An application of proportional hazards model to study the recurrent time between traffic accidents or infractions and subsequent fatal automobile crashes, 1986–1988

Abstract

Drivers who have a record of traffic accidents or infractions may be at increased risk of being involved in a subsequent fatal crash. To develop effective intervention programs, it is important for us to quantify the joint effects of suspected covariates that may affect the recurrent time, which we define as the time interval between the date of the last traffic accident or infraction and the date of the subsequent fatal crash. We analyze the data from the Fatal Accident Reporting System (FARS) for the years 1986 through 1988. Because the FARS includes only drivers who already have been involved in a fatal crash and excludes drivers who have had traffic accidents or infractions but have not yet been involved in any fatal crash, the application of the proportional hazards model directly to the recurrent time considered here is not appropriate. We can, however, apply the proportional hazards model with time-dependent risk set defined by using the reverse hazard function. On the basis of the resulting model, we have found that all the covariates, including age, sex, usage of seat belts, and alcohol involvement during the fatal crashes, have significant effects on the recurrent time. In general, a male, unbelted driver between 16 and 25 years of age, under the influence of alcohol during a fatal crash, and without a valid license tends to have relatively short recurrent times.