«by Johnathon P. Ehsani A dissertation submitted in partial fulfillment of the requirements for the degree of Doctor of Philosophy (Health Behavior ...»
Analytical Method Monthly crash rates per 100,000 population of 16-, 17-, and 18-year-old drivers were analyzed using Auto-Regressive Integrated Moving Average (ARIMA) interrupted time series analysis (McCleary and Hay 1982) for each state. Interrupted time-series analyses compare observations before and after some identifiable event, with the goal of evaluating the impact of the intervention. The transfer function relates an intervention to its effect on crash rates. In this analysis, the transfer function has two parameters. The first parameter, !, is the magnitude of the asymptotic change (rise or fall) in level after the intervention. The second parameter, !, reflects the onset of the change. If the null hypothesis that ! is 0 is supported, there is no impact of the intervention. If ! is significant, the size of the change is ! as a percentage (Tabachnick and Fidell 2007).
For these analyses, ! was fixed at 0, meaning the anticipated change in crash rates would be abrupt and lasting, referred to as a sudden-impact permanent-change model.
Analytical strategy For Michigan and Maryland, analyses were conducted using fatal/disabling injury, non-disabling injury, and possible-injury/PDO crash rates as three separate outcome measures. Florida crash data showed evidence of underreporting that varied across time for ‘possible-injury/PDO’ crashes. Therefore, this crash severity category was excluded from the Florida analyses.
For each state, the models were estimated using the natural logarithm of the monthly crash rates per 100,000 population. Using the natural logarithm, the coefficient representing the intervention effect (!) is directly interpretable (using the formula 100 x [e! – 1]) as the percentage change in the post-intervention series relative to the preintervention series (McDowall, McCleary et al. 1980). Results reported are based on the models using the natural logarithm of crash rates as the outcome variable.
The analyses were conducted in three stages. First, a linear regression model was estimated for the teen driver crash rates and the covariates: adult crash rates, gas prices, and GDL laws. Second, the model for each state was statistically adjusted for trends and seasonal variation. Autoregressive and moving average orders were identified using auto-correlation and partial-auto-correlation functions of the series residuals. Finally, the original regression model was re-estimated with the inclusion of the autoregressive or moving average orders identified in the second stage. Outliers were also detected and controlled for in the final model. Analyses were conducted using the SCA Time Series and Forecasting System, a specialized time-series analysis software package (Scientific Computing Associates 2011).
For each state, crash rates increased with age, across each level of crash severity (Table 4.2). Crash rates were higher prior to the introduction of GDL, than after GDL. This observation held for each state, year-of-age, and crash severity level, with a single exception. In Florida, 18-year-old drivers’ non-disabling injury crashes increased following the second revision of Florida’s GDL program. Fatal/disabling, and nondisabling crash rates were highest among Florida teens relative to teens in Michigan and Maryland. Possible-injury/PDO crash rates were higher for all three age groups of teen drivers in Michigan, compared to Maryland.
16-year-old drivers’ crash rates The results of the analysis partially confirm the first hypothesis: 16-year-old drivers’ crash rates significantly declined in each state following the introduction of GDL or revision of existing GDL programs. However, the decrease was not uniform and varied across each level of crash severity (Table 4.3). The introduction of GDL in Florida in July 1996 was not followed by a significant decline in 16-year-old drivers’ crash rates (Tables 4.3 and 4.4). However, Florida’s second GDL program, effective in October 2000, resulted in significant declines in fatal/disabling injury (-5.4%) and non-disabling injury crash rates (-6.8%). Following the introduction of GDL in Michigan in April 1997, 16-year-old drivers’ crash rates significantly declined in each level of crash severity.
Fatal/disabling injury crashes declined by over one fifth (-21.6%), non-disabling injury crashes were 6.7% lower, and possible-injury/PDO crashes fell 23.1% (Tables 4.3 and 4.5). In Maryland, 16-year-old drivers’ non-disabling injury crash rates were significantly lower (-12.1%) following the introduction of GDL in July 1999. Maryland’s revision of GDL in October 2005 was followed by significant declines in 16-year-old drivers’ nondisabling injury and possible-injury/PDO crash rates, which were 26.4% and 11.2% lower respectively (Tables 4.3 and 4.6).
17-year-old drivers’ crash rates The pattern of results was similar, although weaker, for 17-year-old drivers’ crash rates (Table 4.3). Seventeen-year-old drivers’ crash rates significantly declined in each state following the introduction of GDL or a revision of existing GDL programs, although this decline varied across levels of crash severity. Florida’s first GDL program was not associated with any changes in 17-year-old drivers’ crashes. However, similar to the pattern observed among 16-year-old drivers, Florida’s revised GDL program was followed by significant declines in fatal/disabling and non-disabling injury crashes, falling 9.3% and 7.9% respectively (Tables 4.3 and 4.4). In Michigan, 17-year-old drivers’ fatal/disabling injury crash rates were 10.9% lower following the introduction of GDL, while possible-injury/PDO crashes declined by 2.9% (Tables 4.3 and 4.5). In Maryland, 17-year-old fatal/disabling injury crashes decreased by 14.3% following the introduction of GDL, and were further reduced 25.4% following revision of the program. While there was no effect on non-disabling injury crash rates, Maryland’s revised GDL program resulted in an 8.7% decline in possible-injury/PDO crash rates (Tables 4.3 and 4.6).
18-year-old drivers’ crash rates In Florida, the introduction and revision of GDL was not associated with any significant changes in 18-year-old drivers’ crash rates (Tables 4.3 and 4.4). In Michigan, the introduction of GDL was followed by a statistically significant 3.6% increase in possible-injury/PDO crashes among 18-year-old drivers, but no change in fatal/disabling or non-disabling injury crashes (Tables 4.3 and 4.5). In Maryland, where GDL applied to novice drivers of all ages, the introduction of GDL was not associated with an increase in 18-year-old drivers’ crash rates at any level of crash severity. On the contrary, 18-yearold drivers’ possible-injury/PDO crash rates significantly declined (-6.9%) following the first revision of Maryland’s GDL program (Tables 4.3 and 4.6). These results provide modest support for the third hypothesis of this study, that there will be an increase in 18year-old drivers’ crash rates in states where GDL applies only to 15- to 17-year-old drivers, and that in states where GDL applies to novice drivers of all ages, 18-year-old drivers’ crash rates will remain unchanged or decrease.
This study’s purpose was twofold. The first was to examine the effects of GDL on 18-year-old drivers’ crash rates (while confirming positive outcomes among 16- and 17year-old drivers). The second was to shed light on a potential mechanism responsible for any increase in 18-year-old drivers’ crash rates by comparing states where GDL provisions apply exclusively to 15- to 17-year-old drivers (Florida and Michigan) with a state where the GDL learner permit requirement applies to novice drivers of all ages (Maryland). In the absence of licensing data, it was not possible to determine conclusively whether the observed changes in 18-year-old drivers’ crashes were entirely due to a proportion of teens being licensed for the first time at age 18, or because teens licensed under GDL lack sufficient independent driving experience. However, using this carefully selected sample of states, we tested our hypotheses on three levels of crash severity, employing an analytical approach that accounted for long term trends. After adjusting for confounding factors, we found that 18-year-old drivers’ possible-injury/PDO crash rates increased in Michigan and declined in Maryland, providing some support for the ‘offset effect,’ whereby 18-year-old novice drivers’ crashes may increase following implementation of GDL.
The 3.6% increase in 18-year-old drivers’ possible-injury/PDO crash rates in Michigan supports our hypothesis that 18-year-olds’ crashes increase in states where GDL applies exclusively to 15- to 17-year-old drivers. The absence of a change in crashes of greater injury severity raises issues regarding the differential effects of GDL on crash types, and the extent of GDL’s effect on 18-year-old drivers. Reductions in possible-injury/PDO crashes are consistent with GDL’s objective to protect novice drivers from the consequences of their inexperience as they learn to drive independently. GDL is not a program to control risk-taking behaviors such as speeding and alcohol-impaired driving, which are often involved in the differing etiology of fatal and disabling injury crashes (Lam 2003). Consequently, GDL should influence crashes attributable to lack of experience more than those crashes attributable to deliberate risktaking or neglectful driving behavior (McKnight and McKnight 2003). The absence of a change in 18-year-old drivers’ crashes in Florida has been observed in previous evaluations of Florida’s GDL program (Ulmer, Preusser et al. 2000), and may be due to the unfortunate exclusion in these analyses of possible-injury/PDO crashes, or other state-specific factors.
One such state-specific factor, for example, is that unlike Michigan, Florida’s GDL was incrementally implemented in two stages. States where GDL components have been gradually introduced and the concept of graduated licensing has become an accepted norm, such as Florida, may be less likely to experience an increase in a proportion of teens delaying licensing until age 18 in response to a change in the law. In states that had placed few restrictions on teen driving previously, and subsequently introduced a comprehensive GDL law at a single time point, such as Michigan, GDL may represent a profound shift in the culture of driving for teens, resulting in a greater number of teens delaying licensing until age 18 in order to avoid the provisions of the law. Other factors may be at play, such as required driver education or the cost of driver education.
The 6.9% reduction in 18-year-old drivers’ possible-injury/PDO crash rates in Maryland also provides support for the offset hypothesis that 18-year-old drivers’ crashes decrease in states where GDL applies to novice drivers of all ages. Unlike the declines in 16- and 17-year-old drivers’ fatal/disabling and non-disabling injury crashes following the introduction of GDL in Maryland, possible-injury/PDO crashes were the only crash type that declined for 18-year-old drivers. This finding may be due to the unique provision of Maryland’s GDL that requires all novice drivers, regardless of age, to complete a learner permit phase, although restrictions on passenger and nighttime driving applies only to novice drivers below age 18. The absence of a decline in the most severe crash types may be the consequence of drivers age 18 being unprotected from the high-risk conditions of driving with peer passengers (Chen, Baker et al. 2000) and driving at night (Williams 2003).
Given that a long-term safety benefit of advancing through GDL has been previously reported (Masten and Foss 2010), it is plausible that reductions in 18-year-old drivers’ crashes resulting from safer driving by teens who have completed GDL are being masked by increased crash rates of newly licensed 18-year-old drivers. The increase in 18-year-old drivers’ fatal crashes in California reported by Males (Males
2007) may suggest a greater proportion of teens becoming licensed at age 18 in California, relative to Michigan or Florida. Similarly, Masten’s national findings (Masten, Foss et al. 2011) may have been skewed by the proportion of teens licensed at age in the two largest states, California and Texas, that constitute over 20% of the fatal crashes in the United States.
Our findings build on previous research that examined GDL effects in the states used in this study. Ulmer and colleagues concluded that the introduction of Florida's first GDL program in July 1996, reduced fatal and injury crash rates among 15- to 17-yearold drivers by 9%, but not 18-year-old drivers’ crash rates (Ulmer, Preusser et al. 2000).
By extending the post-GDL evaluation period in Florida and adjusting crash rates for long term trends, we observed significant reductions in 16- and 17-year-old drivers’ fatal/disabling and non-disabling injury crash rates following the enhancement of the existing GDL program in October 2001, and also found 18-year-old drivers’ crash rates remained unchanged for the duration of the study period. Shope and associates concluded that Michigan's GDL reduced the overall crash rate for 16-year-old drivers by 25% (Shope, Molnar et al. 2001), with a similar reduction reported in a four-year follow up study (Shope and Molnar 2004). By including older teen drivers in this study, we demonstrated that the protective effects of Michigan’s GDL extended to significant declines in 17-year-old drivers’ fatal/disabling injury crashes and possible-injury/PDO crashes. Using three levels of crash severity, we exposed the differential effects of Michigan’s GDL on crash outcomes, suggesting that the largest decreases in crashes were concentrated among the most and the least severe crash types. For Maryland, Kirley reported an 18% decline in the overall crash rate, and a 37% decline in the nonfatal crash rate for 16-year-old drivers in Maryland following the introduction of GDL in July 1999 (Kirley, Feller et al. 2008). Using a longer follow-up period, we were able to detect significant declines in 16-, 17- and 18-year-old drivers’ crash rates following the changes in Maryland’s GDL in October 2005, and observed that those reductions were not uniformly distributed by crash severity.