breastfeeding, at least, up to 6 months.

Outcomes

This study tested the hypothesis that breastfeeding cessation among the mothers was associated with maternal factors and in-hospital infant feeding practices. The main outcome measures were the mean and median breastfeeding durations in weeks, because the breastfeeding duration data was negatively skewed, and hazard ratios (HRs) of stopping breastfeeding associated with the variables of interest.

The World Health Organization (WHO) has defined three types of breastfeeding namely, exclusive breastfeeding, breastfeeding, and predominant breastfeeding [35]. In this study we used exclusive breastfeeding and breastfeeding as defined by the WHO.

Statistical analyses

The significance level for all the analyses was set at 5%. Chi-square and Fisher's exact tests were used to test statistical associations for bivariate analyses. Survival analyses techniques were used to estimate the duration of breastfeeding and identify breastfeeding cessation predictors because they provide a good understanding of breastfeeding behaviour over time [27]. Kaplan-Meier (KM) survival analyses were carried out to describe breastfeeding duration by the variables of interest. To quantify the association between these variables and breastfeeding cessation Cox proportional hazards model was used because it is a more powerful statistical technique than single-point prevalence estimates [27]. To identify variables independently associated with breastfeeding cessation Cox multivariate regression analyses were carried out with all the variables included in the model iteratively using the model analyses based on the likelihood ratio test [36]. Variables which did not significantly contribute to the fitness of the model were dropped until the best model was obtained containing ethnicity, parity, formula-feeding and teats-giving status in hospital as predictor variables. In all 2,074 out of 2,107 records were included in the final model. The results from the final multivariate model were reported as the adjusted HRs. StatsDirect (Version 2.6.3) was used for the analyses.

Results

Maternal socio-demographic and obstetric characteristics and in-hospital infant feeding practices

Table 2 shows the distribution of maternal socio-demographic and obstetric characteristics. 1,606 (76.2%) of the mothers were from BwD while 501 (23.8%) were from Hyndburn. Whites were the predominant ethnicity recorded overall (48.0%). They constituted 76.0% among Hyndburn mothers while among BwD mothers they formed 39.1% (p < 0.0001). More Hyndburn mothers were from areas with IMD scores in the 1st and 2nd quartile (35.9% and 31.1% respectively) than those from BwD (22.5% and 22.4% respectively; p < 0.0001). 96.7% of the mothers had partners, 42.7% were primiparae and 73.3% had vaginal deliveries; there were no statistically significant inter-borough differences in the distribution of these factors.

Table 2 Distribution of maternal socio-demographic and obstetric characteristics.
Full size table

Table 3 shows the distribution of in-hospital infant feeding practices. 77.7% initiated breastfeeding within one hour after delivery, 76.5% gave no infant formula, and 84.3% gave no teats while in hospital. There were no statistically significant inter-borough differences in the distribution of these factors.

Table 3 Distribution of in-hospital infant feeding practices.
Full size table

Infant feeding methods

Table 4 shows the distribution of infant feeding methods up to 6 months. Exclusive breastfeeding rates fell as babies grew older. At every assessment point (i.e. 6 weeks, 17 weeks, and 6 months) exclusive breastfeeding rates were higher in BwD than in Hyndburn. The inter-borough difference in the rates at 6 weeks was not statistically significant (p = 0.0594) but attained statistical significance at 17 weeks (48.2% versus 46.2%; p = 0.0011) and 6 months (14.9% versus 8.5%; p = 0.0002). Mixed breast and formula feeding also declined with increasing age but there were no significant inter-borough differences. Formula feeding peaked at 17 weeks by which time some mothers started introducing weaning foods.

Table 4 Infant feeding methods by borough of residence at 6 weeks, 17 weeks and 6 months.
Full size table

Breastfeeding cessation

Effects of maternal socio-demographic and obstetric characteristics on breastfeeding cessation

Table 5 shows breastfeeding durations by maternal factors, and the association between these factors and breastfeeding cessation. KM analyses showed that, overall, the mean breastfeeding duration was 21.6 weeks (95% CI 20.86 to 22.37 weeks) and the median duration was 27 weeks (95% CI 25.6 to 28.30 weeks). BwD mothers breastfed significantly longer than Hyndburn mothers. The breastfeeding duration for White mothers was significantly shorter than that for non-White mothers. Among non-White mothers Indians and Blacks breastfed longest while Pakistanis had the shortest duration but the differences were not statistically significant.

Table 5 Effects of maternal socio-demographic and obstetric factors on breastfeeding cessation.
Full size table

Partnered mothers breastfed longer than single mothers. The mean and median breastfeeding durations for multiparae were about 2 weeks longer than those for primiparae. Sub-group analyses showed that 2nd born babies were breastfed for the shortest duration while 4th born babies were breastfed for the longest duration. Mothers delivered vaginally breastfed longer than those who had operative deliveries. Sub-group analysis showed that those who had instrumental deliveries breastfed for the shortest duration. However all these differences were not statistically significant.

Unadjusted Cox regression HR showed that Hyndburn mothers were 36% more likely to stop breastfeeding than BwD mothers (HR: 1.36; 95% CI, 1.18 to 1.55) but the association became insignificant after adjustment for ethnicity. Unadjusted Cox regression HR showed White mothers were 54% more likely to stop breastfeeding compared with non-White mothers (HR: 0.65; 95% CI, 0.57 to 0.73). After adjustment White mothers were 69% more likely to stop breastfeeding compared with non-white mothers (HR: 0.59; 95% CI, 0.52 to 0.67). Sub-group analysis showed that White mothers were 89% more likely to stop breastfeeding compared with Indian mothers (HR: 0.53; 95% CI, 0.45 to 0.62), 45% more likely to stop compared with Pakistanis (HR: 0.69; 95% CI, 0.60 to 0.81) and 2.9 times more likely to stop compared with Blacks (HR: 0.35; 95% CI, 0.16 to 0.81).

Marital status was not associated with breastfeeding cessation in unadjusted Cox regression analyses (HR: 0.80; 95% CI, 0.58 to 1.11). From unadjusted Cox regression analyses primiparae were 25% more likely to stop breastfeeding compared with multiparae (HR: 1.25; 95% CI, 1.11 to 1.41); the association was slightly attenuated after adjustment (HR: 1.23; 95% CI, 1.09 to 1.39). Replacing parity as a binary variable with birth order in the model showed that mothers with babies of higher order birth were less likely to stop breastfeeding than those with lower birth order babies. For example mothers with 3rd born babies were about 8% less likely to stop breastfeeding than those with 2nd born babies (HR: 0.92; p < 0.0001). Unadjusted HR showed that operatively delivered mothers were 16% more likely to stop breastfeeding than mothers delivered vaginally (HR: 1.16; 95% CI, 1.01 to 1.33) and IMD 3rd quartile mothers were 22% less likely to stop breastfeeding than 1st quartile mothers (HR: 0.78; 95% CI, 0.66 to 0.92). However these associations became insignificant after adjustment.

Effects of in-hospital infant feeding practices on breastfeeding cessation

Table 6 shows in-hospital infant feeding practices and their effects on breastfeeding cessation. KM analysis showed that there was no statistically significant difference between the duration in mothers who initiated breastfeeding within 1 hour after birth and those who initiated later. Giving babies infant formula and teats in hospital were significantly associated with breastfeeding duration. Mothers who gave infant formula to their babies breastfed for a mean duration of 17.7 weeks (median: 17 weeks) compared with 22.2 weeks (median: 27 weeks) for mothers who did not. The duration varied by the reasons for giving formula: those who gave formula based on informed choice had the shortest duration while those who gave because of medical indications breastfed longest. Mothers who gave teats to their babies breastfed for a mean duration of 17.2 weeks (median: 17 weeks) compared with 22.8 weeks (median: 27 weeks) for those who did not.

Table 6 Effects of in-hospital infant feeding practices on breastfeeding cessation.
Full size table

Unadjusted Cox regression results suggested that initiating breastfeeding after 1 hour post-delivery increased the likelihood of breastfeeding cessation but the association became insignificant after adjustment. Unadjusted HRs showed that mothers who gave infant formula while in hospital were 56% more likely to stop breastfeeding compared with those who did not (HR: 1.56; 95% CI, 1.36 to 1.78). The association was slightly attenuated after adjustment (HR: 1.50; 95% CI, 1.26 to 1.79). Adjusted sub-group analyses results also showed that mothers who gave infant formula based on informed choices were most likely to stop breastfeeding (HR: 1.91; 95% CI, 1.49 to 2.44) while those who gave it based on medical indications were least likely to stop (HR: 1.34; 95% CI, 1.09 to 1.65). A similar association was found for the use of teats in hospital. Unadjusted Cox regression results showed that mothers who gave teats while in hospital were 61% more likely to stop breastfeeding compared with those who did not (HR: 1.61; 95% CI, 1.38 to 1.87). The association was markedly attenuated after adjustment (HR: 1.26; 95% CI, 1.04 to 1.53).

Discussion

Our study showed that ethnicity and parity were independent predictors of breastfeeding cessation. The UK 2005 Infant Feeding Survey and other studies have shown that White mothers breastfeed for shorter durations compared with mothers from other ethnic groups [3, 21, 37]. Among mothers from Asian background in the UK Pakistani and Bangladeshi mothers stop breastfeeding sooner than Indian mothers [23]. Many studies showed that breastfeeding duration increases with increasing parity [3, 26, 38, 39] and some evidence suggests that this might be related to previous breastfeeding experiences [3, 22].

Our study also found that giving infant formula and/or teats to babies in hospital were significantly associated with breastfeeding cessation. These in-hospital infant feeding practices, especially when based on informed choices, may not in themselves be predictive of breastfeeding cessation but may be the early manifestations of maternal intentions relating to long-term infant feeding practices. The literature is consistent about the detrimental effects of using teats on breastfeeding outcomes [13, 29–33] for which there may be a physiological explanation. The innate sucking reflex of the infant is satisfied by the teat, decreasing or eliminating the desire for contact with the nipple leading to reduction in breast milk production [13, 30]. However others have suggested that teat use may simply be a marker for socio-demographic, psychosocial and cultural factors that determine both teat use and breastfeeding [30].

Our results suggested that the reasons for giving formula and/or teats may influence breastfeeding duration. Mothers who gave formula based on informed choice were most likely to stop breastfeeding. The timing of the decision to breastfeed influences breastfeeding duration [24] and it is likely this sub-group of mothers had not decided to breastfeed their babies early during the antenatal period. Those who gave formula for medical indications were least likely to stop breastfeeding and this was consistent with the WHO findings [13].

We found in our study that known predictors of breastfeeding cessation such as marital status, socio-economic deprivation, mode of delivery, and the timing of breastfeeding initiation, were not associated with breastfeeding cessation suggesting the peer support these women received might have been beneficial in attenuating the effects of these factors. Partnered mothers are usually more likely to breastfeed longer than single mothers [24–28] and in the developed countries affluent mothers tend to breastfeed longer [3, 4, 24, 29]. Caesarean delivery is a recognised risk factor for breastfeeding initiation failure but there is no consistent evidence on its effect on breastfeeding duration [24, 25]. Recent evidence has suggested that operative deliveries may have detrimental effects on breastfeeding outcomes [40–43] probably mediated by delayed breastfeeding initiation [40, 44] as a possible consequence of anaesthesia and analgesia [45, 46]. However the UK 2005 Infant Feeding Survey found that mothers were equally likely to be breastfeeding at one and two weeks regardless of delivery methods [3]. Our study has found that mode of delivery was not associated with breastfeeding cessation supporting the conclusion by some that if the cultural and hospital environments generally promote breastfeeding, operative deliveries have little negative impact on breastfeeding outcomes [47].

Delayed breastfeeding initiation adversely affects breastfeeding duration and this formed the basis for the WHO/UNICEF recommendation that breastfeeding should be initiated within 30 minutes of birth even though it was admitted it is difficult to make exact recommendation about the timing of initiation [13]. In our study 77.7% of the mothers had initiated breastfeeding within one hour after delivery broadly in keeping with the recommendation but this was not associated with breastfeeding cessation.

Implications

Our study is one of the few to examine the effects of maternal factors and early infant feeding practices in hospital on breastfeeding in a peer support setting in the UK. The results have suggested that infant feeding practices associated with maternal ethnicity and parity may be more difficult to influence by peer support interventions. Peer support programmes, particularly those in multi-ethnic settings, will need to identify the needs of their various client groups in order to appropriately support them breastfeed longer. This could be done through an effective evaluation of such programmes.

We have also identified that in-hospital infant feeding practices which are likely to continue after leaving hospital have a negative impact on breastfeeding continuation. Infant feeding practices are influenced by the socio-cultural milieu and maternal intentions relating to these practices tend to be consolidated during the ante-natal period. For peer support programmes to be more effective we suggest their support start during this period to enable mothers form healthier infant feeding intentions.

Strengths, limitations and biases

In this study we used a routinely collected data. Problems of a routine data source are largely related to completeness and accuracy. Missing dates of breastfeeding cessation affected BwD more than Hyndburn. Missing data resulted in 2,074 records out of 2,107 to be included in the final Cox multiple regression model. These made the study prone to selection bias. A potential confounder not included in the analyses was maternal age because this was not collected in many cases. An additional shortcoming relates to the accuracy of the estimated breastfeeding duration based on the on self-reported dates of breastfeeding cessation and infant feeding methods.

The strength of our study lies in the statistical methods used. The use of survival analysis to determine the breastfeeding duration provides a good understanding of breastfeeding behaviour over time [27]. Furthermore this type of analysis is suitable for data collected from follow-up studies that involve censoring and/or in which the time to an event is regarded as the outcome variable of interest [36]. To quantify the association between the factors and breastfeeding cessation Cox proportional hazards model was used. This model allows joint estimation of the effects of predictor variables on the risk for breastfeeding cessation rather than the duration itself [27].

Conclusion

In this study ethnicity, parity and in-hospital infant feeding practices remained independent predictors of breastfeeding cessation in this peer support setting. However other recognised predictors of breastfeeding cessation such as marital status, mode of delivery, and socio-economic deprivation were not found to be associated with breastfeeding cessation.