Methods

The FARS database contains comprehensive data about crash situations, drivers and passengers, and vehicles, involved in US traffic fatalities from 1975 onward (analyses used 1975–1998 data). It has built-in quality control mechanisms, and allows for the control of many confounders. For further details on the FARS database, visit their website (http://www-fars.nhtsa.dot.gov/).

Partner websites:
* Cars22.com: Allows you to search for automobiles in the Middle East and Africa. Buy second hand cars in Dubai, Egypt, Israel, Jordan, Syria and other cities.
* Morettiwheels.com: Offers chrome wheels, custom racing rims including brand names such as Diablo, Antera, Brabus, Racinghart, Lowenhart, Alba, KMC, Lexani, Azev and others.
* Traxms.com: Offers Polaris ATV parts, racks, snow plow, modular helmets, motorcycle tires and jackets.
* Autopartsgiant.com: Offers auto body parts for domestic and import cars including Mercedes, BMW, Nissan, Volkswagen and others. Parts include brake pads, clutch kits, fuel and water pumps or other accessories.
* Schuckhitches.com: Offers tractor and truck trailer hitches, towing equipment and accessories for your agricultural, commercial and recreational towing needs.

The analyses presented were based on driver fatalities only to factor out the effect of seating position. Seating position affects the risk of a fatal injury, with rear center seats as the safest location in a vehicle. Also, in multiple vehicle crashes, the risk of fatality is dependent on the characteristics of the other vehicle(s). Therefore, to maximize the interpretability of findings, a model focusing on driver fatalities involving single-vehicle crashes with fixed objects was chosen. Focusing on drivers removed the confounding impact of seating position, and focusing on single-vehicle crashes removed the effect of other vehicle characteristics.

Because crashes are included in FARS only if a fatality ensued, all drivers of single-vehicle crashes in which they were the sole occupants were killed, effectively removing any variability, and possibly biasing the results of the analyses. Including single-occupant crashes in the analyses would have led to a sample of 85% fatalities. To alleviate this difficulty, analyses were performed on crashes where at least two vehicle occupants were present. In all cases, the crash was sufficiently serious to kill at least one occupant, but not necessarily the driver. A total of 110 813 drivers were available for the analyses, and approximately 50% of all drivers were fatally injured. However, because of missing data, the analyses did not include all 110 813 drivers.

Driver characteristics studied included age category (<20, 20–29, 30–39, 40–49, 50–64, 65–79, 80+ years), gender, blood alcohol concentration (BAC) (0, 0.01–0.04, 0.05–0.09, 0.10–0.14, 0.15–0.19, 0.20–0.29, 0.30+), and seatbelt use (none, lap only, shoulder only, lap+shoulder). Crash characteristics included impact direction based on the four quadrants (front, right side, rear, left side), vehicle deformity (severe, less severe), and vehicle speed immediately before or at impact (<56 kilometers per hour (kph) (35 mph), 56–95 kph (35–59 mph), 96–111 kph (60–69 mph), 112+ kph (70 mph)). Vehicle characteristics included air bags (deployed, not), weight (kg), wheelbase length (cm), model year, and vehicle age. Injury severity was dichotomized as ‘fatal' or ‘non-fatal'.

Descriptive data were compiled by age category for driver and crash characteristics. To determine whether driver, crash, and vehicle characteristics were associated with fatalities, the odds ratio (OR) along with 99% confidence interval (99% CI) of a fatal injury were calculated for each variable with univariate statistics (unadjusted OR), followed by a multivariate logistic regression using backward variable selection to control for driver, crash, and vehicle characteristics (adjusted OR). Statistical significance was concluded when the 99% CI did not include unity ( P <0.01).