The study population consists of a population-based sample of parents with children aged 0-2 years in the German-speaking part of Switzerland. The birth registries of the City of Zurich and small municipalities in the same region randomly selected 2573 mothers who had given birth in the past 24 months and provided their names and addresses. Urban and rural municipalities were included to represent the urban/rural distribution of Switzerland (75%/25%).
Between January and May 2018 parents received an invitation letter with a link to an online questionnaire; together with the second reminder letter a paper questionnaire and a prepaid envelope were provided. Overall, 842 individuals responded to the survey of which 73 had to be deleted in the data cleaning process. Reasons for exclusion were empty questionnaire (N= 31), missing answers to key questions (N= 40), non-plausibility of key questions (N=1), and double entry (N=1).
A total of 769 questionnaires were completed, which represents a response rate of 30%. 429 participants (56%) responded online, 340 (44%) on paper. Samples were compared regarding the general use of digital media for information on child and adolescent health, main outcome, and the percentage of children with disability, as well as socio-demographic variables. Samples differed significantly by age, gender and socio-economic status, but not in the main outcome variable nor proportion of children with disability (Supplemental table 1). Online and paper samples were merged for this analysis.
The cantonal ethical commission Zürich, responsible for the caption area in which this study was performed, confirmed the exemption from a full ethics review based on the Swiss Federal Act on Research involving Human Beings (Business Administration System for Ethics Committees (BASEC) Req-2017-00817), due to the fully anonymized data collection and anonymized health-related data.
The questionnaire covered different topics, from socio-demographic and health status of survey participant and child, the use of information resources for child general development and acute child’s illness, health information seeking, as well as e-health literacy and attitudes towards online health information of survey participant (not included in this analysis). The questionnaire was developed for this survey. Socio-demographic and health questions were taken from previous surveys on child health (Swiss infant feeding Study, https://www.zhaw.ch/de/gesundheit/institute-zentren/igw/forschung/kinder-und-jugend-public-health/ ; German Health Interview and Examination Survey for Children and Adolescents, KiGGS ), as well as newly constructed based on literature and paediatric consultation. E-health literacy was measured with the eHealth Literacy Scale (eHEALS) [30,31]. Digital information seeking, trust and understanding questions are based on survey items from the Flash Eurobarometer 404 on European citizens’ digital health literacy  and Wainstein et al (6)), respectively. The full questionnaire included 68 questions or items (see Harvard Dataverse: https://doi.org/10.7910/DVN/JI9GIJ.).
Health-information seeking behaviour
Our main outcome of interest is the frequency at which parents use the information resources “digital media” for two different targets: either for seeking information on “general health and development” or on an “acute child’s illness”. Digital media presented in the questionnaire were social media (e.g. Facebook), websites for parents, apps on mobile devices, search engines, websites of paediatricians or children’s hospitals and official websites of health services or health organizations, and an “other” option. Open responses were checked it they could be grouped under one of the listed options, which was the case in all 32 “other” digital sources. Secondary outcomes are the frequency at which parents use the information resources “print media” (books, journals, newspapers and other print media) and “personal contacts”. The latter category included different persons and consultant, with whom a direct, personal contact and response to the individual need is assumed: paediatrician, other health professional, telephone consultation of a children’s emergency service or a children’s hospital, telephone consultation of a health insurance, family members and friends, acquaintances or neighbours.
Frequency scores for digital media summarize all six items, for print media all four items and for personal contact all six items mentioned above. Each item had five response options: never; rarely; sometimes; frequently; very frequently coded 0 (never) to 4 (very frequently). To calculate the sum scores, responses by type of resource were added and standardized to a scale from 1 to 100 to allow a comparison of the frequency between the three information resources.
Parents were further questioned if they had searched for health information before or after their last visit to the paediatrician and if yes, for which reasons. A list of reasons was proposed: I have received too little information from the paediatrician; information from the paediatrician was incomprehensible or contradictory; I needed to check information from the paediatrician; I had the need to exchange with others or to search for experiences and tips; I was looking for other therapies. For each, parents could answer: does not apply at all; does slightly apply; does partly apply; does apply; does apply very much. We created a binary variable summarizing the last three options into the category “does apply” and the first two into “does not apply”.
Child characteristics: Child’s age is reported in months, sex as a binary variable and birthweight in grams. Gestational age is reported in weeks and parity indicates whether a child is the first born from her/his parents. Disability was defined as a binary variable based on the parents’ reporting of either a physical impairment (e.g. malformation), developmental delay, hearing or visual impairment or congenital disability.
Respondent characteristics: We distinguish between three roles of the respondents: Mother, father and other. In the regression analysis, a binary measure for parent’s role is used, excluding “other” respondents (n=4). Respondents’ age is reported in years. Education is reported as no or compulsory education (max. 9 years of education), upper secondary education (e.g. apprenticeship or high school degree) or tertiary education (university or similar degree). Net household income is measured in five categories: less than 4500 Swiss francs (CHF); CHF 4500-6000; CHF 6000-9000; more than CHF 9000 and no indication / don’t know. Citizenship is used as a binary variable distinguishing between Swiss and Non-Swiss. Parents reporting a double citizenship were categorized as Swiss.
First, we ran a descriptive analysis for the overall sample as well asd stratified by the child’s disability status (chi2-tests for categorical variables and independent-samples t-tests for numerical variables). Second, in case of the health information seeking target “acute child’s illness”, we further investigated the time point of the digital health-seeking behaviour and the reasons for doing so by disability, applying chi2-test. Third, we examined the standardized frequencies by type of health information seeking target in parents who provided information on both targets. For each information resource (digital media, print media and “personal contacts”), we compared the frequency of use for general health and development and acute child’s illness by applying box-plot analyses and paired-samples t-tests. We also reported the frequency of single items of information resources used by type of health-seeking target (independent-samples t-test) and disability status (paired-samples t-tests).
Fourth, we carried out ordinary least square (OLS) regression analyses for the primary and secondary outcomes for both health information seeking targets. The following confounders were included in the model: child’s sex, child’s age, parental age, parental sex and parental education.
We ran three sensitivity analyses with and without the following variables: First, due to many missing values for parents’ age, we excluded this variable. Second, we included net household income. Third, sensitivity analyses on the same models but on a sample restricted to parents who had answered the survey questions about both health information seeking targets instead of only one (N=480 for digital media; 419 for print media; 597 for personal contacts). All analyses are carried out with the statistical software STATA/SE 15.1. Statistical significance was established at P<0.01, to take account of potential multiple comparisons.