The second Perth Infant Feeding Study (PIFSII) was conducted between mid-September 2002 and mid-July 2003 to monitor breastfeeding rates and identify changes in breastfeeding practices and the determinants of breastfeeding. The study was conducted in the same hospitals using the same methodology as the first Perth Infant Feeding Study (PIFSI). PIFSI was conducted 10 years previous and results have been reported elsewhere .
Mothers were contacted within the first three days following the birth of their infant. Women were considered eligible for the study if they had delivered a live infant free of any serious health conditions requiring transfer to the neonatal intensive care unit at Perth's major maternity hospital. Mothers whose infants were admitted to the Special Care Nurseries (SCN) of the participating hospitals were eligible for recruitment.
Those women agreeing to participate in the study completed the self-administered baseline questionnaire while in hospital or shortly after discharge. Women declining to participate were asked to provide some basic sociodemographic data in order to determine the representativeness of the sample. All women regardless of their chosen infant feeding method were followed up by telephone interview at 4, 10, 16, 22, 32, 40 and 52 weeks postpartum. The study instruments used were essentially the same as that used in PIFSI, with only minor improvements and additions being made to the instruments used in the PIFSII. Questions relating to smoking were based on the 1989–90 National Health Survey . Mothers were asked if they had smoked before pregnancy and if they had smoked during pregnancy as part of the baseline questionnaire. Mothers who acknowledged that they had smoked before pregnancy but had not smoked during pregnancy were categorised as 'stopping smoking' during pregnancy.
Data were entered and analysed using the Statistical Package for Social Sciences, Version 11.0 (SPSS for Windows, SPSS Inc., Chicago, IL, USA). Risk factors associated with stopping smoking during pregnancy were analysed using the baseline questionnaire. Variables identified in the literature as being associated with breastfeeding initiation and duration were examined and included in the development of each statistical model.
Estimation of odds ratios was performed for univariate analysis testing statistical significance by χ2 test. Adjusted odds ratios were calculated by logistic regression. All variables presented in Table 2 were entered into the model for the multivariate analysis of predicting stopping smoking during pregnancy. The model was reduced manually by excluding those variables with a less significant value.
The difference between duration of breastfeeding in those who stopped smoking during pregnancy and those who did not was initially explored using Kaplan Meier survival analysis. This relationship was further examined using logistic regression to examine breastfeeding duration less than and greater than six months using a variety of sociodemographic, biomedical and psychosocial factors reported to have an effect on breastfeeding duration in the literature. Variables were entered into the model to determine the effect on breastfeeding duration for more than or less than six months. Non-significant variables (p > 0.10) were manually excluded from the final model. Further analysis of this relationship using a Cox proportional hazard model was not attempted as the proportionality assumption had been violated and additional potential analytical techniques were considered beyond the scope of this paper. The six month time period was chosen based on the WHO recommendations for exclusive breastfeeding and was considered to be a significant reference point for infant feeding duration.
A mother's attitude towards infant feeding was measured by the Iowa Infant Feeding Attitude Scale (IIFAS) . The IIFAS is a 17 item scale which measures attitudes towards both breast and formula feeding with regards to the health and nutritional benefits, and the cost and convenience of each method. It has been shown previously to be a valid and reliable measure of infant feeding attitudes amongst women in the USA  and Scotland . Each item is measured on a 5-point scale and total scores could range from 17 (reflecting positive formula feeding attitudes) to a high of 85 (indicating attitudes that favour breastfeeding). For the purposes of the analysis mothers were split into two groups, those with an IIFAS score at or above the median (≥ 65) and those with a score less than the median (
Presented p values are two-sided, and a 5% significance level was used.
The PIFSII was approved by the Human Ethics Committee of Curtin University and the Research Ethics Committees of the two participating hospitals. Signed informed consent was obtained from participants. Confidentiality was assured and mothers were advised that their participation was voluntary and that they could withdraw at any time without prejudice.