Correcting sample selection in FARS data to estimate seatbelt use

Area of Science:

• Traffic safety research
• Public health surveillance
• Statistical modeling

Background:

• Estimating national seat belt usage is crucial for public health and traffic safety initiatives.
• Observational data, such as the National Occupant Protection Use Survey (NOPUS), are commonly used but can be subject to biases.
• The Fatality Analysis Reporting System (FARS) offers a rich dataset but has not been traditionally used for estimating usage rates.

Purpose of the Study:

• To evaluate the Fatality Analysis Reporting System (FARS) as a viable alternative to observational data for estimating national seat belt usage.
• To address and correct for potential sample selection bias in FARS data when estimating seat belt use.
• To determine if corrected FARS data can provide comparable or improved estimates of seat belt usage rates.

Main Methods:

• Developed statistical methods to correct for sample selection bias in FARS data.
• Established lower and upper bounds for seat belt usage rates based on independence assumptions.
• Compared seat belt usage estimates derived from corrected FARS data with existing NOPUS estimates.

Main Results:

• Corrected FARS data provides seat belt usage estimates comparable to NOPUS.
• The methodology successfully accounts for sample selection bias, improving the reliability of FARS-based estimates.
• A lower and upper bound for seat belt usage rates were established, providing a range for accurate estimation.

Conclusions:

• Corrected FARS data is a comparable alternative to NOPUS for estimating seat belt usage in the United States.
• Researchers can leverage corrected FARS data to complement NOPUS, benefiting from its detailed cross-sectional information.
• This advancement allows for more robust analysis of factors influencing seat belt use and related research questions.