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 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 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.
Characteristic of mother and baby pairs and some feeding practices during hospital stay. Figures are number (percentage) of mothers unless stated otherwise
Table 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).
Feeding at discharge and at three and six months in two hospital groups. Figures are number (percentage) of participants
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).