4F.003 Spatially forecasting serious injury into the future

Ben Beck, Belinda J. GabbeAbstracts 2021 · DOI

Abstract

Understanding how injury varies across space and time is critical for informing injury prevention activities at a population level. This study utilized spatiotemporal modelling to investigate spatial and temporal variation in major trauma across Victoria, Australia. Using data from the Victorian State Trauma Registry (2008–2018), researchers mapped 28,630 major trauma events to small statistical areas. Bayesian spatiotemporal models were employed to identify historical patterns and generate forecasted counts through 2023. The results revealed substantial spatial variation, with regional areas generally experiencing higher incidence rates and sharper increases over time compared to metropolitan areas. Furthermore, the study identified distinct geographic patterns based on the cause of injury: motor vehicle collisions were more prevalent in regional zones, while low falls predominated in metropolitan areas. These findings demonstrate that spatiotemporal forecasting can identify high-risk hotspots and inform targeted injury prevention activities.

Introduction

Injury prevention has traditionally relied on reactive reporting—analyzing where accidents have already occurred to justify safety measures. However, a "one-size-fits-all" approach often fails to account for the dynamic nature of population growth, infrastructure development, and shifting demographics. To move toward proactive prevention, public health officials require high-resolution, predictive intelligence that identifies not just where trauma has happened, but where it is likely to increase.

This paper details a retrospective review and forecasting study of major trauma in Victoria, Australia. By leveraging over a decade of registry data, the study aims to provide a granular understanding of the spatial and temporal "fingerprints" of serious injury, allowing for interventions tailored to the specific risks of local neighborhoods and regional corridors.

Methodology

The study analyzed data from the Victorian State Trauma Registry (VSTR), a comprehensive population-based database. The cohort included 28,630 major trauma patients with known event coordinates recorded between 2008 and 2018.

Bayesian Spatiotemporal Modelling

The technical foundation of this research is Bayesian spatiotemporal modelling. Unlike standard heatmaps, which can be susceptible to noise or bias in areas with small sample sizes, Bayesian models utilize "spatial smoothing." This technique borrows statistical strength from neighboring areas to produce more stable and accurate risk estimates.

The methodology involved three primary phases:

  1. Geocoding: Mapping ambulance-attended trauma event coordinates to small statistical areas.
  2. Trend Analysis: Investigating how injury incidence changed over the 11-year historical period.
  3. Forecasting: Projecting established patterns forward to generate predicted trauma counts for each small area through 2023.

The study revealed a significant global increase in the incidence of major trauma across the state of Victoria. However, this increase was not uniform, uncovering a widening "regional gap."

Regional vs. Metropolitan Divergence

Geography serves as a primary determinant of injury type and frequency. Area-specific incidence rates were generally higher in regional areas than in metropolitan zones. More concerningly, the relative increase in trauma incidence over time was greater in regional areas. This suggests that while safety improvements (such as vehicle safety technology or urban road design) may be mitigating risk in cities, these benefits are not reaching rural populations at the same rate.

Cause-Specific Variation

The models highlighted a sharp divide in the mechanisms of injury based on location:

  • Motor Vehicle Collisions (MVCs): These were the primary drivers of trauma in regional areas, likely due to higher travel speeds and different road infrastructure.
  • Low Falls: Conversely, metropolitan areas saw a higher incidence of serious injuries resulting from low-level falls, likely reflecting the demographic density and aging populations in urban centers.

The Role of Predictive Forecasting

By extending the model's reach to 2023, the research transforms from a historical record into a policy tool. Spatiotemporal forecasting enables health authorities to identify "emerging" hotspots—areas where injury rates are not only currently high but are accelerating.

These forecasted counts allow local governments and health providers to "pre-deploy" resources. For example, a regional area flagged for increasing MVCs might receive targeted road safety audits or enhanced emergency response positioning, while a metropolitan area flagged for low falls might see increased investment in geriatric care and home-safety community programs.

Conclusion

Spatiotemporal modelling provides a sophisticated lens through which to view the public health challenge of major trauma. By identifying small areas of high and increasing incidence, this study offers a roadmap for localized, evidence-based injury prevention. As trauma registries continue to evolve, the integration of predictive forecasting will be essential for reducing the burden of serious injury and ensuring that prevention resources are directed where they are needed most.

Summary updated

Abstract

<h3>Background</h3> Understanding how injury varies across space and time is critical for informing injury prevention activities at a population level. The aim of this study was to use spatiotemporal modelling to investigate spatial and temporal variation in major trauma. <h3>Methods</h3> A retrospective review of major trauma was conducted using the population-based Victorian State Trauma Registry (Victoria, Australia) from 2008 to 2018. Coordinates of ambulance attended major trauma event locations were mapped to small statistical areas. Bayesian spatiotemporal modelling was used to investigate spatial and temporal patterns and generate forecasted counts in each small area to 2023. <h3>Results</h3> Over the 11-year period, there were 28,630 major trauma patients with known event coordinates. Substantial spatial variation in the incidence of all major trauma was observed. Generally, area-specific incidence rates were higher in regional areas than metropolitan areas. Global temporal trends in the incidence of major trauma demonstrated a significant increase, with relative increases greater in regional areas compared to metropolitan areas. Differences in spatial and temporal variation were observed between causes of injury. For motor vehicle collisions, area-specific incidence rates were higher in regional areas than metropolitan areas. Conversely, for low falls, area-specific incidence rates were higher in metropolitan areas than regional areas. <h3>Conclusion</h3> Spatiotemporal forecasting enables the identification of small areas of relatively high incidence and of increasing incidence over time. Furthermore, these models can be used to derive forecasted counts of trauma counts that can be used to inform injury prevention activities at the small spatial area.
BibTeX
@inproceedings{beck2021f,
  author = {Ben Beck and Belinda J. Gabbe},
  title = {{4F.003 Spatially forecasting serious injury into the future}},
  booktitle = {Abstracts},
  year = {2021},
  doi = {10.1136/injuryprev-2021-safety.117},
}
Figure