Introduction
Mobility has been identified as an essential need for older adults, allowing individuals to engage in activities, socialize, run errands, and access health care services (Webber et al., Reference Webber, Porter and Menec2010). Driving mobility is understood as the ability to achieve mobility by car and can meet older adults’ practical, emotional, and experiential needs by enabling essential tasks, fostering autonomy, and providing opportunities for enjoyment (Musselwhite & Haddad, Reference Musselwhite and Haddad2010). Although driving rates decline with age (Schouten et al., Reference Schouten, Wachs, Blumenberg and King2022), older adults rely on driving as a main form of mobility at a comparable frequency to other age groups (Scott & Tulloch, Reference Scott and Tulloch2021).
The ability to drive has been associated with increased quality of life (Dickerson et al., Reference Dickerson, Molnar, Eby, Adler, Bédard, Berg-Weger, Classen, Foley, Horowitz, Kerschner, Page, Silverstein, Staplin and Trujillo2007; Musselwhite & Haddad, Reference Musselwhite and Haddad2010) and provides older adults with a feeling of freedom and independence (Stinchcombe et al., Reference Stinchcombe, Hopper, Mullen and Bédard2021). While data suggest that the current generation of older adults is driving longer than earlier generations (Schouten et al., Reference Schouten, Wachs, Blumenberg and King2022), many older adults eventually cease driving. Some older adults who cease driving report decreased feelings of independence and increased reliance on others (Mullen et al., Reference Mullen, Parker, Wiersma, Stinchcombe and Bédard2017), which can lead to significant lifestyle changes and limitations in accessing essential services (Musselwhite & Haddad, Reference Musselwhite and Haddad2010). Additionally, driving cessation has been associated with declines in physical, social, and cognitive health, as well as increased risks of admission to long-term care and mortality (Chihuri et al., Reference Chihuri, Mielenz, DiMaggio, Betz, DiGuiseppi, Jones and Li2016). Specifically, research has found that driving cessation almost doubled the risk of depressive symptoms (Chihuri et al., Reference Chihuri, Mielenz, DiMaggio, Betz, DiGuiseppi, Jones and Li2016) and is associated with higher social isolation (Qin et al., Reference Qin, Xiang and Taylor2020) in older adults. Despite the well-documented consequences of driving cessation, there remains limited understanding of the multifaceted factors leading up to this transition, particularly in the Canadian context. As such, this study aims to identify and examine the range of physical, cognitive, and environmental antecedents that influence the transition from driver to non-driver among middle-aged and older Canadians.
In their comprehensive framework of mobility in older adults, Webber et al. (Reference Webber, Porter and Menec2010) present five factors that affect mobility: cognitive, psychosocial, physical, environmental, and financial factors, with gender, culture, and personal life history identified as important cross-cutting influences (Webber et al., Reference Webber, Porter and Menec2010). The factors identified in Webber et al.’s (Reference Webber, Porter and Menec2010) framework are interdependent; for example, age-related declines in vision may interact with environmental conditions and cause difficulties when driving in certain situations. Cross-cutting influences shape experiences across mobility determinants. For example, changes in vision (i.e., a physical determinant) may limit driving safety in low light, whereas gender norms influence self-restriction and cessation of driving. Indeed, research has shown that men associate driving with feelings of masculinity and pride and are therefore less likely to stop driving compared to women (Ang, Oxley, Chen, & Lee, Reference Ang, Oxley, Chen and Lee2019). Other research identifying factors that influence driving cessation suggests that cognition, notably deficits in attention, executive function, and global cognition, is linked to driving safety (Jacobs et al., Reference Jacobs, Hart and Roos2017). In a systematic review and meta-analysis of self-regulation practices among individuals aged 60 and older, Ang et al. (Reference Ang, Oxley, Chen, Yap, Song and Lee2019) identified that physical fitness, mental well-being, the impact of social networks, and the level of support provided to older adults were key factors motivating changes in their mobility and travel patterns.
The transportation environment also impacts driving status insofar as older drivers with access to alternative modes of transportation may retain mobility following driving cessation, while older adults living in rural areas may choose to continue driving, despite declining fitness to drive (Ang, Oxley, Chen, & Lee, Reference Ang, Oxley, Chen and Lee2019). The negative effects of a lack of mobility associated with driving cessation can be mitigated by the availability of rides from family members and friends or through the availability of alternative and accessible modes of transport (Musselwhite & Shergold, Reference Musselwhite and Shergold2013). In terms of financial factors, a qualitative study by Stinchcombe et al. (Reference Stinchcombe, Hopper, Mullen and Bédard2021) found that financial security was an emergent theme in that participants described the costs associated with a driving (e.g., vehicle maintenance, fuel, etc.) as a barrier to continuing driving, but also weighed the costs associated with alternative modes of transportation (e.g., taxi, bus, etc.) concluding that they would most likely reduce out of the house activities if they had to rely on alternative transportation.
A recent study by Wood et al. (Reference Wood, Bhojak, Jia, Jacobsen, Snitz, Chang and Ganguli2023) found that in a cohort of 1,982 older adults aged 65 and older, poor self-reported health, more depression symptoms, poorer cognition, and being female were associated with a higher likelihood of ceasing driving. In their integrative review of non-cognitive factors related to driving cessation, Maliheh et al. (Reference Maliheh, Nasibeh, Yadollah, Hossein and Ahmad2023) identified physical health, psychological health, interpersonal factors, sociodemographic characteristics, and transportation policies as relevant predictors of stopping driving. In addition, they note that physical health and physical changes are among the most important predictors and that psychological factors remain understudied. It is clear that driving cessation occurs for a variety of reasons, highlighting the importance of considering multiple domains when examining mobility in older adults. Identifying factors that may predict driving cessation can help older adults and their care partners better plan for driving retirement (Schofield et al., Reference Schofield, Kean, Oprescu, Downer and Hardy2023).
While the negative impacts of driving cessation are clear, the process leading up to driving cessation is complex, and the specific covariates associated with it remain understudied. Moreover, much of the existing literature is cross-sectional, precluding insights into temporality. Identifying factors that are associated with driving cessation is crucial for developing interventions and policies to support older adults in transitioning from being a driver to a non-driver. This study leverages longitudinal data to examine how baseline psychosocial, physical, cognitive, and environmental factors are associated with changes in driving status over three years, offering insights into potential causal pathways and the dynamic nature of this transition. The objective of this study is to better understand how the wide range of factors, including age, gender, income, physical health, cognitive function, vision, and urban or rural residence, influence the decision to stop driving among older adults in Canada. The selection of these variables was informed by prior research and reflects the multifaceted determinants of driving status, providing a comprehensive understanding of the interplay between physical, cognitive, and environmental influences on mobility transitions. These insights will inform interventions and policies that support older adults in transitioning from driver to non-driver roles while maintaining mobility and independence.
Method
Data set
The data presented in this analysis are from the Canadian Longitudinal Study on Aging (CLSA), a nationwide study that collects data every three years until the study concludes, or participants pass away. The CLSA recruited a cohort of over 50,000 Canadians aged between 45 and 85 at baseline and will follow them over time. Baseline data collection occurred from 2011 to 2015, and Follow-up 1 data collection occurred between 2015 and 2018 (Raina et al., Reference Raina, Wolfson, Kirkland, Griffith, Balion, Cossette, Dionne, Hofer, Hogan, van den Heuvel, Liu-Ambrose, Menec, Mugford, Patterson, Payette, Richards, Shannon, Sheets, Taler and Young2019). Data collection is expected to continue for 20 years.
Recruitment for the CLSA began in 2010, using a subset of the Canadian Community Health Survey (CCHS) – Healthy Aging, provincial health care registration databases, and random digit dialing of landlines. Certain populations were excluded from the study, including individuals unable to respond in either English or French, those with cognitive impairment at baseline, individuals residing in Canadian territories or specific remote areas, individuals living on First Nations reserves or settlements, full-time members of the Canadian Armed Forces, and institutionalized populations.
The CLSA encompasses two cohorts: a comprehensive cohort (n = 30,097) and a tracking cohort (n = 21,241). Participants in the tracking cohort take part in computer-assisted telephone interviews (CATI), whereas participants in the comprehensive cohort participate in face-to-face in-home interviews, and additional data collection at a data collection site (DCS) (Raina et al., Reference Raina, Wolfson, Kirkland, Griffith, Balion, Cossette, Dionne, Hofer, Hogan, van den Heuvel, Liu-Ambrose, Menec, Mugford, Patterson, Payette, Richards, Shannon, Sheets, Taler and Young2019). To be a part of the comprehensive cohort, the participants must live within 25–50 km of one of the 11 DCS distributed across Canada. All participants provided written informed consent (Raina et al., Reference Raina, Wolfson, Kirkland, Griffith, Balion, Cossette, Dionne, Hofer, Hogan, van den Heuvel, Liu-Ambrose, Menec, Mugford, Patterson, Payette, Richards, Shannon, Sheets, Taler and Young2019). The University of Ottawa Research Ethics Board (REB) granted approval for the analyses presented in this work. In this study, we analyzed Baseline and Follow-up 1 data.
Measures
Driving status change
Participants were asked their driving status twice (at baseline and Follow-up 1) with possible answers including ‘never had a driver’s license’, ‘had a driver’s license at one point in life, but not currently’, ‘have a driver’s license without restrictions’ and ‘have a driver’s license with restrictions’. If participants responded that they had a driver’s license (with or without restrictions), they were then asked how frequently they drive. For this analysis, a current driver was defined as a participant who stated that they had a valid licence and did not respond ‘not at all’ to how frequently they drive. To determine change in driving status, participants were considered to have stopped driving if they were drivers at baseline then reported that they did not have a license at Follow-up 1, or were drivers at baseline and reported that they do not drive at all at Follow-up 1 (response option ‘not at all’). Participants who never had a driver’s license were excluded from analysis.
Demographics
Participants were asked to report their chronological age. For analysis, chronological age was divided by five to examine the risk of stopping driving at five-year increments. Sex (male/female) was recorded by the interviewer. Annual household income was categorized as <$20,000 (referent), $20,000–$49,999, $50,000–$99,999, and ≥$100,000. Statistics Canada’s Postal Code Conversion File (PCCF) was used to determine if a participant lives in an urban or rural area. Participants were also asked how many people they lived with which was then classified as living alone (response of 0) or not living alone (response ≥ 1).
Health variables
Self-reported general health was captured by asking participants, ‘In general, would you say your health is excellent, very good, good, fair, or poor?’. Response options excellent and very good were combined for the analysis. A measure of memory was computed based on two trials of the Rey Auditory Verbal Learning Test (RAVLT) (Rey, Reference Rey1964). Administration of the RAVLT required participants to listen to a list of 15 words and immediately recall them within 90 seconds. Only one of the five RAVLT trials was administered. After five minutes, participants were asked to recall as many of the initial words as they could within 60 seconds (Tuokko et al., Reference Tuokko, Griffith, Simard and Taler2017). Scores for the (first) immediate and delayed recall trials represent the total number of correct responses for each trial, respectively, and have a possible range of 0–15. For this analysis, the immediate and delayed recall scores were combined to create a score ranging from 0 to 30. The RAVLT has been shown to exhibit strong psychometric properties (Lezak, Reference Lezak, Howieson, Bigler and Tranel2012; Tuokko et al., Reference Tuokko, Griffith, Simard and Taler2017). To measure sensory function, participants self-reported the quality of their hearing and vision on a five-point scale: excellent, very good, good, fair, or poor. Participants were classified as having hearing or vision problems if they rated their hearing or vision as fair or poor. A measure of recent depressive symptoms was determined using the Center for Epidemiologic Studies Short Depression Scale (CESD-10) (Andresen et al., Reference Andresen, Malmgren, Carter and Patrick1994). The CESD-10 consists of items designed to evaluate the frequency of depressive symptoms experienced over the past week. Participants responded to each item using a four-point scale with the following options: rarely or never (less than one day), some of the time (1–2 days), occasionally (3–4 days), and all of the time (5–7 days). An overall score between 0 and 30 was calculated by summing the response values, such that a higher score indicates more depressive symptoms. The CESD-10 has demonstrated strong reliability and validity and acceptable sensitivity and specificity in identifying depression (Björgvinsson et al., Reference Björgvinsson, Kertz, Bigda-Peyton, McCoy and Aderka2013; González et al., Reference González, Nuñez, Merz, Brintz, Weitzman, Navas, Camacho, Buelna, Penedo, Wassertheil-Smoller, Perreira, Isasi, Choca, Talavera and Gallo2017). It also reliably measures depressive symptoms across diverse populations, age groups, education levels, and administration language (English or French) (O’Connell et al., Reference O’Connell, Grant, McLean, Griffith, Wolfson, Kirkland and Raina2018).
Activities of daily living (ADLs) were measured using modified questions from the Older Americans’ Resources and Services Multidimensional Functional Assessment Questionnaire (CLSA, 2018; Fillenbaum & Smyer, Reference Fillenbaum and Smyer1981). Participants were asked seven questions about their Basic ADLs and seven questions about their instrumental ADLs. A measure of dependence was determined based on the total number of times a respondent indicated that they needed help with an activity or were unable to do an activity. All ADL and instrumental ADL items were included besides meal preparation. Meal preparation was excluded from the total score as the original creators of the measure deemed meal preparation as independently more detrimental to maintaining independent living than other ADLs (Fillenbaum, Reference Fillenbaum2013). Possible scores ranged from 0 to 13, such that higher scores represent greater dependence.
Analysis
Demographic characteristics were compared between groups using χ 2 (for categorical variables) and t-tests (for continuous variables). Binomial logistic regression was used to examine the relationship between covariates and change in driving status (i.e., having a licence at baseline but not at follow-up vs. maintaining a licence). This multivariable approach calculates odds ratios, which indicate the likelihood of a change in driving status associated with each predictor, while accounting for the effects of other variables in the model. All predictors were simultaneously entered into the model. To analyse differences between participants who maintained their driving status vs. those who transitioned from driver to non-driver, any participants who never had a license or did not have a license at baseline were excluded from the analysis. Missing data were addressed using listwise deletion, wherein cases with missing values for any variable included in the analysis were excluded, an approach that ensures that the results are based on complete data for variables included in the model. After removing missing data (n = 5,582; 15%), the final analytic sample was n = 30,901. An alpha level of 0.05 was used to determine statistical significance. The statistical models met the necessary assumptions, including the absence of multicollinearity. Multicollinearity was assessed by computing Variance Inflation Factors (VIF) for each variable – all VIF values were low and within an acceptable range. All data analyses were conducted in R version 4.2.1 (release date: June 23, 2022).
Results
Characteristics of participants in the analytic sample are presented in Table 1. The mean age of participants at baseline was 61.6 years. The mean age of participants who continued driving at Follow-up 1 was 61 (SD = 10), and the mean age of people who stopped driving at Follow-up 1 was 72 (SD = 10) (p < .001). There was an even representation of males (52%) and females (48%) in the sample, with the majority of participants living in urban areas (86%). 64 per cent of participants who continued driving reported their health to be excellent or very good, compared to 43 per cent of those who stopped driving. Hearing issues were reported by 11 per cent of continuing drivers and 16 per cent of those who stopped driving; vision issues were reported by 6 per cent and 16 per cent, respectively. Supplemental eTable 1 shows participant characteristics broken down by cohort (Comprehensive and Tracking) and driving status (drivers vs. non-drivers).
Participant characteristics
Results of the logistic regression are presented in Table 2. Participants transitioning from driver to non-driver tended to be older (OR = 1.589, 95% CI: 1.505–1.679) and female (OR for males = 0.699, 95% CI: 0.569–0.856). Income was significantly associated with transitioning from driver to non-driver; compared to the lowest income category (<$20,000/year), participants with higher incomes were less likely to cease driving over the study period (e.g., OR = 0.338, 95% CI: 0.242–0.475 for those earning $50,000–$99,999).
Binomial logistic regression results (n = 30,901)
Additionally, participants living in an urban area had increased odds of stopping driving (OR = 1.564, 95% CI: 1.164–2.150) compared to those in rural areas. Living with others was not significantly associated with a participant’s odds of stopping driving (OR = 0.990, 95% CI: 0.801–1.227, p = .926).
Self-rated general health was strongly correlated with transitions to non-driving. Compared to participants reporting excellent or very good health, those who reported good (OR = 1.494, 95% CI: 1.214–1.838), fair (OR = 1.980, 95% CI: 1.459–2.660), or poor (OR = 4.244, 95% CI: 2.604–6.713) health had higher odds of stopping driving.
Hearing problems were not significantly associated with stopping driving (OR = 0.866, 95% CI: 0.665–1.116, p = .278), whereas vision problems were significantly associated (OR = 2.116, 95% CI: 1.624–2.726). With each additional ADL limitation, participants’ odds of stopping driving increased (OR = 1.435, 95% CI: 1.278–1.601).
Memory scores were inversely associated with stopping driving – a one-point increase in memory score was associated with reduced odds of stopping driving (OR = 0.957, 95% CI: 0.931–0.983). Participants reporting more depressive symptoms also had slightly greater odds of stopping driving (OR = 1.031, 95% CI: 1.011–1.051, p = .002), although the effect was modest.
To explore potential age-related differences in the main findings, we conducted a Supplemental analysis comparing participants aged 45–64 and those 65–85 at baseline (see Supplemental eTable 2). While the direction of most associations was broadly consistent across age groups and with the main model, some notable differences emerged. In contrast to the main model, neither age nor sex was significantly associated with driving cessation among younger participants. Several health-related variables that were significant in the full sample (such as memory and vision problems) were not significant in the model with the younger participants. Among participants aged 65–85, associations were more aligned with the full model, including significant effects for age, sex, memory, and vision problems.
Discussion
Enhancing mobility is a critical factor in promoting healthy aging and encompasses various dimensions, including cognitive, physical, psychosocial, and environmental domains. The act of driving plays a pivotal role in fostering mobility and, subsequently, in enhancing overall quality of life for older adults. Facilitating smooth transitions to non-driving requires a comprehensive understanding of the factors that are associated with changes in driving status. The purpose of this study was to shed light on the process of stopping driving and identify variables that were associated with a change in driving status over three years. The study yielded several noteworthy results.
The findings of this study align with and expand upon the Webber et al. (Reference Webber, Porter and Menec2010) framework for mobility in later life, which emphasizes the interplay between physical, cognitive, psychosocial, and environmental determinants of mobility. Our results underscore the importance of physical health (e.g., difficulties with ADL) and cognitive function (e.g., memory) in predicting changes in driver status. Specifically, the observed relationship between memory impairments and change in driver status supports prior evidence suggesting that cognitive health is a critical component of driving fitness, as memory limitations may reduce the ability to navigate complex traffic scenarios or recall essential driving rules (Wood et al., Reference Wood, Bhojak, Jia, Jacobsen, Snitz, Chang and Ganguli2023; Zhang et al., Reference Zhang, Guo, Yuan and Li2023). Our findings showed that when considered together, physical factors were stronger predictors compared to mental health and cognition variables. This is in contrast to previous findings that showed that well-being and cognition are stronger predictors of driving cessation compared to health variables (Anstey et al., Reference Anstey, Windsor, Luszcz and Andrews2006).
Our findings showed that low income was associated with a change from driving to non-driving over the three-year period. Financial stability may also influence older adults’ decision to continue driving or not. Similar to our findings, in an Australian sample of older adults, financial problems predicted driving cessation (Anstey et al., Reference Anstey, Li, Hosking and Eramudugolla2017). It is possible that the cost of owning and maintaining a vehicle may play a role in the decision to stop driving (Stinchcombe et al., Reference Stinchcombe, Hopper, Mullen and Bédard2021), especially in urban areas where there are affordable alternative transport options.
Moreover, the significant influence of environmental factors, such as urban vs. rural residence, reflects the contextual variability in mobility needs and resources. Rural residents, for example, often face fewer transportation alternatives (Hansen et al., Reference Hansen, Newbold, Scott, Vrkljan and Grenier2020) and anticipate more significant impacts from driving cessation, making the decision to stop driving particularly consequential (Strogatz et al., Reference Strogatz, Mielenz, Johnson, Baker, Robinson, Mebust, Andrews, Betz, Eby, Johnson, Jones, Leu, Molnar, Rebok and Li2020). In our study, urban older adults were more likely to stop driving compared to rural drivers. Despite declining health and recommendations to stop driving, rural older adults may decide to drive longer (Hansen et al., Reference Hansen, Newbold, Scott, Vrkljan and Grenier2020). This may be due to the importance that rural drivers place on driving, caused by a lack of alternative transport options (Strogatz et al., Reference Strogatz, Mielenz, Johnson, Baker, Robinson, Mebust, Andrews, Betz, Eby, Johnson, Jones, Leu, Molnar, Rebok and Li2020). These findings emphasize the multidimensional nature of mobility transitions and reinforce the need for a holistic approach to understanding driving cessation, one that considers the dynamic interactions between individual health and environmental constraints.
To explore whether the results from the main analysis differed by age, we conducted a supplemental analysis stratified by age at baseline (separate models for participants aged 45–64 and 65–84 years). While the overall pattern of results was broadly consistent across groups, several distinctions emerged. Among younger participants, higher income was more strongly associated with continued driving, and living alone was linked to lower odds of stopping driving, an association not observed in the older group or the main model. Factors such as age, sex, memory performance, and vision problems were significant factors among older participants only. These findings suggest that while functional and health-related predictors are important across age groups, social and economic factors may play a more prominent role in driving transitions earlier in later life. Driving cessation can happen at any time during the driving lifespan, and these results highlight the value of age-sensitive approaches to understanding and supporting driving continuity and transitions to non-driving.
This study has numerous strengths, including a large population-based sample, longitudinal data that enabled us to examine changes to driving status, and the availability of numerous variables that allowed for the examination of the complexity of mobility change. This study is not without its limitations. Firstly, no data were available regarding the reason a participant went from being a driver to a non-driver. As such, it is not possible to distinguish those who voluntarily stopped driving from those who had their license revoked. Further, CLSA data follows one cohort of individuals and therefore may not be generalizable to other cohorts. Previous research has found cohort differences in driving habits, such that younger cohorts are less likely to limit and stop driving, potentially due to maintaining greater health into older age (Schouten et al., Reference Schouten, Wachs, Blumenberg and King2022). Aside from cognition data, the data analyzed here were self-reports and subject to response bias. While income was significant in the model, its relationship with change in driver status may, at least in part, reflect differences in employment or retirement status between groups, especially since the former driving group was older. Memory was the sole cognitive measure, excluding other important domains such as executive function and processing speed, which are also important for safe driving. Similarly, data on participants’ mobility following a change in driver status were not captured; such data would help delineate the mobility impacts of transition to non-driving.
For many older adults, driving cessation is a critical transition that is associated with a host of negative physical and mental health outcomes (Chihuri et al., Reference Chihuri, Mielenz, DiMaggio, Betz, DiGuiseppi, Jones and Li2016; Missell-Gray & Simning, Reference Missell-Gray and Simning2024). Evidence-based approaches to facilitate smooth transitions to driving cessation while maintaining mobility are needed. The results presented here emphasize the importance of physical and cognitive health in driving transitions and highlight Canadian-specific contextual factors, such as urban or rural residence. The results of this study can help identify individuals who are at risk of mobility change and target supports appropriately.
Supplementary material
The supplementary material for this article can be found at http://doi.org/10.1017/S0714980825100184.
Data availability statement
Data are available from the Canadian Longitudinal Study on Aging (www.clsa-elcv.ca) for researchers who meet the criteria for access to de-identified CLSA data.
Acknowledgements
This research was made possible using the data/biospecimens collected by the Canadian Longitudinal Study on Aging (CLSA). Funding for the CLSA is provided by the Government of Canada through the Canadian Institutes of Health Research (CIHR) under grant reference: LSA 94473 and the Canada Foundation for Innovation, as well as the following provinces: Newfoundland, Nova Scotia, Quebec, Ontario, Manitoba, Alberta, and British Columbia. This research has been conducted using the CLSA data sets Baseline Tracking Dataset version 4.0, Baseline Comprehensive Dataset version 7.0, Follow-Up 1 Tracking Dataset version 3.1, and Follow-Up 1 Comprehensive Dataset version 5.0, under Application Number 2109030. The CLSA is led by Drs. Parminder Raina, Christina Wolfson, and Susan Kirkland.
Competing interests
The opinions expressed in this manuscript are the author’s own and do not reflect the views of the Canadian Longitudinal Study on Aging (CLSA).
Open access
