Methods for analyzing matched designs with double controls: excess risk is easily estimated and misinterpreted when...

Area of Science:

• Epidemiology
• Biostatistics
• Traffic Safety Research

Background:

• Elections have been investigated for their potential impact on traffic risks.
• Previous analyses often did not adequately account for matched controls or distinguish between total and excess risk.
• Accumulating count data in matched studies requires specific analytical approaches.

Purpose of the Study:

• To demonstrate analytic methods for matched studies with double controls and accumulating count data.
• To clarify the difference between total risk and excess risk from matched and unmatched perspectives.
• To re-evaluate the relationship between elections and fatal traffic crashes using refined statistical techniques.

Main Methods:

• Review of past research on elections and traffic risks.
• Analysis of fatal crash counts using matched double controls.
• Application of Poisson regression to both total and excess crash counts.
• Comparison of results from total risk (unmatched perspective) and excess risk (matched perspective) analyses.

Main Results:

• A total of 1,546 individuals were in fatal crashes on 10 election days, versus 2,593 on 20 control days.
• Analysis of total counts showed a relative risk of 1.19 (95% CI: 1.12-1.27).
• Analysis of excess counts revealed a significantly higher relative risk of 3.22 (95% CI: 2.72-3.80).
• The discrepancy between total and excess risk analyses was consistent across models and visualizations.

Conclusions:

• The study provides robust analytical approaches for count data in matched designs with double controls.
• It highlights the critical distinction between total risk and excess risk, with excess risk being a more sensitive indicator.
• The findings underscore the substantial, often underestimated, increase in traffic risks associated with election periods.