Do you think that a cluster randomised design was necessary in the context of this study? Provide a justification for your answer.
Please include a word count at the beginning of your document. The maximum length is 1000 words, so you will need to give brief answers.
The questions on this assignment relate to this articles (available on LMS):
Relton C, Strong M, Thomas KJ, Whelan B, Walters SJ, Burrows J, et al. Effect of Financial Incentives on Breastfeeding: A Cluster Randomized Clinical Trial. JAMA Pediatr 2018;172(2):e174523.
Question 1. What was the primary research question asked in this paper? (2.5 marks)
Question 2. What is the unit of analysis for the primary outcome, and how many units of analyses were included? (1.5 marks)
Question 3. What was the primary outcome of this study, and how was it measured? (2 marks)
Question 4. On page 3/7 (underthe subheading “statistical analysis”) the authors make the following statement in regards to individual level mother-infant feeding status outcome data collected using a questionnaire “… it became clear that this method would lead to poor estimates due to respondent bias”?What do the authors mean by this statement, and in what direction was this bias likely to have influenced the study results? (3 marks)
Question 5. Do you think that misclassification of the primary outcome may have occurred? Give reasons for your response and outline what (if any) consequences this could have had on the final result of the study? (3 marks)
Question 6. What was the rate of loss to follow-up for the primary outcome? What, if any, effect would loss have had on the final study outcomes. (2 marks)
Question 7. What proportion of the children in the intervention group actually received any vouchers as part of the intervention? What impact could this have had on the primary outcomes from this study? (3 marks)
Question 8. Interpret the key results presented in table 3 in less than 50 words. (3 marks)
Question 9. Isthere a difference in the baseline breast feeding prevalencebetween the intervention and control groups listed in table 1 that could have affected the findings? If so, have the authors addressed your concerns in the analyses presented? (4 marks)
Question 10. What justification did the authors provide for undertaking this study as a cluster design? Do you think that a cluster randomised design was necessary in the context of this study? Provide a justification for your answer. (4 marks)
Question 11. You consider re-designing the study, using a) the same cluster design, but using individual participant outcomes, and b) using individually randomised study design. Calculate the number of participants required, assuming a power of .8, a two-sided alpha of 0.05, the same effect (4% absolute difference), and a baseline prevalence of 31.7% in the control groups. For the cluster trial, assume that 100 clusters (wards) are available and the ICC is 0.024, as reported in the paper. When calculating these numbers, please use either Stata 14.1 or Stata 15’s “clustersampsi” command. Include the commands you use in your response. Complete the following table to show your results. (5 marks)
Design Clusters Participants per group Total participants
As published 92 5398 + 4612 10,010
Cluster design (individual patient)
Question 12. Do you think the results from this trial are generalisable to Australia? Why or why not? In formulating your answer, consider the information found here – http://www.abs.gov.au/ausstats/abs@.nsf/Lookup/6664B939E49FD9C1CA257B39000F2E4B (2 marks)