Initially, we defined four specific objectives for the questionnaire: (1) Quantify the amount of information given to ICD patients on driving restrictions following ICD implantation and/or ICD shock; (2) Investigate whether ICD patients adhere to driving restrictions and which factors are associated with adherence to driving restrictions; (3) Determine what proportion of Danish ICD patients have experienced an ICD shock or cardiac symptoms of possible arrhythmia while driving, and whether these symptoms resulted in a motor vehicle accident; and (4) Identify whether driving restrictions influence factors associated with patient QoL.

Based on expert opinion and following review of the existing literature on driving following ICD implantation, an initial conceptual framework was developed which was successively refined in qualitative semi-structured focus group interviews. In summary, through conceptualization, specific variables were defined which could subsequently be transformed into items, thus connecting the research objectives with the content of the questionnaire. Additional file 1: Table S1 presents our research aims, hypotheses, concepts and their corresponding variable definitions as well as questionnaire item numbers. For example, the concept “Information about driving restrictions after ICD implantation” resulted in three variable definitions: (1) Whether the patient held a valid private driver’s license (Group 1: car, motorcycle, tractor) during the 6 months leading up to ICD implantation; (2) Whether the patient held a valid Group 2 driver’s license (Group 2: truck, bus or any vehicle for passenger transportation) during the 6 months leading up to ICD implantation; (3) Whether the patient was informed about driving restrictions following ICD implantation (for both Group 1 and Group 2) by health personnel (doctors, nurses and ICD technicians) during the hospitalization for ICD implantation.

A few concepts and variable definitions merit further elaboration: First, in order to compare questionnaire responses with current guideline recommendations, we chose to define private and professional driving based on the definitions in the European Heart Rhythm Association’s recommendations for driving with ICDs and adapted by the Danish Society of Cardiology [5, 6]. For our aim of investigating the driving restrictions’ influence on QoL factors, we were inspired by the World Health Organization’s six proposed domains of QoL [14]. From the 24 facets of QoL proposed, we predicted that 10 could be affected by driving restrictions: negative feelings, positive feelings and self-esteem (psychological domain); mobility, activities of daily living and work capacity (level of independence domain); personal relationships and social support (social relationship domain); participation in and opportunities for recreation/leisure and transport (environment domain).

To uncover unknown concepts related to driving following ICD implantation, we performed three focus group interviews with ten contemporary Danish ICD patients (20% female, median age 62 years, 50% primary prevention ICD indication). The participants were recruited from the outpatient clinic at a university hospital in The Capital Region of Denmark and purposive maximum sampling was used to achieve variation in pre-specified variables, including sex, age, ICD indication, previous ICD therapy and geographic residence. Briefly, the qualitative setting allowed ICD patients to elaborate on their perceptions and experiences with driving and the driving restrictions. Further, we could observe the social and cultural norms within groups of ICD patients and thus gain essential knowledge on what wording to use and how to approach certain delicate questions, such as adherence to driving restrictions [15].

We followed general technical rules for valid questionnaire design. Initially, the items were grouped into themes using a strategic order, taking into account spill over effects and placement of sensitive questions. Thus, the first questions were considered easy, whereas complexity increased throughout the questionnaire. Respondents were asked when they had resumed driving after ICD implantation or ICD shock, before the questions about what information on driving restrictions they had received. Intentionally, direct questions about adherence to driving restrictions were placed at the end of the questionnaire. Since not all questions were relevant for everyone, e.g. the majority of ICD patients have not experienced an ICD shock, branching methods were applied to guide the respondents through the questionnaire. For instance, following a few background questions (about self-assessed health, educational attainment, and employment status), respondents without a valid driver’s license at time of ICD implantation were guided to the end of the questionnaire. Lastly, due to the sensitive nature of some of the questions, we chose to make all items voluntary including a possibility of skipping questions in the web-based questionnaire.

Question formulations were kept short and words with connotations were excluded. Help texts were included if deemed necessary. In order to minimize recall bias, we defined four distinct recall periods (which were confirmed realistic during focus groups) depending on the objective of the question: “the six months leading up to ICD implantation”, “at the time of ICD implantation”, “in periods with driving restrictions”, and “during the previous month.” Text including recall periods was underscored. Response categories were tailored to the question type: for factual questions, we particularly aimed for exhaustive and mutually exclusive response categories, whereas the focus for opinion questions was to achieve balance in the response categories. For potentially sensitive questions, including questions about driving behaviour and the driving restrictions’ influence on factors related to QoL, we chose to express questions as opinions with corresponding Likert scale response categories (strongly agree to strongly disagree) [16], as opposed to factual questions with Yes/No response categories. Likert scale variation was kept minimal with either three- or five-point scales. Likewise, open response categories were generally avoided, but included in a few “other” response categories. Furthermore, the respondents were given the opportunity to leave comments after selected sections of the questionnaire.

The first method used in evaluating the questionnaire was an informal, individually-based expert review. Four participating cardiologists or cardiology fellows (ACR, MS, JP and LS) and one device technician (JDP) each independently conducted a review and determined whether he/she found the items sufficient relative to the questionnaire’s aims, or if a questionnaire item was problematic. Following adjustments based on these evaluations, the questionnaire was reviewed by a senior questionnaire expert (JC), independent from the study group. Adjustments following this expert review included minor modifications like introducing help texts to emphasize different recall periods, using continuous response categories for questions on time to resumption of driving following ICD implantation and/or shock, and strongly urging respondents to complete the questionnaire with a next of kin.

Further examination of content validity and refinement of items was done by cognitive interviewing of 28 ICD patients in the outpatient clinic at all six ICD implanting centres in Denmark (range: 4–6 patients/centre). We chose to include ICD patients from all five regions of Denmark to capture any geographic variance potentially influencing questionnaire responses (Capital Region: 8 participants; Zealand Region: 4 participants; Southern Denmark: 5 participants; Central Jutland: 6 participants; Northern Jutland: 5 participants). On a given day, ICD patients visiting the outpatient clinic were approached in a random manner and were excluded only if they did not have a valid driver’s licence at time of ICD implantation. A retrospective verbal probing approach was used, and the probe questions were asked following completion of each of the four sections of the questionnaire. Probes were both scripted (e.g. “What does the term ICD shock mean to you?”) and spontaneous (e.g. “I noticed that you hesitated. Can you tell me what you were thinking?”). We utilized Tourangeau’s 4-stage cognitive model, investigating: (1) comprehension of the question; (2) retrieval of information (recall strategy); (3) decision processes; and (4) response processes [17]. Another significant focus in these cognitive interviews was recognizing any reluctance in answering questions about adherence to driving restrictions and further, to identify what wording to use to make respondents trust that information on non-adherence would remain anonymous.

Head-investigator (JB) performed all the cognitive interviews over a period of 16 weeks as an iterative process with alterations of the questionnaire items where necessary after each interview session. The initial five rounds of cognitive interviews were conducted using the web-based questionnaire, while the patients in the last round of interviews were presented with the paper format questionnaire. After the cognitive interviews at each implanting centre were completed, JB reviewed and summarized the results and determined whether, for each tested item, significant problems had been detected. Throughout the interviewing process, the frequency of problems encountered per patient interviewed, declined (Additional file 2: Table S2).

The nationwide Danish Pacemaker and ICD registry was used to identify all individuals with a first-time ICD implantation in the period January 1, 2013 to November 30, 2016. Among these, 50 individuals were randomly selected to participate in the pilot test and were subsequently matched with up-to-date address data from the Danish Civil Person Register. Invitations to participate in the study were mailed on February 23, 2017. Participants were urged to complete the web-based questionnaire, however, an option to request a paper version was presented (Additional file 3, translated from Danish). Following 3 weeks, a reminder was mailed to all non-responders including a paper version of the questionnaire and a prepaid return envelope. The web-based questionnaire software SurveyXact, developed by Rambøll A/S and approved for research purposes by the Capital Region of Denmark and the Danish Data Protection Agency, was used for data collection [18]. The pilot test was terminated on April 13, 2017, after 7 weeks’ data collection. All paper responses were entered into the web-based questionnaire software by double manual data entry to reduce the consequences of potential human errors.

To investigate the reproducibility of the questionnaire items, we invited the first 25 pilot test participants who responded to the questionnaire to complete the questionnaire again at 3 weeks after their first response. This timeframe was deemed appropriate as to avoid memory effects positively influencing the test-retest reliability. No reminders were distributed to test-retest participants.

Results of the pilot test were summarized by reporting responses on selected questionnaire items by available case analysis, thus, the number of responses to each question may vary. For descriptive analyses, categorical variables are reported as percentages and continuous variables are presented as medians with interquartile range. Test-retest agreement was investigated by kappa coefficients or weighted kappa coefficients for categorical variables and intraclass correlation coefficients for continuous variables. All analyses were performed using SAS (version 9.4, Cary, NC, USA).

All concepts already identified from the literature were confirmed in the focus groups interviews, excluding the effect of driving restrictions on self-esteem. In addition, we discovered that many of the participants had changed their driving behaviour in some way after ICD implantation: For example, some patients reported being extra cautious when driving with children (e.g. driving at slower speed, avoiding the overtaking lane on highways) or avoiding highways altogether. Also, a significant proportion of primary prevention patients called for more information about potential driving restrictions in case of future ICD therapy and also being explained the rationale behind the driving restrictions, as they saw these as promoting factors for compliance with the restrictions. Generally, in an anonymous setting, the participants willingly volunteered information about not adhering to the driving restrictions.

Among the 50 ICD patients (9 women (18%)) invited to participate in the pilot test, 30 (60%) either responded electronically or requested a paper version of the questionnaire within 3 weeks. Following only one reminder, 39 participants (78%) completed the questionnaire (Fig. 2). Almost all web-based responses (85%) came within 1 week. Median time for completing the questionnaire was 9.5 min for the web-based responses (IQR: 5.2–13.7 min).

Of the responders, 37 (97%) held a valid private driver’s license for car, motorcycle or tractor (Group 1 drivers) prior to implantation and 11 (30%) held a valid driver’s license for large vehicles or any professional driving (Group 2 drivers) (Fig. 3). However, only two responders had actively used their professional license during the 6 months prior to ICD implantation – both were truck drivers. The most common occasions (> 75%) for driving prior to ICD implantation included practical errands, visits to family and friends and in relation with leisure activities, while eight (22%) drove during work hours (Table 1).

Only half of the respondents remembered being informed of any driving restrictions for private driving. Of these, two thirds reported they had solely been informed verbally (Table 2). The information given by health personnel on driving following implantation as recalled by the patients ranged from resuming to drive immediately after ICD implantation to never to drive again. Still, patients were overall satisfied with the communication of the driving restrictions. Among the responders actively driving professionally prior to ICD implantation, only one (50%) could remember being told never to drive professionally again. Of the ICD patients having experienced an appropriate ICD shock, three out of four remembered having been verbally informed not to drive for 3 months.

Following ICD implantation, 35 (95%) of responders with a driver’s license resumed private driving whereas professional driving was resumed by one responder (Fig. 3). Overall median time to resumption of driving was 14 days (interquartile range: 2–60) while professional driving was resumed at 8 months. Among patients being instructed of a minimum 1-month driving ban, either after ICD implantation or ICD shock, three patients (25%) admitted they had been driving although restricted. None had experienced an ICD shock or loss of consciousness while driving, but two patients (6%) had experienced dizziness and four patients (11%) had experienced palpitations while driving. None of these events necessitated stopping the vehicle or resulted in a motor vehicle accident.

Overall, patients had not altered their driving behaviour after ICD implantation. Only four patients (12%) were afraid of having an ICD shock while driving, and only one patient was nervous about driving or tried to avoid highways as a consequence of ICD implantation. None avoided driving alone or driving with children in the car. Among the 11 patients with a significant driving restriction (defined as ≥1 month), six (55%) stated that the driving restrictions in some way had impeded with their daily life and 45 and 33% responded that they had been very limited in maintaining employment and getting to/from work, respectively.

Of the 25 responders invited to participate in a test-retest questionnaire, 19 (75%) responded within 3 weeks. Kappa statistics found that 69% of the analyzed questionnaire items had substantial agreement, with kappas or weighted kappas above 0.6, whereof 37.5% were above 0.8. (Additional file 4: Table S3). Generally, the kappas with moderate (0.4–0.6) and fair agreement (0.2–0.4) were found in the section on the driving restrictions’ influence on factors associated with QoL. In one item, the only test-retest response was reverse, rendering a kappa of 0. The intraclass correlation coefficient for time to resumption of private driving following ICD implantation and appropriate ICD shock was 0.98 and 1.00, respectively, whereas the time to resumption of professional driving could not be analyzed as the item only had one test-retest response.

Questionnaire studies pose some significant inherent limitations. Many relate to poor question design (e.g. problems with wording, leading questions, scale formats) or questionnaire design (e.g. formatting problems, too long or complex questionnaire) which we have tried to overcome by thorough pre-testing, but some questions will inevitably be misinterpreted. Moreover, though the focus group participants were generally good at recalling specific information from the time of ICD implantation, recall bias cannot be excluded. Also, social desirability may contribute to respondents giving untrue answers, particularly to the questions related to adherence, although focus group and cognitive interviews demonstrated the contrary. Lastly, we naturally only included patients who were alive at the time of questionnaire administration, introducing a healthy participant bias which, similarly to the non-response bias, covers the fact that responders (or survivors) could differ significantly from the non-responders.

The questionnaire was developed specifically for use in a nationwide cohort of Danish ICD patients and the final questionnaire has been administered to more than 3900 individuals. Due to the unique Danish administrative registries, we will be able to link the final questionnaire results with reliable clinical and demographic information, including indication for ICD implantation, socio-economic status, and comorbidities. Importantly, linkage with the nationwide registries also allows comparison between responders and non-responders, a major strength in a questionnaire study.

In this pilot study, however, the results have not been linked with relevant clinical data and we were therefore not able to stratify results on important clinical information such as cardiac resynchronization status, pharmacological treatment and comorbidities such as diabetes. All of these will be relevant to include in the final questionnaire study when investigating factors associated with the risk of ICD therapy while driving, as they have been found to affect the risk of adverse clinical outcomes, including ICD therapy [21, 23‐25]. Other limitations of the current study include lack of information on heart failure biomarker levels and home monitoring status which are also expected to affect clinical outcomes, especially in primary prevention patients with a cardiac resynchronization device [26‐28].