Intended for healthcare professionals

Education And Debate Quality improvement report

Effect on rates of breast feeding of training for the Baby Friendly Hospital Initiative

BMJ 2001; 323 doi: (Published 08 December 2001) Cite this as: BMJ 2001;323:1358
  1. Adriano Cattaneo, epidemiologist (cattaneo{at},
  2. Roberto Buzzetti, medical statistician the Breastfeeding Research and Training Working Group.b
  1. a Unit for Health Services Research and International Cooperation, Istituto per l'Infanzia, Via dell'Istria 65/1, 34137 Trieste, Italy
  2. b Centre for the Evaluation of Effectiveness and Appropriateness of Health Care (CEVEAS), Modena, Italy
  1. Correspondence to: A Cattaneo
  • Accepted 28 June 2001


Problem: Breastfeeding rates and related hospital practices need improvement in Italy and elsewhere. Training of staff is necessary, but its effectiveness needs assessment.

Context: Eight hospitals in different regions of Italy.

Design: Controlled, non-randomised study. Data collected in three phases. Training after the first phase in group 1 hospitals and after the second phase in group 2.

Strategies for change: Training of trainers and subsequent training of health workers with a slightly adapted version of the 18 hour Unicef course on breastfeeding management and promotion.

Key measures for improvement: Hospital practices, knowledge of 571 health workers, and breastfeeding rates at discharge, three, and six months in 2669 mother and baby pairs.

Effects of change: After training hospitals improved their compliance with the “ten steps to successful breast feeding,” from an average of 2.4 steps at phase one to 7.7 at phase three. Knowledge scores of health professionals increased from 0.41 to 0.72 in group 1 (training after phase one) and from 0.53 to 0.75 in group 2 (after phase two). The rate of exclusive breast feeding at discharge increased significantly after training: 41% to 77% in group 1 and 23% to 73% in group 2, as did the rates of full (exclusive plus predominant) breast feeding at three months (37% to 50% in group 1 v 40% to 59% in group 2) and any breast feeding at six months (43% to 62% in group 1 v 41% to 64% in group 2).

Lessons learnt: Training for at least three days with a course including practical sessions and counselling skills is effective in changing hospital practices, knowledge of health workers, and breastfeeding rates.


Exclusive breast feeding for about six months has many advantages,1 but its prevalence in infants less than 4 months is low in many countries.2 There is some evidence that the implementation of the “ten steps to successful breast feeding” of the Baby Friendly Hospital Initiative will lead to an increase in breast feeding.3 Such policy requires changes in healthcare practices to be brought about by properly trained health professionals. Unfortunately, their knowledge and competence was always considered below standard when tested,4-10 and the effects of training have rarely been assessed.11 Of the courses available from Unicef and WHO, 12 13 the latter was shown to improve knowledge, clinical, and counselling skills.14 Little is known, however, on the effect of training on hospital practices.15-17 A recent paper showed the effectiveness of the Unicef course on breastfeeding rates in Belarus, where maternity hospital practices are similar to those in Western Europe 20 to 30 years ago.18

We examined the effects of the Unicef course on hospital practices, knowledge of health workers, and breastfeeding rates at discharge and three and six months later in a high income country.


Outline of problem

In Italy not one hospital has been designated as “Baby Friendly.”19 The available figures on the knowledge of some health professionals are discouraging.9 Not surprisingly, the rates of breast feeding are low,20 certainly below the recommended levels. 1 21 Our initial assessment of the situation confirmed that only one to three of the ten steps were implemented in the assessed hospitals; the mean score attained by health professionals in a knowledge test was low; and the rate of exclusive breast feeding at discharge was far from satisfactory.

Outline of context

Eight hospitals agreed to participate in our project. We allocated the eight hospitals to one of two groups, each with three general hospitals and one teaching hospital and with similar catchment populations. The hospitals of group 1 (southern Italy) were bigger than those of group 2 (central and northern Italy), with 30–80 versus 16–40 maternity beds, and had differences in numbers of annual live births (960-1960 v 374–2957 between 1996 and 1998), caesarean section rate (31-44% v 13-21%), proportion of low birthweight babies (7-15% v 3-9%), stillbirth rate (0-0.9% v 0-0.55%), and neonatal mortality (0.56-2.6 v 0-0.26%). The maternity unit of one hospital in group 2 closed because of regional reorganisation of maternity units soon after the initial assessment.

Assessment of problems and strategy for change

Details of approach taken

After the initial assessment in June 1996 (phase one) we implemented our training programme. Trainers in group 1 underwent training in September 1996. From October 1996 to February 1997 courses were offered to health professionals of the four group 1 hospitals by local trainers. In June 1997 (phase two) we carried out the second assessment after a short period to allow for changes, and in September 1997 the trainers in group 2 received their training. From October 1997 to February 1998 courses were offered to health professionals of the three group 2 hospitals. In June 1998 we carried out the third and final assessment (phase three).

Overall, the four hospitals of group 1 trained 377 of a possible 536 health professionals (71-130 per hospital) who would have needed a course and the three hospitals of group 2 trained 194 of a possible 237 (57-72 per hospital). Training covered 54% (29-100%) of obstetricians, 72% (39-100%) of paediatricians, 84% (45-100%) of midwives, and 68% (20-100%) of nurses.


We chose the Unicef 18 hour course12 because our priority was to change hospital practices. Having realised the importance of counselling, we integrated into the course a two hour session from WHO's 40 hour course.13 We added two chapters to guide course directors and facilitators. The final product included these guides, 17 classroom sessions, three clinical practices, references, glossary, and appendices: the Baby Friendly Hospital Initiative tool for self assessment of the hospital, slides, transparencies, and a questionnaire for evaluation of the course. The recommended schedule covered 18 hours over three days. For training of trainers, however, we conducted the course in 24 hours over four days, asking future trainers to participate in the preparation of some sessions and allowing more time for discussion of specific subjects.

Table 1

Assessment of usual practices by hospital group and by phase. Figures are proportions

View this table:

Measurement of problem

We used three tools to collect data during each phase. Firstly, we used the self assessment tool of the Baby Friendly Hospital Initiative, with a series of criteria (one to four) for each of the ten steps and an assessment of usual hospital practice at a point in time randomly chosen (see table 1). Each of the ten steps was considered fulfilled when all the criteria were met. We gave a self administered questionnaire to the trainees, with eight questions on knowledge and variables on professional characteristics. Mothers were interviewed at discharge, followed by a telephone interview after three and six months. The interview used the questionnaires recommended by WHO to get information on exclusive (no other food or fluids), predominant (non-nutritive fluids allowed), full (exclusive plus predominant), and complementary (food and nutritive fluids, including formula milk, added to breast milk) breast feeding. 22 23

We required 130 consecutive mother and baby pairs for the interviews at discharge and at three and six months for each hospital and assessment. With a projected 20% loss to follow up this would result in groups of about 400 women which would allow us to identify a difference in breastfeeding rates (exclusive at discharge, full at three months, any at six months) of at least 10% from a baseline of 50% with 1-a=95% and 1-b=80%.24 Infants with birth weight under 2000 g or a severe disease that required admission to the neonatal ward were excluded. The data were analysed with EpiInfo and SPSS. Breastfeeding rates were adjusted with direct standardisation by parity (associated with age of the mother and previous breastfeeding experience), type of delivery, and birth weight. We then developed a logistic model to examine which practices would have the best effect on the rate of exclusive breast feeding at discharge. Finally, we used logistic regression to study the factors associated with higher breastfeeding rates.

The ethical committee of the Istituto per l'Infanzia in Trieste, where the research was coordinated, approved the research protocol, and each participant enrolled signed an informed consent.

Effects of change

All hospitals improved their compliance with the ten steps, from one to three steps before to six to ten steps after training. Improvement was easier for steps four to nine—that is, for hospital routines. The performance was worse on steps one (written guidelines), two (training at least 80% of staff), and ten (community support). Table 1 shows the results of the assessment at various times. At phase one we found materials featuring names, logos, or brands of infant food manufacturers in only one hospital (in group 1); these materials were removed after training. We also found messages contrary to the ten steps in two hospitals of group 1. Messages in favour were present in one hospital of group 1 and two of group 2. By the third phase all the hospitals were displaying favourable messages.

Knowledge of health workers improved after training. The mean score, weighted by age, year of graduation, and years working in the same position, went up from 0.41 to 0.66 to 0.72 in group 1 and from 0.53 to 0.53 to 0.75 in group 2 at the three phases, respectively. The response rate was stable in group 2 (52-55% of trainees) but dropped by half (from 66% to 34-36%) in group 1 after phase one. Most respondents were women (between 81% and 91% in different groups); most men were physicians (89%) and most women were nurses and midwives (75%). All health professionals were represented, with more nurses (34%-59%) and midwives (13%-24%) than obstetricians (5%-19%) and paediatricians (4%-26%). The mean age varied between 37 and 43 years, with degrees obtained between 1954 and 1994.

Table 2 shows some features of the enrolled mother and baby pairs, and some feeding practices during their stay in hospital. The difference in sample size between phase one and two in group 2 is due to the reduction from four to three hospitals (but exclusion of data from the fourth hospital did not change the results). The difference between phase two and three is due to the loss of 60 data collection forms. Characteristics of mothers and infants in both groups were similar, except for slight differences in age, parity, and employment. These differences were taken into account in the standardisation of breastfeeding rates and in the logistic regression model.

Table 2

Characteristic of mother and baby pairs and some feeding practices during hospital stay. Figures are number (percentage) of mothers unless stated otherwise

View this table:

Table 3 shows the crude rates of exclusive, predominant, complementary, and no breastfeeding at discharge (recall period since birth) and at three and six months (recall period of 24 hours). The standardised rates do not differ significantly (P=0.34). Exclusion of the hospital withdrawn from group 2 after phase one does not change the results. In both groups the differences before and after training in exclusive breast feeding at discharge, full breast feeding at three months, and any breast feeding at six months are significant (at least P<0.05).

Table 3

Feeding at discharge and at three and six months in two hospital groups. Figures are number (percentage) of participants

View this table:

Four factors were significantly associated with exclusive breast feeding at discharge: first breast feed within one hour, rooming in (where the baby stays in the same room as the mother rather than being kept in a separate nursery), not using a pacifier, and instructions on expressing breast milk. The logistic model showed that absence of these factors would lead to about 29% exclusive breast feeding at discharge. When all these factors were present the expected rate increased to over 82%. These four variables probably represent the effect of training; grouped together, they were significantly associated with exclusive breast feeding at discharge (odds ratio 6.78; 95% confidence interval 5.65 to 8.14). Other factors associated with the same outcome were normal delivery (1.49; 1.21 to 1.84) and previous experience of breast feeding (1.45; 1.21 to 1.74). Full breast feeding at three months was significantly associated with exclusive breast feeding at discharge (1.96; 1.63 to 2.36) and previous experience of breast feeding (1.58; 1.34 to 1.87); training health workers had a positive but not significant association (1.20; 1.00 to 1.44; P=0.0543). At six months any breast feeding was significantly associated only with full breast feeding at three months (12.83; 10.32 to 15.95) and with exclusive breast feeding at discharge (1.33; 1.07 to 1.65).

Lessons learnt and next steps

We found that training had an effect on hospital practices and breastfeeding rates, as has been reported previously from Belarus18 and the United Kingdom.25 Previous research looked only at hospital practices and at knowledge, attitudes, and practices of health workers 5 14-17 or at the effect of changing policies rather than training.26 Our study was not a randomised trial, but the design allowed for control before and after in each group of hospitals. We could also separate the effect of the intervention from the possible effect of uncontrolled variables—a secular trend or a chance improvement after training that does not persist in time. Standardisation took care of other differences. Moreover, data were collected on several outcomes. Confounding, some loss to follow up, and some incompleteness of data may compound the interpretation of the effect of training on single outcomes, but the consistent overall picture suggests that this effect is real.

The rates of exclusive breast feeding at three and six months are still not satisfactory, compared with what is recommended. 1 21 Also, we consider that rates obtained with 24 hour recall may overestimate the real rates that might be obtained with longitudinal follow up every date from birth.27 The falling rate of exclusive breast feeding after discharge may be due to a lack of adequate support: many paediatricians lack the necessary competence,9 and there are no other support groups except for some lactation and La Leche League consultants in some Italian towns. Extended networks of peer counsellors, whose effectiveness in supporting breast feeding has already been described,28-30 could integrate the skilled support and follow up provided by trained health professionals.31

Key learning point

Training for the Baby Friendly Hospital Initiative leads to improved knowledge, better hospital practice, and higher breastfeeding rates at discharge and up to six months in high income countries with a modern healthcare system

To conclude, we recommend that effective training on breast feeding (over least three days, with practical sessions and emphasis on counselling skills) be included in all undergraduate courses leading to healthcare degrees and in all plans for in-service training in high income countries. For in-service training, it is important to target multidisciplinary groups of health workers involved in breast feeding at hospital and community levels. The implementation of effective interventions should not be limited to the healthcare system; it should cover a wider range of activities, aimed at changing the cultural representation of breast feeding32 and at defending breast feeding from the marketing of breast milk substitutes.33


Members of the working group

Mara Baldissera, Tea Burmaz, Adriano Cattaneo, Susanna Centuori, Riccardo Davanzo, Carla Pavan, and Sofia Quintero Romero, Istituto per l'Infanzia, Trieste; Luigi Esposito, Savino Mastropasqua, Francesca Monge Benettina, Crescenza Montenegro, and Michele Pontrelli, Miulli Hospital, Acquaviva delle Fonti; Alfredo Gatto, Giuseppe La Gamba, Santina Procopio, Anna Romano, Anna Maria Santelli, and Gemma Spagnolo, Pugliese Ciaccio Hospital, Catanzaro; Anna Aiello, Gabriele Chiappetta, Carlo Corchia, Annunziata De Risi, and Maria Pia Galasso, Annunziata Hospital, Cosenza; Sergio Conti Nibali, Marina Gemelli, Rosa Manganaro, Giovanna Mangano, and Angela Sicilia, University Hospital, Messina; Valeria Bodega, Emanuela Cosentino, Silvia Morassut, Anna Regalia, and Barbara Zapparoli, San Gerardo Hospital, Monza; Pierpaolo Brovedani, Daniela Sebastianutti, Maria Alberta Nassivera, and Lucia Zamolo, General Hospital, Tolmezzo; Antonio Deganello, Giovanna Zarantonello, General Hospital, Isola della Scala; Lucia Basili Luciani, Stefania Bibbiani, Giuliana Blasi, and Anna Federici, Belcolle Hospital, Viterbo; and Roberto Buzzetti, Centre for the evaluation of effectiveness and appropriateness of health care (CEVEAS), Modena.

Contributors: AC formulated the hypothesis, designed the study, developed the data collection tools, supervised data collection, checked the completeness and consistency of data, and coordinated the whole study. He analysed the results and wrote the paper with RB. All the members of the working group read, discussed, modified, and eventually approved the paper. Riccardo Davanzo, Carla Pavan, and Sofia Quintero Romero translated and adapted the 18 hour Unicef manual and acted as trainers of trainers. Mara Baldissera, Tea Burmaz, and Susanna Centuori helped with data entry, checking and analysis. The members of the working group based in the eight hospitals were trainers for the several courses conducted in each hospital and collected all the data. One of them was appointed as local coordinator and filled in the hospital self assessment tool.


  • Funding Istituto per l'Infanzia, IRCCS Burlo Garofolo, Trieste, and from the Associazione Culturale Pediatri.

  • Competing interests None declared.


  1. 1.
  2. 2.
  3. 3.
  4. 4.
  5. 5.
  6. 6.
  7. 7.
  8. 8.
  9. 9.
  10. 10.
  11. 11.
  12. 12.
  13. 13.
  14. 14.
  15. 15.
  16. 16.
  17. 17.
  18. 18.
  19. 19.
  20. 20.
  21. 21.
  22. 22.
  23. 23.
  24. 24.
  25. 25.
  26. 26.
  27. 27.
  28. 28.
  29. 29.
  30. 30.
  31. 31.
  32. 32.
  33. 33.
View Abstract