Abstract
The Transportation Support Scale (TSS) has the potential to help measure and clarify care partners’ responses—both negative and positive—to driving cessation as well as inform the development and evaluation of services for care partners who are providing transportation to former drivers.
Mobility is a meaningful activity for many older adults. Although the concept of mobility encompasses several environments, from one’s immediate home to geographically distant places (Webber et al., 2010), mobility beyond the home typically is ensured by personal transportation means (Owsley, 2002). Results from the 2001 American National Household Travel Survey indicated that 89% of Americans ages 65 yr and older use a personal vehicle for their travel (Collia et al., 2003). Similarly, information from the Canadian Longitudinal Study on Aging revealed that 84% of women and 94% of men ages 75 yr and older rely on driving or being a passenger in a personal vehicle as their main form of transportation (Vrkljan et al., 2018).
Although most older adults may want to maintain their mobility by driving well into old age (Naumann et al., 2014), it is often not possible; the typical woman and man will live for 10 yr and 7 yr, respectively, after driving cessation (Foley et al., 2002). Despite this predictability, the prospect of driving cessation is often ignored or met with great reluctance. Yassuda et al. (1997) found that most drivers did not even want to consider driving cessation, stating a lack of suitable transportation alternatives, and that driving cessation meant a loss of independence. Shope (2003) also documented that older drivers did not want to consider driving cessation, because their vehicle was an essential element of their lives. In a recent study of 92 older drivers, this same reluctance about driving cessation was reiterated, along with the shortcomings of other mobility options (e.g., walking, public transportation alternatives; Stinchcombe et al., 2021).
This reluctance about driving cessation highlights that the need for, and the many advantages of, a private automobile do not end when one stops driving. Hence, most older adults who cease driving often become dependent on unpaid care partners (family and friends) for transportation (Bonnel, 1999; Siren & Haustein, 2015; Taylor & Tripodes, 2001; Zeitler & Buys, 2015). Notably, Sinha (2013) reported that the most common task provided by care partners was transportation (73% of respondents), ahead of housework (51%), house maintenance and outdoor work (45%), scheduling and coordinating appointments (31%), managing finances (27%), assisting with medical treatments (23%), and providing personal care (22%).
Despite care partners’ leading role in providing transportation, very few studies have been designed specifically to explore the impact that driving cessation has on them. In their seminal study exploring the effects of driving cessation on individuals with dementia and their care partners, Taylor and Tripodes (2001) found that 56% of individuals with dementia depended primarily on a spouse for transportation, whereas 24% depended on adult children (usually daughters and daughters-in-law). They highlighted that care partners “must frequently juggle work and caregiving responsibilities, including chauffeuring the non-driver to shopping, medical appointments, and recreational activities” (p. 520). Not surprisingly, 33% of care partners in their sample missed work, and 13% quit work entirely to fulfill care-partner duties, including providing transportation. These findings were echoed in a recent survey of family members of older drivers, in which 80% of respondents indicated that their relative’s driving cessation would have a negative impact on them, with the most important aspect being spending more time providing transportation (Connor et al., 2021).
The potential impact of driving cessation on care partners is not lost on former drivers, either. In a focus group, former drivers expressed concerns about the additional pressures placed on others after driving cessation (Mullen et al., 2017). Similarly, in a 2-yr longitudinal study examining the impact of driving cessation on both older drivers and primary care partners, older drivers reported concerns about being a burden on their care partners (Schryer et al., 2019).
Despite the emerging evidence that driving cessation may affect care partners, this impact remains challenging to assess (Liddle et al., 2017). Thus, a natural starting point for addressing the pressures that care partners may experience is to better capture data on how care partners are affected by having to provide transportation after someone’s driving cessation. There are numerous generic tools to measure the impact of caring, but there is, to our knowledge, no tool that specifically measures the care partner’s response to driving cessation. Furthermore, although positive aspects of providing transportation have been reported (Cohen et al., 2002; Liddle et al., 2017; Tarlow et al., 2004), generic tools typically measure only the negative aspects of caring. For example, recent data indicated that care partners of older adults who do not drive have higher levels of burden than care partners of drivers (even after controlling for functional status and many potential confounding variables), but the potential positive aspects of caring were not captured (Connors et al., 2020).
Our goal was to develop a tool to measure both the negative and positive effects on care partners that may arise from driving cessation and from assuming transportation responsibilities. The ideal tool would provide clinicians and researchers the means to gather more details in various care-partner life domains. This would allow the evaluation of interventions to support former drivers and care partners through the driving cessation process. Using accepted psychometric practices for tool development (Streiner et al., 2015), we created the Transportation Support Scale (TSS) to evaluate how unpaid care partners are affected by driving cessation and transportation responsibilities. We then used the tool to document this impact on a convenience sample of care partners and provided initial validation results.
Method
Transportation Support Scale Item Development
We took a rational–empirical (sequential construct–oriented) approach to scale development (Jackson, 1984). An initial item pool was developed by all of the authors, who represent expertise in various relevant areas, including transportation, the impact of caring, mental health, psychometrics, community-based research, quantitative methods, and qualitative methods. Items were developed with reference to the literature covering driving cessation and caring, as well as existing tools that measure the impact of caring in other domains. These tools included the Caregiver Burden Inventory (Novak & Guest, 1989), the Caregiver Strain Index (Robinson, 1983), the Screen for Caregiver Burden (Hirschman et al., 2004), and the Zarit Burden Interview–12 (ZBI–12; Bédard et al., 2001). The literature on the positive aspects of caring (e.g., Cohen et al., 2002; Tarlow et al., 2004) was also used to develop items measuring the positive impacts. We reworded existing items from other tools and created new items to cover all potential care-partner domains, including Physical, Social, Time, Emotional, Financial, and Positive domains (e.g., spending more time with the care recipient). The initial 98 items were grouped together to facilitate easier understanding and flow of the tool.
Items were assessed for face validity as well as interpretability (readability, ambiguity, double-barreled questions, jargon, value-laden words, positive and negative wording, and length of items; Streiner et al., 2015). Answers were set on a 5-point Likert scale ranging from 1 (strongly disagree) to 5 (strongly agree; positively worded items were reverse-scored for the analyses; higher scores represent a higher negative impact of caring and lesser positivity). Use of the 5-point Likert scale is consistent with research indicating that the optimum number of options ranges from 4 to 7 (Lozano et al., 2008).
Participants
Participants for this study were a convenience sample of unpaid care partners. The inclusion criteria included being a care partner for and providing transportation to an older adult (age 65 yr or older) who has ceased driving (voluntarily or involuntarily), as well as being able to read and converse in the English language. All participants were recruited using posters and newspaper advertisements in various locations throughout a city with a population of 110,000 people. We did not purposefully recruit participants in long-term care homes, as we expected that there is less need for transportation from care partners for individuals living in long-term care, but care partners who were providing transportation for former drivers living in long-term care were not excluded from this study. All participants were offered a $25 gift certificate in appreciation for their participation. All participants provided informed consent before their involvement. Data collection took place during the period of September 2017 to May 2018.
Procedure
Phase 1
After item development, we pretested the tool with a small sample of care partners (n = 11). Collins (2003) encourages the use of pretesting for questionnaires to examine the question-and-answer process in a systematic way and to determine whether participants understand the items, respond in a consistent manner, and respond in a way envisioned by the researchers. Thus, our primary objective in this phase was to qualitatively assess how participants responded to the tool.
This process involved conducting semistructured cognitive interviews with participants, using both “think-aloud” interviewing and “probing” methods (Collins, 2003). Think-aloud interviewing involves asking the participant to think aloud as they are completing items, whereas probing involves the interviewer using specific probes to elicit how participants approach responding to items (Collins, 2003). Collins (2003) also suggested that cognitive interviewing acts as a diagnostic tool in pretesting and that it “focuses mainly on the questionnaire rather than the survey process, paying explicit attention to the mental processes respondents use to answer survey questions and thus allows covert as well as overt problems to be identified” (p. 235). Participants had the option of completing the interview either at the university or at a location convenient to them (e.g., the participant’s home). We used feedback from these participants to revise the questionnaire before Phase 2.
Phase 2
Phase 2 involved further evaluation of the tool with a larger sample of care partners (n = 66). Participants completed the questionnaire independently (without cognitive interviewing) at one of several possible locations, such as the university, a location convenient to them, or one of the city libraries, during scheduled sessions.
Demographic information was collected, as well as information regarding care partners’ transportation time commitment and circumstances surrounding driving cessation. To account for social desirability bias, participants in Phase 2 completed the Social Desirability Scale (SDS) from the Jackson Personality Research Form (Jackson, 1984). This is a 16-item measure of social desirability with various statements (e.g., “If someone gave me too much change, I would tell him/her”) answered in a true–false format.
To explore the preliminary construct and convergent validity of the TSS, participants completed the ZBI–12 (Bédard et al., 2001). The ZBI–12 is frequently used to assess the generic impact of caring. The ZBI–12 items are answered on a 5-point Likert-type scale ranging from 0 (never) to 4 (nearly always); possible ZBI–12 scores range from 0 to 48 (higher scores represent a greater negative impact of caring). To prevent potential confusion for participants who were not necessarily providing care to an individual because of an illness, the original wording of one ZBI–12 item was changed from “Do you feel that you have lost control over your life since your relative’s illness?” to “Do you feel that you have lost control over your life because of your involvement with your relative?” (Italics in both items are ours.)
In addition to the total score generated by summing all items, the ZBI–12 includes two factors: personal strain (nine items) and role strain (three items). Personal strain refers to how personally stressful an individual finds the caregiving experience to be, and role strain refers to stress arising from role conflict or overload (Whitlatch et al., 1991). Given former drivers’ continued reliance on personal vehicles for transportation, and the prominence of transportation responsibilities for care partners, we expected a relationship between the two tools but not to the extent to which the TSS would be redundant. All aspects of the study were approved by the Lakehead University Research Ethics Board before participant recruitment.
Statistical Analyses
Descriptive statistics for the TSS and other variables are presented using means, standard deviations, frequencies, and percentages. Internal consistency was assessed through Cronbach’s α and item–total correlations. Construct validity was assessed by correlating the TSS with the SDS and the ZBI–12 (Pearson correlations), as well as with demographic items related to transportation intensity (Spearman correlations). We set statistical significance at p < .05; all analyses were conducted using IBM SPSS Statistics (Version 25).
Results
Participant Characteristics
Eleven care partners participated in Phase 1. The mean age was 57.64 yr (SD = 20.29), with a range of 21 to 90 yr. Eight care partners were female, eight were married or common-law partners, six were retired or not employed, eight were Caucasian, and nine were living in an urban setting. Seven care partners were adult children caring for their parents, three were spousal care partners, and one was a friend. None of the 11 care partners in Phase 1 had children younger than age 18 yr living in their household.
Sixty-six care partners participated in Phase 2. The mean age was 61.74 yr (SD = 10.64), with a range of 20 to 83 yr. Most care partners were female (54; 81.8%), married or common-law partners (38/65; 58.5%), retired or not employed (40/64; 62.5%), Caucasian (63; 95.5%), and living in an urban setting (51; 77.3%). Twelve participants (18.2%) were spousal care partners, 39 (59.1%) were adult children who were care partners, and 15 (22.7%) were care partners of another nature (including six friends, three siblings, three grandchildren, one ex-wife, one with power of attorney, and one former sister-in-law). See Table 1 for a further breakdown of Phase 2 care-partner demographic characteristics.
Phase 2 Care-Partner Demographic Characteristics (N = 66)
Information is missing for one other care partner.
Information is missing for one adult child care partner.
Information is missing for one spouse and one other care partner.
Other employment categories included casual, disabled, injured worker, and maternity leave.
Most care partners provided transportation <3 days per wk (58.2%), followed by 3 to 4 days per wk (23.1%), and >4 days per wk (18.5%). On the days when care partners provided transportation, 3 (4.5%) reported spending <1 hr, 26 (39.4%) reported 1 to <2 hr per day, 32 (48.5%) reported 2 to <4 hr per day, and 5 (7.6%) reported ≥4 hr per day. Twenty care partners (30.3%) provided transportation for more than one person. Fifty-three care partners (80.3%) indicated that they had other individuals available to help with providing transportation. However, of care partners who had others available to help, nearly 80% indicated that these other individuals helped “sometimes.” When asked about the primary purposes of transportation provided to former drivers (on a check-all-that-apply basis), care partners reported driving for health appointments (65; 98.5%), grocery shopping (46; 69.7%), leisure and recreational activities (43; 65.2%), and social activities (37; 56.1%).
Most former drivers were female (42; 63.6%), retired or not employed (65; 98.5%), Caucasian (61; 92.4%), and living in an urban setting (55; 83.3%). Twenty-one (31.8%) lived with their care partner. Of those who did not live with their care partner, 17 (25.8%) lived <5 km away, 21 (31.8%) lived 6 to 15 km away, 5 (7.6%) lived 16 to 25 km away, and 2 (3.0%) lived >25 km away.
Fifty-seven former drivers (86.4%) ceased driving >1 yr before the care partner completed the TSS, and 38 (57.6%) ceased driving suddenly (i.e., in response to a sudden health concern or license revocation). Forty (60.6%) ceased driving voluntarily. The most common reasons for ceasing driving (more than one answer was allowed) were a health issue (42; 63.6%), safety concern (16; 24.2%), uncomfortable driving (13; 19.7%), and driver’s license being revoked (13; 19.7%).
Item Reduction
After a careful review of the comments from the participants in Phase 1 on the initial set of 98 items, 30 items were removed from the pool of questions because of perceived redundancy, poor wording, or limited relevance. The remaining 68 items were reviewed to clarify wording and remove ambiguities and constituted the working set of items for Phase 2.
To reduce the number of items assessed in Phase 2, we deleted those with poor psychometric performance from the tool according to several criteria. One item was deleted because there were too many missing answers. Twelve items were deleted because endorsement frequencies on either end of the response continuum (i.e., strongly disagree and somewhat disagree, strongly agree and somewhat agree) were ≥80%; such items offer little in measuring participant response variability (Streiner et al., 2015). Two items were deleted because they had a higher correlation with the SDS than with the total TSS, which suggests that these items are measuring social desirability to a greater extent than our target construct. Five more items were deleted because item–total correlations were <.3; Streiner et al., (2015) contended that items should correlate with the total scale above this value. As a final step, items were deleted if item–domain correlations were <.3 or >.7. Kline (1986) argued that correlations should not exceed .7, because a higher value can indicate that the scale is too specific or repetitive. This led to the deletion of 26 additional items, leaving 22 items in the final TSS, with possible scores ranging from 22 to 110 (for items and domains included in the final TSS after item reduction, see the Supplemental Appendix, available online with this article at https://research.aota.org/ajot). We based all remaining analyses on these 22 items.
Transportation Support Scale Descriptive Statistics
Using the Flesch–Kincaid Grade Level, the final TSS was assessed at a 10th-grade reading level. Scores for all individual items ranged from 1 to 5. The mean total care-partner score on the TSS was 59.87 (SD = 15.14); total scores ranged from 22 to 93. Spousal care partners had an average score of 53.26 (SD = 17.92), whereas it was 61.5 (SD = 13.73) for adult children who were care partners and 60.81 (SD = 15.93) for other care partners. Female and male care partners had a mean TSS score of 53.41 (SD = 15.26) and 61.92 (SD = 15.96), respectively. Among care partners’ scores, 34.8% fell above the middle score of 66. Descriptive statistics for the domains are available in Table 2.
Transportation Support Scale Domain Means
Transportation Support Scale Internal Consistency and Validity
Cronbach’s α for the overall TSS was .88. Cronbach’s α values for the Social, Emotional, and Positive domains were .76, .72, and .82, respectively (the Physical and Time domains were not assessed because they included only two items each). Item–total correlations ranged from .32 to .70 (see Table 3 for TSS item statistics). Interdomain correlations ranged from .15 to .71, with a mean interdomain correlation of .42 (see Table 4 for the interdomain correlation matrix). The TSS was correlated, albeit weakly, with the SDS (r = −.29, p = .012; r 2 = .08).
Transportation Support Scale Item Statistics
Note. For the actual wording of the questions, see the Supplemental Appendix, available online with this article at https://research.aota.org/ajot.
Transportation Support Scale Interdomain Pearson Correlation Matrix
Note. p values are in parentheses.
The mean care-partner score on the ZBI–12 was 16.82 (SD = 10.27). Scores ranged from 0 to 41. Spousal care partners had an average score of 14.00 (SD = 10.18), whereas it was 18.49 (SD = 9.84) for adult children who were care partners and 14.76 (SD = 11.22) for other care partners. Female and male care partners had an average score of 17.23 (SD = 10.64) and 15.00 (SD = 8.55), respectively. The TSS total score was positively correlated with the ZBI–12 total score (r = .75, p < .001; r 2 = .56) as well as with both the ZBI–12 personal factor (r = .71, p < .001; r 2 = .50) and the ZBI–12 role factor (r = .52, p < .001; r 2 = .27).
We identified three questions denoting the intensity of the transportation responsibilities. The TSS was negatively correlated with the care partner having other individuals available to help with driving (r = −.43, p < .001), indicating that higher scores were associated with not having others available to help with transportation duties. Correlations with the other two variables (number of days per wk and time spent providing transportation) were not statistically significant: r = .15, p = .223; and r = .03, p = .784, respectively.
Discussion
We developed a tool to measure both the negative and positive impacts on care partners that may arise from driving cessation. The tool has potential value for clinicians and researchers when they evaluate interventions needed to support former drivers and care partners through the driving cessation process. The TSS displayed good internal consistency for both the overall TSS and the individual domains. Item–total correlations ranged from .32 to .70, values within the recommended range (Kline, 1986; Streiner et al., 2015). This suggests that items on the TSS are measuring the same construct but are not too specific or repetitive. Furthermore, the TSS displayed a weak correlation with the SDS, suggesting that social desirability was not an issue with care partners’ responses on the TSS.
The TSS also demonstrated initial evidence for construct validity. It was positively correlated with the ZBI–12, indicating that the TSS is measuring some form of impact, as intended. The TSS also correlated with both ZBI–12 factors, and the items retained appear to capture aspects of both role and personal strain. The two tools shared 56% of the variance, indicating that they measure similar, but not identical, constructs. This was expected, because the ZBI–12 and other similar tools are generic and do not address transportation specifically. Given the importance of transportation activities in the role of care partners, these findings reinforce our impression that this tool may be valuable to clinicians who are particularly concerned with driving cessation and its potential impact on care partners.
Although we do not currently have benchmark or cutoff scores for the TSS, 34.8% of care partners’ scores fell above the middle possible score. This suggests that some care partners were experiencing a high negative impact from providing transportation. Our results also indicated that care partners in our sample had the highest item mean on the Time domain, illustrating that time commitments may be the most generalized form of role strain. However, an alternative explanation is that less than one-third of care partners lived with the ex-driver, adding to travel time. Nevertheless, such findings should be interpreted with caution because the domains and items they include are preliminary and preclude a formal statistical analysis.
In addition to negative items, the TSS contains positive outcome items, something that the typical generic tool does not and that may help capture the experience of care partners better. It is interesting that the correlations between the Positive domain and the Negative domains were small, suggesting that the two domains may represent different constructs rather than opposite ends of the same continuum. This is reminiscent of a similar argument made regarding the impact of caring and general well-being by others (Stuckey et al., 1996) and may be informative for intervention research in which approaches to enhance the positive aspects may differ from those aimed at reducing the negative aspects. Given that we know little about the needs of care partners to best support older drivers as they transition to nondriving (Dickerson et al., 2019; Liddle et al., 2017), and given the paucity of intervention research related to driving cessation and mobility (Rapoport et al., 2017; Stav, 2014), the development of the TSS appears timely.
Although driving cessation can affect the lives of former drivers in many ways, it can also have impacts, both negative and positive, on the lives of those who take over transportation responsibilities for former drivers. Our results show that the TSS can capture relevant domains of care-partner response to driving cessation, including Physical, Social, Time, Emotional, and Positive impact domains. Although being a care partner has been linked to several negative outcomes (e.g., poorer physical and psychological health; Pinquart & Sörensen, 2003; Turcotte, 2013; Vitaliano et al., 2003), one question to elucidate is the extent to which transportation responsibilities may mediate this link.
Our study has limitations. We used a self-selected convenience sample that is not representative of all older adults who stop driving or of all care partners who provide transportation. However, most respondents were women, and this is consistent with the predominance of women as care partners. Nonetheless, evaluating the TSS with more representative samples, and confirming its validity in different settings (e.g., urban versus rural), is desirable. It is also worth considering whether ceasing driving voluntarily versus involuntarily, and gradually versus suddenly, may affect the experience of care partners, because planning for driving cessation is associated with the presence of more transportation alternatives (Vivoda et al., 2021). Furthermore, different types of care partners (e.g., spouses vs. children, women vs. men, urban dwellers vs. rural dwellers) may have different experiences. We had a heterogenous group of participants, but our sample size was too small to conduct meaningful subgroup analyses. Future work should focus on investigating the tool’s psychometric properties with a larger sample of care partners. Specifically, the items presented here should be examined using a factor analysis. Additional validation of the tool (including divergent validity) and assessment of its reproducibility and ability to detect change over time would be valuable. Ultimately, some norms could be derived from larger, representative samples of care partners.
Implications for Occupational Therapy Practice
Transportation is the predominant form of support provided by care partners. However, we know little about the impact that driving cessation and providing transportation may have on care partners. This study has the following implications for occupational therapy practice: The TSS shows potential to help clinicians and researchers understand the specific impact of driving cessation and transportation responsibilities on care partners. The TSS could help clinicians initiate a conversation about the needs of care partners regarding the driving cessation process and their transportation responsibilities.
Conclusion
We believe that the TSS is a good starting point to measure care partners’ reactions, both negative and positive, to driving cessation and the provision of transportation, contributing to both the informal care and driving fields. After further validation, the TSS could potentially help clinicians and researchers better capture the experience of care partners who are providing transportation and evaluate services and interventions meant to support them.
Footnotes
Acknowledgments
We thank the study participants for their dedication; without their commitment this work would not have been possible. We also thank the reviewers for their constructive comments. This study was supported through funding from the Canadian Institutes of Health Research (Grant No. 152958).