Slippery Road Conditions and Fatal Motor Vehicle Crashes in the Northeastern United States, 1998–2002

Abstract

Objectives. We investigated risk factors for fatal motor vehicle crashes on slippery roads in the Northeastern United States, 1998–2002.

Methods. We analyzed data from the Fatality Analysis Reporting System of the National Highway Traffic Safety Administration.

Results. Rates of crashes on slippery roads, and ratios of crashes on slippery roads to crashes on dry roads, were greatest among the youngest drivers. Among those aged 16 to 19 years, logistic regression analysis showed significant, independent risks associated with excessive speed for conditions (odds ratio [OR]=1.38), time of day (OR=1.80 for 5:00 to 9:00 am vs 10:00 am to 2:00 pm), time of year (OR=6.17 for January vs July), type of road (OR=1.27 for rural vs urban roads ), and age (OR=1.19 for those aged 16 to 17 years vs those aged 18 to 19 years). Licensure from states with graduated licensing programs was protective against crashes attributed to swerving on slippery roads (adjusted OR = 0.63). Risk factors among drivers older than 19 years were similar but peaked at different times of day and included increased risks for women compared with men.

Conclusions. Driver training programs need to better address hazards presented by slippery roads.

Motor vehicle crashes were the leading cause of death among persons aged 16 to 19 years during 1998–2001, both in the United States as a whole and specifically in the Northeastern states.1 Among factors affecting automobile deaths, road slickness has received only limited attention. Several research groups have found that during rainstorms drivers fail to reduce speeds enough to account for increased braking distances on wet road surfaces.2^–4 Researchers in Finland showed that excessive or moderate speed on slippery road surfaces, even in the absence of alcohol consumption, was the background cause of 47% of fatal loss-of-control crashes in that country.5^{(Table 4)} A study of police reports on crashes involving teenaged drivers in California and Maryland showed that drivers aged 16 to 17 years were more likely than those aged 18 to 19 years to have been driving too fast for conditions.6

In an effort to reduce teenaged driving deaths and injuries, the federal government has encouraged states to adopt graduated licensing programs (GLPs). These programs put various restrictions on novice drivers, such as requiring the novice driver to be supervised by a licensed driver, usually a parent, guardian or driving instructor, for a certain number of hours before being allowed to drive solo, or restricting younger drivers from driving at night or with multiple occupants in the car. Seventeen states enacted GLPs in the period 1996–1999.7 These programs have been shown to be successful in reducing crashes among teenaged drivers.8^,9 GLPs do not restrict driving during adverse weather conditions, and perhaps an even-greater reduction in crashes involving teenaged drivers might result if GLPs increased attention to good driving practices on slippery roads.

We ask the following questions: (1) Are younger drivers at greater risk than older drivers for crashes on slippery roads? (2) Is there a seasonal pattern to fatal crashes on slippery roads? (3) Are there risk factors associated with crashes on slippery roads? If risk factors can be identified, they may point to interventions that could reduce teenaged driving deaths and injuries.

METHODS

Data on fatal crashes occurring in the United States during the years 1998–2002 were obtained from the Fatality Analysis Reporting System (FARS).10^,11 The FARS data set includes all crashes involving motor vehicles in which at least 1 person died. Data from the 5 years, 1998–2002, were combined using SAS version 9.00 (SAS Institute Inc, Cary, NC).

We first calculated fatal crash rates, using as numerators the numbers of crashes from the FARS database involving drivers licensed in the Northeastern states (defined as Massachusetts, Connecticut, New Hampshire, New Jersey, New York, Pennsylvania, Rhode Island, Vermont, and Maine) irrespective of where the crashes occurred. Although the environmental variation incurred by including crashes outside the Northeast may have been substantial, denominator data (i.e., the numbers of licensed drivers by age, gender, and state) were available for the calculation of rates.12 We included in the numerators only crashes among persons with full licenses, as that was the base population reflected in the denominators. Crashes were further limited, using the FARS variable SUR_COND, to those that occurred on roadway surface conditions noted as “dry,” “wet,” “snow or slush,” or “ice.” We deleted data for crashes that occurred on unusual roadway surface conditions, noted as “sand, dirt, oil,” “other,” or “unknown.”

We next analyzed the 34501 crashes that occurred in the Northeastern states from 1998 through 2002 irrespective of the places of licensure of the drivers, in order to obtain a group of crashes that occurred under similar climatic and road conditions. From these data we deleted 4 096 (11.9%) crashes in which the driver did not have a valid license or learner’s permit or for which the driver’s license status was unknown, so analyses could address the question of why persons with valid driver’s licenses or learner’s permits were involved in fatal accidents.

We also deleted 203 crashes that occurred on unusual surfaces, 94 crashes in which the hour of the crash was unknown, 9 crashes in which the age of the driver was unknown, 3 crashes in which the driver was younger than 16 years (all were aged 15 years), and 6 crashes in which the gender of the driver was unknown. The final study sample contained data on 30 090 fatal crash records. Of the 30 090 fatal crashes, 3.84% (1154) involved drivers licensed outside the Northeastern states; the most common states of licensure were Florida (194), Maryland (124), Ohio (121), and Virginia (98).

We created dichotomous variables from the FARS variables used to code crash-related factors—driver level (DR_CF1, DR_CF2, DR_CF3, and DR_CF4). Each of these FARS variables can contain a single numeric code selected from a list that includes physical factors, mental factors, extravehicular environmental factors, and intravehicular environmental factors, such as distractions from mobile telephone use. The dichotomous variables we created summarized the information in these 4 individual fields so that they were positive if the driver-related factor code of interest was present in any of the 4 fields, and negative if it was absent. The variables created in this manner were excessive speed for conditions or speed in excess of the speed limit (FARS code 44), possible distractions inside the vehicle (cellular telephones, computers, and other electronic devices inside the automobile; FARS codes 93–98), and driver drowsy, sleepy, asleep, or fatigued (FARS code 1). An alternative dependent variable, crash related to swerving due to rain, snow, or other factors, was created that was positive if any of the 4 FARS variables contained code 87, indicating a driver-related factor of avoiding, swerving, or sliding due to ice, water, snow, slush, sand, dirt, oil, wet leaves in road ).

Another dichotomous variable was created from the FARS variable ROAD_FNC (roadway function class) that indicated either a rural road (ROAD_FNC codes 1–9) or an urban (ROAD_FNC codes 11–19). We also analyzed time of day, month, and year of the crash and the age, gender, and state of licensure of the driver. Crashes were further characterized as those involving drivers who were licensed in states with GLPs introduced before January 1, 2000, including California, Colorado, Delaware, Florida, Georgia, Illinois, Iowa, Louisiana, Maryland, Massachusetts, Michigan, New Hampshire, Nebraska, North Carolina, Ohio, Pennsylvania, and Rhode Island.7

In addition to calculating rates, we analyzed the data by computing odds ratios, or the odds that a risk factor would be present in crashes on slippery roads divided by the odds that the risk factor would be present in crashes on dry roads. We used EGRET version 2.0.3 (Cytel Software Corp, Cambridge, Mass.) to calculate these odds ratios and their 95% confidence limits, and also to create multivariable models by means of logistic regression analysis. Multiple logistic regression models were constructed for crashes on slippery roads among 16- to 19-year-old drivers considering for entry all variables investigated in univariate analyses except those for which the direction of association of the univariate analysis was counterintuitive.

To investigate whether risk factors might differ among those older than 19 years from those aged 16 to 19 years, we also constructed a multiple logistic regression model with the use of data on crashes with drivers of all ages, categorized into 7 age strata (16 to 19, 20 to 23, 24 to 33, 34 to 43, 44 to 56, 57 to 69, and 70 years or older). These strata were chosen by dividing the age distribution of drivers involved in fatal crashes into approximate quintiles and then further subdividing each of the extreme upper and lower quintiles approximately in halves. We considered for entry into this model the factors that had been significant in the model for drivers aged 16 to 19 years as well as gender, gender-by-risk-factor interaction terms, and age-by-risk-factor interaction terms. Finally, we modeled crashes among drivers aged 16 to 19 years that were attributed in the FARS database to swerving because of slippery road conditions compared with all other crashes.

RESULTS

During the study period, there were 30 806 fatal crashes that involved drivers with full licenses from the Northeastern states, of which 30 613 (99.37%) met our criteria for occurring on dry or slippery roads. Of these crashes, 1794 (5.86%) occurred outside the Northeastern states, with the most common states of occurrence outside the Northeast being Florida (301 fatal crashes), Maryland (226), and North Carolina (170).

Crashes on slippery roads accounted for 20.85% (6382/30 613) of the crashes selected for analysis (i.e., those that occurred on slippery or dry roads), and 20.46% (7344/34 225) of the fatalities associated with these crashes. The annual rates of crashes by age averaged across the 5-year study period had a U-shaped distribution for crashes on both dry and slippery roads, with rates declining almost continuously among drivers aged 16 to 70 years and then increasing among drivers older than 70 years (Figure 1 ▶). The distance between the lines for crashes on slippery roads and those on dry roads in this semilogarithmic plot represents the ratio of the rate of crashes on 1 road condition compared with that of the other.

FIGURE 1—

Rates of fatal crashes by age and road condition among drivers licensed in the Northeastern states, 1998–2002, irrespective of the states of occurrence.

This ratio was relatively constant among drivers older than 25 years but narrowed markedly from drivers aged 25 years down to those aged 16 years, indicating a growing ratio of crashes on slippery to crashes on dry roads with decreasing age. Thus, the annual rate of crashes on slippery roads among 16-year-old drivers was 33.0% of the rate of fatal crashes on dry roads (15.8 crashes/100 000 person-years on slippery roads vs 47.9 crashes/100 000 person-years on dry roads) compared with 27.2% among drivers aged 30 to 64 years (3.05 crashes/100 000 person-years on slippery roads versus 11.2/100 000 person-years on dry roads). Among drivers older than 70 years, the ratio of crashes on slippery compared with dry roads remained fairly constant. Similar age patterns of crash rates were observed within each gender (Figure 2 ▶), although crash rates among female drivers declined more sharply than those among male drivers as ages increased from 16 to 59 years and were lower at each age than those observed among male drivers.

FIGURE 2—

Rates of fatal crashes by age and road condition among male (a) and female (b) drivers licensed in the Northeastern states, 1998–2002, irrespective of the states of occurrence.

Crashes among drivers older than 19 years that occurred on slippery roads in the Northeastern states, irrespective of the place of licensure of the drivers, had a clear seasonal pattern with annual peaks in January or December (Figure 3 ▶). The seasonal pattern was less clear among 16- to 19-year-old drivers during 1998, but winter peaks could be discerned during 1999–2002 despite the small numbers of crashes per month in this age group.

FIGURE 3—

Fatal crashes on slippery roads in the Northeastern United States among drivers with full licenses or learner’s permits by month and age, January 1998–December 2002.

The apparently increased risk of crashes on slippery roads among 16- to 19-year-old drivers caused us to focus further analyses on this age group. As shown in Table 1 ▶, among 16- to 19-year-old drivers, the odds ratios for fatal crashes on slippery roads were increased for every month compared with the reference month of July. Odds ratios by time of day were highest from 5:00 to 9:00 am. Other risk factors with significantly increased odds ratios were speeding, rural compared with urban roads, and calendar years 2000 and 2002 compared with the reference year of 1998.

TABLE 1—

Risk Factors for Fatal Crashes on Slippery Roads Among Drivers Aged 16 to 19 Years: Northeastern United States, 1998–2002

Risk Factor	Fatal Crashes on Dry Roads, No.	Fatal Crashes on Slippery Roads, No.	Odds Ratioa (95% Confidence Interval)
Month
January	125	94	5.79 (3.63, 9.29)
February	124	54	3.36 (2.02, 5.58)
March	137	62	3.49 (2.13, 5.71)
April	170	54	2.45 (1.49, 4.03)
May	228	58	1.96 (1.21, 3.18)
June	231	48	1.60 (0.97, 2.64)
July	262	34	1
August	285	50	1.35 (0.83, 2.21)
September	227	48	1.63 (0.99, 2.69)
October	237	72	2.34 (1.47, 3.74)
November	175	63	2.77 (1.71, 4.51)
December	193	90	3.59 (2.27, 5.69)
Hour of the day
12:00 midnight to 4:59 am	455	122	0.99 (0.73, 4.55)
5:00 to 9:59 am	270	120	1.64 (1.20, 2.23)
10:00 am to 2:59 pm	420	114	1
3:00 to 7:59 pm	717	201	1.03 (0.79, 1.35)
8:00 to 11:59 pm	532	170	1.18 (0.89, 1.56)
Driver gender
Male	1697	502	1
Female	697	225	1.09 (0.91, 1.31)
Driver fatigue
No	2307	713	1
Yes	87	14	0.52 (0.28, 0.95)
Excessive speed for conditions
No	1506	399	1
Yes	888	328	1.39 (1.17, 1.77)
Year of crash
1998	490	113	1
1999	497	140	1.22 (0.92, 1.63)
2000	456	161	1.53 (1.16, 2.03)
2001	477	144	1.31 (0.98, 1.74)
2002	474	169	1.55 (1.17, 2.04)
Road class on which crash occurred
Urban	1079	284	1
Rural	1311	443	1.28 (1.08, 1.53)
State of crash
Connecticut	146	40	0.82 (0.55, 1.23)
Maine	113	33	0.88 (0.57, 1.36)
Massachusetts	230	79	1.03 (0.76, 1.40)
New Hampshire	54	15	0.84 (0.44, 1.56)
New Jersey	305	81	0.80 (0.59, 1.08)
New York	686	228	1
Pennsylvania	785	215	0.82 (0.66, 1.03)
Rhode Island	39	15	1.16 (0.60, 2.22)
Vermont	36	21	1.76 (0.97, 3.17)
Driver’s age, y
18–19	1481	425	1
16–17	913	302	1.15 (0.97, 1.37)
Driver licensed in a state with GLPs initiated before January 1, 2000b
No	1269	403	1
Yes	1125	324	0.91 (0.76, 1.08)

Note. GLP = graduated licensing program.

^a The odds of a risk factor being present in crashes on a slippery road divided by the odds of the risk factor being present in crashes on dry roads.

^b See Methods section for a listing of these states.

Risk factors with odds ratios that were not significantly associated with crashes on slippery roads included the driver’s use of drugs, the driver’s use of alcohol, distractions within the vehicle, the day of the week, and the number of occupants in the vehicle (data not shown). Driver fatigue was associated with an odds ratio significantly less than 1. Odds ratios did not vary significantly by state of occurrence of crashes within the Northeastern states (the P value for homogeneity of odds ratios among states was .13). There was a nonsignificant 9% reduction in the odds ratio for having a license from 1 of the 17 states with a GLP adopted before January 1, 2000, compared with having a license from any other place.

Multiple logistic regression modeling considered for entry all variables in Table 1 ▶ for which a causal role in the crashes on slippery roads seemed likely. Driver fatigue thus was excluded from multivariable modeling because it seemed an unlikely protective factor. The final model showed independently significant associations of fatal crashes on slippery roads among 16- to 19-year-old drivers with excessive speed for conditions (odds ratio [OR] = 1.38; 95% confidence interval [CI] = 1.15, 1.64), time of day (OR = 1.80 for 5 to 9 am compared with 10 am to 2 pm; 95% CI = 1.32, 2.46), time of year (e.g., OR = 6.17 for January compared with July; 95% CI = 3.92, 9.70), type of road (OR = 1.27 for rural compared with urban roads; 95% CI = 1.06, 1.51), age (OR = 1.19 for those aged 16 to 17 years compared with those aged 18 to 19 years; 95% CI = 1.00, 1.43); and calendar year (increased odds ratios for all years from 1999 to 2002 compared with 1998). Point estimates of all odds ratios in the multiple logistic regression model remained within 10% of their unadjusted values, with the sole exception of the odds ratio for 12 midnight to 4 am compared with 10 am to 2 pm, which increased 12% to 1.11 (95% CI = 0.82, 1.49) from 0.99. (Odds ratios and 95% confidence intervals for the complete multiple logistic regression model are available from the first author.)

Adding a term for being licensed in a state with a GLP introduced before January 1, 2000, suggested a reduction in risk of 11% compared with other states, but this term was not significant (OR = 0.89; 95% CI = 0.75, 1.06; P = .19) and thus was not included in the final model. Drivers with learner’s permits were not at different risk from drivers with full licenses (P = .6 for entry of a term indicating a learner’s permit).

Crashes Attributed to Swerving Due to Rain, Snow, or Other Factors

Among the 3121 fatal crashes involving 16- to 19-year-old drivers, there were 146 (4.68%) for which the FARS database indicated that a specific contributory driver-related factor was avoiding, swerving, or sliding due to rain, snow, fog, sand, dirt, oil, or wet leaves on the road. Logistic regression analysis with this variable as the dependent variable resulted in a model quite similar to that reported earlier for which all crashes on slippery roads served as the dependent variable. The model for crashes specifically attributed to swerving, however, included significant terms for gender (OR = 2.0 for women compared with men; 95% CI = 1.4, 2.9) and being licensed in a state with a graduated licensing program before January 1, 2000 (OR = 0.63; 95% CI = 0.43–0.90), whereas rural compared with urban road did not enter this model.

Risk Factors for Crashes on Slippery vs Dry Roads Among Drivers Older Than 19 Years

Modeling crashes on slippery roads compared with crashes on dry roads among drivers of all ages resulted in a model that was quite similar to that found for 16- to 19-year-old drivers. There again was an annual cycle peaking in January, an increased risk for rural roads, and an increased risk during the 5 to 9 am period. However, the model included an additional term for gender (OR = 1.55 for female drivers compared with male drivers; 95% CI = 1.32, 1.83), and significant interaction terms were found for gender by excessive speed for conditions and for time of day by age group. In the final model with interaction terms, drivers aged 70 years and older were at increased risk of crashes on slippery roads during 8 to 11 pm (OR = 1.64; 95% CI = 1.09, 2.47) when compared with the reference categories of drivers aged 34 to 43 years during the 10 am to 2 pm period; and drivers aged 24 to 33 years old were at reduced risk compared with these same reference categories at both 5 to 9 am (OR = 0.74; 95% CI = 0.55, 0.98) and 8 to 11 pm (OR = 0.65; 95% CI = 0.48, 0.88).

DISCUSSION

The present analyses show that slippery roads were particularly important causes of crashes among teenaged drivers in the Northeastern United States during 1998–2002. Significant risk factors for crashes on slippery roads at all ages included traveling at speeds too fast for conditions, rural roads compared with urban roads, and driving during the winter months. Our findings thus confirm and expand on analyses showing a greater proportion of crashes during adverse weather among younger (aged 16 to 34 years) and middle-aged (aged 35 to 54 years) drivers compared with older drivers (aged 55 years or older).13

The hours of 5 to 9 am were associated with significantly elevated risks for fatal crashes on slippery roads among teenaged drivers, and drivers aged 16 to 17 years were at increased risk compared with those aged 18 to 19 years. When we examined crashes among drivers of all ages, we also found a clear seasonality to crashes on slippery roads, peaking in the winter and reaching a minimum in the summer months.

Taken together, these findings suggest an intervention to reduce teen deaths and injuries due to crashes on slippery roads: teenagers and perhaps their parents or guardians could be required to attend driver’s education refresher seminars during the late fall or early winter. Such requirements could be enforced by GLPs or driver’s education programs in high schools. Attendees at these seminars could be instructed about the risks of driving on slippery roads.

In-car training to help young drivers learn to better handle cars on slippery roads could also be introduced, although in Finland, where “skid training” has been required as part of driver training, the efficacy of such training has yet to be demonstrated.14 Driving simulators could play important roles in safely teaching automobile-handling techniques applicable to slippery roads if the cost of this technology could be reduced enough to permit widespread implementation. Driving simulators might also be helpful in reducing gender disparities, as our analysis of crashes among drivers > 19 years old suggested a skill deficit in women compared with men.

The importance of the decisionmaking process for crashes on slippery roads is underscored by the finding of progressively lower rates of crashes on slippery roads with increasing age among drivers aged between 16 and 70 years. Driving skills most likely do not improve progressively throughout this extended age range, but decisionmaking about driving in bad weather perhaps does become progressively more cautious. Previous research found that younger teenaged drivers had more crashes than older teenaged drivers because of excessive speed for conditions and wet road conditions.15 In light of our findings, community education programs should urge greater caution in decisions regarding driving when the road conditions are adverse, especially for younger drivers.

The increased odds ratio found for fatal crashes on slippery roads during the morning hours is of concern because much of this driving may be necessitated by work or school. Peaks in all fatal crashes among 16- to 17-year-old drivers before and after school have previously been demonstrated.15^(Figs11,12) The present analyses suggest that the morning peak may be an important factor for teen crashes on slippery roads. Public health announcements noting hazardous road conditions and urging individuals to leave extra travel time when roads are slippery might help reduce the frequency of fatal crashes.

When we analyzed crashes among 16- to 19-year-old drivers that were specifically attributed to swerving on slippery roads, we found a significant protective effect of licensure from a state with a GLP. This is an encouraging finding, but a causal relation between training through a GLP and lessened crashes on slippery roads cannot be conclusively demonstrated from an observational study such as ours. A randomized trial of a GLP versus some “placebo” condition would be the most rigorous method for assessing the efficacy of GLPs. The time to conduct such a trial may have passed, however, as indications that GLPs reduce injuries and deaths have probably made assignments to a placebo condition in such a study unethical.

A number of factors were not associated with crashes on slippery roads, including the driver’s use of alcohol or drugs, multiple passengers in the vehicle, and within-vehicle distractions. These factors have been demonstrated to be important for teenaged drivers in analyses of all crashes irrespective of road conditions,15 and alcohol use has been shown to be a contributor to fatal crashes among persons of all ages.16 The lack of association for these factors in our study does not indicate their lack of importance for crashes on slippery roads. Rather, the findings indicate that these factors are not different in importance for crashes on slippery or dry roads.

This study has several potential shortcomings. One is that many of the analyses are based on odds ratios that contrast the proportion of fatal crashes on slippery roads with and without a risk factor with the same proportion calculated for fatal crashes on dry roads. The null hypothesis is that the risk factor has no impact on the likelihood of the crashes occurring on slippery compared with dry roads. Sometimes, however, a risk factor like driver fatigue may play a substantial causal role in fatal crashes on dry roads, thus resulting in a counterintuitive odds ratio less than 1 for crashes on slippery roads. As fatigue is probably not protective, the likely explanation of the finding is that driver fatigue is a more frequent cause of crashes on dry roads when drivers are more likely to get bored or to fall asleep at the wheel.

To summarize, we have identified several risk factors for fatal crashes among all drivers and in particular among teenaged drivers on slippery roads. Further research is needed to confirm these findings. Case-control studies would permit detailed comparisons of drivers, the environment, and the vehicles involved in fatal and nonfatal crashes on slippery roads with these same factors among persons who had not been involved in crashes. Multicenter, multistate studies of cohorts of newly licensed drivers could also be conducted and might permit accurate and unbiased evaluations of the efficacy of driver’s education programs, graduated licensing programs, and individual patterns of decisionmaking for causing or avoiding crashes on slippery as well as dry roads. The data presented here, however, suggest that certain actions can be taken before the conduct of further research. In particular, we suggest that drivers of all ages be made more aware of the hazards of slippery roads, so that death rates, especially among teenaged drivers, can be reduced.