Instrument development

The academic research team worked with private practitioners from Reckitt Benckiser Group plc (RB) to develop the survey. RB is an international company that produces cleaning products, and so this research fits the Global Public–Private Partnership’s collaborative model. Early on, it was determined that the survey items should be statements that participants could express their agreement with using Likert scales. The initial items were informed by Huijg et al.’s (2014) validated template survey of the Theoretical Domains Framework’s domains [27]. For example, an item to assess knowledge read, “I know that my children should wash their hands with soap and water for at least 20 seconds.” In phase 1, many items were created as poor items could be removed later [28].

Two sets of items were developed in phase 1, see Supplemental Materials 1. The primary set (N = 50 items) was designed to measure the behavioural factors that influence parents’ encouragement of children’s handwashing. The second set (N = 28 items) was designed to measure the behavioural factors that influence surface cleaning. Each set assessed 12 of the 14 theoretical domains. The Optimism and Reinforcement domains were excluded because the items developed for these domains aligned better conceptually with the definition of the Beliefs in Consequences domain. The Intentions and Goals domains were combined, because the items developed for these domains were often about intentions to achieve a goal. Each domain included at least one negatively worded item. Each handwashing domain contained at least four items, and each surface cleaning domain contained at least two items. All items were originally written in the English language and then translated into Hindi for participants in India. The translations were initially conducted by native-level language speakers at Opinion Health. Opinion health is a company with over 50 years of experience conducting market research globally [29]. The item translations were checked for accuracy and accessibility by native-language speakers at RB.

In January 2020, a pilot study was conducted with 100 participants who identified as mums or dads of at least one child 5 to 10 years old (inclusive), 50 from the UK and 50 from India. The survey was disseminated through Opinion Health’s survey panel, which anyone with a valid email address can join by submitting an online form [29]. Participants indicated their informed consent before participating. The items were presented in a non-random order, and participants expressed their agreement using a five-point Likert scale, in which only the end items were labeled, from “strongly disagree” to “strongly agree.” Demographic information was also collected about participants’ gender (male, female, or other/prefer not to say) and their children’s ages. The survey was set up such that participants were required to answer all items and were compensated for their time with the equivalent of one British Pound in their nation’s currency.

The analysis of phase 1 data was conducted to identify items most likely to provide valid measures for each theoretical domain. The identified items would be retained in the final draft survey. We sought to retain three handwashing items for each domain and two surface cleaning items for each domain. Data were analysed in SPSS v.26. Negatively worded items were reverse scored. Descriptive statistics (frequencies and medians) were used to summarise participants’ gender and their children’s ages. Then, item data were considered for the variability of responses, skewness, kurtosis, and internal consistency.

Next, a parallel version of the retained items was created for teachers by adjusting relevant words. For example, an item designed to measure memory attention and decision making read the following for parents, “I forget to remind my children to wash their hands” and read the following for teachers “I forget to remind my pupils to wash their hands.” Lastly, all items were translated from English into the most predominant language of five non-English speaking countries (China, India, Indonesia, Saudi Arabia, and South Africa) by native-level speakers from each country at Opinion Health checked for accuracy by native-level speakers from each country at RB. During these translations, we aimed to make the minimal adjustments necessary to retain each item’s semantic meaning.