Visual Field Defects and the Risk of Motor Vehicle Collisions among Patients with Glaucoma

Abstract

purpose. To evaluate the association between visual field defects in the central 24° field and the risk of motor vehicle collisions (MVCs) among patients with glaucoma.

methods. A nested case–control study was conducted in patients with glaucoma aged 55 or more. Cases were patients who were involved in a police-reported motor vehicle collision (MVC) between January 1994 and June 2000; controls were those who had not experienced an MVC at the time of their selection. For each patient, an Advanced Glaucoma Intervention Study (AGIS) score was calculated on automated visual fields collected with the 24-2 or 3-2 programs.

results. With respect to the better-eye AGIS score, compared with patients with no visual field defect, those with severe defects (scores 12–20) had an increased risk of an MVC (odds ratio [OR] 3.2, 95% CI 0.9–10.4), although the association was not statistically significant. Moderate (6–11) or minor field defects (1–5) in the better eye were not associated with the risk of involvement in a crash. In the worse eye, patients with moderate or severe field defects were at significantly increased risk of an MVC (OR 3.6, 95% CI 1.4–9.4 and OR 4.4, 95% CI 1.6–12.4, respectively) compared with those with no defects. Minor field defects in the worse eye did not increase risk of MVC (OR 1.3, 95% CI 0.5–3.4).

conclusions. Patients with glaucoma who have moderate or severe visual field impairment in the central 24° radius field in the worse-functioning eye are at increased risk of involvement in a vehicle crash.

Open-angle glaucoma is a chronic progressive optic neuropathy characterized by changes of the optic disc, thinning of the retinal nerve fiber layer, and gradual loss of visual function beginning in the peripheral field. An estimated 6.7 million people worldwide are bilaterally blinded by glaucoma, and another 67 million are affected by glaucoma, with approximately half of those who are unaware they have the disease. ¹ The loss of peripheral vision due to glaucoma is associated with decreased health-related quality of life. ²  

Motor vehicle collisions (MVCs) are among the most potentially adverse mobility-related outcomes that have been reported to be associated with a diagnosis of glaucoma, ^{3 4 5} although not all agree. ⁶ Several limitations of these studies were self-reported diagnosis of glaucoma, ^{3 4 5} failure to take driving exposure into account when assessing crash risk, ³ and a small number of glaucoma cases in the study sample. ^{3 5} Recently, we reported a cohort study that overcame these methodological problems, finding that adults aged 55 years or more with a diagnosis of glaucoma drive at least as safely as older persons without glaucoma. ⁷  

What was not addressed in this earlier report was whether the subpopulation of drivers who have glaucoma with visual field impairment are at increased risk for vehicle crash involvement compared with those with no field loss. There is reason to believe that this could be the case, because prior studies suggest that substantial visual field impairment elevates crash risk and impairs driving performance, ^{4 8} although these earlier studies did not focus on persons with glaucoma per se. This issue has clear clinical and policy relevance, because information about how severe visual field loss has to be before negatively affecting driver safety is not well understood. 

The objective of the current study was to examine the relationship between the severity of visual field defects in the central 24° radius field in drivers aged ≥55 years with glaucoma and their risk of involvements in a vehicle crash. 

Methods

Study Cohort

The study cohort consisted of individuals aged 55 or more who had been seen at least once between January 1994 and December 1995 in any of three university-affiliated ophthalmology and optometry practices specializing in the diagnosis and treatment of glaucoma. The International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) codes 365.1 and 365.2 were used to identify all potentially eligible patients with glaucoma seen at each of these locations. The medical records of each potentially eligible patient were abstracted to verify the diagnosis of glaucoma though information regarding the basis for that diagnosis was not obtained. Patients were excluded if (1) their primary cause of visual impairment was an ocular disorder other than glaucoma (e.g., macular degeneration, diabetic retinopathy, or clinically significant cataract for which surgery was recommended). Persons with diagnoses of refractive error, dry eye, and early cataract were eligible for the study. (2) Automated visual field data (either a 30-2 or 24-2 test) for both eyes were not in the medical record during the study period; and (3) patients were not legally licensed to drive by the State of Alabama. Information on licensure status was obtained by cross-referencing each subject’s demographic and residential information obtained from the medical record with the Alabama Department of Public Safety (ADPS) database. 

Data Collection

In addition to confirming the diagnosis of glaucoma, medical records were also used to obtain information on use of glaucoma medication, best corrected visual acuity in both eyes, and visual fields in both eyes. All patient visits during the period of January 1994 through December 1995 were abstracted. The visual field reports were then used to calculate a visual field defect score for each eye, based on the Advanced Glaucoma Intervention Study (AGIS) scoring system. ⁹ AGIS scores were categorized into four categories based on previously described cutoff points ⁹ : no defect (score 0), mild defect (scores 1–5), moderate defect (scores 6–11), and severe defect (scores 12–20). 

To obtain additional information on demographic, driving, general health, smoking and alcohol use, we conducted a telephone survey between February and June 2000. Demographic information was gathered by using standard questions regarding age, gender, and race. History of cigarette smoking and alcohol use was assessed with standard questions. ¹⁰ The Short Portable Mental Status Questionnaire, ¹¹ modified for telephone administration, ¹² was used to assess cognitive status. Respondents were asked to respond to a general health questionnaire and questions on driving habits, using 1995 as the reference point. The Driving Habits Questionnaire (DHQ), ¹³ previously shown to be reliable and valid among older drivers, was used to collect information on driving exposure defined in terms of estimated weekly mileage. ¹⁴ The DHQ was also used to calculate a driving avoidance score to estimate the extent to which a respondent avoided certain driving situations that are known to be especially problematic for older drivers. Items addressed driving at night, in fog, in the rain, alone, during rush hour, on the highway/freeway, with children, in high-density traffic, when passing other cars, when changing lanes, when making left-hand turns at intersections, and when parallel parking. The possible responses were “Always,” “Often,” “Sometimes,” “Rarely,” or “Never.” For analytic purposes, subjects who reported “Always” or “Often” avoiding a specific situation were defined as avoiders and those who reported “Sometimes,” “Rarely,” or “Never” were defined as nonavoiders. A composite variable to reflect overall driving avoidance was created by summing the binary avoider and nonavoider variables. Given the 12 situations, the composite variable ranged from 0 to 12, with larger values representing more avoidance. 

Information regarding all MVCs that occurred between January 1994 and June 2000 wherein the study subject was the driver was obtained from the Alabama Department of Public Safety. Information of specific relevance to the study was abstracted from hard-copy accident reports, including the date of the accident and whether the study subject was deemed to be at fault according to the officer at the scene. 

The Institutional Review Board for Human Use at UAB approved the study protocol. The study followed the tenets of the Declaration of Helsinki, and informed consent was obtained from the subjects after explanation of the nature and possible consequences of the study. 

Study Design

Within the study cohort, a nested case–control study was conducted. Cases were those enrollees who had experienced an MVC during the observation period. Incidence density sampling was used to select a single control for each case. In general, incidence density sampling refers to a situation wherein one (or more) control subjects are selected from those members of the cohort who have not experienced the event of interest as of the time the case event occurred. Thus, the control subjects have been under observation in the cohort for as long as the given case, yet have not experienced the relevant outcome of interest. In the present study, control subjects were those who, at the time of their selection, had not been involved in an MVC. For a control subject to be eligible for selection for a given case, the enrollee must have had the first clinic visit during the chart abstraction period (i.e., January 1994 and December 1995) before the date of the MVC. This criterion ensured that both the case and the matched control had contemporaneous data. For each case and matched control, the collision date for the case was used to identify the visual field measurement in closest proximity before the collision, and that measurement was used in the current analysis. Among the eligible control subjects for a given case, a single patient was randomly selected. A proportion of the cases were, according to the police report, responsible for the collision. For this subset of at-fault cases, incidence-density sampling was used to select the controls for at-fault cases from those individuals who have not experienced an at-fault MVC as of the time the at-fault case event occurred. 

A total of 406 patients met the inclusion criteria for the study. During the observation period (January 1994 and June 2000), 120 collisions were observed and selected as cases for study. Of these, 84 (70%) were determined at the collision scene to be caused by the patient (at-fault cases). Regarding the number of collision each patient had, 75 had one collision, 18 had two collisions, and 3 had three collisions. For these 120 cases, a total of 120 controls were selected. 

Statistical Analysis

Demographic, behavioral, driving, and general health characteristics were not available for 40% of the selected cases and 37% of the selected controls because of failure to complete the telephone survey. Reasons for failing to complete the survey included death of the study subject, failure to contact the study subject after multiple attempts, and refusal to participate. Those with complete and incomplete survey data were similar with respect to age (72.2 vs. 74.0, P = 0.06), gender (56.4% female vs. 45.7% female, P = 0.11), and AGIS score in the better (3.54 vs. 3.10, P = 0.45) and worse (7.10 vs. 7.44, P = 0.65) eyes. To prevent the exclusion of these subjects from the analysis, we used multiple imputation to create values for the missing observations using a Markov Chain Monte Carlo (MCMC) method. ^{15 16 17} This method was appropriate because the pattern of missing data tended to be monotone (i.e., for a given subject, groups of variables were missing information). The following characteristics were imputed: age, gender, race, smoking status, alcohol consumption, self-reported medical conditions, number of glaucoma medications, cognitive impairment, visual acuity, driving avoidance, and annual miles driven. Not all these variables had missing values because some information was available from the medical record; however, they were used in the imputation process to yield improved estimates for the missing values. To prevent the imputation of values outside the boundary of real observed values, limitations were set for each variable. 

Descriptive statistics were generated for demographic, behavioral, driving and clinical characteristics. These variables were then compared between the case and control groups using χ² and t-tests for categorical and continuous variables, respectively. Crude and adjusted odds ratios (ORs) and associated 95% confidence intervals (CIs) for the association between field defects and the risk of MVC involvement were calculated by using generalized estimating equations (GEEs). GEEs were used to account for the dependence among those subjects who contributed multiple cases. Separate analyses were conducted using the AGIS score from the better and worse eyes (defined according to the AGIS score). Adjusted analyses were used to account for the possible confounding effects of demographic and medical characteristics. A parsimonious approach retained in the model only those variables that appeared to be confounders for the association between crash involvement and AGIS score. By including fewer variables in the model, this approach provides an opportunity to obtain more precise estimates. The determination of which variables should be retained as confounders was based on the change-in-estimate criteria ¹⁸ using a value of 10%. Finally, analyses were also conducted with the at-fault cases and their matched control subjects. 

Results

The patients representing the cases and at-fault cases did not differ from the control subjects in age, race, ever having smoked, or number of glaucoma medications used. All cases and controls were similar with respect to cognitive impairment, and visual acuity (Table 1) . However, compared with the control subjects, cases and at-fault cases were significantly more likely to be male (P = 0.003 and P = 0.001, respectively). Regarding alcohol, while cases did not differ compared with controls, at-fault cases were more likely than control subjects to have consumed alcohol during their lifetime (P = 0.01). When self-reported medical conditions were compared between cases and control subjects, cases were more likely to have cataract than control subjects (P = 0.02). No difference was noted for diabetic retinopathy, age-related macular degeneration, hearing aid use, or history of falls. However, when at-fault cases were compared to control subjects, the at-fault cases were more likely to have cataract (P = 0.006), diabetic retinopathy (P = 0.03), or age-related macular degeneration (P = 0.07) and to have incurred a fall (P = 0.06). There was no difference in hearing aid use (P = 0.14). At-fault cases had higher cognitive impairment scores than control subjects (P = 0.04); they also had poorer visual acuity in the better and worse eyes. AGIS scores were elevated in cases, though only significantly so for the worse eye. In at-fault cases, AGIS scores were significantly elevated in the better and worse eyes. Cases had lower driving-avoidance scores and lower annual mileage (P = 0.0270 and P = 0.0298, respectively) compared with controls. There were no differences in these variables between at-fault cases and controls. 

Table 2presents the crude and adjusted OR for MVCs according to AGIS categories for the cases and controls. Overall, as the AGIS score increased (visual field defect was more severe) for both the better and worse eye, there was a corresponding increase in the odds of an MVC. Regarding the better eye, patients with a severe defect were more likely to have had an MVC than were patients with no field defect, but this association was not statistically significant after adjustment. Regarding the worse eye, patients with a moderate or severe defect were more likely to have a subsequent MVC than were patients with no field defect, for both the crude (OR 3.0, 95% CI 1.3–7.1 and OR 4.3, 95% CI 1.8–10.3, respectively) and adjusted analyses (OR 3.6, 95% CI 1.4–9.4 and OR 4.4, 95% CI 1.6–12.4, respectively). 

Table 3presents the crude and adjusted ORs for MVC according to AGIS score categories for the at-fault cases and controls. Similar to Table 2 , as the AGIS score increased for the worse eye, there was a corresponding increase in the odds of an MVC, but only for the better eye. Patients with a moderate or severe defect in the worse eye were more likely (OR 3.3, 95% CI 1.1–9.6 and OR 6.9, 95% CI 2.3–20.3, respectively) to have an at-fault crash than patients with no field defect. After adjustment, these associations were still significant for both moderate and severe defects (OR 4.2, 95% CI 1.2–15.0 and OR 9.0, 95% CI 2.4–33.2, respectively). 

Discussion

Older adults with glaucoma with moderate to severe central field loss in their worse functioning eye are at increased risk of involvement in collisions than are those with glaucoma who have no field loss. In this case–control study those with AGIS scores indicating severe central field impairment were six times more likely to cause an at-fault crash and four times more likely to cause a crash, regardless of fault than were those with AGIS scores indicating no impairment. Previous studies on drivers with visual field loss are consistent with these findings, although they did not focus on drivers with glaucoma per se and used measurement methods quite different from ours. Drivers with severe binocular field loss, as determined by a screening test administered 40° nasally and 60° temporally at motor vehicle licensing offices, are approximately two times more likely to be involved in a crash than are those with no field loss. ⁴ In our study, the visual field variable was not based on a screening test but rather on a full-threshold procedure in the central field. A study of driving performance on a closed road course found that the avoidance of obstacles was impaired in normally sighted persons with simulated field loss that reduced the diameter of the field to a 30° radius. ⁸  

In the present study, visual field loss was defined in terms of each eye separately, as is the clinical convention when managing glaucoma and monitoring its potential progression. One might argue, based on face validity, that it is the binocular visual field assessed by conventional automated perimetry methods or by the Esterman grid, ¹⁹ not the monocular central field of each eye, that is the most direct way to evaluate risk for the adverse events such as falling, problems locating objects, and vehicle crashes; however, this conjecture remains to be proved. In clinical practice each eye’s health and functionality is routinely evaluated separately. In view of this, it is important to recognize based on this study’s results that the convention of assessing each eye’s field by itself, even just the central field, is informative with respect to crash risk. It is further interesting that the worse eye’s field sensitivity was more strongly related to crash risk than that for the better eye. There is a conventional notion in clinical practice that the eye with better function dictates visual performance. However, the worse eye’s visual field characteristics were significantly associated with crash involvement, whereas those of the better eye were not. These findings are reminiscent of our earlier findings on contrast sensitivity and crash risk in drivers with cataract, in which severe contrast-sensitivity impairment in one eye only elevated crash risk. ²⁰ However, it should also be noted that despite the lack of statistical significance of the better eye results, a general pattern of increased risk with worsening impairment was observed. 

Strengths of this study are as follows: The medical records of all study subjects were reviewed to confirm glaucoma as the primary eye disorder and cause of visual dysfunction. This factor is important because other eye diseases common in the elderly (e.g., diabetic retinopathy, age-related macular degeneration) can cause central visual field impairment. Second, the AGIS score was used to define severity of field defect. The AGIS score is an already-established metric of field loss severity in glaucoma and is computed based on the output of the most common automated perimetry programs (30-2, 24-2) used in glaucoma management. Third, an independent and impartial source (the Alabama Department of Public Safety) was used to obtain information on MVCs for the study participants. Crash data were not based on driver self-reports, known to be unreliable. ²¹ Procurement of the accident report also allowed us to obtain information on who was responsible for the MVC and therefore refine our case definition to focus on those MVCs in which the study subject was deemed at fault. 

Study limitations must also be acknowledged. General health and driving habits were collected in 2000 by using a telephone survey that relied on participants’ ability to recall these characteristics, with 1995 as the reference point. However, there is little reason to suspect differential bias among the cases and controls in the ability to recall the requested information accurately, and thus any misclassification is likely to result in a conservative bias. Second, the response rate for the telephone survey was not ideal (approximately 61% overall due), yet it did not differ between the cases (60%) and controls (63%). For those who did not complete the survey, multiple imputation was used to create values for the missing information. Fortunately, the primary independent variable (i.e., visual field defect) was obtained from each patient’s medical record and therefore not vulnerable to survey nonresponse. Moreover, when the adjusted analyses were performed excluding patients with imputed data, the overall results were highly consistent with the results based on all patients, suggesting that little bias resulted from the imputation process. 

In conclusion, this study suggests that drivers with glaucoma aged 55 or more with moderate to severe impairment in the central 24° field of the worse-functioning eye are at increased risk for vehicle crash involvement. This finding has clinical relevance for ophthalmologists caring for patients with glaucoma, because it provides some guidance as to the point in disease progression at which it is prudent to begin a dialog with the patient about driver safety. What is practically useful is that the marker for increased crash risk identified in this study can be computed from the automated perimetry tests commonly used in the management of glaucoma, and thus there is no additional patient or economic burden. 

 Supported by National Institutes on Aging Grant P50-AG11684 and National Eye Institute Grant R21-EY14071, Research to Prevent Blindness, Inc., and the EyeSight Foundation of Alabama. CO is a Research to Prevent Blindness Senior Scientific Investigator.

 Submitted for publication June 14, 2005; revised July 26, 2005; accepted September 20, 2005.

 Disclosure: G. McGwin, Jr, None; A. Xie, None; A. Mays, None; W. Joiner, None; D. K. DeCarlo, None; T. A. Hall, None; C. Owsley, None

 The publication costs of this article were defrayed in part by page charge payment. This article must therefore be marked “advertisement” in accordance with 18 U.S.C. §1734 solely to indicate this fact.

 

Table 1.

 

Demographic, Medical, and Visual Function Characteristics among Glaucoma Patients Involved in an MVC (Cases) Versus Those Not (Controls) and Those at Fault for an MVC Versus Control Subjects

	Cases (n = 120)	Controls (n = 120)	P	At-Fault Cases (n = 84)	Controls (n = 84)	P
Mean age (y)	73.4	72.3	0.23	74.3	72.2	0.07
Gender (%)			0.003			0.001
Male	56.9	38.3		65.5	40.2
Female	43.1	61.7		34.5	59.8
Race (%)			0.29			0.99
White	61.0	70.0		66.7	65.9
African-American	34.2	25.0		26.2	26.8
Other	4.9	5.0		7.1	7.3
Ever smoked (%)	34.2	25.0	0.12	22.6	28.1	0.42
Ever consumed alcohol (%)	47.5	40.0	0.27	54.8	35.4	0.01
Medical conditions (%)
Cataract	88.6	77.5	0.02	95.2	81.7	0.006
Diabetic retinopathy	32.5	23.3	0.11	58.3	30.5	0.03
Age-related maculopathy	29.3	30.8	0.79	42.9	29.3	0.07
Hearing aid use	33.3	33.3	0.99	44.1	32.9	0.14
Fall	49.6	48.3	0.84	63.1	48.8	0.06
Mean glaucoma medications (n)	4.03	3.89	0.52	3.94	3.99	0.87
Mean cognitive impairment	3.13	3.35	0.62	4.11	2.97	0.04
Mean visual acuity (logMAR)
Better eye	0.24	0.22	0.48	0.30	0.21	0.02
Worse eye	0.25	0.21	0.13	0.31	0.21	0.007
Mean AGIS score
Better eye	3.90	2.83	0.06	3.89	2.41	0.02
Worse eye	8.91	5.63	<0.0001	9.39	5.40	<0.0001
Mean driving-avoidance score	2.20	2.87	0.03	2.23	2.33	0.76
Mean miles driven per year	7479	9784	0.03	10,407	8,932	0.24

Table 2.

 

Crude and Adjusted OR according to AGIS Score Categories for Cases and Controls

	Cases (%)	Controls (%)	Crude OR (95% CI)	Adjusted OR* (95% CI)
Better eye
No defect	33.3	44.2	Reference	Reference
Mild defect	38.2	35.8	1.4 (0.8–2.5)	1.5 (0.7–2.8)
Moderate defect	17.9	15.0	1.6 (0.7–3.3)	1.4 (0.5–3.4)
Severe defect	10.6	5.0	2.8 (1.0–8.0)	3.2 (0.9–10.4)
Worse eye
No defect	9.8	21.7	Reference	Reference
Mild defect	25.2	38.3	1.5 (0.6–3.3)	1.3 (0.5–3.4)
Moderate defect	30.9	22.5	3.0 (1.3–7.1)	3.6 (1.4–9.4)
Severe defect	34.2	17.5	4.3 (1.8–10.3)	4.4 (1.6–12.4)

  n = 120.

*  Adjusted for alcohol consumption, cataract, diabetic retinopathy, and worse eye visual acuity.

Table 3.

 

Crude and Adjusted OR according to AGIS Score Categories for At-Fault Cases and Controls

	Cases (%)	Controls (%)	Crude OR (95% CI)	Adjusted OR* (95% CI)
Better eye
No defect	33.3	47.6	Reference	Reference
Mild defect	36.9	35.4	1.5 (0.7–3.0)	1.7 (0.7–3.7)
Moderate defect	20.2	13.4	2.2 (0.9–5.3)	2.0 (0.7–5.4)
Severe defect	9.5	3.7	3.7 (0.9–15.3)	4.2 (0.9–19.8)
Worse eye
No defect	8.3	23.2	Reference	Reference
Mild defect	26.2	39.0	1.9 (0.7–5.1)	1.9 (0.6–6.1)
Moderate defect	26.2	22.0	3.3 (1.1–9.6)	4.2 (1.2–15.0)
Severe defect	39.3	15.9	6.9 (2.3–20.3)	9.0 (2.4–33.2)

  n = 84.

*  Adjusted for alcohol consumption, cataract, diabetic retinopathy, and worse eye visual acuity.