A Look at Cycling Safety in a Southern Municipality of the Netherlands

Objectives Cycling is an important means of transportation in the Netherlands. Unfortunately, the number of cycling accidents and their adverse outcomes (injury and death) are on the rise. We set out to observe the nature of these accidents in Maastricht from 2001 to 2015 and analyzed the recommendations of stakeholders on ways to improve cycling safety. Methods An explanatory sequential mixed methodology was used for this population-based study. In the first phase, a retrospective quantitative analysis of the VIA® accident database for Maastricht was done. This was followed by a thematic analysis of data from five semi-structured interviews. Integration was at the Interpretation stage. Result The first phase showed males (54%) and under-25s (59.9%) had the most cycling accidents, while a larger percentage of females (50.7%) and people >65 years (67%) had adverse outcomes with accidents. More accidents occurred at intersections (52.6%), on shared roads (61.4%), and involved motorized vehicles (95.6%). Bivariate analysis and multivariable logistic regression showed that cycling accidents involving elderly people, women, wet weather or road surfaces, an innocent cyclist, the northeastern district, and morning hours had a higher chance of injury or death. Thematic analysis summarised stakeholder opinions under four themes: role in cycling safety; partners of cycling safety; the importance of accurate data; and investing in safety. Most of the respondents felt improvements in the city’s accident database, cycling policy, and infrastructure were needed. Conclusion Our findings suggest that there has been a decrease in the overall number of cycling accidents in the 15-year period studied. However, differences in sociodemographic variables still determine the distribution and severity of accidents in Maastricht. The existing cycling database at the time of the study needed improvements with data collection and the city needs to involve more stakeholders in its policy-making process.


Introduction
In the world today, most injuries and fatalities from Road Traffic Accidents (RTAs) are in low-and-middleincome countries (LMICs). Yet, RTAs still occur in more advanced countries where, like anywhere else in the globe, they are a cause of immense sorrow and suffering [1][2][3][4]. The Netherlands has a low record of RTAs and fatalities bested only by countries like Sweden, Switzerland, and the United Kingdom among a few others [5]. Programs like Sustainable Safety Vision (Duurzam veilig), launched in the early 1990s as a collaboration between the Institute for Road Safety Research and the Dutch Central Government [6,7], have played a big role in guiding policy. The Sustainable Safety Vision had two objectives: preventing crashes from occurring and preventing severe injury if a crash could not be avoided [8].
Of growing importance in the Netherlands, now, is cycling safety due to a steady increase in the numbers and severity of cycling accidents when compared to other road vehicles [9][10][11][12]. Currently, there are almost as many cycling fatalities -about 200 cyclists killed each year since 2009 -as there are amongst car users [11,13]. This rise is mainly driven by deaths among men aged 65 and above who use electric bikes and made up three-quarters of the 206 cyclists killed in 2017 [14].
And so, we set out to assess the general trend and patterns of cycling accidents in Maastricht from 2001 to 2015 and to investigate risk factors for cycling injuries and deaths over this time span. Our ultimate goal was to evaluate how the obtained empirical findings could contribute to future policies while assessing the effectiveness of cycling accident data collection. Understanding trends in the city, during this period, was relevant as it was an era that witnessed the commencement of two major cycling safety plans, namely, the Strategic Plan for Road Safety [6] and the Safe Cycling Network project [13], thereby helping monitor their implementation in Maastricht.
We also sought to understand the perceptions of major cycling stakeholders on the identified trends, the quality of data, and future cycling policies. Stakeholder analysis has been shown to generate information about important actors in a field of endeavor, thereby, "understanding their behavior, intentions, interrelations, agendas, interests, and the influence or resources they have brought -or could bring -to bear on decision-making processes" [31].
Thus, the study utilized both a systematic approach to analyzing data and stakeholder opinions in understanding the nature of cycling accidents in Maastricht, incorporating the principles of quantitative and qualitative research.

Materials And Methods
We combined data from Maastricht city's approved traffic accident database, called VIA® Traffic Solutions Software (Vught, the Netherlands) database, with the everyday experiences of different working groups using the concept of Complementarity [32] and an Explanatory Sequential design. Both quantitative and qualitative findings were then integrated at the interpretation (discussion) stage of the research.

Data collection
Quantitative Retrospective data on cycling accidents in the city of Maastricht, from 2001 to 2015, were obtained from the VIA online database (http://www.via.nl). This database is a product of collaboration between the Dutch police, the Dutch Association of Insurers, and VIA traffic data specialists [33]. It contains details of accidents known to the Dutch police or insurance companies.
A total of 1656 cycling accident victims were recorded in the database for Maastricht. We included all cycling accidents which occurred between January 2001 and December 2015, leaving us with the following variables: accident severity; type of injury sustained; the types of vehicles involved; whether the cyclist was at fault or not; the main cause of the accident; the time period of the accident; the day; the month; the district; the part of the road involved; the weather, state of the road, and light condition; and type of road on which cyclist was riding.
The outcome measure, accident severity, was operationalized as 0=no injuries and 1=injuries, with 1 consisting of cases classified as "type of injury sustained" -mild, severe, and fatal. The variable month was recoded into seasons and used for analysis that way. Cases occurring in December, January, and February made up Winter (1); March, April, and May -Spring (2); June, July, and August -Summer (3); and September, October, and November -Autumn (4). Age was recoded into a categorical variable Age group-from 0 to 4 (0= <12 years, 1= 12-18 years, 2= 18-25 years, 3= 25-65 years, and 4= >65 years). Also, the different points of the road involved in an accident were recoded into three categories -straight road (0), intersection (1), and roundabout (2). Finally, the variable "time period of the accident" was recoded into three categories: Morning (0), Afternoon (1), and Evening/Night (2).
We excluded all variables with more than 25% missing values like street name, alcohol use by non-cycling parties, and drug use by the cyclist or not, among others. Accidents not involving a cyclist as well as those which occurred outside of Maastricht were also excluded.

Qualitative
This phase involved seven semi-structured interviews with informants in the field of road traffic safety at the Gemeente Maastricht (Municipality), Fietsersbond (Dutch Cyclists' Union), SWOV (Institute for Road Safety Research), Maastricht ambulance services, and the Politie (Police). The organizations visited were important stakeholders in road traffic safety and had policies and action plans on cycling safety. In turn, the informants within these organizations were chosen specifically for their experience and versatile knowledge of cycling issues in Maastricht.
The semi-structured interviews were done in English, a language that all informants were proficient in, and they lasted between forty minutes to an hour. They were mostly face-to-face interviews, except for one occasion where it occurred via Skype®. The questions asked (see Appendices) were broadly grouped into four sets: questions about the informant's organization and the policies on cycling of importance to them; questions on the organization's partners for safe cycling in Maastricht; questions about data collection on cycling accidents in Maastricht and their perceptions of the first phase quantitative analysis; and finally, questions about their opinions on how to improve cycling safety in Maastricht.

Quantitative data analysis
The data collected from VIA® statistics was transferred into SPSS (Statistical Package for the Social Sciences) version 20.0 (IBM Corp., Armonk, USA) for analysis. Descriptive statistics with the distribution of variables based on demographics (age -continuous, age group -categorical, and sex), physical characteristics (weather, season, time of day, month, year, state of the road, light condition, type of road), and accident circumstances (alcohol use by a cyclist, cyclist at fault or not, accident severity, type of injuries sustained, type of vehicles involved, and main cause).
Logistic regression was used to investigate links between the independent variables and the dichotomous outcome of accidents with injuries. Visual inspection of the data suggested that the proportion of the latter changed over time in a nonlinear fashion. Therefore, polynomials (Year2 and Year3) were created to model the temporal pattern of injuries over time. Odds ratios (OR), and 95% confidence intervals (CI) were significant at a p-value of < 0.05.

Qualitative data analysis
For the qualitative component, each interview was recorded and then transcribed. Each transcript was subsequently explored using thematic analysis. Themes were derived from the questions asked, and then coding of keywords and phrases was done to create categories. The categories from each organization were then tabulated and compared to identify areas of similarities or differences.

Ethical considerations
Maastricht University declared that ethical approval was not applicable for this study. Permission to access the VIASTAT database was granted by the Gemeente Maastricht and VIA, and informed consent was obtained from all informants for the semi-structured interviews. Care was taken to ensure the anonymity of all participants and data was handled in a secured manner.

Descriptive Statistics
In 2003, cycling accidents peaked in Maastricht, with a 70% drop noted in the numbers between 2009 and 2010 ( Figure 1). Most of the accidents were injury-free or minor (92.2%, n=1515) with only 7.8% being fatal or requiring some sort of hospitalization ( Figure 2).

FIGURE 2: Percentage distribution of injuries sustained in cycling accidents in Maastricht.
The severity of a cycling accident varied depending on the variable analyzed.  Based on regions, the Eastern districts were the most dangerous (northeast -35% and southeast -26%), however, there was a general decline in all districts with time ( Figure 3).

Unadjusted Analysis
Bivariate analysis as seen in Table 2 of the appendices shows that the age group of an accident victim was important in determining the severity of an accident. All age groups less than 65 years had lower chances of injuries after an accident (<12 years -OR=0.

Multivariable Analysis
All significant variables in the unadjusted analysis were added to the multivariable logistic regression model, which entailed sex, age (categorical), year (continuous) and its polynomials, time of day, district, weather, parties involved, and interaction between year and district of accident occurrence.
Adjusted effects are presented in terms of ORs in Figure 4 for the variables showing an independent link to the outcome. Fewer odds of sustaining an adverse effect for cyclists aged below 65 years as against those older than 65 were observed. Sex, Weather, and Period of day variables had non-referent variables with higher -injuries were more likely in the morning, in wet weather, and for female cyclists. Age had the larger impact, as results indicate that the odds were substantially smaller for younger compared to the elderly (65+).

FIGURE 4: Forest plots showing the Odds Ratios of the independent effects in the final model.
A significant interaction between district and year indicates that either the difference in the likelihood for adverse outcomes, seen among districts, depended on the year of occurrence or that changes in the odds of injury over the years were not the same for all districts. To facilitate the interpretation of this interaction we present the estimated probabilities of having an adverse outcome, according to the multivariable logistic model in Figure 5.

Qualitative
The following themes: role in cycling safety; partners of cycling safety; importance of accurate data; and investing in safety were identified from thematic analysis.

Role in Cycling Safety
Most respondents said their contributions to cycling safety, involved primary prevention and centered on cycling-friendly policy creation and implementation, and infrastructural development.

We develop and design cycling infrastructure…and formulate cycling-friendly policies. (Municipality) …We want more cyclists and need to have a safe and comfortable infrastructure to achieve this. (Cyclists' Union)
We do research on road safety -road user behavior, infrastructure, research, and laws concerning epidemiology...

(SWOV)
They also felt that an enabling environment for cycling depends on plans which are detailed and productive and which will result in more people choosing cycling as a regular mode of commute: …cyclist-friendly infrastructure leads to more cycling… (SWOV) Educating cyclists on policies and responsible road use was of importance to most organizations. The ambulance service was involved in secondary prevention, being a first responder in serious accidents.
…no direct role to play in public health prevention of road or cycling accidents…We are involved only in treating injuries and not necessarily in taking precautions on the road or any other preventive measure… (Ambulance service)

Partners of Cycling Safety
All five organizations worked together on numerous occasions; however, their relationships were not always warm: …the relationship between stakeholders is dynamic so there are moments when disagreements exist. We also work closely with the police who in most cases are the first to get to an accident scene… we work with the fire service whom we call after accidents where the victims are entangled and trapped by motor vehicle wreckage, also when there is leaking fuel around the scene of an accident…bystanders. (Ambulance service) This network of stakeholder interaction shows the complexity needed for transforming the constantly revised cycling policies into sustained safe cycling.

Importance of Accurate Data
Regarding data source, the VIA database was of common interest to the Municipality, Cycling Union, and SWOV: …our most important source of data is from VIASTAT. (Municipality) … The 70% drop was due to a reduction in registration, as can be seen from the national summary of VIA.

(Cyclists' Union) … SWOV is also involved in the STAR (Smart Traffic Accident Reporting) initiative of VIA… (SWOV)
The VIA database depends on the police and insurance companies for its data. Some stakeholders highlighted that a drop in the number of accidents, after 2009, was due to under-reporting, a situation that affects the accuracy of the database.

The reduction between 2009 and 2010 is due to a reduction in the registration of accidents… data was better 10 years ago than it is now because the police collected data on all accidents… (Municipality) …Police registration…is weak in the field of injury data collection for all road users, as well as cycling fatalities. (SWOV) …we stopped registering accidents unless they were severe. Less severe accident victims were referred to their insurance companies instead. (Police)
Other sources of cycling data for the different organizations, in Maastricht, were as follows:

…the media and cyclists who write letters to the organization…we also do surveys, watch and observe traffic situations in the city and infrastructure. (Municipality)
In terms of injuries, the hospital data are better and these are a major source for SWOV -although it becomes less accurate as one gets to the regional levels… (SWOV)

…in the last 2 -3 years we restarted partial registration; recording names, types of injuries, and not many other variables… (Police)
We collect accident data at the scene of the accident, the biodata of the patient, and the kinds of injuries sustained… (Ambulance service) Every respondent saw an accurate database of cyclists' numbers, and accident records, as vital for effective planning and monitoring purposes. And the period after 2009, when data collected by the police and insurance companies was reduced, was a source of concern to all the stakeholders.

Investing in Safety
The call to improve and continue investing in cycling infrastructure resonated at the Municipality, Cycling Union, and SWOV.

…the government has to invest more because every dead man on the street is one too many…The design and capacity of infrastructure should be upgraded as the number of cyclists increase…Educational programs should be emphasized…Data collection can be improved if people involved in accidents can input the information directly into the VIA database… (Municipality)
…we should do more for the safety of the elderly cyclist…make the infrastructure more friendly for them…there should be cycling classes for the young… (Cyclists' Union)

Removing obstructions, and making people wear helmets should reduce cyclist-only accidents… Maastricht should have better data collection initiatives using innovative computer applications and automatic data collection… add data on the number of cyclists and bicycle types in Maastricht. (SWOV)
…make the insurance companies contribute more to the collection of cycling accident data. (Police)

There is the need for the use of protective gear like helmets and reflectors, and the promotion of their use in the young and old... (Ambulance service) Informant's Opinions on Quantitative Results
The opinions of the different stakeholders on important phase one findings can be seen in Table 3.

Discussion
This study set out to assess the trend of cycling accidents in the city of Maastricht from 2001 to 2015 and to understand the opinions of major stakeholders on the quality of cycling data, the quantitative findings from the study, and future policymaking. No such study had yet been done in the city of Maastricht.
It concluded that major improvements in the database used by the city for planning and monitoring its cycling safety operations are a priority. A reduction in the number of accidents across the fifteen-year period might have been interpreted to mean Maastricht city was succeeding with cycling safety; however, major stakeholders sounded a note of caution by ascribing this reduction to a decrease in the quality of data collection from 2009, instead. All the stakeholders, thus, concurred on the need to strengthen and modernize data collection and proffered various ways to achieve this.
The same conclusions were drawn for an observed decrease in the number and risk of getting an injury across the years was observed, which was at variance with what was seen at the national level where a steady increase in the number of injuries from cycling accidents was a source of concern [34]. This shows that the importance of reliable and accurate data cannot be overemphasized as it helps to "identify road safety problems, risk factors, and priority areas" [35].
Most accidents we analyzed were between cyclists and motorcars or vans with increased severity amongst cyclist-only and bicycle-pedestrian accidents, as well as with bicycle and two-wheeler accidents. However, a number of studies show that cyclist-only accidents are the most prevalent form of cycling accidents in the Netherlands. These studies also point out that accidents involving motorized vehicles often result in more severe injuries and fatalities than do single bicycle counterparts [13,34,[36][37][38]. Understanding the types of accidents experienced by cyclists in the city and the nature of their severity will be important in future infrastructural design; indeed, our respondents highlighted various links between the nature of cycling infrastructure in an area and the types of accidents happening there.
Another observed trend was that most of the cycling accidents involved people who were young and males.
On the other hand, more accidents with adverse outcomes (injuries or death) were seen amongst elderly and female cyclists. Statistics by SWOV, at the national level, agreed with this trend [34]. Our respondents ascribed the situation to the inexperience of young cyclists, the larger number of them who cycle, and the high use of old bicycles by university students. They also opined that the carefree nature of male cyclists and their proneness for riding while drunk played a role; as did the delicate physical nature of females and the elderly, most of whom ride with no form of protection (common in the Netherlands). Risky behaviors like using electronic devices while cycling is also more likely amongst the young, with old Dutch studies showing phone use as a culprit in 3-4% of cycling accidents with injuries [22,39].
Intersections are the most dangerous points on the road for a cyclist; roundabouts being the safest. In a lot of cities in the Netherlands, there is a push towards converting intersection points to roundabouts [13,34,35], a view shared by some of our respondents who explained that in roundabouts, separation of traffic was usually maintained with bicycles kept away from motorized traffic, a situation that is harder to replicate at an intersection. However, a note of caution on the application of these data was sounded by a respondent, who said some of the current research on the subject showed the positive effects were majorly for cars.
Furthermore, designing infrastructure properly does not guarantee safety alone. Maintaining them and educating users on their safe use are also important factors. Road surfaces that are uneven (with potholes or cobblestones, for example), unclean, and improperly marked tend to be more accident-prone [13]. In a similar vein, the study showed that autumn and wet weather made cyclists more prone to accidents and injuries. Respondents linked the slippery nature of roads to the wet fallen leaves of autumn and wet cobblestones during rains or winter.
Maastricht central and the Southwest, combined, had fewer accidents and adverse accident outcomes. The unique road design espoused by Sustainable Safety Vision (Duurzaam veilig) is credited for this situation; as the area has benefited from "street hierarchy" and separation of traffic which reduce the exposure of cyclists to motorized vehicles. These are also areas with a high number of residential houses and a road network designed for such. Therefore, this shows that the policy of Sustainable Safety and the bicycle plan of the city, which follows its principles, can succeed in improving safety if well implemented [8,40].

Limitations
The study had a three-month limit which resulted in fewer stakeholders being interviewed, also, because the data had to be manually transferred from VIA's database to SPSS, imputation errors may have occurred. Furthermore, a change in the police protocol for recording cycling accident data in Maastricht, from 2010, resulted in a greater emphasis on serious accidents a possible source of instrumentation bias and a threat to the internal validity of the study. Finally, by applying multidimensional qualitative coding to onedimensional quantitative codes [41,42], the richness of the qualitative data and the precision of the quantitative data will both have been affected at the point of complementary integration.

Conclusions
The study tried to understand the nature of cycling accidents in Maastricht using the VIA database and major stakeholder analysis. And we conclude that major improvements in the database used by the city for planning and monitoring its cycling safety operations need to be made. Also, most stakeholders agree that for Maastricht to become safer for cyclists than it already is, there needs to be greater coordination and harmony between the various stakeholders.
Therefore, the city of Maastricht should strive to improve its cycling accident database by utilizing other data sources like those of hospitals, coroners, or ambulance services, in addition to VIA. The new STAR partnership between VIA, the Police, and the insurance companies, promises to improve the accuracy of this database, but VIA should also find a way to incorporate direct reporting by the individual cyclist. This will go a long way to reduce underreporting by those cyclists who do not sustain grievous injury to self or damage to property, who may therefore feel unobligated to make contact with the police or insurance company.
VIA should also improve the consistency of entries, and eliminate variables that are difficult to collect. Also, the city should seek to make its own cycling accident database easy to access in the future, by providing them in formats that can easily be read using different statistical software programs (like SPSS). Furthermore, when the quality of data collection has been improved, it will be helpful to include population distribution figures for each district, cycling distance traveled, and cycling volume figures to enable the accurate calculation of prevalence and incidence rates.
A more detailed stakeholder analysis prior to finalizing the new cycling safety policy of the city is recommended in order to arrive at an understanding by all cycling stakeholders in a way that would make the new policy easily implementable. The stakeholder analysis should also look into the possibility of educating young cyclists on the benefits of helmet use and the avoidance of riding under the influence of alcohol or drugs. But this will not be easy as the policy process is complex and interactive -both at the formulation and implementation levels -with many groups to engage with. The World Health Organization tried to classify the various road safety groups of importance into providers, enforcers, and users. In the case of Maastricht, applying this classification might look like this: Providers will be the provincial government and the municipality's traffic safety, management, and transport arms; Enforcers will be the police, educators, media, SWOV, ROVL, health professionals and the ambulance service; and users will be the individual cyclists and their advocacy group the Cyclists' Union.
The stakeholder analysis may also want to consider which policy process will best suit the city's needs. A top-down pattern, in which there is a clear division between those who formulate policy and the implementers, might seem more acceptable to the Providers, while a bottom-up approach will favor Enforcer and User groups as it gives them "some discretion to reshape the dictates of higher levels in the system". In contemporary Public Health, the bottom-up approach is most preferred as it takes into account local influences without necessarily negating the role of the policymaker.
The North-Eastern and South-Eastern parts of the city will require that more attention be paid to their cycling infrastructure. We share the views of some of our respondents that the districts can be made safer if they become more compliant with the Sustainable Safety Vision plan (Duurzaam veilig). Upgrading existing infrastructure across the city must continue to occur with the changing times. For example, converting more intersections to roundabouts, replacing cobblestone streets with less slippery surfaces, and making cycling lanes wider and more "forgiving" so that they become more suitable for the elderly and E-bikes. Finally, there are a number of studies on ways to make motorcars and heavy vehicles less dangerous to cyclists, in the event of a collision. Introducing some of the recommendations of such studies into the city's bicycle plan will be helpful in improving the safety lot of the city's cyclists.
submitted work. Financial relationships: All authors have declared that they have no financial relationships at present or within the previous three years with any organizations that might have an interest in the submitted work. Other relationships: All authors have declared that there are no other relationships or activities that could appear to have influenced the submitted work.