Matched-pair cohort methods in traffic crash research

Area of Science:

• Epidemiology
• Traffic Safety Research
• Statistical Analysis

Background:

• Standard matched-pair analysis often excludes pairs without outcomes, limiting traffic fatality studies.
• Crash databases may lack data on non-fatal incidents, necessitating specialized analysis for matched occupants.
• Confounding by vehicle and crash factors persists even with matched occupants due to differing seat positions.

Purpose of the Study:

• To review and compare methods for estimating relative risks in matched-pair traffic crash data.
• To address confounding issues arising from seat position and other crash-related variables.
• To evaluate the strengths and limitations of different analytical approaches for crash data.

Main Methods:

• Review of Mantel-Haenszel stratified methods, noting potential bias with seat position association.
• Examination of the double-pair comparison method designed to mitigate seat position confounding.
• Analysis of conditional Poisson and Cox proportional hazards regression, requiring interaction terms for unbiased estimates.

Main Results:

• Mantel-Haenszel methods can yield biased relative risk estimates when seat position influences the outcome.
• The double-pair comparison method offers improved control for seat position confounding.
• Regression models provide unbiased estimates but necessitate careful specification of interaction terms.

Conclusions:

• The choice of method for analyzing matched-pair crash data depends on the specific confounding factors present.
• Conditional Poisson and Cox regression models offer robust estimation but require complex model specification.
• Accurate risk estimation in traffic safety necessitates addressing seat position and other potential confounders.