JMIR Publications

ABSTRACT

Dear Editor, We are responding to the Letter to the Editor regarding our recent article, “Effect of short, animated video storytelling on maternal knowledge and satisfaction in the perinatal period in South Africa: randomized controlled trial.”1 Unfortunately, we believe the letter’s authors (refer to hereafter as LAs), may not have read the paper as carefully as they have claimed and, because of this, we find their criticisms to be unfounded. Please see our detailed responses below. The LA’s comment that the “study may not truly reflect the situation of patients in the real world” because we did not choose to use propensity score matching, is unfounded. Propensity score matching has many applications2, one of them is matching to conduct quasi-RCTs. Large-scale and ad hoc population trials could benefit from a propensity score matching approach, but this was not the aim of our study. Importantly, one applies this type of matching to ensure that the background variables are similarly distributed among the comparison groups; in our study, we conducted proper random assignment of the participants and found that, quoting the letter, “no significant differences were seen between the SAS Intervention group and the control group” (Table 1). Thus, while we did not control for all the potential covariates, for the purpose of identifying the intervention effects, our design provides the highest level of confidence. We do not see a link between the statistical test used in this study and the LA’s statement that the study “may not truly reflect the situation of patients”. We respectfully refer the LAs to the Statistical Analysis section of our paper1 where we detail why we used the beta-binomial model to describe the score distribution and control for covariates effects instead of non-parametric, univariate analysis of the knowledge score. The LA’s comment advising us to employ generalized estimating equations (GEEs)3 has several unfounded statements: (i) there are no longitudinal data so GEE’s would not be advisable here, (ii) GEEs are a form of multivariate regression analysis, (iii) generalized linear model (GLM) was already used in our analysis, (iii) simply using GLM does not address whether a curvilinear relationship exist or not, and vice versa it is not necessary to use GLM to model curvilinear relationships. Finally, while we agree that environmental factors might affect marital satisfaction, this is not related to our paper: the LAs misread and did not realize that we measured maternal satisfaction with the video intervention, however our main outcome was not maternal satisfaction, but the knowledge score. The LAs also erroneously state that “subgroup analyses were not conducted based on factors such as age, race, education level, and income. Hence, this paper would be more meaningful if these factors were analyzed in layers.” In Figure 3 of our paper1 we show the results of our multivariable regression analysis, including the factors listed by the LAs. Thank you for allowing us to respond to this letter. Kind regards, The Authors 1. Adam M, Kwinda Z, Dronavalli M, Leonard E, Nguyễn VK, Tshivhase V, Bärnighausen T, Pillay Y. Effect of Short, Animated Video Storytelling on Maternal Knowledge and Satisfaction in the Perinatal Period in South Africa: Randomized Controlled Trial. J Med Internet Res. 2023;25:e47266. 2. Caliendo M, Kopeinig S. Some practical guidance for the implementation of propensity score matching. Journal of economic surveys. 2008;22:31-72. 3. Hanley JA, Negassa A, Edwardes MDd, Forrester JE. Statistical analysis of correlated data using generalized estimating equations: an orientation. Am J Epidemiol. 2003;157:364-375.