Abstract
Spatiotemporal modelling techniques allow for the prediction of injury across time and space, yet these methods remain underutilised in injury research. This study employs statistical spatiotemporal modelling to identify areas of significantly high injury risk and those witnessing increasing risk over time in Victoria, Australia. Analyzing 31,317 hospitalised major trauma patients from the Victorian State Trauma Registry (2007–2019), we utilized Bayesian models to quantify patterns across 79 local government areas (LGAs). Our findings reveal a significant 2.1% annual increase in injury incidence, with distinct geographical profiles: motor vehicle collisions (MVCs) predominate in regional areas, while low-level falls are surging in metropolitan centers. Crucially, the greatest increases in injury incidence were observed in the most socioeconomically disadvantaged areas, highlighting a growing inequity in injury risk.
Introduction
Injury surveillance is a cornerstone of public health, providing the data necessary to design and evaluate prevention strategies. However, traditional surveillance often relies on "snapshots"—static maps or tables that display injury counts at a single point in time. These methods frequently fail to account for the dynamic nature of risk, which evolves across both space and time. Furthermore, simple descriptive statistics often lack the robustness to handle sparse data or provide reliable measures of uncertainty.
This research addresses these limitations by applying sophisticated Bayesian spatiotemporal modeling to major trauma data. By understanding not just where trauma occurs, but how the risk in specific areas is trending relative to the rest of the state, policymakers can transition from broad, "one-size-fits-all" prevention strategies to geographically targeted interventions that address specific local risks and emerging inequities.
Methods and Data Sources
The study performed a retrospective review of major trauma patients recorded in the Victorian State Trauma Registry (VSTR) between 2007 and 2019. Major trauma was defined by an Injury Severity Score (ISS) > 12, admission to an intensive care unit for more than 24 hours, urgent surgery, or death from injuries. A total of 31,317 patients met the inclusion criteria.
To analyze this data, the researchers employed Bayesian spatiotemporal models across Victoria’s 79 Local Government Areas (LGAs). Key technical components included:
- Besag-York-Mollie (BYM) Model: This framework accounts for spatial autocorrelation—the principle that neighboring areas share similar risk profiles. This allows the model to "borrow strength" from adjacent regions, producing stable estimates even in areas with low population counts.
- Integrated Nested Laplace Approximations (INLA): The study utilized the
R-INLApackage for Bayesian inference, providing a computationally efficient alternative to traditional Markov Chain Monte Carlo (MCMC) methods. - Covariates: The models were adjusted for geographical remoteness (using the ARIA+ index) and socioeconomic status (using the IRSAD index) to identify underlying drivers of trauma trends.
Key Findings
The "Two Victorias" of Trauma
The analysis revealed a stark geographical divide in the causes of major trauma. Regional and outer-regional LGAs face a significantly higher risk from motor vehicle collisions (MVCs), with a median relative risk of 1.48 compared to 0.52 in major cities. Conversely, metropolitan Melbourne is characterized by a high risk of "low falls" (falls from less than one meter), with a median relative risk of 1.41 compared to 0.82 in outer regional areas.
The Surge in Low-Level Falls
While the incidence of MVCs remained relatively stable over the 13-year study period, injuries resulting from low falls increased by an average of 4.0% annually. This trend likely reflects an aging population and suggests that urban infrastructure and public health initiatives may need to prioritize "age-proofing" environments to mitigate fall risks.
Widening Socioeconomic Inequities
One of the most significant findings is the correlation between socioeconomic disadvantage and trauma trends. The most disadvantaged LGAs (Quintile 1) showed a median relative risk for MVCs of 1.44, while the most advantaged (Quintile 5) stood at 0.43. More alarmingly, the greatest year-on-year increases in overall trauma incidence were concentrated in these disadvantaged communities, suggesting that the "injury gap" is widening over time.
Conclusion
Spatiotemporal modeling provides a powerful lens through which to view public health challenges. By identifying a study-wide 2.1% annual increase in major trauma and pinpointing the specific areas where this risk is "heating up," this research provides a roadmap for targeted intervention. The findings underscore the need for injury prevention initiatives that are not only geographically specific—addressing road safety in regional areas and fall prevention in cities—but also socially conscious, focusing resources on the most disadvantaged populations to reduce growing health inequities.
Limitations
While robust, the study notes that spatial aggregation at the LGA level can obscure finer-grained details of where injuries occur. Additionally, as an ecological study, the results describe area-level risks rather than individual-level probabilities. Future research could benefit from incorporating exposure-based denominators, such as vehicle kilometers traveled or pedestrian traffic counts, to further refine risk assessments.
Summary updated