Design and participants
This quasi-experimental study was conducted online between August 10 and October 14, 2016. Participants were recruited through a study notification published on 1) a website managed by Medicus Shuppan Publishers Co., Ltd, a large Japanese publisher in nursing and medical sciences, 2) a job search site for nurses, and 3) Facebook. Macromill, Inc., an online survey company, posted the study invitation on the abovementioned media, and eligibility was confirmed by screening questions when the participant responded. The inclusion criteria were possession of a national nurse license and previous work experience in obstetrics or pediatrics, including neonatal intensive care units and growing care units.
Participants were allocated to one of three groups: Group 1 received a “benefits of breastfeeding” text, Group 2 received a “risks of formula feeding” text, and Group 3 received a control text. Participants answered the questionnaire before and after reading the intervention text. The group allocation was not random; instead, participants were allocated by a computer system that balanced the number of participants in each group. Specifically, a new participant was assigned to the group with the smallest number of participants who had completed the pre-test questionnaire. After completing the pre-test questionnaire, the text of the allocated group was presented to the participants, after which they were invited to a post-test questionnaire. The whole process took approximately 30 minutes. Although participants were encouraged to read the text before proceeding to the post-test questionnaire, skipping the reading was possible and not monitored.
Blinding
As the intervention, participants were asked to read the text assigned to their group. Therefore, participants were not blinded to group allocation and could likely determine the allocation from the content of the text. The authors were blinded to the group allocation until the outcomes were analyzed and could not influence the responses.
Texts
The texts for Group1 and Group 2 were created in Japanese by using materials developed in English (9). Wallace and Taylor developed two texts to convey the same information on infant feeding: one using benefit language and the other using risk language. For example, the message “breastfed children are less likely to suffer from infectious illnesses and their symptoms” in the benefits text was converted to “formula-fed children are more likely to suffer from infectious illnesses and their symptoms” in the risks text. The present authors created the Japanese version of these texts with the permission of Wallace and Taylor, and the back-translation method was used for the translation. Specifically, the first author translated the original English version into Japanese, and a Japanese–English bilingual graduate student of public health who was blinded to the original version back-translated it into English. The authors compared the back-translated version and original English version. In the event of a discrepancy in meaning, the first author revised the Japanese translation. The process was repeated until the authors agreed that the Japanese translation was semantically equivalent to the original.
Before translation, the authors checked the relevance of the content of the texts. First, the authors assessed if the textual information was consistent with existing evidence in 2016. Systematic reviews and a policy statement published after the development of the original English version were examined (14-21). No corrections were made to the texts, as all statements were consistent with evidence discussed in the reviews. After the Japanese translation was completed, the content of the texts was reviewed by a pediatrician and a breastfeeding researcher in Japan who were not part of the research team. They confirmed that the texts were relevant for nurses in Japan. The final versions of the back-translated texts are shown in Supplementary Material 1.
The control group (i.e., Group 3) received a brochure on healthy pregnancy and delivery developed by the Ministry of Health, Labour and Welfare (22). The brochure was published online on the Ministry’s website for free use. The authors chose it as a control text because it did not mention infant feeding and was about the same length as the intervention texts.
Measurement
Attitude toward breastfeeding
Attitude toward breastfeeding was measured by the Iowa Infant Feeding Scale-Japanese version (IIFAS-J) (23)—a Japanese version of the Iowa Infant Feeding Attitude Scale developed by de la Mora and Russell in the United States (24). The IIFAS-J was tested for reliability and validity among 673 mothers, and one item was omitted from the original 17-item version (23). The IIFAS-J thus consists of 16 items rated on a 5-point Likert-type scale (1 = strongly disagree to 5 = strongly agree). The total IIFAS-J score ranges from 16 to 80; lower scores indicate a more positive attitude toward formula feeding, while higher scores indicate a more positive attitude toward breastfeeding (23). In this study, the Cronbach alpha values for pre-test and post-test surveys were 0.78 and 0.80, respectively, indicating that the scale was sufficiently reliable.
Nurse reactions to the texts
The reaction of participants to the texts was assessed by their response to the three statements: “I can agree with the content of the text,” “The text makes me uncomfortable,” and “I'm interested in the text.” For all items, participants were asked to respond by using a 5-point Likert-type scale, and responses were dichotomized after checking the distribution of answers (Supplementary Table 1). The responses “strongly agree” and “agree” to the first statement were categorized as agreement with the text; all other responses were categorized as disagreement with the text. The responses “strongly agree,” “agree,” and “neither” to the second statement were categorized as comfort with the text; all other responses were categorized as discomfort with the text. The responses “strongly agree” and “agree” to the last statement were categorized as interest in the text; all other responses were categorized as lack of interest in the text.
Background characteristics of participants
The pre-test questionnaire assessed the sociodemographic and employment characteristics of the nurses. In addition to age, gender, total number of years of work experience as a nurse, possession of an advanced license in addition to the nursing license, education level, having a child (children), and recognition of the benefits of breastfeeding and risks of formula feeding were measured (25-27).
Sample size estimation
A previous study of pregnant women reported a mean IIFAS-J score of 62 for those who intended to breastfeed exclusively. We aimed to increase the IIFAS-J score by 3 points by using benefit language to present the latest infant feeding information. We expected that use of risk language to present the same information would not have an equivalent effect. Specifically, we assumed that the control group would have a mean score of 62 at both the pre-test and post-test and designed the intervention for Group 1 to increase the mean score from 62 to 65 (SD 5). To detect a difference in pre-test score between Group 1 and the control with 80% power and a two-sided alpha of 5%, we estimated that a minimum sample size of 44 for each group would be necessary (http://clincalc.com/Stats/SampleSize.aspx). To account for the possibility of drop-outs, we enrolled a total of 147 participants.
Data analysis
Four assumptions were tested: 1) the texts using benefit language and risk language would increase IIFAS-J scores, 2) the improvement in IIFAS-J score would be greater for Group 1 (i.e., those who read the benefits text) than for the other groups, 3) benefit language would be received more favorably than risk language, and 4) nurses who reacted favorably to a text would have a more positive attitude toward breastfeeding in post-test measurement than did those who read the risks text. Analysis of Variance (ANOVA) or the chi-square test of independence was used to assess between-group differences in nurse characteristics. The paired t-test was used to compare mean differences in pre-test and post-test IIFAS-J scores within each group. ANOVA was used to compare mean IIFAS-J scores among groups at baseline and after the intervention, and change in IIFAS-J score after the intervention. The chi-square test of independence was used to compare how nurses in the three groups reacted to the texts. The t-test was used to assess the association between receiving the text favorably and post-test IIFAS-J scores. Stata version 13.1 was used to analyze all the data (Stata Corp. LLC, College Station, Texas, USA), and a p-value less than 0.05 was considered to indicate statistical significance.