Skip to main content
NCBI home page
The Journals of Gerontology Series B: Psychological Sciences and Social Sciences logoLink to The Journals of Gerontology Series B: Psychological Sciences and Social Sciences
. 2011 Dec 1;67(4):437–446. doi: 10.1093/geronb/gbr122

Patterns of Level and Change in Self-Reported Driving Behaviors Among Older Adults: Who Self-Regulates?

Melissa L O’Connor 1,✉, Jerri D Edwards 2, Brent J Small 2, Ross Andel 2,3
PMCID: PMC3391073  PMID: 22138270

Abstract

Objectives.

The present study investigated patterns of self-reported driving habits among healthy older adults over 5 years, as well as characteristics that distinguished subgroups with different patterns.

Methods.

Participants were drivers from the control group of the Advanced Cognitive Training for Independent and Vital Elderly study at the baseline assessment (N = 597). The outcome was a composite of driving frequency, driving space, and perceived driving difficulty. Growth mixture models identified classes of participants with different baseline scores and change trajectories, and classes were compared in terms of baseline sensory, physical, and cognitive functioning.

Results.

A 3-class model was indicated, consisting of 2 classes with intercept differences and stability over time, “above-average stable” (39%) and “average stable” drivers (44%), and 1 class with a lower intercept and negative slope, “decreasers” (17%). Relative to stable drivers, decreasers exhibited significantly more depressive symptoms and poorer self-rated health, balance, everyday functioning, and speed of processing after controlling for age and education (p < .05).

Discussion.

The majority of older drivers maintained their driving over time at different levels, whereas a subgroup of individuals with poorer baseline functioning self-regulated by reducing their driving. Future studies should determine how such patterns affect driving safety.

Keywords: Driving mobility, Driving self-regulation, Growth mixture models, Older drivers

For Americans aged 65 and older, driving is important for maintaining autonomy (Shope, 2003). Age-related declines in sensory, physical, and cognitive functioning may affect older adults’ abilities to drive safely. However, individuals with age-related impairments may compensate by adjusting, or self-regulating, their driving behaviors (Anstey, Wood, Lord, & Walker, 2005). There is some controversy over whether older adults self-regulate appropriately, and a population of older drivers may contain heterogeneous subgroups with different average patterns of self-regulation (Ackerman, Crowe, et al., 2010). Studies are needed to examine which subsets of older drivers reduce their driving appropriately and whether there are subgroups of older drivers who maintain their driving over time. The present study addresses these needs.

BACKGROUND

Self-regulation of driving can be measured by self-reported avoidance of complex driving situations, such as night driving, as well as the perceived difficulty of such situations (Charlton et al., 2006; Molnar & Eby, 2008). Older adults also self-regulate by restricting their driving space, driving less frequently and more slowly, and ceasing driving (Edwards et al., 2008; Kostyniuk & Molnar, 2008). According to the psychological compensation model of Bäckman and Dixon (1992), older drivers may restrict their driving when they perceive a mismatch between their driving capabilities and environmental demands.

Accordingly, cross-sectional studies have found significant relationships between level of driving self-regulation and older age (Betz & Lowenstein, 2010; Vance et al., 2006), vision problems (Lyman, McGwin, & Sims, 2001; West et al., 2003), poor health (Donorfio, D’Ambrosio, Coughlin, & Mohyde, 2009), physical limitations (Anstey et al., 2005), low education (West et al., 2003), greater depressive symptoms (Keay et al., 2009), and poor cognitive functioning (Freund & Szinovacz, 2002; Vance et al., 2006). These findings suggest that many older drivers with reduced capabilities self-regulate appropriately. However, drivers with severe cognitive deficits may fail to self-regulate, possibly due to impaired judgment (Baldock, Mathias, McLean, & Berndt, 2006; Freund, Colgrove, Burke, & McLeod, 2005).

Most longitudinal studies of driving self-regulation have focused on driving cessation. Increased age, female sex, poor vision, poor balance, health problems, poor everyday functional performance, and diminished cognitive abilities have emerged as predictors of driving cessation over periods ranging from 3 to 10 years (Ackerman, Edwards, Ross, Ball, & Lunsman, 2008; Anstey, Windsor, Luszcz, & Andrews, 2006; Edwards, Bart, O’Connor, & Cissell, 2010; Edwards et al., 2008; Kostyniuk & Molnar, 2008). Many of these variables, particularly age, self-rated health, and cognitive speed of processing, may also predict lower baseline driving mobility, declines in driving space and frequency, and increases in perceived driving difficulty (Ross et al., 2009). However, findings are mixed, perhaps due to population heterogeneity in patterns of driving self-regulation (Ackerman, Vance, Wadley, & Ball, 2010).

O’Connor, Edwards, Wadley, and Crowe (2010) and Ross and colleagues (2009) used random effects models to examine mean trajectories of older adults’ self-reported driving behaviors across 4–5 years. On average, participants reported declines in their driving mobility, or increased self-regulation, over time. On the other hand, Ackerman and colleagues (2010) found that 64% of community-dwelling older adults (N = 426) experienced stability in self-ratings of their driving across 3 years. The previous studies suggest that a normative population of older drivers may contain unobserved subgroups with different longitudinal trajectories, as well as different baseline levels, of driving self-regulation. Prior studies have not investigated this possibility empirically, so the characteristics of such subgroups are unknown. Certain older adults may drive less at baseline and reduce their driving over time to compensate for age-related impairments, whereas other drivers may maintain their driving over time.

Previous studies have either examined aggregate scores for driving self-regulation across the entire samples or grouped participants according to cutoff scores on cognitive tests (Ross et al., 2009). These approaches may mask meaningful differences among naturally occurring subgroups (Wang & Bodner, 2007). Determining normative patterns of driving self-regulation among older adults will be helpful for developing models of driving behaviors among older adults, comparing normal driving patterns to those observed in clinical populations, and evaluating safety outcomes for drivers with different patterns.

Current Objectives and Hypotheses

The current study examined patterns of self-reported driving habits in a large community sample of older adults across 5 years. There were two main objectives. First, we used growth mixture models (GMMs) to identify latent clusters of participants who could be empirically distinguished in terms of their baseline driving or change trajectories. We hypothesized the existence of at least two clusters: a subgroup showing below-average baseline driving and increases in self-regulation over time and a subgroup displaying stability. Second, we examined whether cluster membership could be differentiated by baseline sensory, physical, and cognitive functioning. Given previously cited findings, we expected that participants with increased self-regulation would be characterized by older age and poorer self-rated health, balance, depressive symptoms, vision, everyday functioning, memory, reasoning, and speed of processing.

Method

Participants and Procedure

We used data from the Advanced Cognitive Training for Independent and Vital Elderly (ACTIVE) study, a clinical trial that examined the effects of cognitive training interventions on older adults’ cognitive and everyday functioning (for details, see Jobe et al., 2001). Beginning in 1998, community-dwelling older adults were recruited from sites in Alabama, Michigan, Massachusetts, Indiana, Pennsylvania, and Maryland. Prospective participants first completed in-person screening and baseline visits, during which assessments of cognitive and functional abilities were given. Individuals meeting the following inclusion criteria were accepted into the study: (a) aged 65 or older, (b) no significant functional impairment, (c) Mini-Mental State Examination score ≥ 23, (d) no medical conditions with a high probability of functional decline, (e) far visual acuity of at least 20/50, and (f) no communication difficulties. Participants (N = 2,802) were randomly assigned to a no-contact control condition or one of three intervention conditions (reasoning, speed of processing, or memory training). Follow-up assessments were conducted approximately two months, one year, two years, three years, and five years after baseline. Driving mobility was assessed at baseline and each annual visit.

The present study utilized all available data from participants in the control group who reported driving at baseline (N = 597). Analyses were limited to the control group due to the evidence that some of the cognitive training programs used in ACTIVE lead to improved driving capabilities and protect against declines in driving mobility (Edwards, Delahunt, & Mahncke, 2009; Edwards, Myers, et al., 2009; Roenker, Cissell, Ball, Wadley, & Edwards, 2003); extending prior research on training effects was beyond the scope of the current study. Current participants had an average age of 73 to 60 years (standard deviation [SD] = 5.78), and the majority were women (71.40% of the sample) and Caucasian (74.50% of the sample). Years of education ranged from sixth grade to the doctoral level, with a mean of 13.63 years (SD = 2.69). A total of 116 participants were lost to attrition before the last follow-up assessment. Of these study dropouts, 25 died, 74 refused further participation, and 17 could not be contacted.

Measures

Demographics and attrition.—

Relevant measures were baseline age, years of education, sex (coded as women = 0 and men = 1), race (coded as Caucasian = 0 and other = 1), and attrition (coded as nondropouts = 0 and dropouts = 1).

Balance.—

Balance was measured by the Turn 360 Test (Steinhagen-Thiessen & Borchelt, 1999). Examinees were asked to stand and turn in a complete circle for two trials. Observers recorded the number of steps required to complete each turn, and fewer steps indicated better performance. The average number of steps across the two turns was used in analyses.

Depressive symptoms.—

A 12-item version of the Center for Epidemiological Studies-Depression scale (Liang, van Tran, Krause, & Markides, 1989; Radloff, 1977) was used to measure depressive symptoms. Respondents rated how often they experienced various symptoms over the week preceding the assessment, ranging from 0 (none of the time) to 3 (most of the time). Higher scores signified more depressive symptoms.

Driving behaviors.—

Overall driving self-regulation was measured by a composite of driving frequency, driving space, and driving difficulty at each time point (O’Connor et al., 2010; Ross, 2007). These variables were assessed via the Driving Habits Questionnaire (DHQ; Owsley, Stalvey, Wells, & Sloane, 1999; Stalvey, Owsley, Sloane, & Ball, 1999). Driving frequency was defined as the number of days (ranging from 0 to 7) that participants reported driving during a typical week. For driving space, participants completed six dichotomous items that assessed whether they personally drove beyond their property, neighborhood, or town during the past week and whether they drove beyond their county, state, or region during the past two months. Total scores could range from 0 to 6, with higher scores indicating greater driving space.

On the DHQ, participants reported whether they avoided eight challenging driving situations (e.g., driving at night and driving alone) and how much difficulty they experienced with each situation (on a 4-point scale from 1 = no difficulty to 4 = extreme difficulty). An administrative skip pattern was used in ACTIVE, such that each participant had data for difficulty or avoidance, but not both. Therefore, participants who reported ceasing driving or avoiding a driving situation were coded as having extreme difficulty, whereas those who did not avoid the situation were coded as having no difficulty (Ross, 2007). The difficulty items were reverse scored and summed to create a total difficulty variable.

Then, the variables for driving frequency, space, and total difficulty were converted to z scores and summed to form a global weighted composite for each time point. Lesikar, Gallo, Rebok, and Keyl (2002) used a similar composite to assess driving habits. The calculation of a global composite was desirable because driving space, frequency, and difficulty are significantly correlated and have shown similar overall patterns of change for older adults (Ross, 2007); we were interested in driving self-regulation as a whole; and GMMs require ample outcome variance to generate reliable solutions (B. O. Muthén, 2004). At baseline, the driving composite ranged from −7.96 to 3.80 (M = 0.02, SD = 2.20, skewness = −0.85, standard error [SE] of skewness = 0.10).

Everyday functioning.—

Everyday functioning was measured by the Everyday Problems Test (EPT), the Observed Tasks of Daily Living Test (OTDL), and the Timed Instrumental Activities of Daily Living Test (TIADL). Total scores on these tests were standardized and summed to create an everyday functioning composite. The EPT assessed practical problem-solving skills involving medication management, shopping, finances, household activities, meal preparation, transportation, and telephone use (Willis, 1996). Participants viewed 14 stimuli, such as medication labels and recipes, and answered two multiple-choice questions about each one. Total scores could range from 0 to 28 items correct. The OTDL involved behavioral simulations of actual tasks of daily living (Diehl, Willis, & Schaie, 1995). There were nine tasks with a total of 13 questions that assessed medications, telephone use, and finances; total scores could range from 0 to 28 correct responses.

The TIADL assessed participants’ speed and accuracy at completing everyday tasks involving real-world stimuli, such as finding items on a shelf and reading medication labels (Owsley, McGwin, Sloane, Stalvey, & Wells, 2001). Each task was timed and had a maximum time limit, and errors made during the tasks resulted in a time penalty. Total scores were global time composites in which lower scores indicated better performance (Ackerman et al., 2008; Edwards, Wadley, Vance, Roenker, & Ball, 2005). TIADL scores were reverse scored before being standardized and combined with EPT and OTDL scores.

Far visual acuity.—

A GoodLite Model 600A illuminated cabinet with a standard Early Treatment Diabetic Retinopathy Study chart was used to measure far visual acuity (Good-Lite, 2010). Examinees read the chart from a 10-foot distance, wearing corrective lenses if applicable. Ten points were given for each of nine lines read correctly. Total scores could range from 0 (a Snellen score of 20/125) to 90 (a Snellen score of 20/16).

Memory.—

Memory was assessed via the Hopkins Verbal Learning Test (HVLT; Brandt, 1991), the Rivermead Behavioral Memory Test (RBMT; Wilson, Cockburn, & Baddeley, 1985), and the Auditory Verbal Learning Test (AVLT; Jobe et al., 2001). Total scores on these tests were standardized and summed to form a memory composite. On the HVLT, a list of 15 words was read aloud across five consecutive trials. Following each presentation of the list, respondents recalled as many words as possible, and total scores were the number of words correctly recalled. Prose memory was assessed with the stories subtest of the RBMT. On this test, respondents listened to a passage of prose read aloud (54–65 words) and, in a 2-min time limit, wrote down as much of the story as they could recall. Words and phrases were “blocked together” and scored as individual units, with possible scores ranging from 0 to 21 correct answers. The AVLT involved the auditory presentation of 15 words, repeated across five trials. After each trial, participants were given 3 min to write down as many of the words as they could recall. Total scores were the number of words correctly recalled across trials.

Reasoning.—

Inductive reasoning was measured by the Letter Series test (Thurstone & Thurstone, 1949), Word Series (Gonda & Schaie, 1985), and Letter Sets (Ekstrom, French, Harman, & Derman, 1976) tests. Total scores on these tests were standardized and summed to create a reasoning composite. In the Letter Series task, respondents were shown rows of 10–15 letters, and each row contained a pattern. Respondents discerned the pattern and chose, from five possible options, which letter came next in each row. There were 30 items with 6 min allowed for completion, and higher scores indicated more correct answers. The Word Series test was similar to the Letter Series task, but respondents discerned patterns among words instead of letters. In the Letter Sets task, respondents were presented with 15 rows, each comprised of five sets of letters with four letters per set. Four of the letter sets shared a similar pattern, and respondents eliminated the one letter set that did not fit. Seven minutes were allocated to complete the task, with higher scores indicating more correct answers.

Self-rated health.—

Participants rated their health on a 5-point scale ranging from 1 = excellent to 5 = poor (Jobe et al., 2001).

Speed of processing.—

Cognitive speed of processing was measured via the Wechsler Adult Intelligence Scale–Revised (WAIS-R) Digit Symbol Substitution Test (DSS; Wechsler, 1981) and the personal computer (PC), touch, 4-subtest Useful Field of View Test (UFOV; Edwards, Vance, et al., 2005). Total UFOV and DSS scores were z scored and summed to generate a speed of processing composite. DSS measured motor and perceptual processing speed. Participants received a grid of 93 empty squares with the numbers 1 through 9 above each square, as well as a key in which each number was paired with a symbol. In 90 s, participants filled in the empty squares with the corresponding symbols; total scores were the number of substitutions completed correctly.

UFOV measured speed of processing for visual attention tasks (Edwards, Vance, et al., 2005). Central targets (a car or a truck) were presented at durations ranging from 16.67 to 500 ms, and the subtests became progressively more difficult, requiring identification of the central target as well as localization of a peripheral target embedded in distracters. Total scores could range from 66.68 to 2000 ms, and smaller scores indicated faster speed of processing (i.e., shorter display durations needed to correctly identify and localize the targets). UFOV scores were reverse scored before being standardized and combined with DSS scores.

Analyses

GMMs were used to examine patterns in the driving composite across the five assessment points (B. O. Muthén, 2004). Conventional growth curve modeling assumes that individuals within a population vary around a single mean intercept and growth curve, but GMMs allow latent subgroups of individuals to vary around different mean intercepts and growth curves. In a GMM context, the term “subgroup” refers to a cluster or class of individuals within a heterogeneous population, and each individual’s class membership is unobserved. GMMs can be represented as structural equation models with latent factors for intercept and slope, as well as a categorical latent variable representing class membership that loads on the intercept and slope factors. Parameters within each class, as well as the most likely class membership of each participant, are estimated by the models (Li, Duncan, Duncan, & Acock, 2001; Wang & Bodner, 2007).

In the current study, GMM analyses were conducted via Mplus (Version 5; L. K. Muthén & Muthén, 2007). Parameter estimates were obtained using full-information maximum-likelihood estimation with the expectation-maximization algorithm, which allowed cases with missing data to be included. Time was centered at baseline, driving composites from each time point were standardized, and intercept and slope factors were allowed to covary. First, we tested two single-class models, one with a linear slope term and another with linear and quadratic slope terms. Differences in fit between these models were evaluated using the −2 Log Likelihood (−2LL) ratio test, which is computed as a chi-square statistic, in which degrees of freedom equal the difference in model parameters. Then, the best-fitting model was run with additional numbers of classes specified. The −2LL ratio test is not appropriate for comparing models with different numbers of classes because such models are not nested (B. O. Muthén, 2004). Therefore, the optimal model was determined by the following fit indices: Akaike information criteria (AIC), Bayesian information criteria (BIC), and the Lo–Mendell–Rubin (LMR) likelihood ratio test, which tests the null hypothesis that the model of interest was generated by a model with one less class (Lo, Mendell, & Rubin, 2001). In order to differentiate classes by slope alone, the previous analyses were repeated with baseline driving as a covariate.

Class membership for each participant was assigned by the best-fitting GMM models. Next, multivariate analysis of covariance (MANCOVA) was used to examine between-class differences in terms of continuous baseline variables. Covariates were age and education, and dependent variables were depressive symptoms, balance, self-rated health, vision, everyday functioning, memory, reasoning, and speed of processing. Chi-square statistics were used to examine sex, race, and attrition differences between classes. This 2-step approach, in which GMMs are followed by other statistical tests, was supported by B. O. Muthén, Jo, and Brown (2003). Researchers have used MANCOVA and χ2 following GMMs in order to compare multiple classes and explore the meaningfulness of GMM results in a different context (Jobe-Shields, Cohen, & Parra, 2011; Mora et al., 2009; Schwartz, Mason, Pantin, & Szapocznik, 2009).

Results

Models Unadjusted for Baseline Driving

Extracting classes.—

In the single-class GMM with linear slope and without baseline driving as a covariate (−2LL = −2,809.10; BIC = 5,656.56, AIC = 5,630.21), the slope was significant and negative (estimate = −0.06, SE = 0.01, p < .05). The single-class model with linear and quadratic slopes did not show improved fit over the linear-only model, χ2(4) = 9.45, p > .05. Thus, quadratic slope was not included in subsequent models. A 2-class model (−2LL = −2,583.70; BIC = 5,244.11; AIC = 5,191.41) exhibited significantly better fit than the single-class model, LMR = 439.34, p < .01. In turn, a 3-class model (−2LL = −2,561.10; BIC = 5,237.20; AIC = 5,158.26) showed improved fit over the 2-class model, LMR = 44.06, p < .01. A 4-class model repeatedly failed to converge.

For the 3-class model, Class 1 comprised 17.10% of the sample (N = 102), Class 2 comprised 38.86% (N = 232), and Class 3 comprised 44.05% (N = 263). See Table 1 and Figure 1. Class 1 had a negative intercept and slope, indicating that this group generally reported below-average scores on the driving composite at baseline and also declined over time. This class was called “decreasers.” Class 2 exhibited a significant positive intercept, indicating above-average scores on the driving composite at baseline, and a flat slope, indicating stability over time. Thus, this class was called “above-average stable.” Finally, Class 3 had an intercept close to the sample mean and a nonsignificant slope, so this class was called “average stable.” To examine the robustness of the 3-class model, we replicated it after reducing floor effects by excluding participants who ceased driving during the study, drove less than 10 miles per week at baseline, or, if missing data for miles per week, did not drive beyond their neighborhood (total excluded N = 54). Additionally, the driving composite at each time point was log transformed to reduce skewness. The 3-class structure held, and less than 5% of the remaining participants changed the class membership. Therefore, the 3-class model was retained for interpretation.

Table 1.

Summary of Parameter Estimates for a Growth Mixture Model with Three Latent Classes Defined by Patterns of Driving Self-Regulation

Decreasers Above-average stable Average stable
Variable Estimate SE Estimate SE Estimate SE
Fixed effects
    Intercept −0.93* 0.14 0.73* 0.09 −0.17* 0.10
    Slope −0.28* 0.05 0.02 0.01 0.07 0.02
Random effects
    Residual 0.36* 0.05 0.36* 0.05 0.36* 0.05
    Variance (intercept) 1.26* 0.20 −0.05 0.04 0.26* 0.09
    Variance (slope) 0.10* 0.02 −0.02* <0.01 −0.01 0.01
    Cov (intercept, slope) −0.11* 0.05 0.03* <0.01 −0.01 0.02
Open in a new tab

Notes. Cov = covariance; decreasers N = 102; above-average stable N = 232; average stable N = 263.

*p < .05.

Figure 1.

Figure 1.

Open in a new tab

Observed and estimated standardized driving composite scores within three latent classes defined by patterns of driving self-regulation.

Differences between classes.—

An omnibus MANCOVA with age and education as covariates indicated that there were significant differences between the classes in terms of the dependent variables, Wilks λ = 0.87, F(8,565) = 5.04, p < .01. Using an alpha level of .001 to test homogeneity assumptions (Tabachnick & Fidell, 2001), Box’s M test was not significant, p = .01. The following main effects were significant: vision, F(2,572) = 3.99, p = .02; depressive symptoms, F(2,572) = 7.37, p < .01; balance, F(2,572) = 11.73, p < .01; self-rated health, F(2,572) = 12.74, p < .01; reasoning, F(2,572) = 5.88, p < .01; everyday functioning, F(2,572) = 7.03, p < .01; and speed of processing, F(2,572) = 6.54, p < .01. Pairwise comparisons showed that decreasers exhibited more depressive symptoms and worse vision, balance, self-rated health, everyday functioning, and speed of processing relative to average stable and above-average stable drivers (p < .05). Decreasers also had lower reasoning than above-average stable drivers, and above-average stable drivers had better self-rated health and reasoning than average stable drivers (p < .05). Additionally, above-average stable drivers were significantly younger and more educated than decreasers and average stable drivers (p < .05), and above-average stable drivers were also more likely to be men, χ2(2) = 59.67, p < .01. Memory, attrition, and race did not significantly differ by class. See Table 2.

Table 2.

Baseline Characteristics by Membership in Three Latent Classes Defined by Patterns of Driving Self-Regulation

Decreasers Above-average stable Average stable
Baseline characteristic M (%) SD M (%) SD M (%) SD
Age 77.42 6.34 71.79a 4.67 73.70a 5.73
Years of education 13.19 2.39 14.60a b 2.68 12.94 2.55
Sex (% women) (81.40) (53.40)a b (83.30)
Race (% Caucasian) (65.70) (80.20) (73.00)
Attrition (% dropout) (17.60) (20.70) (19.00)
Balancec 7.98 2.32 6.51a 1.84 6.70a 1.65
Visual acuity 68.10 12.85 75.46a 11.32 74.37a 10.78
Self-rated healthc 2.94 0.88 2.31a b 0.80 2.70a 0.86
CES-Dc 6.45 5.18 3.74a 4.03 5.19a 5.02
Everyday functioning composite −1.37 2.90 0.72a 2.19 −0.10a 2.32
Memory composite −1.08 2.61 0.51 2.40 −0.04 2.39
Reasoning composite −1.11 2.46 1.02a b 2.65 −0.47 2.54
Speed of processing composite −1.02 1.87 0.52a 1.51 −0.06a 1.61
Open in a new tab

Note. CES-D = Center for Epidemiological Studies-Depression scale.

a

Significantly different from decreasers at p < .05.

b

Significant difference between above-average stable and average stable at p < .05.

c

Smaller scores reflect better performance.

Models Adjusted for Baseline Driving

Extracting classes.—

To extract classes differing in slope only, GMMs were estimated in which the intercept and linear slope factors were defined by the driving composite at the last four time points. Participants with sufficient follow-up data (N = 499) were included in these GMMs, and baseline driving was a covariate. For a single-class model, −2LL = 1,921.66, AIC = 3,859.33, and BIC = 3,893.03. A 2-class model (−2LL = −1,618.17, AIC = 3,268.33, BIC = 3,335.73) had significantly better fit than the single-class model, LMR = 595.02, p < .01. There were no additional fit improvements with a 3-class model, LMR = 79.71, p = .09, so the 2-class model was interpreted. The first class (decreasers) comprised 18% of the sample (N = 91) and had a significant negative slope, and the second class (stable) comprised 82% of the sample (N = 408) and had a nonsignificant positive slope. Thus, average stable and above-average stable drivers were collapsed into one class, whereas a class of decreasers remained distinct (Table 3 and Figure 2).

Table 3.

Summary of Parameter Estimates for a Growth Mixture Model with Baseline Driving as a Covariate and Two Latent Classes

Decreasers Stable
Variable Estimate SE Estimate SE
Fixed effects
    Intercept −0.84* 0.21 0.15* 0.04
    Slope −0.22* 0.06 0.06 0.02
Random effects
    Residual 0.30* 0.05 0.30* 0.05
    Variance (intercept) 2.35* 0.49 <0.01 0.05
    Variance (slope) 0.19* 0.04 −0.02* <0.01
    Cov (intercept, slope) −0.52* 0.14 0.03 <0.01
Baseline driving
    Loading on intercept 0.77* 0.15 0.57* 0.04
    Loading on slope −0.09* 0.04 −0.03 0.01
Open in a new tab

Notes. Cov = covariance; decreasers N = 91; stable N = 408.

*p < .05.

Figure 2.

Figure 2.

Open in a new tab

Observed and estimated standardized driving composite scores within two latent classes adjusted for baseline driving.

Differences between classes.—

An omnibus MANCOVA with age and education as covariates was significant, Wilks λ = 0.95, F(8,473) = 3.30, p < .01. At an alpha level of .001 (Tabachnick & Fidell, 2001), Box’s M was not significant, p = .01. The univariate tests showed that, relative to the stable class, decreasers had significantly more depressive symptoms, F(1,480) = 5.63, p = .02, worse balance, F(1,480) = 17.69, p < .01, poorer self-rated health, F(1,480) = 6.25, p < .01, slower speed of processing, F(1,480) = 6.21, p < .01, and worse everyday functioning, F(1,480) = 3.99, p < .01. Decreasers were also significantly older (p < .05). Education, memory, reasoning, vision, attrition, race, and sex did not differ between the groups; see Table 4.

Table 4.

Baseline Characteristics by Membership in Two Latent Classes Defined by Changes in Driving Self-Regulation

Decreasers Stable
Baseline characteristic M (%) SD M (%) SD
Age 76.92 6.06 72.86* 5.40
Years of education 13.33 2.24 13.75 2.77
Sex (% women) (79.10) (72.30)
Race (% Caucasian) (70.30) (77.50)
Attrition (% dropout) (7.70) (8.80)
Balancea 7.88 2.05 6.56* 1.80
Visual acuity 70.51 12.31 74.89 11.20
Self-rated healtha 2.82 0.86 2.51* 0.82
CES-Da 5.85 4.71 4.37* 4.44
Everyday functioning composite −0.82 2.87 0.36* 2.30
Memory composite −0.72 2.76 0.33 2.38
Reasoning composite −0.76 2.50 0.37 2.67
Speed of processing composite −0.73 2.01 0.26* 1.52
Open in a new tab

Notes. Sample size for these analyses was 499. CES-D = Center for Epidemiological Studies-Depression scale.

a

Smaller scores reflect better performance.

*Significant between-group difference at p < .05.

Discussion

The current study used GMMs to examine whether a sample of older adults contained unobserved subgroups with different initial levels and growth trajectories of driving self-regulation. Without adjusting for baseline driving, models revealed three latent classes that were distinguishable by intercept or slope parameters: decreasers, average stable, and above-average stable drivers. Thus, our hypothesis regarding the presence of at least two subgroups was supported. Decreasers, the smallest class, drove less at baseline than the stable groups and showed declines in driving over time. above-average stable drivers had higher baseline levels of driving than average stable drivers, but these groups did not change significantly over time. In models that adjusted for baseline driving, the above-average stable and average stable classes combined to form a single class, resulting in a 2-class solution. The 3-class solution is most relevant because it captures group differences in both level and change. However, the 2-class solution is useful for examining slope differences alone.

When the overall sample was considered (i.e., the unrestricted single-class model), the average slope for driving self-regulation was negative, so it would appear that older drivers reduce their driving over time. This finding has been reported by previous studies (O’Connor et al., 2010; Ross et al., 2009). However, the current study suggests that stable driving habits characterize the majority of older drivers, whereas only a minority of drivers self-regulate over time. The current analyses illustrate how GMMs capture systematic population heterogeneity that cannot be modeled by other statistical methods, which may represent longitudinal data more realistically (Wang & Bodner, 2007).

Our second hypothesis that sensory, physical, and cognitive characteristics would significantly differentiate the classes was generally supported. Decreasers, the group exhibiting the greatest amount of self-regulation, were significantly older and showed more depressive symptoms, poorer self-rated health, worse balance, worse vision, poorer everyday functional performance, poorer reasoning, and slower speed of processing at baseline relative to average stable and above-average stable drivers. Thus, drivers with age-related functional deficits tended to reduce their driving accordingly. These findings corroborate previous research (e.g., Anstey et al., 2006; Edwards et al., 2008) and support the Bäckman and Dixon (1992) model of psychological compensation.

Of the dependent variables, everyday functioning was not previously studied in relation to driving self-regulation, except for a study on driving cessation (Ackerman et al., 2008). In light of the present findings, everyday functioning may predict reductions in driving as well as cessation. The main effects of visual acuity and reasoning on class membership were significant for the 3-class solution, but not the 2-class solution that controlled for baseline driving. These discrepancies may have occurred because the 2-class solution was limited to participants with adequate follow-up data, which may have reduced between-group variance for vision and reasoning. Vision and reasoning may also be more strongly associated with one’s level of driving self-regulation than one’s slope. Memory was not significantly related to class membership in any of the models. This finding is consistent with Ackerman and colleagues (2010) and Edwards and colleagues (2008), who found that memory was not related to driving cessation or self-rated driving, respectively.

Relative to average stable drivers and decreasers, above-average stable drivers were better educated, had better self-rated health, and were disproportionately men. Some studies have reported that women reduce their driving and cease driving more than men (Kostyniuk & Molnar, 2008; Ross et al., 2009), but sex differences are not consistently found (Edwards et al., 2008; Ross, 2007). It may be that sex and education influence one’s overall level of driving more than one’s slope, such that healthy, highly educated men drive more than average throughout their lives. In the present cohort of older adults, men are likely to have more driving experience than women (Hakamies-Blomqvist & Siren, 2003). Later cohorts may not show sex differences in terms of driving expertise or behaviors.

Although the present study yielded informative findings regarding self-regulation and older adults, there are some limitations. First, we maximized outcome variance and reliability by creating a composite of driving behaviors. When we conducted analyses of separate driving behaviors in addition to the composite, the solutions either did not converge or did not contribute new information. However, previous studies have analyzed driving space, frequency, and difficulty as separate outcomes. Any concerns are lessened by the knowledge that these variables have shown similar overall patterns of change for older adults (Ross et al., 2009). GMMs also carry a heavy computational load, so there may not have been sufficient power to detect quadratic slope effects (Wang & Bodner, 2007). Despite these limitations, the classes generated in the present study appeared well differentiated and representative of the observed means (Figures 1 and 2).

Another limitation is that driving habits were measured via self-report, and objective assessments of driving skills were not examined. These concerns are lessened by the knowledge that there are significant positive correlations between self-reported and objectively measured driving patterns and that the DHQ is a reliable and valid measure of driving for older adults (Marshall et al., 2007; Owsley et al., 1999; Stalvey et al., 1999). However, it is important to corroborate the present findings with objective assessments in future studies because older drivers may underestimate their actual driving frequency (Blanchard, Myers, & Porter, 2010; Freund et al., 2005). It is also important to examine how patterns of self-regulation affect driver safety, as Ross and colleagues (2009) found that self-reported driving restrictions did not attenuate crash risk.

The ACTIVE data set lacks information about other factors that may influence driving self-regulation, such as alternate transportation opportunities, self-efficacy, feedback on driving performance, and interpersonal relationships (Ackerman, Crowe, et al., 2010; Ackerman, Vance et al., 2010; Freund & Szinovacz, 2002). Further research should examine how these factors affect driving mobility across time. The present findings should generalize to community-dwelling, healthy older adults residing in the Eastern half of the United States, considering ACTIVE sample characteristics (Jobe et al., 2001). However, clinical populations of older adults, such as those with dementia, may show different, less appropriate patterns of self-regulation (Baldock et al., 2006). Additionally, future work should use GMMs to examine the impact of cognitive training on driving self-regulation.

In conclusion, we found distinct patterns of self-reported driving self-regulation among older adults across 5 years. Most drivers maintained their driving over time at different levels, whereas a minority of drivers with lower baseline functioning self-regulated by reducing their driving. Thus, some older drivers with age-related impairments appeared to adjust their driving accordingly.

Funding

National Institute on Aging ; Cognitive Training Gains and Mobility Outcomes in Advanced Cognitive Training for Independent and Vital Elderly (5 R03 AG23078 -02), Karlene Ball, PhD (principal investigator); The Advanced Cognitive Training for Independent and Vital Elderly study, National Institute on Aging; National Institute of Nursing Research to Hebrew Senior Life (U01 NR04507); Indiana University School of Medicine (U01 NR04508); Johns Hopkins University (U01AG14260); New England Research Institutes (U01 AG14282); Pennsylvania State University (U01 AG14263); University of Alabama at Birmingham (U01 AG14289); University of Florida (U01AG14276).

Acknowledgments

The authors would like to acknowledge the ACTIVE principal investigators, including the following: Hebrew Senior Life—J. N. Morris, PhD, Indiana University School of Medicine—F. W. Unverzagt, PhD, Johns Hopkins University—G. W. Rebok, PhD, New England Research Institutes (Data Coordinating Center)—S. L. Tennstedt, PhD, Pennsylvania State University—S. L. Willis, PhD, University of Alabama at Birmingham—K. K. Ball, PhD, University of Florida/Wayne State University—M. Marsiske, PhD, National Institutes of Health—K. M. Koepke, PhD, and National Institute on Aging—J. King, PhD.

Conflict of Interest

Dr Rebok is an investigator with Compact Disc Incorporated for the development of an electronic version of the ACTIVE memory intervention. Dr Marsiske has received support from Posit Science in the form of site licenses for the Insight program for another research project. Dr Ball owns stock in the Visual Awareness Research Group and Posit Science, Inc., the companies that market the UFOV and speed of processing training software (now called Insight). Posit Science acquired Visual Awareness, and Dr Ball continues to collaborate on the design and testing of these assessments and training programs as a member of the Posit Science Scientific Advisory Board. Dr Edwards has worked in the past as a limited consultant to Posit Science.

References

  1. Ackerman ML, Crowe M, Vance DE, Wadley VG, Owsley C, Ball KK. The impact of feedback on self-rated driving ability and driving self-regulation among older adults. The Gerontologist. 2010;51:367–378. doi: 10.1093/geront/gnq082. doi:10.1016/j.trf.2010.06.003. [DOI] [PMC free article] [PubMed] [Google Scholar]
  2. Ackerman ML, Edwards JD, Ross LA, Ball KK, Lunsman M. Examination of cognitive and instrumental functional performance as indicators for driving cessation risk across 3 years. The Gerontologist. 2008;48:802–810. doi: 10.1093/geront/48.6.802. doi:10.1093/geront/48.6.802. [DOI] [PMC free article] [PubMed] [Google Scholar]
  3. Ackerman ML, Vance DE, Wadley VG, Ball KK. Indicators of self-rated driving across 3 years among a community-based sample of older adults. Transportation Research Part F. 2010;13:307–314. doi: 10.1016/j.trf.2010.06.003. doi:10.1016/j.trf.2010.06.003. [DOI] [PMC free article] [PubMed] [Google Scholar]
  4. Anstey KJ, Windsor TD, Luszcz MA, Andrews GR. Predicting driving cessation over 5 years in older adults: Psychological well-being and cognitive competence are stronger predictors than physical health. Journal of the American Geriatrics Society. 2006;54:121–126. doi: 10.1111/j.1532-5415.2005.00471.x. doi:10.1111/j.1532-5415.2005.00471.x. [DOI] [PubMed] [Google Scholar]
  5. Anstey KJ, Wood JM, Lord SR, Walker JG. Cognitive, sensory, and physical factors enabling driving safety in older adults. Clinical Psychology Review. 2005;25:45–65. doi: 10.1016/j.cpr.2004.07.008. doi:10.1016/j.cpr.2004.07.008. [DOI] [PubMed] [Google Scholar]
  6. Bäckman L, Dixon RA. Psychological compensation: A theoretical framework. Psychological Bulletin. 1992;112:259–283. doi: 10.1037/0033-2909.112.2.259. doi:10.1037//0033-2909.112.2.259. [DOI] [PubMed] [Google Scholar]
  7. Baldock MRJ, Mathias JL, McLean AJ, Berndt A. Self-regulation of driving and its relationship to driving ability among older adults. Accident Analysis and Prevention. 2006;38:1038–1045. doi: 10.1016/j.aap.2006.04.016. doi:10.1016/j.aap.2006.04.016. [DOI] [PubMed] [Google Scholar]
  8. Betz ME, Lowenstein SR. Driving patterns of older adults: Results from the second injury control and risk survey. Journal of the American Geriatrics Society. 2010;58:1931–1935. doi: 10.1111/j.1532-5415.2010.03010.x. doi:10.1111/j.1532-5415.2010.03010.x. [DOI] [PubMed] [Google Scholar]
  9. Blanchard RA, Myers AM, Porter MM. Correspondence between self-reported and objective measures of driving exposure and patterns in older drivers. Accident Analysis and Prevention. 2010;42:523–529. doi: 10.1016/j.aap.2009.09.018. doi:10.1016/j.aap.2009.09.018. [DOI] [PubMed] [Google Scholar]
  10. Brandt J. The Hopkins Verbal Learning Test: Development of a new memory test with six equivalent forms. The Clinical Neuropsychologist. 1991;5:125–142. doi:10.1080/13854049108403297. [Google Scholar]
  11. Charlton JL, Oxley J, Fildes B, Oxley P, Newstead S, Koppel S, et al. Characteristics of older drivers who adopt self-regulatory driving behaviours. Transportation Research Part F. 2006;9:363–373. doi:10.1016/j.trf.2006.06.006. [Google Scholar]
  12. Diehl M, Willis SL, Schaie KW. Everyday problem solving in older adults: Observational assessment and cognitive correlates. Psychology and Aging. 1995;10:478–490. doi: 10.1037//0882-7974.10.3.478. doi:10.1037//0882-7974.10.3.478. [DOI] [PubMed] [Google Scholar]
  13. Donorfio LK, D’Ambrosio LA, Coughlin JF, Mohyde M. Health, safety, self-regulation and the older driver: It's not just a matter of age. Journal of Safety Research. 2009;39:555–561. doi: 10.1016/j.jsr.2008.09.003. doi:10.1016/j.jsr.2009.04.002. [DOI] [PubMed] [Google Scholar]
  14. Edwards JD, Bart E, O’Connor ML, Cissell GM. Ten years down the road: Predictors of driving cessation. The Gerontologist. 2010;50:393–399. doi: 10.1093/geront/gnp127. doi:10.1093/geront/gnp127. [DOI] [PMC free article] [PubMed] [Google Scholar]
  15. Edwards JD, Delahunt PB, Mahncke HW. Cognitive speed of processing training delays driving cessation. The Journals of Gerontology, Series A: Biological Sciences and Medical Sciences. 2009;64:1262–1267. doi: 10.1093/gerona/glp131. doi:10.1093/gerona/glp131. [DOI] [PMC free article] [PubMed] [Google Scholar]
  16. Edwards JD, Myers CA, Ross LA, Roenker DL, Cissell GM, McLaughlin AM, et al. The longitudinal impact of cognitive speed of processing training on driving mobility. The Gerontologist. 2009;49:485–494. doi: 10.1093/geront/gnp042. doi:10.1093/geront/gnp042. [DOI] [PMC free article] [PubMed] [Google Scholar]
  17. Edwards JD, Ross LA, Ackerman MA, Small BJ, Ball KK, Bradley SL, et al. Longitudinal predictors of driving cessation among older adults from the ACTIVE clinical trial. The Journals of Gerontology, Series B: Psychological Sciences and Social Sciences. 2008;63:6–12. doi: 10.1093/geronb/63.1.p6. [DOI] [PubMed] [Google Scholar]
  18. Edwards JD, Vance DE, Wadley VG, Cissell GM, Roenker DL, Ball KK. The reliability and validity of the Useful Field of View Test as administered by personal computer. Journal of Clinical and Experimental Neuropsychology. 2005;27:529–543. doi: 10.1080/13803390490515432. doi:10.1080/13803390490515432. [DOI] [PubMed] [Google Scholar]
  19. Edwards JD, Wadley VG, Vance DE, Roenker DL, Ball KK. The impact of speed of processing training on cognitive and everyday performance. Aging and Mental Health. 2005;9:262–271. doi: 10.1080/13607860412331336788. doi:10.1080/13607860412331336788. [DOI] [PubMed] [Google Scholar]
  20. Ekstrom RB, French JW, Harman H, Derman D. Kit of factor referenced cognitive tests. Revised ed. Princeton, NJ: Educational Testing Service; 1976. [Google Scholar]
  21. Freund B, Colgrove LA, Burke BL, McLeod R. Self-rated driving performance among elderly drivers referred for driving evaluation. Accident Analysis and Prevention. 2005;37:613–618. doi: 10.1016/j.aap.2005.03.002. doi:10.1016/j.aap.2005.03.002. [DOI] [PubMed] [Google Scholar]
  22. Freund B, Szinovacz M. Effects of cognition on driving involvement among the oldest old: Variations by gender and alternative transportation opportunities. The Gerontologist. 2002;42:621–633. doi: 10.1093/geront/42.5.621. doi:10.1093/geront/42.5.621. [DOI] [PubMed] [Google Scholar]
  23. Gonda J, Schaie KW. Schaie-Thurstone Mental Abilities Test: Word Series Test. Palo Alto, CA: Consulting Psychologists Press; 1985. [Google Scholar]
  24. Good-Lite. Far visual acuity chart. 2010 Retrieved April 16, 2010 from www.good-lite.com. [Google Scholar]
  25. Hakamies-Blomqvist L, Siren A. Deconstructing a gender difference: Driving cessation and personal driving history of older women. Journal of Safety Research. 2003;34:383–388. doi: 10.1016/j.jsr.2003.09.008. doi:10.1016/j.jsr.2003.09.008. [DOI] [PubMed] [Google Scholar]
  26. Jobe JB, Smith DM, Ball KK, Tennstedt SL, Marsiske M, Willis SL, et al. ACTIVE: A cognitive intervention trial to promote independence in older adults. Controlled Clinical Trials. 2001;22:453–479. doi: 10.1016/s0197-2456(01)00139-8. doi:10.1016/S0197-2456(01)00139-8. [DOI] [PMC free article] [PubMed] [Google Scholar]
  27. Jobe-Shields L, Cohen R, Parra GR. Patterns of change in children's loneliness trajectories from third through fifth grades. Merrill-Palmer Quarterly. 2011;57:25–47. doi:10.1353/mpq.2011.0003. [Google Scholar]
  28. Keay L, Munoz B, Turano KA, Hassan SE, Munro CA, Duncan DD, et al. Visual and cognitive deficits predict stopping or restricting driving: The Salisbury Eye Evaluation Driving Study (SEEDS) Investigative Ophthalmology and Visual Science. 2009;50:107–113. doi: 10.1167/iovs.08-2367. doi:10.1167/iovs.08-2367. [DOI] [PMC free article] [PubMed] [Google Scholar]
  29. Kostyniuk LP, Molnar LJ. Self-regulatory driving practices among older adults: Health, age, and sex effects. Accident Analysis and Prevention. 2008;40:1576–1580. doi: 10.1016/j.aap.2008.04.005. doi:10.1016/j.aap.2008.04.005. [DOI] [PubMed] [Google Scholar]
  30. Lesikar S. E., Gallo J. J., Rebok G. W., Keyl P. M. Prospective study of brief neuropsychological measures to assess crash risk in older primary care patients. Journal of the American Board of Family Medicine. 2002;15(1):11–19. [PubMed] [Google Scholar]
  31. Li F, Duncan TE, Duncan SC, Acock A. Latent growth modeling of longitudinal data: A finite growth mixture modeling approach. Structural Equation Modeling. 2001;8:493–530. doi:10.1207/S15328007SEM0804_01. [Google Scholar]
  32. Liang J, van Tran T, Krause N, Markides KS. Generational differences in the structure of the CES-D scale in Mexican Americans. Journal of Gerontology. 1989;44:110–120. doi: 10.1093/geronj/44.3.s110. doi:10.1093/geronj/44.3.S110. [DOI] [PubMed] [Google Scholar]
  33. Lo Y, Mendell NR, Rubin DB. Testing the number of components in a normal mixture. Biometrika. 2001;88:767–778. doi:10.1093/biomet/88.3.767. [Google Scholar]
  34. Lyman JM, McGwin GJ, Jr., Sims RV. Factors related to driving difficulty and habits in older drivers. Accident Analysis and Prevention. 2001;33:413–421. doi: 10.1016/s0001-4575(00)00055-5. doi:10.1016/S0001-4575(00)00055-5. [DOI] [PubMed] [Google Scholar]
  35. Marshall SC, Wilson KG, Molnar FJ, Man-Son-Hing M, Stiell I, Porter MM. Measurement of driving patterns of older adults using data logging devices with and without global positioning system capability. Traffic Injury Prevention. 2007;8:260–266. doi: 10.1080/15389580701281792. doi:10.1080/15389580701281792. [DOI] [PubMed] [Google Scholar]
  36. Molnar LJ, Eby DW. The relationship between self-regulation and driving abilities in older drivers: An exploratory study. Traffic Injury Prevention. 2008;9:314–319. doi: 10.1080/15389580801895319. doi:10.1080/15389580801895319. [DOI] [PubMed] [Google Scholar]
  37. Mora PA, Bennett IM, Elo IT, Mathew L, Coyne JC, Culhane JF. Distinct trajectories of perinatal depressive symptomatology: Evidence from growth mixture modeling. American Journal of Epidemiology. 2009;169:24–32. doi: 10.1093/aje/kwn283. doi:10.1093/aje/kwn283. [DOI] [PMC free article] [PubMed] [Google Scholar]
  38. Muthén BO. Latent variable analysis: Growth mixture modeling and related techniques for longitudinal data. In: Kaplan D, editor. The sage handbook of quantitative methodology for the social sciences. Thousand Oaks, CA: Sage Publications; 2004. pp. 345–368. [Google Scholar]
  39. Muthén BO, Jo B, Brown H. Comment on the Barnard, Frangakis, Hill, & Rubin article, principal stratification approach to broken randomized experiments: A case study of school choice vouchers in New York City. Journal of the American Statistical Association. 2003;98:311–314. doi:10.1198/016214503000071. [Google Scholar]
  40. Muthén LK, Muthén BO. Mplus User's Guide. 5th ed. Los Angeles, CA: Author; 2007. [Google Scholar]
  41. O’Connor ML, Edwards JD, Wadley VG, Crowe M. Changes in mobility among older adults with psychometrically defined mild cognitive impairment. The Journals of Gerontology, Series B: Psychological Sciences and Social Sciences. 2010;65:306–316. doi: 10.1093/geronb/gbq003. doi:10.1093/geronb/gbq003. [DOI] [PMC free article] [PubMed] [Google Scholar]
  42. Owsley C, McGwin GJ, Jr., Sloane ME, Stalvey BT, Wells J. Timed instrumental activities of daily living tasks: Relationship to visual function in older adults. Optometry and Vision Science. 2001;78:350–359. doi: 10.1097/00006324-200105000-00019. doi:10.1097/00006324-200105000-00019. [DOI] [PubMed] [Google Scholar]
  43. Owsley C, Stalvey BT, Wells J, Sloane ME. Older drivers and cataract: Driving habits and crash risk. The Journals of Gerontology, Series A: Biological and Medical Sciences. 1999;54:203–211. doi: 10.1093/gerona/54.4.m203. doi:10.1093/gerona/54.4.M203. [DOI] [PubMed] [Google Scholar]
  44. Radloff L. The CES-D scale: A self-report depression scale for research in the general population. Applied Psychological Measurement. 1977;1:385–401. doi:10.1177/014662167700100306. [Google Scholar]
  45. Roenker DL, Cissell GM, Ball KK, Wadley VG, Edwards JD. Speed of processing and driving simulator training result in improved driving performance. Human Factors. 2003;45:218–233. doi: 10.1518/hfes.45.2.218.27241. doi:10.1518/hfes.45.2.218.27241. [DOI] [PubMed] [Google Scholar]
  46. Ross LA. Does speed of processing training impact driving mobility in older adults? Birmingham: University of Alabama; 2007. (Doctoral dissertation) [Google Scholar]
  47. Ross LA, Clay OJ, Edwards JD, Ball KK, Wadley VG, Vance DE, et al. Do older drivers at-risk for crashes modify their driving over time? The Journals of Gerontology, Series B: Psychological Sciences and Social Sciences. 2009;64:163–170. doi: 10.1093/geronb/gbn034. doi:10.1093/geronb/gbn034. [DOI] [PMC free article] [PubMed] [Google Scholar]
  48. Schwartz SJ, Mason CA, Pantin H, Szapocznik J. Longitudinal relationships between family functioning and identity development in Hispanic adolescents: Continuity and change. The Journal of Early Adolescence. 2009;29:177–211. doi: 10.1177/0272431608317605. doi:10.1177/0272431608317605. [DOI] [PMC free article] [PubMed] [Google Scholar]
  49. Shope JT. What does giving up driving mean to older drivers and why is it so difficult? Generations. 2003;2:57–59. [Google Scholar]
  50. Stalvey BT, Owsley C, Sloane ME, Ball KK. The Life Space Questionnaire: A measure of the extent of mobility of older adults. Journal of Applied Gerontology. 1999;18:460–478. doi:10.1177/073346489901800404. [Google Scholar]
  51. Steinhagen-Thiessen E, Borchelt M. Morbidity, medication, and functional limitations in very old age. In: Baltes PB, Mayer KU, editors. The Berlin Aging Study: Aging from 70 to 100. New York, NY: Cambridge University Press; 1999. pp. 131–166. [Google Scholar]
  52. Tabachnick BG, Fidell LS. Using multivariate statistics. 4th ed. Needhan Heights, MA: Allyn & Bacon; 2001. [Google Scholar]
  53. Thurstone L, Thurstone T. Examiner manual for the SRA Primary Mental Abilities Test (form 10–14) Chicago, IL: Science Research Associates; 1949. [Google Scholar]
  54. Vance DE, Roenker DL, Cissell GM, Edwards JD, Wadley VG, Ball KK. Predictors of driving exposure and avoidance in a field study of older drivers from the state of Maryland. Accident Analysis and Prevention. 2006;38:823–831. doi: 10.1016/j.aap.2006.02.008. doi:10.1016/j.aap.2006.02.008. [DOI] [PubMed] [Google Scholar]
  55. Wang M, Bodner TE. Growth mixture modeling: Identifying and predicting unobserved subpopulations with longitudinal data. Organizational Research Methods. 2007;10:635–656. doi:10.1177/1094428106289397. [Google Scholar]
  56. Wechsler D. WAIS-R manual. New York, NY: The Psychological Corporation; 1981. [Google Scholar]
  57. West CG, Gildengorin G, Haegerstrom-Portnoy G, Lott LA, Schneck ME, Brabyn JA. Vision and driving self-restriction in older adults. Journal of the American Geriatrics Society. 2003;51:1348–1355. doi: 10.1046/j.1532-5415.2003.51482.x. doi:10.1046/j.1532-5415.2003.51482.x. [DOI] [PubMed] [Google Scholar]
  58. Willis SL. Everyday problem solving. In: Birren JE, Schaie KW, editors. Handbook of the psychology of aging. 4th ed. San Diego, CA: Academic Press; 1996. pp. 287–307. [Google Scholar]
  59. Wilson B, Cockburn J, Baddeley A. The Rivermead Behavioral Memory Test. Reading, England: Thames Valley Test; 1985. [Google Scholar]

Articles from The Journals of Gerontology Series B: Psychological Sciences and Social Sciences are provided here courtesy of Oxford University Press

RESOURCES