Some limitations of this study should be discussed. First, there was an important number of mothers lost to follow-up, probably due to the long time of the follow-up period. Losses to follow-up are common in population-based studies involving youth living in the peripheries of developing countries. In order to minimize a possible selection bias due to these losses, the statistical model employed included variables showing different prevalence rates in the intervention and control groups.
Another possible limitation is the large age range (4 to 7 years) at follow up assessment. This happened because we did not determine a specific age for the follow up evaluation. Since we spent almost two years to recruit the sample and ten months to locate all families for the follow up assessment, we ended up having this wide age range. Nevertheless, we believe that this fact has not affected the results, especially as the child's age was considered in the multivariable analyses.
And finally, we can not disregard the possibility of recall bias, as data on the duration of breastfeeding were collected retrospectively for mothers breastfeeding for over 12 months (53.2 % of the sample). However, this type of bias is more relevant when investigating duration of exclusive breastfeeding [24], as mothers tend to recall the duration of breastfeeding with relative accuracy. According to a study conducted in the United States, breastfeeding duration was only slightly overestimated at 1 to 3.5 years after the outcome [25]. Moreover, in our study, the outcome did not focus on a specific date, but rather on a period (2 years or more), which probably reduces the potential negative impact of a recall bias.
Some limitations of this study should be discussed. First, there was an important number of mothers lost to follow-up, probably due to the long time of the follow-up period. Losses to follow-up are common in population-based studies involving youth living in the peripheries of developing countries. In order to minimize a possible selection bias due to these losses, the statistical model employed included variables showing different prevalence rates in the intervention and control groups.Another possible limitation is the large age range (4 to 7 years) at follow up assessment. This happened because we did not determine a specific age for the follow up evaluation. Since we spent almost two years to recruit the sample and ten months to locate all families for the follow up assessment, we ended up having this wide age range. Nevertheless, we believe that this fact has not affected the results, especially as the child's age was considered in the multivariable analyses.And finally, we can not disregard the possibility of recall bias, as data on the duration of breastfeeding were collected retrospectively for mothers breastfeeding for over 12 months (53.2 % of the sample). However, this type of bias is more relevant when investigating duration of exclusive breastfeeding [24], as mothers tend to recall the duration of breastfeeding with relative accuracy. According to a study conducted in the United States, breastfeeding duration was only slightly overestimated at 1 to 3.5 years after the outcome [25]. Moreover, in our study, the outcome did not focus on a specific date, but rather on a period (2 years or more), which probably reduces the potential negative impact of a recall bias.
การแปล กรุณารอสักครู่..
