The association of breastfeeding self-ecacy with breastfeeding duration and exclusivity: assessment of the psychometric properties of the Greek version of the BSES-SF tool.

Background: Breastfeeding self-ecacy (BSES) is an important modiable determinant of BF initiation, duration and exclusivity. The use of valid measurement tools of BSES is important in clinical and community practice. Thus, we aimed to assess the properties of the Breastfeeding Self-Ecacy Scale (BSES-SF) to measure the self-ecacy of mothers in Cyprus and its association with BF and EBF up to the sixth month. Methods: A cross-sectional and longitudinal descriptive study with a nationwide consecutive sample of 586 mother-infant dyads, recruited from all state maternity wards and 24 (of 30) private clinics within 24-48 hours of birth and followed up by telephone up to the 6 th month postpartum. BSES was assessed at the clinic and at the rst month with the Greek version of the 14-item 5-point Likert BSES-SF scale. Information on infant feeding practices was collected at all four time points. The association between self-ecacy and breastfeeding outcomes was estimated in logistic regression models. Results: With Mean=3.55 (SD=0.85), BSES was only moderate, and lower among Cypriot women, primiparas, those who delivered by C/S and did not intent to breastfeed exclusively. There was good internal consistency across the 14 items (alpha coecient=0.94). While higherlevels of self-ecacy were not strongly associated with the likelihood of breastfeeding initiation, there was a stepwise association with exclusivity. The association between in-hospital assessment of self-ecacy and long-term EBF persisted in multivariable models adjusting for socio-demographic characteristics. Women in the upper quartile of BSES at 48 hours were more likely to breastfeed exclusively at the 1 st and at the 4 th month. Similar, if not stronger, associations were observed between self-ecacy at the 1 st month and BF outcomes at subsequent time-points. High BSES at the rst month (>3.96) had 58.9% positive and 79.6% negative predictive value for breastfeeding at 6 months. Conclusions: The Greek version of the BSES-SF showed good metric properties. The generally low prevalence of breastfeeding among mothers in Cyprus, community support screening at higher risk for premature BF discontinuation. and N = 284 respectively. Participants with missing socio-demographic information were excluded from the statistical analysis. The percentage of missing values was generally low and ranged between 0–7% (N = 488–504 and 262–284 respectively), with the exception of family income (N = 447 and 244 respectively); ‡ p-values as estimated using independent sample t-test or one-way ANOVA, as appropriate.

The 33-item Breastfeeding Self-E cacy Scale was developed based on this theoretical model with the aim to identify mothers at risk to discontinue breastfeeding [7]. Due to item redundancy, a shorter 14-item form of the scale with similar psychometric properties was proposed [13]. The BSES-Short Form has been translated and validated in many languages, including Polish [14], Croatian [15], Hong Kong Chinese [16], Turkish [17], Swedish [18], Portugese [19] and Spanish [20]. It has been extensively used in studies among the general population of mothers or speci c population groups; adolescents [21], primiparas [22][23][24], ethnic minorities [25][26][27], low income [28] and many more. Only recently, a 9-item exclusive breastfeeding speci c self-e cacy scale was developed on the basis of Dennis BFSE-SF scale to assess EBF where breastfeeding is common, with very good psychometric properties [29].
Research evidence suggests a positive predictive association of breastfeeding self-e cacy and breastfeeding initiation, duration and exclusivity [14-15, 17, 19-21, 28]. Assessment of maternal breastfeeding self-e cacy prenatally or/and postnatally can identify women at high risk of breastfeeding discontinuation and thus in need of breastfeeding support.
The BSES scale, either in its original or short form, is most widely used scale in the literature. Even though the BSES is available in Greek (personal contact with the developer), no study assessing the breastfeeding self-e cacy of Greek or Greek-Cypriot mothers has been identi ed in the literature. The purpose of this study was to assess the psychometric properties of the Greek version of the Breastfeeding Self-E cacy Scale -short form (BSES-SF) among a sample of women giving birth in Cyprus and investigate its association with breastfeeding initiation (at 48 hours), exclusivity and likelihood of breastfeeding up to the sixth month.

Study design
This methodological study was part of the wider research program "BrEaST start in life" aimed at strengthening the evidence-base around breastfeeding in Cyprus. The programme explored various determinants of breastfeeding, including the degree of implementation of the Baby-Friendly Hospital Initiative's "10 steps for successful breastfeeding", which form the basis of the National Strategy and Policy of the Cyprus National Breastfeeding Committee. All maternity clinics in Cyprus were formally invited to participate. Maternity wards in state hospitals (5 in total) and 24 of 30 private clinics agreed to participate. A convenience consecutive sample of mother-infant dyads was recruited during stay at the maternity clinics based on pre-de ned criteria. The study design was described in detail previously [6]. In brief, trained eld workers approached the women between 24 and 48 hours after birth and asked them to complete a battery of self-administered questionnaires while waiting outside. Mothers who consented to participate at the next phase of the study, were followed up with a telephone interview at the rst, fourth and sixth month.
Sample size calculation and, consequently the period of recruitment, was based on precision analysis with nite population correction (birth cohort around 10000 annually) to estimate the prevalence of breastfeeding with 95% con dence interval not wider that ± 5%. The recruitment period was constant across all sites (6-8 weeks) in order to approximate the correct distribution of births across settings, as there is no o cial clinic-level record of the number of births in the private sector, which nevertheless accounts for over 70% of births. In terms of detecting the association between breastfeeding self-e cacy and breastfeeding outcomes, the sample size was considered more than satisfactory based on power analysis using the range of estimates observed in previous studies.

Eligibility Criteria
Mothers were eligible to participate if they gave birth to a live infant in the participating clinics during recruitment, irrespective of whether they had a single or multiple pregnancy, they were at least 18 years of age, could read or speak Greek or English, had no health problems precluding them from breastfeeding, as recorded in the medical le and/or communicated to the team by the clinic staff (e.g. bilateral mastectomy, postpartum maternal complications) and were not separated from their infants after birth for medical reason, which would not allow breastfeeding initiation within one hour, e.g. transferred to NICU at the same hospital or at a different location (low birth weight < 2500gr, gestational period < 37 weeks).

Measurement tools
At baseline, mothers were asked to complete the Perceived Breastfeeding Self-e cacy scale -short-form (BSES-SF), developed by Dennis & Faux [7], along other tools including the WHO/ UNICEF questionnaire -Sect. 4 [31] on the self-reported experience of the "10 Steps for Successful Breastfeeding". Participating mothers were also asked to provide information on sociodemographic factors, parity, breastfeeding history, intention to breastfeed, lifestyle factors (e.g. smoking, alcohol) and other. At the rst month telephone follow-up, only breastfeeding mothers completed the BSES-SF. Information on infant feeding practices was collected at each contact, such as the type, time of introduction and frequency of supplemental feeding including formula, other liquids, solids, medication, vitamin, mineral drops or Oral Rehydration Solution (ORS). Self-reported current status, 24-hour recall as well as a retrospective event calendar method were used to estimate the prevalence of breastfeeding and exclusive breastfeeding, as previously reported [6].
Breastfeeding Self-E cacy Scale Short Form (BSES-SF) In its short-form, the BSES-SF contains 14 items with a ve-point Likert scale response-set. All statement are phrased positively, and begin with the phrase 'I can always…'. The response scale ranges from 1 = not at all con dent and 5 = very con dent. The theoretical range of the scale is 14-70, with higher scores indicative of higher levels of breastfeeding self-e cacy. Commonly, the total score is calculated by aggregating the responses on all 14 statements. In this study, the average (rather than the overall) score was used i.e. dividing the total score by the number of items. Other than allowing the inclusion of a small number of questionnaires (N = 27) for which an answer was not provided on all 14 items, this allows to express the score of on a scale of 1-5. There are no set cut-off values for the scale. However, it has been suggested that a 5-10 unit difference (approximately 0.5-1.0 SD) is clinically signi cant in terms of predicting breastfeeding success in the long-run. A 10-unit difference in the aggregate score would correspond to 0.7 on a 1-5 scale.

Translation of the tool
The tool was used with permission by the developer who provided an existing Greek translation of the scale. A number of grammatical changes were deemed necessary, after a forward-backward translation process, in order to improve readability. The main change was in terms of the correct use of the tense for the key verbs to refer to a continuing process. Other small syntactical changes were also necessary.

Ethical considerations
All necessary approvals were obtained from all involved bodies: Cyprus National Bioethics Committee, Research Promotion Committee of the Ministry of Health, which also grants access permission to state hospitals, and equivalently from the administration of all participating clinics. Furthermore, noti cation was sent to the Commissioner of Personal Data Protection.
Separate written consent was obtained for participating at each phase of the study in order to ensure higher participation at baseline. Mothers were informed that participation was volunteer and they could withdraw their participation at any time point of the study. Con dentiality and anonymity was assured.

Statistical Analysis
Maternal Breastfeeding Self-E cacy was expressed as a continuous variable (i.e. average score) as well as an ordinal variable (i.e. quartiles of participants based on the score distribution). A statistical criterion was used due to the lack of generally accepted and uniform cut-off points across populations. Differences in mean BSES between subgroups of participants based on their sociodemographic and other characteristics were explored in one-way analysis of variance (ANOVA) and independent t-test as appropriate. Bonferroni post hoc comparisons were performed were necessary. Construct validity was evaluated through exploratory factor analysis with a principal components extraction with a varimax rotation to identify the dimensionality of the scale. Reliability was assessed using the Cronbach's alpha coe cient for internal consistency. Concurrent and predictive validity of BSES was explored with the investigation of the association of breastfeeding self-e cacy as measured at baseline and at rst month with infant feeding practices postpartum. Odds ratios (and 95% CI) of BF/EBF at each time point across quartiles of increasing BSES were estimated in logistic regression models before and after adjusting for important covariates. The diagnostic ability of the scale was also assessed by calculating the receiver operating characteristics (ROC) curve and the area under the curve. SPSS for Windows Version 21(SPSS Inc., Chicago, IL, USA) was used for the analyses.

Participant characteristics
The baseline sample consisted of 586 mother-infant dyads (response rate 73.5% among those eligible to participate), approximating the expected national distribution of births across districts and 70:30 split between private and public sector. Of those, 372 (response rate: 63.5%), 383 and 340 mothers respectively participated at the rst, fourth and sixth month follow-up.
The total number of mothers who completed the BSES-SF scale at 48 hours and 1st rst month was 504 and 284 respectively, since with a few exceptions, those who did not initiate breastfeeding or were not breastfeeding at the time of assessment did not respond to the scale. As shown in Table 1, the majority of women were aged 25-29 years (46.3%) and 30-34 (21.0%). Only 6.3% of participating mothers were older than 35 years of age. The percentage in the sample who were of Cypriot origin was 77.2%, which is consistent with o cial statistics. In terms of educational attainment, 57.3% had either undergraduate or postgraduate tertiary education. For 48.9% of mothers, this was their rst child. Among the rest, 44.7% reported previous breastfeeding experience. Even though the true prevalence of exclusive breastfeeding was only 18.8% at 48 hours, 81.9% of mothers reported their intention to breastfeed exclusively and 76.6% to do so for at least six months. It is of note that as many as 55.8% delivered by C/S, which is consistent with o cially published data. 3.53 (0.85) †Total number of breastfeeding mothers who completed the BSES-SF scale at 48 hours and 1st rst month: N = 504 and N = 284 respectively. Participants with missing socio-demographic information were excluded from the statistical analysis. The percentage of missing values was generally low and ranged between 0-7% (N = 488-504 and 262-284 respectively), with the exception of family income (N = 447 and 244 respectively); ‡p-values as estimated using independent sample t-test or oneway ANOVA, as appropriate.

Internal consistency of BSES-SF
The internal consistency of the scale was estimated by Cronbach's α at 0.94, which is identical to the one reported in the original study [13] (Dennis, 2003). There was no increase by more than 0.1 in the alpha coe cient in response to the deletion of any items. The inter-item correlations ranged between 0.26 to 0.82, with a mean of 0.55.

Breastfeeding Self-E cacy by Socio-Demographic Characteristics
The mean breastfeeding self-e cacy score was 3.55 (SD = 0.85) within the rst 48 hours and 4.01 (SD = 0.79) at the rst month assessment. Table 1 shows mean levels (SD) of BSES scores according to sociodemographic characteristics. A statistically signi cant difference in mean BSES scores was observed by mode of delivery. Mothers who had vaginal delivery were more likely to report higher levels of BSES at 48 hours (3.57, SD = 0.87) compared to those who gave birth by C/S with (3.37, SD = 0.93) and without general anesthesia (3.23, SD = SD: 0.89; p-value < 0.001). Even though BSES scores were generally higher for all by the rst month, a signi cant difference across different types of delivery was still apparent, corresponding to a moderate effect size in the magnitude of 0.5 SD.
Mothers who intended to EBF for six months reported signi cantly higher levels of BSES from the rst 48 hours (M = 3.37, SD = 0.82 vs M = 2.85, SD = 0.93; p-value < 0.001). With regards to nationality, Cypriot mothers (M = 3.29, SD = 0.86) were more likely to report lower levels of BSES than non-Cypriot mothers (M = 3.83, SD = 0.88; p-value < 0.001). This difference appeared smaller by the rst month, but remained statistically signi cant (p-value = 0.026).
Surprisingly, mothers with postgraduate education had the lowest BSES scores (M = 3.22, SD = 0.85) compared to both mothers with University or College education (M = 3.51, SD = 0.91) as well as those with at most secondary education (M = 3.51, SD = 0.91; p-value = 0.025). By the rst month, this difference was no longer apparent and the mean scores of the three groups appear similar.

Factor validity
The construct validity of the BSES-SF was assessed in Exploratory Factor Analyses -see Table 2. The correlation matrix was adequate with the majority of the correlation coe cients exceeding 0.3. The Kaiser-Meyer-Olkin coe cient for sampling adequacy (KMO) was 0.96, which is above the recommended value of 0.60 and the Bartlett's test of Sphericity was signi cant (pvalue < 0.001), both suggesting that the data are appropriate for factor analysis. Based on the default criterion of eigenvalues greater than 1.0, the analysis yielded a one-factor solution even though the scree plot was suggestive of a two-dimensional structure. of the variance. With a few exceptions (e.g. "Tell when my baby is nished breastfeeding"), most items were more likely to tap on Breastfeeding Self-e cacy in terms of the cognitive aspects, such as "Keep wanting to breastfeed", "Comfortably breastfeed with my family members present" and "Deal with the fact that breastfeeding can be time-consuming". The second component included the remaining six items, and explained 32.5% of the variance. These were more likely to tap on the technical aspects of breastfeeding management. For example, "Determine that my baby is getting enough milk", "Ensure that my baby is properly latched on for the whole feeding" and "Manage to breastfeed even if my baby is crying". One of the items ("Finishing feeding my baby on one breast before switching to the other") cross-loaded in both components with similar factor loadings.
The tool appears to be tapping on two different aspects of breastfeeding self-e cacy. While these two aspects have been described previously in the literature, calculating an overall score seems to be the most standard approach. Thus, for comparability, the overall score was used for further analyses.
Known-group validity According to Bandura, previous experiences have a signi cant in uence on self-e cacy [9]. Thus, multiparous mothers with previous breastfeeding experience would be expected to be more likely to have higher BSES compared with primiparous mothers.
A known-group comparison analysis was conducted to assess this assumption. As shown in Table 1 p-value < 0.001) -see Table 3. A similar stepwise pattern of association with breastfeeding status and BSES scores was observed across all follow-up phases of the study, irrespective of whether the analysis looked at the in-hospital or the 1st month assessment of BSES. Consistently, mothers who were still exclusively breastfeeding at a speci c time-point of investigation were those who reported higher on average BSES scores (in-hospital or at 1st month), with progressively lower mean scores observed among mother who were either breastfeeding but not exclusively or not breastfeeding by that point. and about twice as many as those in the second and third quartile (30.7% vs 14.1% vs 15.4%, respectively). Differences in the prevalence of EBF at the fourth month widened further and only appear to converge by the sixth month, due to a sharper decline in the prevalence of EBF among mothers with the highest in-hospital BSES scores. Similar stepwise patterns were observed when using BSES as reported at the rst month to track BF and EBF continuation beyond the rst month. In the multivariable model, the association appears even stronger with adjOR = 9.94 (95% CI: 3.72-26.58; p-value < 0.001) in the upper quartile. In relation to EBF continuation, an association with BSES scores at 48 hours was observed throughout the study period, attenuating slightly at the sixth month, possibly due to the small number of women exclusively breastfeeding by that point. Even though fewer than 10% of the mothers at the upper quartile were exclusively breastfeeding at the sixth month, this gure was still about four, three and two times higher compared to the respective gure observed among mothers at the lowest (2.3%), the second (3.3%) and the third quartile (4.7%) of BSES scores respectively. The observed stepwise pattern across increasing quartiles of BSES appeared consistent at all time-points of investigation irrespective of whether the analysis looked at 48-hour or 1stmonth BSES scores.
Diagnostic Ability of the Tool Table 5 shows the results of the ROC analysis for the predicting ability of the BSES -SF tool to identify mothers likely to successfully continue breastfeeding in the long term. In terms of diagnostic ability, the tool appears to perform better when used at the rst month and less well when used at 48 hours in terms of predicting BF continuation at 4th and 6th month. For instance, at a cut-off value of 3.96, the Sensitivity and Speci city of the BSES-SF at 1st month for BF continuation at 4th month is 79.7% and 63.7%, respectively. The positive and negative predictive value are 74.8% and 71.3%, respectively. Against a generally low prevalence of breastfeeding among mothers in Cyprus, this study showed that low self-e cacy in the early period is associated with non-exclusivity and earlier discontinuation of breastfeeding. In general, breastfeeding self-e cacy among women in Cyprus was only moderate. If expressed as a sum (instead of average) score, it corresponds to a score of 49.7.
With a some exceptions [16,22], this is lower than what is commonly reported among other populations in the international literature using the same tool (13-15, 17, 19, 32].

Strengths and limitations
This is the rst study to evaluate the psychometric properties of the Greek version of the BSES-SF and describe the breastfeeding self-e cacy of women giving birth in Cyprus. A clear strength is the prospective design which facilitated the assessment of feeding practices over the rst six months, avoiding the recall bias of a retrospective design. In fact, it is among a few studies that measured the predictive validity of BSES measured on two occasions on BF duration and exclusivity up to the sixth month, suggesting that the BSES -SF can be a useful tool for the identi cation of mothers who are more likely to succeed their breastfeeding goal. Even though a number of private clinics opted to self-exclude from the study, the generalizability of ndings, at least in a national context, is supported by the fact that the sample is largely representative of the cohort of mothers giving birth in Cypriot maternity clinics. With a response rate of 73.5% at baseline and 63.5% at follow-up, selection bias cannot be ruled out and it is likely that women who intended to breastfeed might be overrepresented in the sample. Even so, the prevalence of exclusive breastfeeding was particularly low while the observed variability in BSES scores among mothers in Cypriot maternity clinics is within the range, if not somewhat higher, than the variability observed in populations elsewhere, since a SD of 0.85 (on a 1-5 scale) corresponds to a SD of 12 on a 14-70 scale.
It is also acknowledged that the association between BF and BFSE is likely to be bi-directional (i.e. successful BF establishment positively in uences BFSE in the long-run). Like previous similar studies, this study look at the extent to which BSES is predictive of breastfeeding outcomes in the long term. Ever though the literature is limited with regard to the potential bi-directional association between BF and BSES, there is evidence to suggest that successful early initiation of BF within the rst hour after birth, when acknowledged as a positive personal experience, might be associated with higher BFSE levels at rst week postpartum [32]. Finally, intention to breastfeed was measured only with a single-item and other potential covariates related to motivation (e.g. beliefs and attitudes related to breastfeeding, maternal personality characteristics, etc) have not been considered.

Psychometric Characteristics of the BSES-SF scale
The observed internal consistency of the BSES-SF items was consistent with the original [13] as well as most similar studies in the literature. The BSES-SE seems to be tapping on two aspects of self-e cacy, namely breastfeeding technique and intrapersonal thoughts. Both the original study [13] as well as other translated versions of the scale [14-15, 17, 19] identify the scale as unidimensional. Our ndings are in agreement with a recently published study by Brandão et al. among Portuguese pregnant women which found a similar two-dimensional structure of the BSES-SF scale [31]. In that study, the second component explained only 7.6% of the variance, compared to 53.2% for the rst component, whereas in the present study, the percentage of variance explained by the two components was more equally distributed.
As expected, large differences were observed in terms of in-hospital BSES-SF scores according to breastfeeding status, demonstrating the concurrent and predictive validity of the tool. This nding is not surprising and, with a few exceptions [23,34], it is in agreement with the majority of previous studies. However, only a few studies investigated the association of BSES with BF/EBF up to the sixth month [16,[22][23][24][25][34][35] as studies commonly investigate the association of BSES with BF within shorter time periods [12,27,[35][36][37][38][39][40]. Similar, if not even larger differences were observed between BF/EBF and breastfeeding self-e cacy as reported at the 1st month, and this is also consistent with studies which measured BSES postnatally [16,22,[24][25][33][34].
The ROC analysis showed that the in-hospital BSES-SF scale at a cut-off value of 3.6 (corresponding to a sum score of 50.4) has acceptable diagnostic ability that a mother would still breastfeed at the 4th month and 6th month. A study by Ip et al. among Hong Kong Chinese mothers found that the BSES-SF at a cut-off value = 45.5 during hospital stay (48-72 hrs) could be used as a screening tool to identify mothers most likely to discontinue breastfeeding before six months with Sn = 73%, Sp = 73%, PPV = 92% and NPV = 42% [16]. In the present study, the predictive value of the scale appeared somewhat better when BSES was assessed at the rst month, rather than within the rst 48 hours. The sensitivity and speci city of the 1st month BSES-SF to identify breastfeeding continuation at four months were 79.7% and 63.7% respectively. The sensitivity of the 1st month assessment for breastfeeding at six months is very good (81%); however the speci city is average (56.1%) but not surprising as only one in twenty women breastfeed exclusively at six months. The positive predictive value of a relatively high score at the rst month is 74.8% and 58.9% for breastfeeding at the 4th and 6th month respectively. Thus, one in four and one in two women with high scores will be false positives, and they will discontinue breastfeeding suggesting that several other factors are at play. In terms of the negative predictive value, it is encouraging that a relatively low BSES score at the rst month will correctly identify 63.7% and 56.1% of the women who will discontinue breastfeeding by the 4th and 6th month respectively.
Sociodemographic Characteristics and Breastfeeding Self-E cacy No differences were observed in in-hospital breastfeeding self-e cacy in relation to maternal demographic characteristics (13-15, 19-20, 28], with the exception of educational attainment. In fact, mothers with postgraduate education appear to have the lowest breastfeeding self-e cacy levels. This comes in contrast with other ndings that suggest a positive relationship between breastfeeding self-e cacy and maternal education [15]. This association diminished by the rst month. This may suggest that mothers with higher education are more likely to be aware of the di culties risen during BF initiation which might result to negative beliefs towards their perceived ability to initiate BF, but are more likely to seek support and overcome the challenges in the long run. In fact, this is consistent with the fact that mothers with postgraduate education appear to have the largest increase in breastfeeding self-e cacy between the two time-points. This also appears consistent with the nding, that, even though there was no difference in the likelihood to initiate exclusive breastfeeding according to educational attainment, in the long term those with the highest educational attainment were 1.8-times (1st month), 2.3-times (4th month) and 3.7-times (6th month) more likely to be exclusively breastfeeding compared to mothers with primary or secondary education -results not shown in detail. It is also interesting to note that Cypriot women appear to have lower on average breastfeeding self-e cacy than non-Cypriot women. The extent to which this is re ective of differences in breastfeeding culture or other breastfeeding determinants between the two groups is not clear.
However, there is evidence to suggest that a negative or neutral previous experience may affect breastfeeding self-e cacy negatively [32]. This study did not explore the characteristics of the previous or current experience, which may determine the continuation of the behaviour [47][48].
Intention to breastfeed was also associated with breastfeeding self-e cacy [42,45]. This might be explained by the fact that mothers that intent to BF are more likely to be aware about the bene ts of exclusive breastfeeding, had attended antenatal educational sessions [45] and seeked formal or informal support [45,49]. In a recent study, Kronborg et al. found that both intention and self-e cacy are signi cant mediators of EBF and BF duration even among second-time mothers. It is interesting to note that in this study, even though the actual prevalence of exclusive breastfeeding was only 18.8% at 48 hours, 73.2% of mothers reported their intention to breastfeed exclusively [50].
The present study also con rmed the association between breastfeeding self-e cacy and mode of delivery, with mothers who gave birth vaginally having higher levels of breastfeeding self-e cacy [13,42]. There is evidence to suggest that intention to breastfeed is lower among mothers who give birth by C/S [51]. Furthermore, there is evidence to suggest that women who give birth by C/S are less likely to experience or request the implementation of "good practices" [52]. Experience of the "10 Steps to Successful Breastfeeding" are thought to facilitate the development of breastfeeding skills [53] and thereafter the strengthening of breastfeeding self-e cacy [32], which in turn strengthens maternal commitment to breastfeed [54]. In addition, women who deliver by C/S are more likely to experience breastfeeding di culties [51] including latching di culties, perceived lack of infant satiation and perceived lack of infant interest towards breastfeeding [55]. Early BF initiation within one hour after birth [32] (Koskinen et al. 2014), skin-to-skin [56] and rooming-in [32], all of which there is evidence to suggest are not widely implemented in Cyprus [5] have all been positively associated with higher breastfeeding self-e cacy levels. This is consistent with the nding that women who had a C/S without general anesthesia have somewhat higher levels of breastfeeding self-e cacy than those that gave birth with general anesthesia.

Implications for research and practice
In line with the aims, this study used only quantitative methods to explore breastfeeding self-e cacy and subsequent breastfeeding outcomes. However, future studies should focus on an in-depth exploration of the perceptions and attitudes of women in Cyprus with regards to breastfeeding, and perceived reasons for premature discontinuation using qualitative methods.
Further research is also required to disentangle the bi-directional association of BSE and BF since personal experience of BF, and the extent to which this is negative or positive, is likely to be the most important source of self-e cacy.
A systematic and structured assessment of BSES is not standard practice while the mother is at the clinic, let alone after discharge since in Cyprus there is no continuation of care in the postnatal period. As this study suggests, the BSES could be adopted in clinical practice as a screening tool to facilitate the identi cation of the mothers at higher risk to discontinue breastfeeding prematurely. However, for this to be effective, it is important to reconsider the current structure of maternal health care services in the community, either through the widening of existing roles or the establishment of new roles such as Community Midwifery.
Future research should be focused on the development and the evaluation of breastfeeding community support programmes which aim to enhance maternal breastfeeding self-e cacy. These programmes could include both formal and peer mother-tomother support groups [57][58][59]. A number of intervention studies have been designed based on self-e cacy theory and/or investigated the effect of breastfeeding education and/or support programmes explicitly through the enhancement of self-e cacy. For instance, a pre-and post-test experimental study with the participation of 74 Chinese primiparas [60], showed signi cant differences in BSES between the intervention and the control group at four and eight weeks after birth, while enhancement of BSES in this period was found to be signi cantly higher in the intervention group. The study showed that the positive impact of the intervention on BF duration and exclusivity, was mediated by the enhancement of BSES. However, a number of breastfeeding self-e cacy studies did not show a positive effect on breastfeeding outcomes [61] or the observed effect was short of statistical signi cance [62][63].

Conclusions
The Greek version of the BSES-SF showed good metric properties and it can be considered a valid and reliable measure of breastfeeding self-e cacy among new mothers in Cyprus. Concurrent and predictive validity of the scale was supported by the observed association of BSES with breastfeeding exclusivity at 48 hours and with breastfeeding outcomes at the rst, fourth and sixth month of the infants' life. The adoption of the BSES scale as a screening tool in clinical and community practice will assist in the targeted identi cation of women at higher risk for premature BF discontinuation. Of course, the generally low prevalence of breastfeeding among mothers in Cyprus suggests the wider need for the design and evaluation of community support interventions based on self-e cacy theory.

Consent for publication Not Applicable
Availability of data and materials The datasets generated or analyzed during the study are not publicly available because data analysis is still ongoing.

Competing interests
The authors declare no competing interests

Funding
The study was part of the research program "The BrEaST start in life: addressing social inequalities and supporting breastfeeding through inclusion activities", which was funded by Iceland, Liechtenstein, and Norway through the EEA Financial Mechanism and the Republic of Cyprus. The funders played no part in any of the stages of the study including study design, collection, analysis and interpretation of data and in writing the manuscript.
Authors' contribution NM, IPD, OK, CK, EL and EH conceived and designed the study and secured the funding. IPD was the project leader and NM was the scienti c coordinator of the BrEaST start in life programme. ME was the study's research coordinator and the study forms part of her doctoral dissertation. NM was the main PhD advisor of the rst author (ME) and CK and EL were members of the PhD advisory committee and were all involved in the overall supervision of the study. NM, VH and ME trained the eld workers and monitored the data collection at phase I. VH and FT performed data collection at phase I of the study and contributed to data management and analysis and to the editing of the manuscript. ME conducted the telephone follow-up data collection at phase II, managed the dataset, performed the statistical analysis and wrote the rst draft of the manuscript. NM guided the statistical analysis and NM and OK assisted in drafting and editing the rst draft of the manuscript. All authors assisted in interpreting the data and critically revised the manuscript. All authors have read and approved the nal version of the manuscript.