The primary question in this study was: does offering financial incentives for breastfeeding increase breastfeeding at 6 – 8 weeks post-partum in areas with low (<40%) breastfeeding prevalence?
The unit of analysis for the primary outcome was the electoral ward area while the prevalence of breastfeeding was treated as the continuous outcome. Four units of analysis were included.
The primary outcome of this study was the intervention effect on breastfeeding prevalence based on controlled baseline. The primary outcome was measured by a weighted multiple linear regression model where the weights for the outcomes were calculated using Donner and Klar method. The result of this analysis was based on a correlation between interclass coefficients which were estimated using Fleiss and Cuzick method. This method of analysis was intentional as it treated the electoral ward area at cluster level. When calculating for the primary data, the breast feeding prevalence at the electoral ward area level were analysed on a case basis where the number of infants with documented history of breastfeeding status was used as the denominator. Infant without breastfeeding outcome data were excluded in the analysis.
Respondent bias refers to a tendency by the respondents in a study to give false and misleading information about the information being sought. Respondent bias are often prevalent where the participants are expected to offer self-reports about themselves, or in situations where structured interviews or survey designs are adopted. A wide range of factors can induce response bias, nevertheless, all area related to the fact that human beings respond to stimuli based on an integration of multiple source of information. In this case, the motivation for incentives may have affected the mothers to lie about the status of breastfeeding.
Where respondents give false information about surveys, these responses have a major impact on the validity of the questionnaires and therefore the results of the study. In this case, the false information would have resulted in a conclusion offering financial incentives for breastfeeding increase breastfeeding in areas with low breastfeeding prevalence.
I do not think misclassification of the primary outcome occurred. The researchers designed the study in a way that allowed precise acquisition of relevant data for appropriate answering of the question raised. The classifications in the primary outcomes captured all the aspect required to answer the research question. If misclassification of the primary outcome would have occurred, the final results would not be reliable and therefore leading to a wrong conclusion.
The rates for lost of follow up in the primary outcome were 7.9% for the intervention group, and 8.2% for the control group. The loss to follow up in this case was insignificant and did not have any impact on the final outcome of the study. However, if any major loss would have been experienced, the final result would be significantly skewed.
Within the six-month period, a total of 2179 representing (40.4 %) of all the eligible infants claimed for vouchers in the intervention group. Claiming for voucher positively impact on the primary outcome because it suggested that the intervention (financial incentive) was influencing the decision to improve the level of breastfeeding in the intervention group as compared to the control group and baseline statistics.
There was a 6.2 % increase in breastfeeding after the intervention in the six-month period. The impact of the intervention efforts was more in quarter 3 and 4. In the nutshell, financial incentives incrementally result in increased rates of breastfeeding within the six-month period.
Based on the parameters compared as the baseline characteristics in the intervention and control group in table 1, there were no significant population differences that could have resulted to major impact on the final results. Furthermore, there was no significant differences in the number of infants due, the baseline breastfeeding prevalence, and deprivation scores. However, the total tally of births within the six month period in both the control and intervention group differs significantly. The intervention group had more births than the control group, meaning the rates of breastfeeding reports were higher in the intervention group than the control group. The researchers have not addressed this concern in the analysis provided yet the difference in tally could have affected the findings.
The author argues that a cluster design would assess an area level impact of financial incentives on breastfeeding. The cluster randomised design was necessary in the context of the study because of its ability to eliminate bias during the randomization process. The researcher selected an expert blinded to ward names to use computer generated random sequence allocation method to allocate the clusters as either intervention or control group.
Clustersampsi 0.317 1, alpha(0.8) power(.05) ratio(0.04)
Estimated sample size for two-sample comparison of proportions
Test Ho: p1 = p2, where p1 is the proportion in population 1
and p2 is the proportion in population 2
alpha = 0.8000 (two-sided)
power = 0.0500
p1 = 0.3170
p2 = 1.0000
n2/n1 = 0.024
Estimated required sample sizes:
n1 = 5867
n2 = 5013
Design Clusters Participants per group Total participants
As published 92 5398 + 4612 10,010
Cluster design (individual patient) 100 5867+5013 10880
Individual design 86 4959+4398 9357
The results from this trial cannot be generalisable to Australia because of differences in contexts and test factors, however, it is apparent from the data on breastfeeding presented by the Australian Bureau of Statistics that there is urgent need for research to determine which factor can improve the prevalence of breast feeding in Australia.