Design and population
We conducted a survey study from December 2015 to January 2016 among individuals in Denmark, the Netherlands, and Switzerland using hypothetical case scenarios. The study formed part of a project undertaken by the European Research Network for Out-Of-Hours Primary Health Care (EurOOHnet) [14]. In addition to the present paper, two other papers have been written: one on clinical out-of-hours help-seeking behaviour in Denmark, the Netherlands and Switzerland [13] and one on the impact of alternative healthcare plans on intended out-of-hours help-seeking in Switzerland [15].
We included individuals of three age groups: parents of children aged 0–4 years, adults aged 30–39 years, and adults aged 50–59 years. We chose these age groups because a previous study found differences in the use of out-of-hours care between Danish and Dutch young children and young adults [16]. We added the age group 50–59 years to include a broader range of age groups. We aimed to get 600 respondents per age group per country, to gain enough power for one of the other studies [13]. Due to different data collection methods and expected variations in response rates, we selected a different number of individuals for each country. For Denmark, we used the Danish Civil Registration System, which holds information on all Danish individuals, to select 1200 individuals per age group. Individuals living in institutions and individuals with unknown address were excluded. For the Netherlands and Switzerland, a nationally representative consumer panel was used for each country. For the Netherlands, we used the consumer panel of TNS NIPO, a professional organisation for market research, to select 950 individuals per age group. This consumer panel consists of a representative group (over 200,000 members) of citizens (www.tnsglobal.com, 2017). For Switzerland, 6093 representative German-speaking members of two consumer panels (Respondi and Bilendi) were used to select 600 respondents for the two adult age groups.
Setting
The organisation of the healthcare system in Denmark and the Netherlands is quite similar. Almost all Danish and Dutch citizens are listed with a GP, who acts as a gatekeeper to secondary care. Outside office hours, a GP cooperative can be contacted by telephone. Self-referral to the ED is possible, but is generally discouraged as it is mostly preferable to first contact primary care. Primary care is free of charge in both daytime and outside office hours. An ED visit is free of charge in Denmark, whereas residents in the Netherlands must pay an annual (tax-deductible) fee of at least EUR 375 (2015 figures).
In Switzerland, patients can freely access the ED and specialist care. However, patients may choose another healthcare insurance plan, which reduces the costs but also obligates them to first contact a gatekeeper (for example a GP) for healthcare. The organisation of out-of-hours care varies between the regions. GP and emergency care is covered by the mandatory health insurance plan, except for an annual (tax-deductible) fee of at least CHF 300 (approx. EUR 275) and 10% co-payment [15].
Questionnaires
We developed two questionnaires for the study: one for parents of young children and one for adults. Both questionnaires consisted of predefined cases describing situations involving specific symptoms and signs of disease. After the cases, respondents were asked to answer questions concerning factors related to help-seeking behaviour based on Andersen’s Behavioural Model. The cases for parents and adults differed, but all cases described situations that could prompt a need for acute healthcare. All cases involved frequently occurring health problems at different levels of urgency. An English version of the questionnaires are presented in Additional files 1 and 2.
Development of case descriptions
To ensure that the presented cases constituted sufficient content validity, the development process consisted of several steps. We selected previously used cases from other studies [17,18,19] and added new cases at different levels of urgency inspired by common reasons for encounter in the three included countries. Each case described a situation, including a specific weekday and time. For cases involving children, we stated a specific age of the child as even small age differences in this group can change the intended help-seeking behaviour in the parents (even for the same illness). For cases involving adults, we did not state a specific age, but we gave an age range (30–39 years or 50–59 years) to ensure that the respondents were able to see themselves in the described situation. The cases underwent several feedback cycles (both face-to-face and by email) with researchers, GPs, and laypersons. Finally, we ended up with 20 cases concerning children and 32 cases concerning adults.
To get an overview of the urgency levels of the cases and to check the representativeness and clarity of formulations, an expert panel of 29 GPs reviewed the cases. These GPs had to meet the following inclusion criteria: ≥ 2 years GP, ≥ 6 out-of-hours shifts per year, coming from different regions within the countries, and fair knowledge of English. Cases classified as ‘unclear’ according to the expert panel were excluded. In a research meeting, we selected 11 cases concerning children and 13 concerning adults with different levels of urgency. The included cases were translated from English into Danish using a backward-forward translation procedure and a consensus meeting [20]. The cases were sent to 150 Danish individuals per age group and tested for variations in help-seeking behaviour. We performed a Rasch analysis and selected cases across the whole range, and cases without response variation were excluded. This resulted in a final selection of cases representing varying responses; six cases for children and six cases for adults.
Outcome measure: Intended help-seeking behaviour
The six cases were used to measure our outcome measure “intended help-seeking behaviour outside office hours”. For each case, we dichotomised the individual responses concerning intended behaviour into “Yes/1” and “No/0” categories: ‘Contacting out-of-hours care’ (‘Contact out-of-hours primary care’, ‘Contact the ED’, ‘Contact 112/144 ambulance care’) and ‘Not contacting out-of-hours care’ (‘Wait-and-see’, ‘Self-care’, ‘Ask my partner, a relative, or others for advice’, ‘Check a guidebook, the internet, or an app’, ‘Contact my own GP on the next working day’). Intended help-seeking behaviour was estimated by combining the dichotomised scores of the six cases for each respondent.
Theoretical framework and development of model
The study was guided by Andersen’s Behavioural Model of Health Services Use [10], which defines population characteristics (predisposing characteristics, enabling resources, and need), health behaviour, and outcomes that may affect the use of health services (Fig. 1).
The following predisposing characteristics were included: age, gender, education level, medical education, ethnicity, work status, living status, number of children (for parents of children aged 0–4 years), social support, health literacy (navigating the system and finding information), self-efficacy, anxiety, and attitude towards use of out-of-hours primary care. The following enabling factors were included: travel time, problems with planning, organising childcare (for children), and accessibility and availability of own GP. We included one need factor (self-assessed health of adult and child), two behavioural factors (frequency of contacts with own GP and frequency of contacts with out-of-hours care), and one environment factor (country). For some of the determinants, the following validated questionnaires were used: Generalized Anxiety Disorder scale (GAD-2) [21], General Self-Efficacy Scale (GSE-10) [22], two scales from the Health Literacy Questionnaire (HLQ) [23], and the self-assessed health item from the 36-item Short-Form Health Survey (SF-36) [24]. Questions from previous studies were used; sometimes in adjusted form (i.e. on education, employment [25], and social support [26]).
We also added newly developed questions (medical education, living status, attitude towards use of out-of-hours primary care, and perceived problems). For the Netherlands and Switzerland, standard questions for the internet panels were used (age, gender, education, and employment). The data preparation of these factors is described in Additional file 3.
Interviews and pilot study
The readability and feasibility of the original Danish version of the questionnaire were tested in two steps. First, cognitive interviews with eight patients from one GP practice were conducted to see if they understood the questions. Second, we performed a pilot study by sending the questionnaire to 50 Danish individuals per age group, including one reminder. The pilot study resulted in minor layout adjustments and showed good feasibility with a response rate of 38% for children, 28% for adults aged 30–39 years, and 50% for adults aged 50–59 years.
Data collection
The Dutch and Swiss questionnaires were each translated from the Danish source text by using the backward-forward procedure and a consensus meeting [20]. The Danish individuals received a paper questionnaire in January 2016 with the option to complete the questionnaire online, and a reminder was sent three weeks later. The Dutch consumer panel members received an email with a link to the questionnaire in December 2015, and a reminder was sent for age groups 0–4 and 30–39 years (the aimed response rate was met for age group 50–59 years). The data collection stopped after one week as the minimum of 600 respondents had been reached for all groups. The Swiss consumer panel members were invited by an email link in December 2015; all 6093 individuals in the age groups 30–39 and 50–59 years were contacted. The data collection stopped after five days as the minimum of 600 respondents per age group had been reached. The datasets received from the consumer panel organisations included only anonymous data. The Danish respondents participated in a draw for three sets of two cinema tickets, whereas the Dutch and Swiss consumer panel members each received a small financial compensation as a standard procedure.
Statistical analysis
We checked the representativeness of our data. For Denmark, we compared respondents with non-respondents for age, gender, region, education level, ethnicity, living status, and employment as our sample was selected randomly from the entire population. For all countries, we compared respondents with the general population (age, gender, region, education level, ethnicity, living status, and employment) using 95% confidence intervals (CI).
All analyses were done separately for children and adults (adults consisted of two age groups). Descriptive analyses were used to show the distribution of factors affecting help-seeking behaviour. Two multiple binomial regression analyses were performed to assess the influence of all factors on the inclination to contact out-of-hours care (one for parents and one for adults). Odds ratios (ORs) were calculated and presented in forest plots with 95% confidence intervals. For all analyses, we combined data from all participating countries (Denmark and the Netherlands for cases based on children; all three countries for cases based on adults). We performed the binomial regression analyses separately for each country to check the robustness of our data (data not presented). All analyses were conducted using Stata 14.2 (StataCorp LP, College Station, TX, USA).