First, a time-only model was constructed to estimate the overall breastfeeding cessation risk across time and to indicate when breastfeeding cessation was most likely to occur (Aim 1).
Next, background and intrapersonal variables were added to the time-only model to determine the effects of these predictor variables on breastfeeding cessation (Aim 2).
The “beginning of time “ was breastfeeding initiation.
All women initiated breastfeeding and were therefore at risk of breastfeeding cessation at any point in time (measured in months).
There were 12 discrete time intervals, and each time interval had a unique risk set (pool of women who entered each successive time interval) of women who were eligible to stop breastfeeding during that particular interval.
For 238 women, an actual date of breastfeeding cessation was recorded, and these women were coded as 1 for “stopped breastfeeding” at that point in time.
For the remaining 71 women, no actual date of breastfeeding was recorded ; therefore, the last known breastfeeding date was used to compute survival history.
They were coded as 0 for “unknown if stopped breastfeeding” and treated as censored data (Singer & Willett,2003).
Of the 71women without a breastfeeding cessation date,62 were censored at their last known breastfeeding dates and nine were censored at the end of the 12-month study period.