Projects per year
Abstract
Increasing the mode-share of the bicycle poses a solution to many issues in modern society. It has the potential to alleviate urban congestion, reduce noise and air pollution (NOx and PM25) while also providing health benefits for the individual users through increased physical activity. However, the safety of cyclists in traffic remains a major concern. Bicyclists are considerably more vulnerable in traffic than most other road users. Thus, increasing the cycling mode share might have undesired impacts on road safety as the number of cycling injuries may increase. To address the current safety concerns regarding cycling and ensure the future attractiveness of cycling for transport, informed decisions concerning safety measures and preventive efforts to promote cycling safety are necessary. These informed decisions rely on a thorough understanding of the factors related to bicycle crashes and their potential outcomes, not to mention the subsequent experiences. This PhD thesis concerns several aspects of modelling and analysing bicycle crashes to supply new methodologies and findings to support future mitigating efforts and promote cycling safety. The work of this thesis is split into four parts: Part I concerns the use of model-based approaches to improve the current standard for the estimation of disaggregated bicycle exposure, which is essential to the investigation of cycling crashes. Part II concerns the risk assessment that is possible when presented with disaggregated exposure data. Following this, Part III covers the estimation of the injury severity of bicycle crashes. Finally, Part IV deals with health consequences reported by cyclists having suffered bicycle crashes.
Information concerning bicycle exposure is crucial to analysing factors influencing bicycle crashes. Therefore, the first part of the thesis addresses the issue of accurately estimating bicycle exposure. In this part of the thesis, the importance of having accurate measures of bicycle exposure to improve cycling safety is highlighted. Meanwhile, bicycle traffic data at the segment level is sparse and seldomly available. This issue leads many studies to either leave out exposure variables or use aggregated variables. To alleviate this, the paper in Part I presents a novel deep-learning method to estimate hourly bicycle flow in a
network conditional on mean cycling flow, weather and time. The model presents a superior alternative to the calibration-factor method applied by the Danish Road Directorate, yielding significantly more accurate cycling flow estimates. Furthermore, bicycle crash frequency models are estimated using cycling flow estimates of varying degrees of disaggregation and quality to quantify the impact of improved exposure data in crash analyses.
This comparison provides statistical evidence for improved model accuracy given better exposure data. The second part of the thesis investigates the analysis and inference concerning bicycle crash risk that is possible when presented with disaggregate cycling flow data. The paper presented in this part deals with the bicycle crash risk assessment in Copenhagen through a study that applies a recently developed disaggregated and non-parametric method for crash risk evaluation. The method presents researchers with a flexible framework for crash risk analysis that compares the distribution of conditions as seen by an arbitrary cyclist (the Palm distribution) with those seen by a cyclist subject to a crash (the accident distribution). The paper uses this Palm distribution method to reveal the relative changes in bicycle crash risk given adverse weather and time conditions. The study reveals several findings concerning the relation between weather conditions and the bicycle crash risk. Furthermore, the disaggregate risk assessment approach identifies correlational patterns that disappear with aggregation.
The third part of the thesis investigates the injury severity outcomes of bicycle crashes seeking to identify circumstances around crashes statistically associated with the resulting injury severity. The paper presented in this part specifically investigates the factors influencing the injury severity of single-bicycle crashes. These crashes are of special interest due to their high prevalence, reported to make up more than half of all bicycle crashes. Also, with the efforts to increase the bicycle mode share, this crash type is expected to increase. Nevertheless, the factors that influence the injury severity in single-bicycle crashes
are still rather unexplored. The analysis relies on a Latent Class Ordered Probit model estimated using hospital data on bicycle crashes combined with infrastructure information for the time and place of the respective crashes. By allowing the probabilistic class assignment to vary according to cyclists’ age and gender, this study yields beneficial insight into the behavioural groups of crashed cyclists. Furthermore, the study provides valuable results concerning the impact of bicycle-specific infrastructure, and its maintenance, on injuries.
The final part of the thesis examines the possible health consequences of suffering bicycle crashes with respect to the injuries and non-injury factors. The paper presented in this section addresses the issue by investigating distress symptoms reported by cyclists and their relation to potential crash involvement. The paper first investigates the frequency of reported distress symptoms, comparing the reported frequency among crashed cyclists with a control group of non-crash cyclists. Secondly, structural equation modelling (SEM) is applied to identify underlying distress constructs and investigate their relation to injuries
and non-injury factors. The results show that the crashed cyclists report fewer distress symptoms than the non-crash cyclists. Only crashed cyclists who considered their crashes severe reported more distress symptoms than the non-crash cyclists. Three latent distress constructs are identified: ”General stress & exhaustion”, ”Depression & anxiety”, and ”Physical impairment”. The SEM reveals several injuries positively associated with the distress structures and highlights the importance of the lower-extremity injuries to cyclists’ health. Furthermore, several non-injury factors are significantly associated with the latent
distress constructs. Thereby, highlighting the importance of accounting for non-injury factors when assessing the health consequences of crashes. Lastly, the strong relation between ”Depression & anxiety” and a poorer perceived quality of life among the crashed cyclists signifies the importance of considering the potential psychological and mental aspects of suffering a crash.
In summary, this PhD has contributed to the research on bicycle crash modelling and the investigation of crash consequences, covering topics related to cycling exposure and the impact on accident analysis, bicycle crash risk, injury severity outcomes, and distress following bicycle crashes. Through the development and application of novel methodologies and data in the context of Danish cycling, this thesis seeks to provide practitioners, researchers, and policymakers alike with tools and findings relevant for the continued promotion of cycling safety. Specifically the results from this thesis suggest a great need for better bicycle traffic estimation methods to overcome the lack of cycling monitoring. The thesis contributes important knowledge concerning the impact that better bicycle traffic data can have on cycling safety analysis, adding that accurate cycling flow estimates are paramount to accurate analysis of crash factors. Furthermore the thesis exemplifies the detailed analysis of crash factors that is enabled through the improved cycling flow estimates in an analysis assessing the relative risk of bicycle crashes in adverse weather conditions.
This analysis contributes findings of previously unexplored correlations such as the effects of impaired visibility on the bicycle crash risk. Regarding the injury severity outcome of bicycle crashes, this thesis presents new knowledge on the factors influencing the injury severity of single-bicycle crashes. It specifically highlights the importance of bicycle lanes and their maintenance to mitigate the severity of single-bicycle crashes.
Finally, the thesis expands the knowledge of the health impacts of bicycling and crashes, providing additional evidence of the complex interactions of the health benefits from cycling and the adverse health effects of crashing. Furthermore it presents findings revealing that both injuries and demographic factors influence various aspects of cyclists health after crashing, and shows how these aspects relate to their perceived quality of life.
Information concerning bicycle exposure is crucial to analysing factors influencing bicycle crashes. Therefore, the first part of the thesis addresses the issue of accurately estimating bicycle exposure. In this part of the thesis, the importance of having accurate measures of bicycle exposure to improve cycling safety is highlighted. Meanwhile, bicycle traffic data at the segment level is sparse and seldomly available. This issue leads many studies to either leave out exposure variables or use aggregated variables. To alleviate this, the paper in Part I presents a novel deep-learning method to estimate hourly bicycle flow in a
network conditional on mean cycling flow, weather and time. The model presents a superior alternative to the calibration-factor method applied by the Danish Road Directorate, yielding significantly more accurate cycling flow estimates. Furthermore, bicycle crash frequency models are estimated using cycling flow estimates of varying degrees of disaggregation and quality to quantify the impact of improved exposure data in crash analyses.
This comparison provides statistical evidence for improved model accuracy given better exposure data. The second part of the thesis investigates the analysis and inference concerning bicycle crash risk that is possible when presented with disaggregate cycling flow data. The paper presented in this part deals with the bicycle crash risk assessment in Copenhagen through a study that applies a recently developed disaggregated and non-parametric method for crash risk evaluation. The method presents researchers with a flexible framework for crash risk analysis that compares the distribution of conditions as seen by an arbitrary cyclist (the Palm distribution) with those seen by a cyclist subject to a crash (the accident distribution). The paper uses this Palm distribution method to reveal the relative changes in bicycle crash risk given adverse weather and time conditions. The study reveals several findings concerning the relation between weather conditions and the bicycle crash risk. Furthermore, the disaggregate risk assessment approach identifies correlational patterns that disappear with aggregation.
The third part of the thesis investigates the injury severity outcomes of bicycle crashes seeking to identify circumstances around crashes statistically associated with the resulting injury severity. The paper presented in this part specifically investigates the factors influencing the injury severity of single-bicycle crashes. These crashes are of special interest due to their high prevalence, reported to make up more than half of all bicycle crashes. Also, with the efforts to increase the bicycle mode share, this crash type is expected to increase. Nevertheless, the factors that influence the injury severity in single-bicycle crashes
are still rather unexplored. The analysis relies on a Latent Class Ordered Probit model estimated using hospital data on bicycle crashes combined with infrastructure information for the time and place of the respective crashes. By allowing the probabilistic class assignment to vary according to cyclists’ age and gender, this study yields beneficial insight into the behavioural groups of crashed cyclists. Furthermore, the study provides valuable results concerning the impact of bicycle-specific infrastructure, and its maintenance, on injuries.
The final part of the thesis examines the possible health consequences of suffering bicycle crashes with respect to the injuries and non-injury factors. The paper presented in this section addresses the issue by investigating distress symptoms reported by cyclists and their relation to potential crash involvement. The paper first investigates the frequency of reported distress symptoms, comparing the reported frequency among crashed cyclists with a control group of non-crash cyclists. Secondly, structural equation modelling (SEM) is applied to identify underlying distress constructs and investigate their relation to injuries
and non-injury factors. The results show that the crashed cyclists report fewer distress symptoms than the non-crash cyclists. Only crashed cyclists who considered their crashes severe reported more distress symptoms than the non-crash cyclists. Three latent distress constructs are identified: ”General stress & exhaustion”, ”Depression & anxiety”, and ”Physical impairment”. The SEM reveals several injuries positively associated with the distress structures and highlights the importance of the lower-extremity injuries to cyclists’ health. Furthermore, several non-injury factors are significantly associated with the latent
distress constructs. Thereby, highlighting the importance of accounting for non-injury factors when assessing the health consequences of crashes. Lastly, the strong relation between ”Depression & anxiety” and a poorer perceived quality of life among the crashed cyclists signifies the importance of considering the potential psychological and mental aspects of suffering a crash.
In summary, this PhD has contributed to the research on bicycle crash modelling and the investigation of crash consequences, covering topics related to cycling exposure and the impact on accident analysis, bicycle crash risk, injury severity outcomes, and distress following bicycle crashes. Through the development and application of novel methodologies and data in the context of Danish cycling, this thesis seeks to provide practitioners, researchers, and policymakers alike with tools and findings relevant for the continued promotion of cycling safety. Specifically the results from this thesis suggest a great need for better bicycle traffic estimation methods to overcome the lack of cycling monitoring. The thesis contributes important knowledge concerning the impact that better bicycle traffic data can have on cycling safety analysis, adding that accurate cycling flow estimates are paramount to accurate analysis of crash factors. Furthermore the thesis exemplifies the detailed analysis of crash factors that is enabled through the improved cycling flow estimates in an analysis assessing the relative risk of bicycle crashes in adverse weather conditions.
This analysis contributes findings of previously unexplored correlations such as the effects of impaired visibility on the bicycle crash risk. Regarding the injury severity outcome of bicycle crashes, this thesis presents new knowledge on the factors influencing the injury severity of single-bicycle crashes. It specifically highlights the importance of bicycle lanes and their maintenance to mitigate the severity of single-bicycle crashes.
Finally, the thesis expands the knowledge of the health impacts of bicycling and crashes, providing additional evidence of the complex interactions of the health benefits from cycling and the adverse health effects of crashing. Furthermore it presents findings revealing that both injuries and demographic factors influence various aspects of cyclists health after crashing, and shows how these aspects relate to their perceived quality of life.
Original language | English |
---|
Publisher | Technical University of Denmark |
---|---|
Number of pages | 118 |
Publication status | Published - 2022 |
Fingerprint
Dive into the research topics of 'Statistical modelling of cycling accidents: Investigating Risk, Severity and Consequences'. Together they form a unique fingerprint.Projects
- 1 Finished
-
Cyclist accident modelling and the long-term societal costs of cyclist accidents
Myhrmann, M. S. (PhD Student), Dozza, M. (Examiner), Fountas, G. (Examiner), Lima de Azevedo, C. M. (Examiner), Mabit, S. E. (Main Supervisor), Janstrup, K. H. (Supervisor) & Møller, M. (Supervisor)
01/12/2018 → 30/09/2022
Project: PhD