Helmet (Health impact of e-bikes and e-scooters) study: Data collection methods and information gathered for the evaluation of the introduction of share hire schemes

Patient and Public Involvement

The Applied Research Collaboration West (NIHR ARC West) facilitated the recruitment of a combination of e-bike and e-scooter users and non-users for the focus group held during the planning phase of this project. This informed our understanding of what was important to both users and non-users when designing the larger project. We also had a Patient and Public Involvement (PPI) member embedded in the research team who provided important input during research team meetings and advised on documentation, especially that which was publicly facing.

We worked with members from the transport divisions of the local authorities to choose and design the proposed data collection methods for this project. For example, some data collection methods that were originally suggested were deemed likely to be less successful and therefore were not included in the final project. With respect to survey development, local authority members from the Sheffield City Council, Bristol City Council, and Plymouth City Council provided input to improve it. These included adjustments to word choice, restructuring of the landing page, and adjustments to some of the questions.

Study design and setting

This study employed a repeated cross-sectional design. Although the data include 1316 repeated measures, this study focused on a cross-sectional comparison. The full protocol was registered in the Open Science Framework (https://osf.io/gq9s8/, DOI 10.17605/OSF.IO/GQ9S8). Reporting follows the STROBE checklist for cross-sectional studies where applicable. We collected baseline data during August and September 2023 prior to the implementation of the two e-bike share-hire schemes. Resurvey data were collected a year later between August and the first week of October 2024. The main project included four major English cities, with baseline and resurvey data. Leeds had no share-hire scheme at baseline, but launched an e-bike share-hire scheme (EB) in mid-September 2023 (baseline data collection was stopped the day before the launch of the scheme). Bristol had an e-scooter share-hire scheme only at baseline but launched a combined e-bike and e-scooter share-hire scheme (EB+ES) in October 2023. Bradford and Sheffield were non-intervention control areas without share-hire schemes. For triangulation purposes, we added two additional cities to the resurvey phase. These were Plymouth, which launched an EB scheme in late March 2023, and Bath, which had an e-scooter only share-hire scheme but extended it to an EB+ES scheme in late September 2023.

Survey data collection

Online survey. The baseline survey and resurvey, developed in consultation with council communication and transport teams from the four local authorities, contained 32 questions for the baseline survey and 41 questions for the resurvey. The full questionnaire information can be found in the protocol (https://osf.io/gq9s8/, DOI 10.17605/OSF.IO/GQ9S8). They inquired about demographics (age, gender, ethnicity, employment, income, and postcode), physical activity volume (self-reported minutes of moderate-to-vigorous physical activity (MVPA)), use of e-bikes or e-scooters, active travel and other travel, distance travelled by different transport modes, substitution of travel modes, quality of life, and access to venues and services. Written informed consent was obtained from all participants, as outlined in the ethics and consent sections below.

Baseline Survey Data Collection. Baseline recruitment targeted residents aged 16 years and over of the two intervention local authorities, Leeds and Bristol, and the two control local authorities, Sheffield and Bradford. The intervention and control sites were similar with respect to the proportion reporting very good health (50%, 51%, 47%, 46%) and income deprivation (14%, 14%, 16%, 19%)³¹.

The primary mode of distribution for our baseline survey was through advertisements with a link to our survey distributed via council-administered newsletters and/or mailing lists and council-administered social media. We supplemented this with secondary methods. First, paid-for Facebook advertising, which has extensive reach with 79% of UK adults who go online using Facebook and 45% using it on a daily basis³². Second, relevant local stakeholder groups and gatekeepers of relevant social media groups in each study area facilitated distribution to their members to improve reach to seldom heard voices. Third, we used a prize draw in each region as an incentive, as including entry into prize draws has been shown to improve response rates to surveys³³.

Resurvey survey data collection.During the resurvey, we included the original sites and extended them to collect resurvey data in two further cities: Plymouth (EB) and Bath (EB+ES) to triangulate data with Leeds (EB) and Bristol (EB+ES), respectively. For the resurvey, we employed the same survey data collection methods used at baseline (as outlined above). However, we found that engagement with the Bradford population at baseline was considerably more challenging than that at other sites, as evidenced by the lower survey response rate. Given that the Bradford site contains much higher ethnic diversity, it is important to increase its representation to ensure diversity and inclusion and allow a sufficient sample size to explore inequalities. Therefore, we partnered with CNET Bradford (https://cnet.org.uk/), a charity and company that supports the voluntary, community, and social enterprise sectors in Bradford, to engage the local community and improve response rates at resurvey. Additional channels included information events in communities with lower response rates at baseline, facilitated by trusted community advocates, and trusted advocates attending places the target population visited (e.g., supermarkets, libraries, youth groups) to promote the survey and offer support in completing the survey if required. As Bath had a much smaller population size than the other sites, we also supplied physical posters containing a QR code link to the online survey in public places to increase survey reach.

Sample size. For the purposes of the later full evaluation (NIHR163726; https://www.fundingawards.nihr.ac.uk/award/NIHR163726), we estimated the sample size required to detect a difference in self-reported MVPA between the evaluation arms with and without EB schemes of between 5% and 15%, based on 80% power and 5% significance level. These represent differences of approximately 15–50 min in total weekly MVPA, which are achievable and in line with published research for achieving health benefits³⁴. For example, a difference of 30 min is equivalent to an extra 10 min of MVPA three times a week. Originally, we used published data (M (mean) = 258 min, standard deviation (SD) = 214 min) to estimate a sample size of 1080 per arm required to detect a 10% increase in weekly MVPA (26 min), assuming normally distributed data. Using the observed baseline data collected in August and September 2023, we produced revised estimates (M = 305 min, SD = 257 min per week) for an analysis based on a lognormal distribution, which better reflects the highly skewed nature of the data. Sample sizes of 1400–2400 in each arm were calculated to be sufficient to detect a difference of 25–30 min of MVPA.

Survey data cleaning. To ensure the validity of the survey respondents residing in one of the six study sites (Bristol, Leeds, Bradford, Sheffield, Bath, or Plymouth) and to remove potential contamination effects, for example, Leeds respondents accessing the Bradford survey (or vice versa), participants were assigned a site based on the local authority of their home postcode. Respondents with postcodes outside the local authority areas were removed. Respondents who provided no postcode but reported that they 'lived in the city' remained allocated to the site based on the completed site survey.

Duplicate entries were classified as respondents who provided identical names, email addresses, or telephone numbers. All entries after the first (based on the time of survey completion) were removed if the number of duplicates exceeded two. For only two duplicate entries, data available in entry two but not in entry one were added to entry one, and the second entry was then removed.

Implausible demographic data values (considered for age>110 years, weight <30 kg or >400 kg and height <1.2m or >2.2m³⁵) were set to missing³⁶. Physical activity levels were grouped according to meeting (≥ 150 minutes per week) or not meeting (<150 minutes per week) the UK Chief Medical Officer Physical Activity Guidelines¹⁴. Respondents were assigned to the 2019 English Index of Multiple Deprivation (IMD) quintiles based on their home postcode³⁷. Finally, missing demographic data at resurvey for those with repeated measures were replaced with baseline data, with age increased by one year.

Survey data analysis. Categorical data are presented as n (%) and continuous data as mean (SD) if approximately normally distributed and median (IQR: Q1, Q3) otherwise. Data are presented for both the baseline and resurvey, where applicable. Basic descriptive analyses were performed to gain an understanding of typical e-bike and e-scooter use by each local authority. Sample demographic characteristics are described for all respondents at each site and overall. Demographics included: gender (male, female), age (16–34, 35–44, 45–54, 55–64, 65+), ethnicity (White, Black, Asian, Mixed, Other), employment (working (employed or self-employed part- or full-time), not working), education (degree level or above, below degree level), household income (<£7000, £7000–£19,999, £20,000–£29,999, £30,000–£39,999, £40,000–£59,999, £60,000+), IMD quintiles, household type (single, couple, family, other), health conditions (yes, no), physical activity level (meeting guidelines, not meeting guidelines), BMI (<25, 25–29.9, 30+ kg/m²), ever e-bike use (yes, no), and ever e-scooter use (yes, no). Sample demographic characteristics including gender, age, and health conditions were also presented for those reporting ever having used an e-bike/e-scooter previously (‘ever users’). It is important to note that the purpose of the data collection was not to produce representative estimates of the demographics of e-bike or e-scooter users, but to allow a specific evaluation of these schemes, which will be analysed in detail in a subsequent NIHR project (NIHR163726; https://www.fundingawards.nihr.ac.uk/award/NIHR163726). However, to explore how representative the sample is of the wider population, sample demographic characteristics were compared descriptively against census 2021 data³¹ for age, gender, ethnicity, and highest education level for each site and for all sites combined.

We asked respondents to provide information on transport use during a typical week and presented the frequency (five or more times per week, 3–4 times per week, 1–2 times per week, less than once per week, never) of type of transport use (car, walking, bus, train, conventional bicycle, e-bike, e-scooter) for each site and overall. Of the sites we considered, e-scooter use is only legal in Bath and Bristol; we have only reported the frequency of e-scooter use data for Bristol due to the very low numbers in other sites.

We asked the e-bike and e-scooter users how they had accessed them: through a share scheme, loan scheme, hire on a day-to-day basis, or a personal e-bike or e-scooter. We also asked them to indicate which venues they accessed on a weekly basis using an e-bike or e-scooter (work/education, healthcare appointment, public transport, shopping/errands, visiting friends, leisure/leisure venues, and job interviews/job centres). Finally, we asked participants to report the distance they would be willing to walk or wheel to access a share-hire scheme e-bike or e-scooter. Frequencies were reported according to the following categories: I would not use a share scheme, less than 100 meters (about 1 min), 100 to 250 meters (about 2.5 minutes), 250 to 500 meters (about 4 minutes), 500 meters to 1 km (about 7 minutes), more than 1 km (more than 10 minutes), and ‘don’t know’.

Interviews

Data collection. Respondents to the baseline data collection survey were invited to express their interest in a qualitative interview to discuss their views on the introduction of EB and EB+ES share-hire schemes and their experience. Four respondents living in Leeds and four living in Bristol were selected to provide a set with a mix of characteristics based on age, ethnicity, gender, disability, and whether they were users or non-users of the schemes. Interviews of approximately one hour in duration were conducted online (n=3) or in person (n=5) between three and four months after the introduction of the share-hire schemes. The interviews were semi-structured and investigated household travel behaviour, awareness of and attitudes to e-bike/e-scooter schemes in the local context, and prospects for travel behaviour change in the following 12 months (incorporating e-bike and/or e-scooter use and under what conditions). All interviews were audio-recorded. The opportunity was also taken to ask participants about their experience of completing the baseline survey to enable researchers to consider any adjustments that may be required. Informed consent was obtained from all participants as outlined below.

Data analysis. Audio recordings of the interviews were transcribed using Otter.ai and then checked for accuracy by the researcher responsible for conducting the interviews. These were sent to participants, who were asked to confirm if there were any inaccuracies within ten days of receiving them. The researcher then wrote a series of vignettes that involved compiling biographical summaries and accounts of the topics covered during semi-structured interviews, illustrated with verbatim quotes. These were ‘sense-checked’ by a coresearcher who read the researcher vignettes in tandem with the source transcripts.

Survey data

At baseline, 4271 online surveys were completed. Following data cleaning and duplicate removal, N=3771 (n=1048 for Bristol, n=1096 for Leeds, n=521 for Bradford, and n=1106 for Sheffield) remained for the baseline data analysis. In the resurvey, 6160 online surveys were completed. After data cleaning and duplicate removal, N=5370 (n=1153 for Bristol, n=1192 for Leeds, n=1006 for Bradford, n=1024 for Sheffield, n=363 for Bath, and n=632 for Plymouth) remained for the resurvey data analysis. Figure 1 illustrates the flow of participants from recruitment to data analysis.

Figure 1. Study flowchart.

Descriptive characteristics. Generally, there were slightly more women than men in both the baseline (~52% vs. ~46%) and resurvey (60% vs. 38%) samples (Extended Data Tables 1 and 2). At baseline, older age groups (>55 years) were more frequent in Leeds and Sheffield, and younger age groups were more common in Bristol and Bradford (<45 years) (Extended Data Table 1). At resurvey, older people were more frequent in Leeds, Bradford, and Sheffield, with younger people in Bristol and Bath (16–34) (Extended Data Table 2). The sample was predominantly of white ethnicity (92%, baseline; 90% resurvey overall) (Extended Data Tables 1 and 2). Most sites reported 91-95% white ethnicity but this was lower in Bradford respondents (baseline: 83%; resurvey: 76%), who reported 11% Asian ethnicity at baseline and 17% Asian ethnicity at resurvey (Extended Data Tables 1 and 2).

The majority of both the baseline and resurvey samples were in work (baseline: 66%; resurvey: 64%) (Extended Data Tables 1 and 2). The sample was skewed towards higher education, with 71% at baseline (75% at resurvey) educated to degree level or above (Extended Data Tables 1 and 2). The sample was also skewed towards higher household income categories at both the baseline and resurvey (>£40k) (Extended Data Tables 1 and 2). IMD showed a fairly even spread overall; however, deprivation levels were high in Bradford and Plymouth, with lower deprivation levels in Bath and Sheffield (Extended Data Tables 1 and 2).

Across all sites, people predominantly reported living as a couple (baseline: 38%; resurvey: 40%) or as a family with children under 18 years of age (baseline: 28%; resurvey: 23%) (Extended Data Tables 1 and 2). Approximately two-thirds of the respondents reported no health conditions (baseline: 67%; resurvey: 63%) (Extended Data Tables 1 and 2). Slightly more participants (73%) reported meeting the physical activity guidelines in the baseline sample than in the resurvey sample (68%; Extended Data Tables 1 and 2). Approximately one-fifth of the overall sample reported living with obesity (baseline: 22%; resurvey: 21%), and approximately one-third reported living with overweight (baseline: 31%; resurvey: 33%) (Extended Data Tables 1 and 2).

The majority of the sample reported having never used an e-bike, ranging from 50% in Bradford to 71% in Leeds at baseline (61% overall), and 64% in Sheffield/Plymouth to 79% in Bradford at resurvey (69% overall). Similar percentages had never used an e-scooter (baseline 77%: 55% in Bristol, 89% in Sheffield; resurvey: 84%: 70% in Bristol, 93% in Bradford) (Extended Data Tables 1 and 2). Among those who had ever used an e-bike, 53% were men and 45% were women (baseline), with similar proportions of ever e-bike users in the resurvey sample (48% vs. 50%, respectively) (Table 1 and Table 2). Ever e-bike users comprised more 16–35-year-olds (baseline: 32%; resurvey: 28%) and fewer aged 65+ (baseline:12%; resurvey:14%; Table 1 and Table 2). Among those who had ever used an e-bike, 28% reported living with health conditions in the baseline sample (Table 1), with similar percentages in the resurvey sample (30%; Table 2).

Among those who had ever used an e-scooter, 54% were men and 43% were women in the baseline sample. The proportion of ever e-scooter users in the resurvey sample was slightly higher in women (51%) than in men (46%) overall (Table 3 and Table 4). There was a higher proportion of young ever e-scooter users and a lower proportion of old ever e-scooter users in both the baseline and resurvey samples (48% were aged 16–35 years vs 3% aged 65+ years at baseline, and 48% were aged 16–35 years vs 6% aged 65+ years at resurvey) (Table 3 and Table 4). Among the ever users of e-scooters, 22% reported having a health condition in the baseline sample (Table 3) and 25% reported having a health condition in the resurvey sample (Table 4).

There were more women in both the baseline (53%) and resurvey (61%) samples compared to census 2021 data (52%). Our sample had fewer young people (16-34 years) overall (baseline: 21%; resurvey: 20% in the sample vs. 35% in census 2021 data). There was a higher proportion of those with white ethnicity in our baseline (92%) and resurvey (90%) samples compared to census 2021 data (81%). The sample was skewed towards higher education, with 71% of the baseline sample and 75% of the resurvey sample reporting being educated to degree level or above versus the expected 34% in the census 2021 data (Extended Data Table 3).

Typical transport use. In all the cities combined, the most common transport modes typically used each week were cars and walking. At baseline, car use of five or more times per week was reported most frequently in Bradford (41%) and Leeds (40%), with lower proportions in Sheffield (31%) and Bristol (22%) (Table 5). Trends were similar in the resurvey: Bradford (48%), Leeds (37%), Plymouth (32%), Sheffield (27%), Bath (21%), and Bristol (17%) (Extended Data Table 4). In the baseline sample, the proportion of participants reporting walking on five or more days per week ranged from 67% in Sheffield to 59% in Bradford (64% overall) (Table 5). In the resurvey sample, 79% of the Bath sample reported walking five or more days per week down to 57% in Bradford (67% overall) (Extended Data Table 4).

The majority of respondents reported never using a conventional bicycle in both the baseline (53%) and resurvey (67%) samples (Table 5 and Extended Data Table 4). E-bike use was less common than conventional bicycle use, with 72% of the baseline sample and 85% of the resurvey sample reporting never using one in a typical week.

Overall, 8% reported using e-bikes 1–2 times per week and 7% reported using them 3–4 times per week in the baseline sample, with the corresponding proportions in the resurvey sample being 3%. In the baseline sample, lower proportions of e-bike use in a typical week were reported in Leeds (19%) and Sheffield (24%), with higher proportions reported in Bradford (38%) and Bristol (37%) (Table 5). In the resurvey sample, proportions were lower and similar across all cities, with around 15% overall reporting ever using an e-bike in a typical week (Extended Data Table 4).

Regular e-scooter use in Bristol was uncommon, with 55% at baseline and 85% in the resurvey sample reporting never using one in a typical week. When respondents reported using an e-scooter, they most commonly reported using one less than once per week (17% baseline and 10% resurvey) (Table 5 and Extended Data Table 4).

E-bike access and destinations. At baseline, the most common e-bike access route was the use of one’s own e-bike (45% overall; Bristol 37%; Leeds 49%; Bradford 48%; Sheffield 51%). Prior access via a share-hire scheme was 25% overall at baseline, with higher proportions reporting this route in Bristol (38%) and Bradford (32%), with respondents from Leeds (17%) and Sheffield (11%) indicating lower proportions (Table 6). In the resurvey, the overall proportions reporting access via a share-hire scheme (38%) or their own personal e-bike (36%) were similar (Table 7). However, at the individual city level, Leeds was the only site to report similar levels of access via a share-hire scheme (38%) or their own personal e-bike (39%). Bradford (41%) and Sheffield (40%) still indicated high access via their own personal e-bike at resurvey, whereas for respondents from Bristol (49%), Bath (42%), and Plymouth (65%), the most commonly reported access route was via a share-hire scheme. Note that this refers to access at any point in time over their lifetime, and the access routes are not mutually exclusive.

Overall, the most common weekly e-bike destination was leisure/leisure venue at both baseline (50%) and resurvey (37%), followed by work/education (baseline: 43%; resurvey: 27%), and shopping/errands (baseline: 41%; resurvey: 31%) (Table 6 and Table 7). Note that the destinations are not mutually exclusive.

E-scooter access and destinations. The most common e-scooter access route was via a share-hire scheme (e.g., Beryl/Voi/Tier/Dott), with 60% overall at baseline and 74% overall at resurvey (slightly higher in Bristol) (Table 8 and Table 9). However, this differed for Bradford, where the most common e-scooter access route was the use of either their own or a borrowed personal e-scooter (baseline: 51%; resurvey: 47%), followed by a share-hire scheme (baseline: 33%; resurvey: 32%) (Table 8 and Table 9). In Bristol, the most common weekly e-scooter destination was for leisure/a leisure venue (baseline: 61%; resurvey: 30%), followed by work/education (baseline: 48%; resurvey: 18%), and shopping/errands (baseline: 46%; resurvey: 18%) (Table 8 and Table 9).

Share scheme access distances. Half of the respondents indicated that they would not use an e-bike share-hire scheme, and 63% indicated that they would not use an e-scooter share-hire scheme. Of those who indicated that they would potentially use a share-hire scheme, 34% were willing to walk/wheel up to 500 m (32% for an e-scooter) and 28% were willing to walk/wheel up to 1 km to access an e-bike (25% for an e-scooter). Very few respondents were willing to walk/wheel more than 1 km to access an e-bike (14%) or e-scooter (12%) via a share-hire scheme (Table 10).

Interview data

Overall, the interviewees indicated support for these e-bike and e-scooter share schemes as they saw them as a good addition to the wider transport offer. They identified multiple potential benefits such as reduced pollution and congestion, more opportunities to access employment sites, health benefits, and journey reliability. Support was strongest for e-bike schemes. However, they felt that e-scooter schemes required more regulation, control, and enforcement. They provided a range of views on the balance between barriers to the use of these schemes and ensuring safety. These included suggestions regarding training, the need for a driving licence prior to use, and whether helmets could or should be provided to improve safety. There was the perception that these types of schemes were of good value for local authorities under financial pressure because, with the permission of the highway authority, operators owned and operated the transport service and therefore carried the financial risk. The interviewees highlighted the negative impact of privately owned e-scooters and delivery e-bikes. They indicated that the poor behaviour of some of these users affected the perception of these transport modes by others, particularly car drivers. Users of the hire schemes reported the challenges of e-bike and e-scooter use, including poor quality highway infrastructure, animosity towards e-bike and e-scooter users by other road users, and the challenge of sometimes finding points to hire or park e-bikes/e-scooters and frustration with geo-fenced areas within city centres where use is restricted.

We asked them their opinion on how easy it was to complete the survey. In general, they indicated that it was unproblematic to complete, but found it difficult to remember because of the time elapsed since the survey (approximately 3 to 4 months). They were also asked for more details on disabilities to be included. For the resurvey questionnaire, we considered this and added ‘wheeling’ as an option whenever there were questions with transport options. However, respondents to the questionnaire used this option infrequently, with numbers too small to report separately in this paper. Vignettes for each participant, from which these summary results were derived, can be found in the Open Science repository for this project (https://osf.io/gq9s8/, DOI 10.17605/OSF.IO/GQ9S8).