(1) The American Academy of Pediatrics, the National Highway Traffic Safety Administration, and other governmental, private, and professional organizations have recommended legislation to institute a program of graduated driver licensing (GDL). These programs remove driving restrictions in stages as the TD gains experience and maturity, until full licensure is granted a minimum of 18 months after the learner's permit is obtained. (17) This type of program has been shown to decrease the crash rate of TDs in New Zealand, Canada, and several states in the United States. (18-22) Each of these programs incorporates different features of the National Highway Traffic Safety Administration recommendations, yet it has not been well established which components are the most effective. In the present study, we determined the population-based rate of crashes that involved TDs in a single state, hospital charges associated with all people involved in those crashes, and the characteristics of these crashes to determine the potential effectiveness of various components of GDL programs. We hypothesized that TDs are overrepresented in crashes, especially at night, that TDs wear seatbelts less often, and that the presence of passengers significantly increases the risk of a crash that results in fatal or nonfatal injury.

METHODS

Data Sources

Four data sets were used in these analyses: 1) Utah MVC records; 2) the Utah hospital discharge database; 3) the Utah Emergency Department database; and 4) the Utah Division of Motor Vehicles driver license file. The 1992 to 1996 MVC records were obtained from the Utah Department of Transportation, Division of Traffic and Safety. This crash database contains information on all reportable crashes (defined as occurring on a public road and resulting in at least 1 injury or at least $1,000 in property damage). These data are collected on reports that are filled out by police officers at the scene of the crash. Data points include detailed information on the time, location, type of crash, vehicles, drivers, and any other people involved (eg, passengers, pedestrians).

The 1992 to 1996 hospital discharge database was acquired from the Utah Health Data Committee/Office of Health Data Analysis, Utah Department of Health. The Utah Department of Health mandates that all licensed hospitals report all inpatient admissions. The hospital discharge database includes demographic information and up to 9 International Classification of Diseases-Ninth Revision diagnosis codes and 5 procedure...

The Department of Health mandates that all Utah licensed hospitals report information on ED patient encounters. The ED database includes demographic information, up to 9 International Classification of Diseases-Ninth Revision diagnosis codes and 5 procedure codes, up to 2 E. codes, discharge status, and hospital charges. Patients in this database were by definition treated and released from the ED.
The driver license file includes all information required for a driver's license, as well as the original and current issue dates and expiration date. The Institutional Review Board of the University of Utah approved use of these databases for this study.

Probabilistic Linkage

The database that was used for analysis was created by probabilistically linking the crash and inpatient databases for 1992 to 1996. ED data were included in the 1996 linkage. Jaro (23) described probabilistic linkage. By comparing several common fields, probabilistic linkage uses an iterative approach to mathematically linking databases. The variables used for linkage in this study included date of incident, date of birth, hospital code, county code, city code, gender, and age. Successful linkage is affected by errors in the data sources as well as the effectiveness of the method used to achieve the linkage. Because the crash database includes noninjured persons, a substantial portion of the crash database is expected not to link with an inpatient or ED record.

To eliminate 16- or 17-year-old teenagers who were not yet licensed, as well as older unlicensed drivers, the driver license file was linked to the crash database and any unmatched drivers were eliminated from our analyses. Probabilistic record linkage was performed using Automatch 4.2 (Matchware Technologies, Inc., Boston, MA).

Measurements

To study the group of inexperienced drivers that would be most affected by interventions such as a GDL program, we compared 16- to 17-year-old drivers with 18- to 59-year-old drivers. Sixteen- to 17-year-old drivers are referred to as TDs, and 18- to 59-year-old drivers are referred to as adult drivers. A person who is "hospitalized" is one who is probabilistically linked to a hospital discharge record, and a person who is seen in the ED is one who is linked