Risk of eye injury and associated 95% confidence intervals (per 1000 occupants) by year.
Distribution of known sources of eye injuries overall and by vehicle model year.
McGwin G, Owsley C. Risk Factors for Motor Vehicle Collision–Related Eye Injuries. Arch Ophthalmol. 2005;123(1):89-95. doi:10.1001/archopht.123.1.89
To evaluate the association between specific occupant, collision, and vehicle characteristics and the risk of motor vehicle collision (MVC)–related eye injury.
The 1988-2001 National Automotive Sampling System Crashworthiness Data System files were used. The Crashworthiness Data System is a national probability sample of passenger vehicles involved in police-reported tow-away MVCs. The risk of eye injury was calculated according to specific occupant (eg, age, seat belt use) and collision (eg, ΔV [estimated change in velocity], vehicular intrusion) characteristics. The association between eye injury and these characteristics was calculated using risk ratios and associated 95% confidence intervals.
The incidence of eye injuries in MVCs has progressively increased since 1998. Frontal air bag deployment was associated with a statistically significant, 2-fold (risk ratio, 2.13 [95% confidence interval, 1.56-2.91]) increased risk of eye injury, whereas seat belt use was associated with a 2-fold (risk ratio, 2.17 [95% confidence interval, 1.89-2.44]) reduced eye injury risk. In late-model vehicles, frontal air bags are the most common cause of MVC-related eye injury. Older age, female sex, seat position, vehicle weight, and collision severity were also associated with eye injury risk.
Seat belt use is the most effective means of occupant protection against MVC-related eye injury. For front-seated occupants in frontal collisions, the adverse effect of frontal air bags on the risk of eye injury should be considered against their protective effect for fatal injury.
Eye injury is a leading cause of monocular blindness in the United States and is second only to cataract as the most common cause of visual impairment.1,2 Among all causes of blindness and visual impairment, eye injury is among the most preventable. It has been noted that perhaps 90% of eye injuries are preventable with the appropriate use of protective eyewear.3,4 A large proportion of eye injuries occur in settings (eg, workplace) or during activities (eg, recreational sports) wherein protective eyewear is often mandated or advocated.5 However, the correct use of appropriate protective eyewear is often infrequent.6,7 Internationally, eye injury is also a significant problem.8 The pattern of eye injuries in industrialized nations is generally consistent with that of the United States, while eye injuries in developing nations tend to reflect underlying social and economic circumstances.9- 13
Motor vehicle collisions (MVCs) are responsible for a significant number of eye injuries.5 However, unlike other common causes of eye injury such as work and sports activities, the automobile is not a setting wherein protective eyewear is used. Thus, the prevention of MVC-related eye injuries will require the identification of characteristics associated with their occurrence. However, the current literature on this topic is limited and replete with case series and reports, the majority of which focus on the role of frontal air bags.14,15 To date, there has been no population-based study wherein the independent effect of occupant, collision, and vehicle characteristics on the risk of MVC-related eye injuries has been evaluated. Understanding the interrelationship between these characteristics and the risk of MVC-related eye injury has important implications for the public health and automotive communities.
The data for this study were obtained from the National Automotive Sampling System (NASS).16 The Crashworthiness Data System (CDS) is part of NASS and is operated by the National Highway Traffic Safety Administration National Center for Statistics and Analysis, Washington, DC. The CDS is a nationwide annual probability sample of approximately 5000 light passenger vehicles (passenger cars, light trucks, vans, sport utility vehicles) that were involved in police-reported tow-away collisions. To be included in the CDS, a crash must fulfill the following requirements: must be police reported, must involve a harmful event (property damage and/or personal injury) resulting from the crash, and must involve at least 1 towed passenger car or light truck or van in transport on a traffic way. For each CDS case, trained investigators collect information on occupant, vehicle, and injury characteristics using 3 sources of data: (1) official documents (eg, police traffic crash reports and vehicle, highway, and medical records); (2) physical evidence (eg, scene characteristics and vehicle damage profile); and (3) interviews. Because the CDS uses a multistage probability sampling design, the data from each case must be weighted to produce national estimates.
The MVCs selected by the CDS are identified from police traffic crash reports at 24 sites across the United States based on a 3-stage sampling system. In the first stage, the country is divided into geographic areas called primary sampling units (PSUs). Each PSU consists of a central city, a county surrounding a central city, an entire county, or a group of contiguous counties. The PSUs are defined so that their minimum population is approximately 50 000. The PSUs are grouped into strata based on geographic region and type (eg, central cities, suburban counties). The PSUs to be sampled are allocated to each stratum roughly proportional to the number of crashes in each stratum. Each PSU contains a number of police jurisdictions that process reports of crashes that occur within the PSU’s boundaries. These police jurisdictions form the frame of the second stage of sampling. Each jurisdiction is assigned a measure of size based on the number, severity, and type of its crashes. The final stage of sampling is the selection of crashes within the sampled jurisdictions. Each crash occurring within a jurisdiction is classified into a stratum based on type of vehicle, most severe police-reported injury, disposition of the injured, tow status of the vehicles, and model year of the vehicles.
A cohort study design was used to evaluate the association between the risk of eye injury and select occupant-, vehicular-, and collision-related characteristics. All occupants in the 1988 to 2001 NASS data files were selected for inclusion in the cohort. Using this approach, it is possible to identify independent risk factors for the etiology of MVC-related eye injury.
The primary outcome of interest for the present study was the occurrence of at least 1 eye injury. The CDS obtains information on injuries from both official (eg, autopsy records, medical records, hospital discharge summaries, emergency department records) and unofficial sources (eg, lay coroner reports, interviews with emergency medical services personnel, police, or occupant) with the former taking precedence. During the period of interest (ie, 1988-2001), the CDS changed the system with which injuries are classified. Prior to 1993, the Occupant Injury Classification17 was used to code injuries while in the 1993 and later data sets, the Abbreviated Injury Scale was used.18 For the purposes of this study, in the pre-1993 data sets, occupants who sustained an eye injury were identified using the Occupant Injury Classification system/organ variable. For the 1993 and later data sets, eye injuries were identified based on Abbreviated Injury Scale codes 230202.2 to 230206.2, 230299.1, 240402.2 to 241699.1, and 240499.1. For this reason, description of specific eye injuries will be limited to data in the 1993 and later files.
The CDS personnel also obtain information on the suspected source of each injury. A level of confidence is assigned to each source reflecting the investigators’ certainty that the identified entity was truly the injury source.
Information on occupant (age, sex, height, weight, seat belt use, frontal air bag deployment, seating position), vehicle (model year, curb weight, vehicle type), and collision characteristics (type of collision, ΔV [ie, the vector velocity change in kilometers per hour (km/h) during the collision phase of the crash]) was also obtained from the NASS data.
With respect to occupant characteristics, age was categorized into 4 groups (<17 years, 17 -28 years, 29-43 years, and >43 years) based on the quartiles of the age distribution of the overall study population. Height and weight were categorized into tertiles. The CDS includes several variables regarding the use and availability of restraint systems within each vehicle and for each occupant. This information is obtained from crash investigations, police and ambulance reports, personal interviews, and occupant injury patterns. A final determination is made based on the preponderance of the evidence. Manual (or active) seat belt use is defined as the use of shoulder belt, lap belt, lap and shoulder belt, or any combination of seat belt use with a child safety seat. Automatic (or passive) seat belt use is coded as either yes, no, or unknown. For purposes of this analysis, seat belt use is defined by the use of either of these restraint systems. For each occupant, information on frontal air bag availability and deployment is also provided in the data files, which is collected in a similar fashion as the seat belt information. With respect to seating position, occupants were classified as being seated in the front or rear of the vehicle. For front-seated occupants, information pertaining to the track position of the seat, which indicates the track position of the occupant’s seat relative to the dashboard (forward or middle/back) was also obtained.
Vehicles were classified into 4 groups: 1986 and earlier, 1987 to 1990, 1991 to 1994, and 1995 and later. Vehicle curb weight was used to classify vehicles into 3 categories using the standard CDS definitions: small (<2500 lbs), midsize (2500-3000 lbs), and large (>3000 lbs). Vehicles were also classified according to their body type, that is, passenger vehicle, sport utility vehicles, trucks, and minivans. Finally, the type of collision was classified using information regarding the primary area of vehicular damage associated with the highest ΔV of the collision. Six collision types were defined: frontal, near side, far side, rollover, and rear end. ΔV was classified into 4 groups: <14 km/h, 14 to 21 km/h, 22 to 30 km/h, and >30 km/h.
Many variables that were of interest for this study contained a substantial number of missing observations, a problem noted by other authors using this database.19 Multiple imputation was used to create values for missing observations using a Markov Chain Monte Carlo method.20,21 A total of 5 imputations were used. This technique was chosen over several alternatives given that the pattern of missing data tended to be predominantly arbitrary rather than monotone.21,22 Values were imputed for all variables of interest (excluding eye injury) using known values for occupant, vehicle, and collision characteristics. The SAS System release 8.02 (SAS Institute, Inc, Cary, NC) was used for the imputation process.
As mentioned previously, the CDS is a probability rather than a random sample. Thus, to account for the multistage sampling of the NASS, SUDAAN version 7.5.6 (Research Triangle Institute, Research Triangle Park, NC) was used for all statistical comparisons based on the appropriately weighted data. Relative risks (RRs) and 95% confidence intervals (CIs) were then used to quantify the strength and precision of the association between the risk of eye injury and occupant, vehicle, and collision characteristics using the SURVIVAL procedure in SUDAAN. A multivariable model and associated adjusted RRs and 95% CIs were computed such that the independent effect of each potential risk factor could be evaluated. A separate model that is limited to front-seated occupants involved in frontal collisions with a ΔV of 15 or greater was generated to evaluate the association between frontal air bag deployment and eye injury risk. Such collisions represent those wherein frontal air bag deployments would be expected, and thus, this restricted analysis focuses on only those occupants who would be potentially “at risk” for frontal air bag deployment. For the etiologic analyses, the unit of analysis was the occupant. Thus, occupants who sustained more than 1 eye injury were treated similarly to those occupants who sustained a single eye injury. For the descriptive analyses, data regarding all eye injuries is presented. P values of ≤.05 (2-sided) were considered statistically significant.
Between 1988 and 2001, there were an estimated 66 941 420 occupants involved in police-reported tow-away MVCs in the United States. During this time, 1 200 131 occupants sustained 1 231 554 eye injuries (18 per 1000 occupants). Figure 1 presents the risk of eye injury by year between 1988 and 2001. Albeit variable from year to year, a general trend of increasing risk can be observed (P<.001). Considering all eye injuries, skin and soft tissue injuries (ie, eyelid) were most frequent (88.4%); however, excluding such injuries, cornea abrasions (49.8%) and conjunctiva injuries (28.6%) were most common (Table 1).
Table 2 presents the crude and adjusted RRs and 95% CIs for the association between occupant, vehicle, and collision characteristics and risk of eye injury for all types of occupants and collisions. There was an elevated risk associated with increasing age; however, none of the associations were statistically significant. This was true for both the crude and adjusted associations. In the multivariable model, women demonstrated a significantly increased risk of eye injury. Although taller occupants (>64 inches) demonstrated an increased risk of eye injury, this association diminished in the multivariable model. Independent of other risk factors, heavier occupants had a reduced risk of eye injury. Lack of seat belt use was associated with a 2-fold increase in the risk of eye injury. There was no association for seating position, seat track position, or vehicle model year. Eye injury risk was consistent across vehicle body types and curb weights. The risk of eye injury was highest for frontal collisions; however, near-side impacts demonstrated a statistically similar risk, whereas other collision types were associated with lower risk. Relative to low-ΔV collisions (<14 km/h), increasing ΔV was associated with an increased risk of eye injury; however, the incremental increase in the RRs diminishes suggesting a plateau in eye injury risk for ΔVs of 22 km/h or greater.
As mentioned, a separate model limited to front-seated occupants involved in frontal collisions with a ΔV of 15 or greater was generated to evaluate the association between frontal air bag deployment and eye injury risk. In the unadjusted model, frontal air bag deployment was associated with a 50% increased risk of eye injury (RR, 1.49 [95% CI, 1.14-1.94]). Following adjustment for age, sex, seat belt use, seat track position, vehicle model year, vehicle body type, vehicle curb weight, and ΔV, the RR increased to 2.13 (95% CI, 1.56-2.91).
Overall, the most common injury mechanism was the windshield followed by the frontal air bag, steering wheel, and flying glass (Figure 2). However, injury mechanism differed according to vehicle model year. Starting in 1993, automobile manufacturers were required to begin phasing in frontal air bags in cars and light trucks. Thus, for 1992 and earlier models, frontal air bags represented an infrequent cause of eye injuries whereas the windshield and rearview mirror were a more common cause. For 1993 and later models, frontal air bags were responsible for more than half of the eye injuries.
Motor vehicle collisions are responsible for the majority of injuries that occur in the United States annually. Of these injuries which are considered minor, a small proportion (<5%) involve the eye (G.M., unpublished data, September 24, 2003).17,23 However, given the enormous number of MVCs that occur each year (more than 6 million in 2001), it can be expected that the absolute number of MVC-related eye injuries will also be large, perhaps 80 000 to 100 000.24 Understanding the etiology of these injuries may help identify strategies to reduce their likelihood of occurrence.
To date, the majority of research on MVC-related eye injuries has been descriptive in nature. Of this research, the role of frontal air bags has received the most attention. However, no study has attempted to estimate the independent association between specific occupant, vehicle, and collision characteristics and the risk of eye injury. Duma et al,17 also using the NASS data, reported that among occupants involved in frontal collisions, 3% of those whose frontal air bag deployed sustained an eye injury compared with 2% of those who did not have a frontal air bag deployment; this difference was not statistically significant. In the current study, we observed nearly identical risks (albeit statistically significant). The difference in statistical significance between their results and ours may be a function of sample size; their study used 7 years (1993-1999) of NASS data, whereas we used 14 years (1988-2001). This yielded an unweighted sample size of 22 236 cases in their study compared with 150 920 cases in the current study. Following adjustment for potential confounding characteristics, the association between frontal air bag deployment and eye injury increased from a 1.5-fold to a 2.1-fold increased risk. This indicates that the smaller increased risk reported herein and by Duma et al is likely confounded by occupant, vehicle, and collision characteristics. Thus, independent of these characteristics, front-seated occupants involved in frontal collisions are twice as likely to sustain an eye injury if their frontal air bag deploys.
Despite the lack of prior empirical research on the topic, that frontal air bags are associated with an increased risk of eye injury is not surprising. This association was observed only in the subgroup of occupants at risk of air bag deployment. The vast number of case reports and series that strongly implicate the frontal air bag as the cause of eye injuries portends that such an association would exist.14,15 In the current study, frontal air bags were the most commonly noted source of eye injury among occupants of 1993 and later model year vehicles who sustained eye injuries. In pre-1993 model year vehicles, the windshield was the most common injury source.
The potential adverse effect of frontal air bags on the risk of eye injuries cannot be interpreted in isolation. Frontal air bags have been shown to reduce the risk of death in frontal MVCs.25- 30 Thus, reductions in fatal injuries associated with frontal air bag deployment may come at the cost of increases in less severe injuries. Given that the majority of eye injuries in the current study were, in fact, minor and likely resulted in no permanent loss of visual function, such a trade-off is likely to be considered acceptable. However, as frontal air bag technology progresses, the potential exists for these safety devices to prevent more than just life-threatening injuries. Lack of seat belt use is associated with a similar increased risk of eye injury as frontal air bag deployment. Thus, legislative and public health initiatives directed toward increasing seat belt use will have a meaningful impact on the risk of eye as well as other types of injuries.
We also found an increased risk of eye injury associated with age, sex, seat position, vehicle size, and collision severity (ie, ΔV). It has been suggested that the increased risk associated with older age is related to increasing stiffness of the lens with aging.31 That older occupants are at increased risk of eye injury is consistent with other research demonstrating a positive association between age and other types of MVC-related injuries.32 Why men have a reduced risk of eye injury in MVCs is not clear. This finding is counter to population-based studies of all-cause eye injuries that report higher incidence among men.33,34 Perhaps sex differences in the use of eyewear, which has been hypothesized to increase the risk of eye injuries in MVCs,14 can explain this association. Unfortunately, information on eyewear use in the CDS is limited to those occupants whose air bags deploy; for all other occupants, eyewear information is not available. This prohibits the evaluation of this characteristic as a general risk factor for eye injury. Occupants who rode with their seats in the middle or rearward track position had a lower risk of eye injury than those whose seats were in the forward position. This association may reflect the fact that occupants whose seats are in the forward-most position are closer to and therefore more likely to strike objects on the dashboard. Occupants of heavier vehicles had a lower risk of eye injury. The effect of vehicle mass in MVCs has been well established.35- 38 Large vehicles provide more opportunity for energy dissipation thereby reducing forces on the occupant and potentially reducing the risk of eye injury. Finally, the risk of eye injury increased with increasing ΔV. This should not be surprising as ΔV reflects the severity of the collision and has been previously correlated with injury severity.38
The results of the present study should be interpreted in light of several strengths and limitations. The NASS is a probability sample and therefore represents an approximation of the true occurrence and characteristics of MVCs that occur in the United States each year. However, provided that the appropriate sampling weights are applied when analyzing NASS data, as was done in this study, population-based estimates and associations can be obtained. Eye injuries were identified using information derived from a variety of sources; however, it is unlikely that the records of ophthalmologists were a common source. Thus, the reliability of the eye injury diagnoses is questionable, yet there is little reason to suspect bias with respect to the risk factors of interest. Multiple imputation was used to address missing information for several variables used in our analyses. Multiple imputation is a well-established technique for addressing the problem of missing data and has been used in other studies using this same data source.19 Moreover, when the analyses reported herein were repeated without imputed observations, the results were highly consistent with those reported herein. We did not have information on the use of eyewear for all CDS occupants and thus could not evaluate this for its association with eye injury. While eyewear may be independently associated with the risk of eye injury, it is unlikely that it would explain the results observed in the current study, particularly the association between airbag deployment and injury risk. For this to occur, the use of eyewear would need to be associated with the likelihood of airbag deployment, and it is difficult to envision why this would be the case. However, as noted earlier, the increased injury risk among women may be attributable to eyewear.
To our knowledge, this is the first study to report associations between occupant, vehicle, and collision characteristics and the risk of eye injury. For all types of MVCs, the results of this study implicate lack of seat belt use as the strongest modifiable risk factor for eye injury. For the subgroup of front-seated occupants involved in frontal MVCs, air bag deployment demonstrated an association of equivalent magnitude. The increased risk associated with frontal air bag deployment must be interpreted in light of the protection they offer against fatal injuries. As frontal air bag technology evolves, it will be important to evaluate whether they reduce injury risk for a wider range of injury severities. Regardless of the future of frontal air bag efficacy, occupants of motor vehicles should heed the ubiquitous message that proper seat belt use is a highly effective means to not only reduce the risk of injury in general but specifically reduce the risk of eye injury, a message that ophthalmologists should be sharing with their patients.
Correspondence: Gerald McGwin, Jr, MS, PhD, Department of Ophthalmology, School of Medicine, University of Alabama at Birmingham, 700 S 18th St, Suite 609, Birmingham, AL 35294-0009 (firstname.lastname@example.org).
Submitted for Publication: October 1, 2003; final revision received June 24, 2004; accepted June 24, 2004.
Financial Disclosure: None.
Funding/Support: This research was supported by grant R21-EY14071 from the National Eye Institute, Bethesda, Md; the EyeSight Foundation of Alabama, Birmingham; and Research to Prevent Blindness, New York, NY.
Previous Presentation: The results of this study were presented at The Eye and the Auto World Congress 2003; June 20, 2003; Detroit, Mich.
Additional Information: Dr Owsley is a Research to Prevent Blindness senior scientific investigator.