Case-control analysis in highway safety: Accounting for sites with multiple crashes

Area of Science:

• Epidemiology
• Transportation Safety
• Statistical Analysis

Background:

• Epidemiological methods, such as case-control and cohort studies, are increasingly used in highway safety analysis.
• These methods identify risk factors by examining characteristics associated with an outcome, analogous to disease in medical research.
• A key limitation in applying these methods to highway safety is the handling of sites with multiple crashes, unlike single occurrences of disease.

Purpose of the Study:

• To explore the application of the case-control method in highway safety analysis.
• To address the limitation of not accounting for sites with multiple crashes.
• To present and compare two alternative case-control methods for handling multiple crashes.

Main Methods:

• The study compares a "typical" case-control approach with two novel methods for defining cases.
• Method 1: Defines a separate case for each crash, increasing the weight of roadway characteristics.
• Method 2: Defines entire crash categories (e.g., sites with one crash, two crashes) as separate cases for analysis.

Main Results:

• The "typical" case-control design, which groups all crashes, may underestimate the safety effects of factors compared to methods accounting for multiple crashes.
• Results indicate that the "typical" design yields less pronounced effects (closer to unity) than the first alternative method.
• The second alternative method provides evidence that grouping single and multiple crash sites together is inappropriate.

Conclusions:

• There is a clear need to differentiate between sites with single and multiple crashes in case-control analyses for highway safety.
• Case-control designs can be adapted to account for multiple crashes at a site.
• Further research is required to determine the optimal method for addressing multiple crashes in highway safety studies.