Driver, Collision and Meteorological Characteristics of Motor Vehicle Collisions among Road Trauma Survivors

Melita J. Giummarra, Rongbin Xu, Yuming Guo, Joanna F. Dipnall, Jennie Ponsford, Peter Cameron, Shanthi Ameratunga, Belinda J. GabbeInternational Journal of Environmental Research and Public Health 2021 · DOI

TL;DR

This study used latent class analysis to identify five distinct types of motor vehicle collisions leading to serious injury in Victoria, Australia. While these collision types were identified, they ultimately did not predict long-term recovery outcomes. Instead, individual patient characteristics like age, sex, socioeconomic status, pre-injury health, and injury severity were the primary drivers of recovery.

What Problem Does This Paper Address?

Road trauma remains a significant public health issue globally, imposing substantial economic and personal costs. While much research has focused on individual risk factors for collisions, there's a need to better understand the complex interplay of circumstances that lead to serious injury. Specifically, this paper aims to identify distinct "classes" or subgroups of motor vehicle collisions based on various characteristics (driver fault, BAC, time of day, weather, etc.) and to determine if these collision types are associated with different long-term recovery outcomes for survivors. The goal is to see if understanding the "how" and "where" of a crash can help predict the "how well" someone recovers.

Methods

The core methodological innovation here is the use of Latent Class Analysis (LCA). LCA is a statistical technique used to identify unobserved subgroups (latent classes) within a population based on patterns in observed categorical variables. In this study, LCA was used to group collisions based on a set of observed characteristics:

  • Driver fault (whether another vehicle was at fault)
  • Blood Alcohol Concentration (BAC) ≥ 0.05
  • Single vs. multi-vehicle collision
  • Whether other claimants were seriously injured
  • Time of day (sunrise, sunset, daytime, evening)
  • Day of the week (weekday vs. weekend)
  • Inclement weather conditions
  • Collision location (metropolitan vs. regional/remote)

After identifying these collision classes, mixed linear and logistic regression models were employed to examine the association between collision class membership and various outcome variables (return to work, health status via EQ-5D-3L, and independent function via GOS-E) at 6, 12, and 24 months post-injury. These analyses were adjusted for demographic, health, and injury-related covariates.

Datasets

The study utilized data from the Victorian State Trauma Registry (VSTR) for the period 2010-2016. This registry captures information on all major trauma admissions in Victoria. Data were linked with information from the Transport Accident Commission (TAC), Victoria's no-fault transport injury compensation system, to obtain claimant information and fault attribution. Additionally, meteorological data from the Australian Bureau of Meteorology and geospatial data (Index of Relative Socioeconomic Advantage and Disadvantage - IRSAD, and Accessibility/Remoteness Index of Australia - ARIA) were linked to the collision events.

The study focused on drivers aged 16 years and older who were admitted to hospital with serious transport-related injuries, survived at least 12 months post-injury, and had a compensation claim. Intentional injuries and severe head injuries were excluded.

Key Insights and Ideas

  • Collision Circumstances Don't Predict Recovery: The most striking finding is that despite identifying distinct collision "classes," these classes themselves did not predict how well a driver would recover in terms of health, function, or return to work. This suggests that the specific circumstances of the crash, while important for understanding risk, are less critical for predicting long-term outcomes than the individual's characteristics.
  • Individual Factors Reign Supreme: Recovery is primarily driven by individual patient characteristics. Younger age, being male, higher socioeconomic status, better education, good pre-injury health (fewer comorbidities, no pre-injury disability), and less severe injuries were consistently associated with better outcomes.
  • Proxy Indicators: The study suggests that collision characteristics might act as proxy indicators for underlying demographic and health factors that are more directly linked to recovery. For example, certain collision types might be more prevalent in specific age groups or socioeconomic strata, which then influence outcomes.
  • Focus on Patient Screening: The findings strongly advocate for a shift in focus from crash circumstances to patient-specific factors during hospital admission. Identifying at-risk individuals early through their demographic and health profiles can lead to more targeted clinical assessments and interventions.

Key Results

  • Five Collision Classes Identified: The LCA identified five distinct collision classes:
  • Daytime multi-vehicle collisions, no other at fault (largest class, older drivers, predominantly female, often not working).
  • Daytime single-vehicle predominantly weekday collisions (middle-aged, male, higher proportion in regional/remote areas).
  • Evening single-vehicle collisions, no other at fault, with a significant proportion (36%) having BAC ≥ 0.05 (younger, male, higher prevalence of substance use and comorbidities).
  • Sunrise or sunset weekday collisions (middle-aged, male, often multi-vehicle).
  • Dusk and evening multi-vehicle in metropolitan areas with BAC < 0.05 (middle-aged, male, higher proportion of serious injury to other claimants).
  • No Association Between Collision Class and Outcomes: After adjusting for covariates, there was no statistically significant association between collision class membership and health status (EQ-5D-3L), independent function (GOS-E), or return to work at 6, 12, or 24 months post-injury.
  • Predictors of Poor Recovery: Factors consistently associated with poorer outcomes included older age, female sex, lower socioeconomic disadvantage, lower education levels, pre-existing health conditions (CCI, pre-injury disability), and higher injury severity.

Limitations and Future Work

  • Exclusion of Fatalities and Less Serious Injuries: The study focused only on major trauma cases who survived at least 12 months. This means the findings do not characterize collisions resulting in death or less severe injuries, limiting the generalizability to the entire spectrum of road trauma.
  • Missing Data on Key Factors: Important collision-related characteristics such as license type, vehicle make/model, road speed limits, cell phone use, and drug use were not available. These factors are known to influence crash risk and injury severity.
  • LCA Subjectivity: The identification of latent classes is inherently data-driven and relies on researcher evaluation of model fit and substantive meaning. The findings may be specific to this cohort and setting.
  • Dynamic Nature of Road Safety: Road safety measures evolve, and collision characteristics might change over time, potentially impacting the replicability of these findings in the future.
  • Future Work: The authors suggest extending this approach to datasets that include fatal and less serious injury collisions to inform broader road safety strategies. They also highlight the need for future research to incorporate the missing collision-related characteristics to gain a more comprehensive understanding.

Summary updated

Abstract

Road trauma remains a significant public health problem. We aimed to identify sub-groups of motor vehicle collisions in Victoria, Australia, and the association between collision characteristics and outcomes up to 24 months post-injury. Data were extracted from the Victorian State Trauma Registry for injured drivers aged ≥16 years, from 2010 to 2016, with a compensation claim who survived ≥12 months post-injury. People with intentional or severe head injury were excluded, resulting in 2735 cases. Latent class analysis was used to identify collision classes for driver fault and blood alcohol concentration (BAC), day and time of collision, weather conditions, single vs. multi-vehicle and regional vs. metropolitan injury location. Five classes were identified: (1) daytime multi-vehicle collisions, no other at fault; (2) daytime single-vehicle predominantly weekday collisions; (3) evening single-vehicle collisions, no other at fault, 36% with BAC ≥ 0.05; (4) sunrise or sunset weekday collisions; and (5) dusk and evening multi-vehicle in metropolitan areas with BAC < 0.05. Mixed linear and logistic regression analyses examined associations between collision class and return to work, health (EQ-5D-3L summary score) and independent function Glasgow Outcome Scale - Extended at 6, 12 and 24 months. After adjusting for demographic, health and injury characteristics, collision class was not associated with outcomes. Rather, risk of poor outcomes was associated with age, sex and socioeconomic disadvantage, education, pre-injury health and injury severity. People at risk of poor recovery may be identified from factors available during the hospital admission and may benefit from clinical assessment and targeted referrals and treatments.
BibTeX
@article{giummarra2021driver,
  author = {Melita J. Giummarra and Rongbin Xu and Yuming Guo and Joanna F. Dipnall and Jennie Ponsford and Peter Cameron and Shanthi Ameratunga and Belinda J. Gabbe},
  title = {{Driver, Collision and Meteorological Characteristics of Motor Vehicle Collisions among Road Trauma Survivors}},
  journal = {International Journal of Environmental Research and Public Health},
  year = {2021},
  doi = {10.3390/ijerph182111380},
}
Figure