Behind the Wheel: Process Evaluation of a Safe-Transport Program for Older Drivers Delivered in a Randomized Controlled Trial

Abstract

This process evaluation explores relationships between program outcomes and intervention implementation in a trial evaluating “Behind the Wheel,” an education-based safe-transport program for older drivers. Participants (intervention group) were 190 Sydney drivers aged ⩾75 years (M = 80 ± 4years). Process measures included fidelity, dose delivered, and received. Outcomes were self-reported driving regulation and objectively measured driving exposure. Relationships were explored using regression models. Older drivers who took ownership of driving retirement and self-regulation by developing plans were more likely to reduce their weekly driving, (β = 38 km, 95% confidence interval (CI) = [7.5,68.7]), and night driving (β = 7 km, 95% CI = [3.5, 10.4]). Drivers of older age (odds ratio [OR] = 1.1/year older, 95% CI = [1.05, 1.3]) had greater odds of developing driving retirement plans. Female drivers (OR = 2.7,95% CI = [1.1, 6.9]), drivers with poorer function (OR = 1.2/5-point decrease on DriveSafe, 95% CI = [1.04, 1.4]), and worse health (OR = 1.2/additional medication, 95% CI = [1.02, 1.5]) had greater odds of developing safe mobility plans. This program had greatest impact with older, lower functioning drivers. A stronger message was delivered and received, as intended, to older drivers with lower function and poorer health. Our logic model can help channel resources to drivers who benefit most.

Keywords

driving aging naturalistic driving education community mobility safety logic models process evaluation

Introduction

Evaluation of intervention fidelity and implementation in clinical trials has gained momentum in recent decades (Bellg et al., 2004; Oakley, Strange, Bonell, Allen, & Stephenson, 2006; Saunders, Evans, & Joshi, 2005). Process evaluations aim to safeguard against discarding ineffective interventions that were poorly executed or adopting effective interventions impractical to deliver in the real world (Broekhuizen, Althuizen, van Poppel, Donker, & van Mechelen, 2012). A one-on-one education-based safe-transport program for older drivers was recently evaluated in a randomized controlled trial (RCT; Coxon et al., 2017)). While the education program was found to increase readiness to engage in the process of self-regulation and retirement from driving, this did not translate to reduced driving exposure (Coxon et al., 2017).

It is possible this program, like many complex health interventions, had several “active ingredients” responsible for the null finding or failure of subgroup analyses to reach significance (Hawe, Shiell, & Riley, 2004; Oakley et al., 2006). For example, a process evaluation of an RCT evaluating the impact of a peer-led sex education program found peer-led education was more effective when it was participatory and skills based, but without this, teacher-led education was more effective (Oakley et al., 2006). In another RCT, a peer-led education program to reduce risk of HIV infection in homosexual men, found no impact, however; the process evaluation revealed that the education program itself was unacceptable to the peer educators (Elford, Sherr, Bolding, Serle, & Maguire, 2002). A nested process evaluation exploring relationships between intervention outcomes and quality of intervention implementation was conducted to tease out these “active ingredients” (Hawe et al., 2004; Oakley et al., 2006; Saunders et al., 2005).

While substantial resources have been invested to evaluate programs to support older driver safety, little evidence exists to direct future spending in this area (Korner-Bitensky, Kua, von Zweck, & Van Benthem, 2009). Considering the resources employed to conduct large trials, it is imperative to drill down further into seemingly ineffective interventions to extract and retain any elements important to program outcomes. This process evaluation represents an important step in understanding outcome pathways, uptake of program messages, and characteristics of older drivers most likely to act and engage in driving self-regulation post program.

Method

The process evaluation involved data from participants in the intervention group (190 drivers aged ⩾75 years) of an RCT evaluating a safe-transport program for older drivers; “Behind the Wheel.” Trial results have been published (Coxon et al., 2017). This process evaluation aimed to (a) evaluate relationships between what was taught (treatment fidelity, timing of intervention, dose delivered), what was learnt (dose received), and what actually changed (treatment enactment/outcomes; Bellg et al., 2004); (b) explore participant characteristics associated with program uptake; and (c) explain the inputs, outputs, and outcomes of the safe-transport program. Ethics approval was granted by the Human Research Ethics Committee of the University of Sydney (Protocol Number:14235) and written consent was obtained from all participants before taking part in the study.

Program Description

The “Behind the Wheel” (Coxon & Keay, 2015) program aimed to help older people drive safely for longer while preparing for retirement from driving. Self-regulation was the central strategy, with content based on the Knowledge Enhances Your Safety (KEYS) curriculum (Stalvey & Owsley, 2003). Two education sessions, held approximately 1 month apart, were delivered face-to-face in an informal one-on-one format with each participant, in their home.

Session 1 focused on helping participants drive safely for longer through self-regulation. Many strategies may be used by older drivers at life-goal, strategic, or tactical levels to self-regulate their driving (Molnar et al., 2013). Educational strategies of raising awareness, building skills, and planning action were employed to help older drivers self-regulate by matching their driving exposure to their driving skills now and in the future (Coxon & Keay, 2015). Results from functional assessments administered at baseline were discussed with participants to raise awareness of driving skills and abilities, and match driving exposure to ability. Eight high-risk driving situations were included in the curriculum: driving at night, in the rain, in heavy traffic, during peak hour, alone, on high-speed highways/freeways, in school zones, and turning right across oncoming traffic.

The educator customized the education, while preserving program fidelity, in line with growing consensus that complex interventions have greater potential for success if an agreed level of flexibility is permitted to tailor content to individual needs (Craig et al., 2008; Dickerson et al., 2017; Hawe et al., 2004; Oakley et al., 2006). Messages were tailored to meet current self-regulation practices of individual participants using the precaution adoption process model (PAPM; Neil D. Weinstein, Lyon, Sandman, & Cuite, 1998; N. D. Weinstein & Sandman, 1992; N. D. Weinstein, Sandman, & Blalock, 2008). Stage-appropriate language, content, and key messages were employed in an effort to move participants to higher stages of adoption and sustained use of self-regulatory driving practices (Coxon & Keay, 2015). For example, older drivers unaware of safety issues (Stage 1) were provided with information to raise their awareness of driving hazards and precautions, while older drivers who have decided to act (Stage 5) were given practical instructions for implementing relevant driving precautions. Older drivers who have decided not to act (Stage 4) were considered “hard-to-reach.” These drivers received specific messages targeted toward tipping the risk–benefit ratio toward adoption of relevant driving precautions (Coxon & Keay, 2015).

Session 2 focused on planning for retirement from driving. As many older drivers are unprepared for their retirement from driving (Oxley & Charlton, 2009; Rudman, Friedland, Chipman, & Sciortino, 2006), all participants were challenged to plan ways to stay mobile and active within their community when driving is no longer a safe option. Individual strategies, based on personal preferences, usual trips and community activities, combined with community mobility packages, prepared by local government area of residence by the research team, were used to facilitate this planning.

Process measures

Participants were interviewed over the telephone post program, by a researcher not involved in program delivery. Data were concealed from the educator until trial results were known. Educator notes recorded development of safe mobility and/or retirement from driving plans for each participant. Process measures (Table 1) were guided by Linnan and Steckler’s (2002) framework, later extended by Saunders et al. (2005). Using the six steps outlined by Saunders et al. (2005), a process evaluation plan was developed, and program fidelity, timing of education delivery, dose delivered, and dose received were included as key process measures. Program context was considered and further explored in a logic model, as per recommendation by Saunders et al. (2005).

Table 1.

Variables.

Variable	Type of data	Units	Data collection
Outcomes
Proximal outcome
Self-regulation stage/profile	Continuous	Stages 1-17	At 12-month assessment
Distal outcomes
Average total distance driven/week	Continuous	Kilometers	Up to 12 months in-vehicle monitoring
Average furthest distance traveled from home/week	Continuous	Kilometers	Up to 12 months in-vehicle monitoring
Average distance driven outside daylight hours/week	Continuous	Kilometers	Up to 12 months in-vehicle monitoring
Process measures
Fidelity	Binary	No/yes	Educator notes
Timing of education delivery
In-vehicle monitoring posteducation	Continuous	Number of weeks posteducation/total weeks in study	Weeks summed from first Monday posteducation sessions
Dose delivered	Binary	Incomplete/complete	Educator notes
Dose received
Developed a safe mobility plan	Binary	No/yes	Educator notes
Developed a retirement from driving plan	Binary	No/yes	Educator notes
Content understood	Binary	No/yes	Participant interview
Education helped participant drive safely for longer	Binary	No/yes	Participant interview
Education helped participant plan for retirement from driving	Binary	No/yes	Participant interview
Other factors of interest
Age	Continuous	Years	Baseline assessment
Sex	Binary	Female/male	Baseline assessment
Education	Continuous	Years	Baseline assessment
DriveSafe score	Continuous	Points (0-128)	Baseline assessment
DriveAware score	Continuous	Points (3-26)	Baseline assessment
Comorbidities	Count	Number	Baseline assessment
Prescription medications	Count	Number	Baseline assessment
Contrast sensitivity	Continuous	Per letter	Baseline assessment
Rural	Binary	Urban/rural	Baseline assessment
Responsible for driving others	Binary	No/yes	Baseline assessment
Availability of another driver	Binary	No/yes	Baseline assessment
Prefers to drive	Binary	No/yes	Baseline assessment
Distance to essential services	Continuous	Kilometers	Baseline assessment

Program fidelity

Fidelity measured quality of intervention delivery. This was measured against the program protocol that was developed and documented in the program manual prior to delivery of the education program. Any deviations in program delivery from the protocol, for example, omission of content or messages not consistent with the self-regulation stage, were considered to reduce program fidelity and were recorded in file notes by the educator. Participants who received partial or no education were also considered to have deviated from protocol.

Timing of education sessions

The timing of sessions may deviate from protocol in large trials for several logistic and participant-related reasons. The proportion of time in the study post program was calculated for each participant. Participants who received no education had all weeks counted as post intervention.

Dose delivered

Dose delivered measured extent to which the program was delivered as intended (Linnan & Steckler, 2002; Saunders et al., 2005). Dose delivered was complete when participants received two sessions; incomplete when participants received no or partial sessions.

Dose received

Dose received measured each participant’s level of engagement with and understanding of program content (Linnan & Steckler, 2002; Saunders et al., 2005). Participants were encouraged to develop both safe mobility and retirement from driving plans before program completion. Plans were not prescribed, instead negotiated in collaboration with participants. The development of a (a) safe mobility and (b) retirement from driving plan were used as measures of dose received.

Participant responses gauged extent to which program purpose and content was understood by participants (Saunders et al., 2005). Participants were asked two questions about program content “What did you think the main aim of the program was?” and “What was the main thing you learnt?” Content was deemed understood if participants discussed safe mobility, planning for retirement from driving, or provided a specific strategy for driving self-regulation or accessing alternative transportation, for either question. Two authors were involved in data coding (KC and LK). One author (KC) coded all responses initially. A senior author (LK) reviewed these codes. Any discrepancies were discussed and codes negotiated until consensus achieved.

Participants were asked if the program (a) helped them drive safely for longer and (b) prepared them for retirement from driving (yes or no/not sure).

Participants who did not receive any education were coded as no dose received on these measures.

Outcomes

As measurement of self-regulation may differ between self-reported and objective measures, two outcomes of driving self-regulation, using both self-reported and objective measures, were included. Outcomes were self-reported regulation of driving measured by self-regulation stage using the PAPM (proximal outcome) and driving exposure (distal outcomes) measured objectively by in-vehicle monitoring. Participants were asked to self-report their level of engagement with driving self-regulation using a staging algorithm (Coxon et al., 2017) based on the PAPM (Neil D. Weinstein et al., 1998). Participants were assigned one of 17 ordered stage and behavior profile combinations on the pathway to adoption of self-regulatory driving practices (Coxon et al., 2017; Coxon & Keay, 2015; Gielen et al., 2007). These ranged from no awareness of driving self-regulation (Stage 1, Profile 1) to consistent use of self-regulatory driving practices >6 months (Stage 7, Profile 4), and were assigned a numerical value from 1 to 17 in ascending order. Higher scores equate to greater self-reported adoption of self-regulatory driving practices at 12 months.

Participants had an in-vehicle monitoring device (C4D, Mobile Devices Ingenierie, Villejuif), with global positioning system and data logger hardwired into their vehicle for up to 12 months. Time-stamped, second-by-second GPS location was transmitted over the telecommunications network to a secure server. The average distance driven, furthest distance traveled from home, and distance driven outside daylight hours were calculated per week for each participant. Driving data were scrutinized for reliability before analysis using range checks (Coxon et al., 2017). Furthest radial distances of 4,000 km or more were considered geographically implausible and excluded from analysis. Trip distances of 385 km or greater were manually examined for reliability. Driving exposure data were considered missing if an in-vehicle device malfunctioned. If a participant was unwell, on holiday, or stopped driving during the study period, driving exposure was preserved as zero for the corresponding weeks.

Other Factors

Years of education were obtained through interview. Function was assessed using DriveSafe, a computerized assessment of visual attention to the driving environment (Kay, Bundy, & Clemson, 2009a). Higher scores on DriveSafe equate to better attention. DriveAware measures participants’ awareness of their driving ability (Kay, Bundy, & Clemson, 2009b). Higher scores indicate less awareness. Comorbidities were self-reported using the functional comorbidity index (FCI; Groll, To, Bombardier, & Wright, 2005). Prescription medications were self-reported and tallied. Addresses were coded as rural or urban based on local government classification. Distance to essential services was estimated in kilometers by road from the participant’s address to their nearest shopping complex/town center. The Driving Habits Questionnaire (Owsley, Stalvey, Wells, & Sloane, 1999) was administered to identify preference to drive for transportation and availability of another driver. Participants were asked if they were responsible for driving others. Binocular contrast sensitivity was assessed using The Mars Perspectix letter chart (Dougherty, Flom, & Bullimore, 2005). Higher log contrast sensitivity scores indicate better vision.

Statistical Analysis

Linear regression was used to examine relationships between process measures, self-regulation stage, and driving exposure. As the proportional odds assumption was not met for self-regulation stage as an ordinal outcome in a cumulative logit model, linear regression was used. Given that self-regulation stage was not normally distributed, we undertook a sensitivity test of the estimates from the final selected model using Bootstrapping (with 5,000 samples) to estimate a bootstrap sample median and 95% coverage interval. The median estimate and coverage interval from the Bootstrapping was very similar to the least squares estimates that confirm that the original regression model was robust to the distribution of self-regulation stage.

Driving exposure was modeled as a continuous variable with a linear model. Possible collinearity of variables was examined using a correlation matrix prior to modeling. Process measures with p < .25 were included in an initial adjusted model and removed using stepwise backward elimination until only measures with p < .05 remained. Plausible interactions were considered before final models were identified. Residuals were scrutinized and results compared using different distributions. To help understand education uptake and participant engagement, process measures such as development of plans were further explored using logistic regression modeling to reveal participant characteristics predictive of program engagement. ProcReg and ProcLogistic in SAS Enterprise Guide Version 5.1 (SAS Institute Inc., Cary, NC) were used for all analyses.

A logic model (McLaughlin & Jordan, 1999) explaining the “Behind the Wheel” program, and its context, was constructed. Process measures and other factors were categorized as outputs (activities and participation; Taylor-Powell & Henert, 2008) and inputs (resources; Taylor-Powell & Henert, 2008). Starting with outcomes and working backward, causal relationships between global, distal, and proximal outcomes were established. Linear regression modeling using stepwise backward elimination was used to confirm which external factors influenced driving exposure. One investigator drafted the initial model (KC) using experience and knowledge of program delivery. The model was reviewed, modified, and discussed at team meetings to reach consensus.

Results

High program fidelity was achieved (Table 2). As one educator (KC) delivered all sessions, a homogeneous program was delivered in a uniform manner per protocol. Errors in stage messages were not made. Safe mobility plans and retirement from driving plans were developed by 79% (n = 140) and 71% (n = 126) of participants, respectively, indicating dose received was high (Table 2). Of the 190 intervention group participants, 183 received the complete program as intended. Of the seven remaining participants, five received no education (n = 3 withdrew, n = 1 declined education, n = 1 unwell) and two received the first session only (n = 1 unwell, n = 1 declined second session). As program fidelity and dose delivered were consistent in the vast majority (96%, 183/190), their impact on program outcomes was not investigated further.

Table 2.

Demographic and Functional Characteristics of Intervention Group Participants.

Characteristics (n = 190)
Demographic information
Age (years), mean ± SD^a	80 ± 4
Sex (male), n (%)	128 (67)
English language spoken at home, n (%)	188 (99)
Years of education, mean ± SD^a	13.5 ± 3.9
Resident in a rural area, n (%)	20 (11)
Distance to essential services (km), mean ± SD^a	2.6 ± 1.4
Restricted driver’s license, n (%)	5 (3)
Type of accommodation, n (%)
Home	131 (69)
Unit	13 (7)
Independent living unit^b	45 (24)
Hostel or assisted living	1 (1)
People living with participant, n (%)
Living alone	64 (34)
Spouse only	104 (55)
Other^c	22 (12)
Functional and clinical characteristics
Comorbidities, mean ± SD	5.3 ± 2.7
Prescription medications (number), mean ± SD (%)	3.8 ± 2.7
Contrast sensitivity(units), mean ± SD^a,d	1.65 ± 0.10
DriveSafe (0-128), mean ± SD^a,d	84 ± 15
DriveAware (3-26), mean ± SD^a,d	13 ± 2
Process evaluation measures
Fidelity (deviations from delivery protocol) n (%)	7 (4)
Timing of education delivery
In-vehicle monitoring posteducation, mean ± SD^a,e	0.8 ± 0.1
Dose delivered (received two education sessions as intended), n (%)	183 (96)
Dose received
Developed a safe mobility plan,^f n (%)	140 (79)
Developed a retirement from driving plan,^f n (%)	126 (71)
Content understood,^g n (%)	137 (73)
Education helped participant drive safely for longer,^g n (%)	132 (71)
Education helped participant plan for retirement from driving,^g n (%)	153 (82)

SD = standard deviation.

Includes town house, villa, duplex, granny flat.

Includes living with a spouse and children, children, relatives, and friends.

One participant did not complete this assessment.

n = 180, 10 participants did not have in-vehicle monitoring.

n = 178, data for this variable unavailable for 12 participants.

n = 187, data for this variable unavailable for three participants.

Self-regulation stage was available for 183/190 participants at 12 months (n = 7 lost-to-follow-up). Average self-regulation stage was Stage 2/Profile 2 (numeric stage/profile combination 7 ± 6) at 12 months. Three process measures were associated with self-regulation stage in unadjusted models (Table 3). Development of a retirement from driving plan was omitted using backward elimination, leaving development of a safe mobility plan (t₁₆₅ = 3.1, p = .003) and understanding program content (t₁₆₅ = 2.1, p = .03) as independent factors explaining self-regulation stage at 12 months. After controlling for these covariates (development of a safe mobility plan and understanding content) in the adjusted model, participants who developed a safe mobility plan were 3.3 stages further toward adoption of driving self-regulation on average than participants without a safe mobility plan (95% confidence interval [CI] = [1.2, 5.5]) and participants who understood program content were 2.1 stages further toward adoption of driving self-regulation on average than participants who did not understand program content (95% CI = [0.2, 4.1]).

Table 3.

Influence of Dose Delivered and Dose Received on Behavior Change and Driving Exposure Outcomes.

Outcome variable	Process measures	Unadjusted model		Adjusted model
Outcome variable	Process measures	Beta coefficient	p value	Beta coefficient	p value	95% CI
Proximal outcome Self-regulation Stage/profile (Stages 1-17) M = 6.9 Median = 5 Interquartile range = 9 km (2-11)	In-vehicle monitoring post education	−3.6	.37	—	—	—
	Developed a safe mobility plan	3.8	<.001	3.3	.003	[1.2, 5.5] ^a
	Developed a retirement from driving plan	2.2	.03	—	—	—
	Content understood	1.9	.049	2.1	.03	[0.2, 4.1] ^b
	Education helped participant drive safely for longer	0.7	.46	—	—	—
	Education helped participant plan for retirement from driving	0.6	.59	—	—	—
Distal outcomes Average total distance driven/week (km) M = 134 km Median = 112 km Interquartile range = 106 km (70-177 km)	In-vehicle monitoring post education	−14.7	.82	—	—	—
	Developed a safe mobility plan	−27.6	.13	—	—	—
	Developed a retirement from driving plan	−38.1	.02	–38.1	.02	[−7.5, −68.7] ^c
	Content understood	9.3	.55	—	—	—
	Education helped participant drive safely for longer	−6.4	.68	—	—	—
	Education helped participant plan for retirement from driving	−6.0	.74	—	—	—
Average furthest distance traveled from home/week (km) M = 32 km Median = 16 km Interquartile range = 25 km (9 to 34 km)	In-vehicle monitoring posteducation	−16.12	.56	—	—	—
	Developed a safe mobility plan	−8.6	.29	—	—	—
	Developed a retirement from driving plan	−10.9	.12	—	—	—
	Content understood	3.9	.58	—	—	—
	Education helped participant drive safely for longer	1.2	.87	—	—	—
	Education helped participant plan for retirement from driving	1.0	.90	—	—	—

Average distance driven outside daylight hours/week (km)M = 9 kmMedian = 5 kmInterquartile range = 11 km(2 to 13 km)	In-vehicle monitoring post education	5.1	.50	—	—	—
	Developed a safe mobility plan	–5.5	.009	—	—	—
	Developed a retirement from driving plan	–7.0	<.001	–7.0	<.001	[–3.5, –10.4] ^c
	Content understood	−0.8	.66	—	—	—
	Education helped participant drive safely for longer	−1.9	.31	—	—	—
	Education helped participant plan for retirement from driving	−1.0	.63	—	—	—

Note.1 mi = 1.61 km; CI = confidence Interval.

Controlled for understanding content.

Controlled for development of a safe mobility plan.

Only variable significant in the model after stepwise backward elimination.

Bold indicates p<0.05.

Of the 190 participants, 180 participants had vehicles successfully instrumented with in-vehicle monitoring devices. Driving data from 6/180 (3%) devices were excluded after data were scrutinized for reliability. In total, driving data were available for 174/190 (92%) participants. Our driving exposure data best fit a gamma distribution (Coxon, Chevalier, Lo, Ivers, Brown, & Keay, 2015); however, as conclusions were unchanged when modeling with linear regression, we report linear regression for ease of interpretation.

On average, participants drove 134 km/week (median = 112 km; Table 3). Development of a retirement from driving plan was the only factor significantly associated with reduction in average distance driven/week in unadjusted and adjusted models (t₁₆₀ = −2.5, p = .02). Participants who developed a retirement from driving plan reduced their average distance driven/week by 38.1 km on average compared with participants who did not devise a plan (95% CI = −7.5 to −68.7 km).

Average furthest distance traveled from home/week was 32 km (median = 16 km). No process evaluation factors were found to be associated with this (Table 3).

Average distance driven outside daylight hours/week was 9 km (median = 5 km). Similar to total distance driven, development of a retirement from driving plan was predictive of reduced night driving (t₁₆₀ = −4.0, p < .001). Participants who developed a retirement from driving plan reduced their average kilometers driven outside of daylight hours by 7 km on average compared with participants who did not devise a plan (95% CI = −3.5 to −10.4 km).

Several process measures explained differences in the mechanisms of treatment effect (Table 4). Exploration of participant characteristics found females had 3.4 times greater odds, on average, of understanding program content compared with males (χ²₁ = 8.4, p = .004, 95% CI = [1.5, 7.8]).

Table 4.

Driver Characteristics Influencing Likelihood of Safe Mobility Plan, Retirement From Driving Plan, and Understanding Content.

		Unadjusted model		Adjusted model
Outcome variable	Participant characteristics	Odds ratio	p value	Odds ratio	p value	95% CI
Development of a safe mobility plan	Older age	1.1	.10	—	—	—
	Female	2.5	.04	2.7	.03	[1.1, 6.9] ^a
	Less education (years)	1.1	.045	—	—	—
	Lower DriveSafe score^b	1.2	.008	1.2	.01	[1.04, 1.4] ^c
	Higher DriveAware score	1.1	.24	—	—	—
	More comorbidities	1.1	.14	—	—	—
	More prescriptions medications	1.2	.03	1.2	.03	[1.02, 1.5] ^d
	Better contrast sensitivity^e	0.9	.31	—	—	—
Development of a retirement from driving plan	Older age	1.1	.004	1.1	.004	[1.05, 1.3] ^f
	Female	3.0	.007	2.5	.04	[1.1, 5.8] ^g
	Less education (years)	1.2	<.001	1.1	.005	[1.04, 1.2] ^h
	Lower DriveSafe score^b	1.0	.52	—	—	—
	Higher DriveAware score	1.1	.61	—	—	—
	More comorbidities	1.0	.92	—	—	—
	More prescriptions medications	1.0	.78	—	—	—
	Better contrast sensitivity^e	1.0	.81	—	—	—
Content understood	Older age	0.9	.07	—	—	—
	Female	3.4	.004	3.4	.004	[1.5, 7.8] ⁱ
	Less education (years)	1.0	.33	—	—	—
	Lower DriveSafe score^b	1.0	.94	—	—	—
	Higher DriveAware score	1.0	.91	—	—	—
	More comorbidites	1.0	.57	—	—	—
	More prescriptions medications	1.0	.71	—	—	—

Note. CI = confidence interval.

Controlled for having a lower DriveSafe score and more prescription medications.

Per 5-point decrease.

Controlled for being female, and having more prescription medications.

Controlled for being female and having a lower DriveSafe score.

Per letter increase.

Controlled for being female and having less years of education.

Controlled for older age and having less years of education.

Controlled for older age and being female.

Only variable significant in the model after stepwise backward elimination.

Bold indicates p<0.05.

Female sex (χ²₁ = 4.5, p = .03), lower score on DriveSafe (χ²₁ = 6.0, p = .01), and more prescription medications (χ²₁ = 4.8, p = .03) were independently predictive of developing a safe mobility plan in the adjusted model (Table 4). After adjusting for these covariates, females had 2.7 times greater odds of developing a safe mobility plan, on average, compared with males (95% CI = [1.1, 6.9]); for each additional prescription medication, participants had 1.2 times greater odds, on average, of developing a safe mobility plan (95% CI = [1.02, 1.5]), and for every 5-point reduction in DriveSafe score (worse function), participants had 1.2 times greater odds, on average, of developing a safe mobility plan (95% CI = [1.04, 1.4]).

Female sex, older age, and less years of education were significantly associated with developing a retirement from driving plan in the adjusted model. After adjusting for age and years of education, older female participants had 2.5 times greater odds, on average, of developing a retirement from driving plan than older males (χ²₁ = 4.3, p = .04, 95% CI = [1.1, 5.8]). For each year increase in age, participants had 1.1 times greater odds, on average, of devising a retirement from driving plan (χ²₁ = 8.5, p = .004, 95% CI = [1.05, 1.3]), after adjusting for sex and years of education in the model. For every year less education, participants had 1.1 times greater odds, on average, of developing a retirement from driving plan (χ²₁ = 7.8, p = .005, 95% CI = [1.04, 1.2]), after adjusting for age and sex.

The logic model of the “Behind the Wheel” program includes proximal, distal, and global outcomes (Figure 1). Global program outcomes included the hypothesized long-term societal benefits. Distal outcomes were action orientated and included change in driving behavior and alternative transport use, while proximal outcomes involved increased knowledge, changes in attitude, and raised level of engagement. Pathways for achieving program outcomes emerged from data. For example, increased engagement in driving self-regulation was influenced by participants’ understanding of program content and development of a safe mobility plan. Participants who engaged to the level of devising a retirement from driving plan made greater reductions in both total kilometers driven/week and kilometers driven outside of daylight hours/week. Rural residence was important to average kilometers driven/week and average kilometers driven outside of daylight hours/week (Table 5) and found to be a barrier to reduction of driving (Figure 1). Participants living in rural areas were found to drive 51 km/week more on average (t₁₇₂ = 2.36, p = .02, 95% CI = [8.3, 93.8]) and 7 km more at night/week on average (t₁₇₂ = 2.75, p = .007, 95% CI = [2.0, 12.0] km) than participants living in urban areas. Responsibility to drive others, availability of another driver, distance to essential services, and preference to drive were not associated with driving exposure outcomes in unadjusted or adjusted models.

Figure 1.

Logic model for the “Behind the Wheel” program.

Table 5.

Influence of External Factors on Driving Exposure Outcomes.

		Unadjusted model		Adjusted model
Distal outcome variables	External factors	Beta coefficient	p value	Beta coefficient	p value	95% CI
Average total distance driven/week (km)^a	Rural residence	51.0	.02	51.0	.02	[8.3, 93.8] ^b
	Responsible for driving others	22.2	.11	—	—	—
	Has another driver available	−40.4	.07	—	—	—
	Prefers to drive	16.5	.50	—	—	—
	Distance to essential services	−0.9	.86	—	—	—
Average furthest distance traveled from home/week (km)^a	Rural residence	9.8	.57	—	—	—
	Responsible for driving others	5.9	.93	—	—	—
	Has another driver available	−16.9	.09	—	—	—
	Prefers to drive	4.9	.66	—	—	—
	Distance to essential services	2.2	.67	—	—	—
Average distance driven outside daylight hours/week (km)^a	Rural residence	7.0	.007	7.0	.007	[2.0, 12.0] ^b
	Responsible for driving others	−2.1	.21
	Has another driver available	−0.6	.81
	Prefers to drive	0.8	.79
	Distance to essential services	−0.4	.49

Note. CI = confidence interval.

Driving exposure data unavailable for 16 participants.

Only variable significant in the model after stepwise backward elimination.

Bold indicates p<0.05.

Discussion

An RCT evaluating the “Behind the Wheel” safe-transport program for older drivers found increased engagement in self-regulation and planning for driving retirement, but no reduction in driving exposure between groups (Coxon et al., 2017). Our process evaluation revealed high fidelity, confirming homogeneous program delivery across participants. Variations in timing of program delivery did not affect program outcomes. However, a variation in program uptake, in particular, developing a retirement from driving plan, was isolated as an important process to reduce driving. Both understanding program content and achieving a safe mobility plan were important to becoming more engaged in the process of self-regulation and planning for driving retirement. This suggests reduced driving exposure involves an active process of identifying, planning, and having access to alternative transport, whereas greater engagement in self-regulation involves careful consideration of the match between driving exposure, level of ability, and on-road safety. The results suggest that the safe mobility plan is an intermediate step, but the retirement from driving plan is more definitive and linked to actual reductions in driving as evidenced by reduced mileage and night driving. In short, taking ownership of the process appeared key to enhancing self-regulation in older drivers.

Our findings reiterate those of Owsley and colleagues (Owsley, McGwin, Phillips, McNeal, & Stalvey, 2004; Owsley, Stalvey, & Phillips, 2003). Both studies conclude education helped older drivers become more engaged in adopting safer driving practices, but behavior changes and on-road safety benefits were limited (Coxon et al., 2017; Owsley et al., 2004; Owsley et al., 2003). While changing behavior is difficult, this evaluation adds a piece to the puzzle by highlighting the behavior change pathways in our program (Figure 1). Participants, who took ownership of the process to the point of devising plans, had higher levels of engagement and made more changes to their driving exposure. Understanding these “active ingredients” is important, particularly given the null result (Coxon et al., 2017). The logic model (Figure 1), constructed from these data, illustrates the behavior change pathways and is offered as a blueprint for development of future interventions and spending in the area of older driver safety.

Process evaluation results show greater uptake of driving self-regulation practices in participants where messages were immediately relevant to their driving due to increased age, worse function, and/or poor health. The overall null trial finding may be explained by the inclusion of many high functioning drivers who did not need to change their driving exposure. The study was likely underpowered to detect differences in driving exposure in subgroups.

Dose received was high with 79% of participants developing a safe mobility plan and 71% developing a retirement from driving plan. Lower functioning older drivers who did not achieve behavior change plans may be considered “hard-to-reach.” These drivers may benefit from different strategies, such as motivational interviewing, face-to-face peer testimonial/instruction, multifaceted educational media (driving simulation or on-road assessment), or additional sessions.

Women, compared with men, had 2.7 and 2.5 times greater odds of developing safe mobility and retirement from driving plans, respectively. Previous research (Brabyn, Schneck, Lott, & Haegerstrom-Portnoy, 2005; D’Ambrosio, Donorfio, Coughlin, Mohyde, & Meyer, 2008; Kostyniuk & Molnar, 2008; Molnar & Eby, 2008; Ross et al., 2009; Vance et al., 2006), including one systematic review (Wong, Smith, Sullivan, & Allan, 2016), found older women were more likely than men to report self-regulatory driving practices. While this lends support to our finding of greater engagement and self-regulation planning among women in our study, no gender difference in driving exposure was found (Keay et al., 2018). These results provide some evidence that women and men engage with this content differently, perhaps due to gender differences in perception of driving importance. While women have been found to view driving as a means to achieving everyday activities, driving for men is interwoven with identity and sense of freedom (Musselwhite & Shergold, 2013). This raises the question of whether different approaches for males and females would yield greater self-regulatory driving practices and on-road safety. Our finding that women had 3.4 times greater odds compared with men of understanding the content of the “Behind the wheel” program adds further weight to this argument. Any influence the gender of the educator may have had on program outcomes remains unknown, but it is possible women related more to the female educator than the men in the study, leading to greater understanding and levels of engagement.

Rural residence emerged as a barrier to reducing driving exposure, while other factors, such as preference to drive and responsibility to drive others were not. It stands to reason that rural residents drive move kilometers/week than urban residents, and our data confirm this. People living in rural areas have been found to drive further distances from home (Coxon et al., 2015), presumably to access their town center. This, coupled with scarce alternative transport options in rural regions, means the impact of driving reduction or cessation may be more problematic for those living in rural areas. Dependence on driving for transport in rural areas may override decisions to regulate or cease driving based on declining function. Education needs to target specific strategies to support driving reduction or cessation in rural areas, while preserving community mobility. Rural communities need to take an active role in supporting their older members while governments need to provide better infrastructure with more support services.

Limitations

Objective measures of driving exposure were a study strength. While participants were encouraged to report if another driver drove their vehicle, it was not possible to determine all occasions when other people drove an instrumented vehicle. Although this is a study limitation, it was likely to have comprised only a small proportion of data (Coxon et al., 2017).

Although this study recruited healthy community-living older drivers, there was a range of function and ability among participants (Allan, Coxon, Bundy, Peattie, & Keay, 2016). All participants were encouraged, but not coerced, to develop a safe mobility and retirement from driving plan relevant to their level of function, with messages targeted at their self-regulation stage. Participants with poorer function were encouraged to make changes to their driving and consider alternatives now, while those with better function were encouraged to consider safest driving conditions and self-regulation strategies for the future, as well as plan and familiarize themselves with transport alternatives in preparation for retirement from driving. High functioning drivers may not need to reduce or make immediate changes to their driving exposure, which may limit observed changes to driving exposure. The program was designed to deliver a stronger, more urgent message when indicated. Safe mobility plans were more likely to be developed for drivers with lower function and/or poorer health.

Self-regulation is a multidimensional construct, which includes different strategies at life-goal, strategic, and tactical levels of driving behavior (Davis, Conlon, Ownsworth, & Morrissey, 2016; Molnar et al., 2013). Reducing mileage and avoiding night driving represent just two strategies older drivers may have included in their safe mobility plan and employed to regulate their driving. While capturing any changes in driving exposure is important, it is possible participants regulated their driving in other ways not evaluated in this study. Future research exploring other changes in self-regulation at life-goal, strategic, and tactical levels is warranted.

Conclusion

Older drivers, who took ownership over the process of driving self-regulation and retirement from driving to the point of developing plans, were more likely to reduce their driving exposure than those who did not. A stronger, more targeted program message was delivered and received, as intended, to older drivers with lower function and poorer health. This analysis suggests “Behind the Wheel” has greatest impact with older, lower functioning drivers. To achieve the best behavior change outcomes for resources invested, these drivers should be targeted. Older, lower functioning drivers are also the most in need of strategies to improve on-road safety and support community mobility when they cease driving. These findings, along with the program logic model, will inform future education programs and help channel resources to those who benefit most.

Footnotes

Acknowledgements

We would like to acknowledge the in-kind support from National Roads and Motorists’ Association (NRMA Motoring) for participant recruitment through letters of invitation send to members.

Authors’ Note

The lead author (KC) affirms this article is an honest, accurate, and transparent account of the study being reported. No important aspects of the study have been omitted. KC and LK led the methodological design of the study, supported by AC, KH, JB, EC, SB, KR, and RI. KR provided support with the statistical analysis. KC drafted the paper and logic model, and all authors contributed to revisions. All authors read and approved the final manuscript. KC is the guarantor of this research. All authors had full access to the data in the study and can take responsibility for the data integrity and accuracy of the analysis. Ethics approval for this trial was granted by the Human Research Ethics Committee of the University of Sydney (Protocol: 14235) and written consent was obtained from all participants before taking part in the study. This research was conducted independent from funders. The study funders were not involved in any stage of this research including study design, analysis, interpretation of data, report writing, or in the decision to submit the article for publication. The funder was provided a copy of the manuscript prior to submission.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This research was funded by an Australian Research Council Discovery Project (DP110101740), University of Sydney Equipment Grant, IRT Foundation, and Center for Road Safety at Transport for New South Wales.

ORCID iD

Kristy Coxon

References

Allan

Coxon

Bundy

Peattie

Keay

(2016). DriveSafe and DriveAware assessment tools are a measure of driving-related function and predicts self-reported restriction for older drivers. Journal of applied gerontology, 35, 583-600.

Bellg

A. J.

Borrelli

Resnick

Hecht

Minicucci

D. S.

Ory

. . .Czajkowski

(2004). Enhancing treatment fidelity in health behavior change studies: Best practices and recommendations from the NIH behavior change consortium. Health Psychology, 23, 443-451. doi:10.1037/0278-6133.23.5.443

Brabyn

J. A.

Schneck

M. E.

Lott

L. A.

Haegerstrom-Portnoy

(2005). Night driving self-restriction: Vision function and gender differences. Optometry and Vision Science, 82, 755-764.

Broekhuizen

Althuizen

van Poppel

M. N. M.

Donker

van Mechelen

(2012). From theory to practice: Intervention fidelity in a randomized controlled trial aiming to optimize weight development during pregnancy. Health Promotion Practice, 13, 816-825. doi:10.1177/1524839912447190

Coxon

Chevalier

Brown

Clarke

Billot

Boufous

Ivers

Keay

(2017). Effects of a safe transportation educational program for older drivers on driving exposure and community participation: a randomized controlled trial. Journal of the American Geriatrics Society, 65, 540-549.

Coxon

Keay

(2015). Behind the wheel: community consultation informs adaptation of safe-transport program for older drivers. BMC Research Notes, 8, 764. doi: 10.1186/s13104-015-1745-0

Coxon

Chevalier

Ivers

Brown

Keay

(2015). Behind the wheel: predictors of driving exposure in older drivers. Journal of the American Geriatrics Society, 63, 1137-1145.

Craig

Dieppe

Macintyre

Michie

Nazareth

Petticrew

(2008). Developing and evaluating complex interventions: The new Medical Research Council guidance. British Medical Journal, 337, 979-983.

D’Ambrosio

L. A.

Donorfio

L. K. M.

Coughlin

J. F.

Mohyde

Meyer

(2008). Gender differences in self-regulation patterns and attitudes toward driving among older adults. Journal of Women & Aging, 20, 265-282. doi:10.1080/08952840801984758

10.

Davis

Conlon

Ownsworth

Morrissey

(2016). Measuring situational avoidance in older drivers: An application of Rasch analysis. Accident Analysis & Prevention, 87, 68-77. doi:10.1016/j.aap.2015.11.018

11.

Dickerson

A. E.

Molnar

Bedard

Eby

D. W.

Classen

Polgar

(2017). Transportation and aging: An updated research agenda for advancing safe mobility. Journal of Applied Gerontology. doi:10.1177/0733464817739154.

12.

Dougherty

B. E.

Flom

R. E.

Bullimore

M. A.

(2005). An evaluation of the Mars Letter Contrast Sensitivity Test. Optometry and Vision Science, 82, 970-975

13.

Elford

Sherr

Bolding

Serle

Maguire

(2002). Peer-led HIV prevention among gay men in London: Process evaluation. AIDS Care, 14, 351-360.

14.

Gielen

A. C.

McKenzie

L. B.

McDonald

E. M.

Shields

W. C.

Wang

M.-C.

Cheng

Y.-J.

. . .Walker

A. R.

(2007). Using a computer kiosk to promote child safety: Results of a randomized, controlled trial in an urban pediatric emergency department. Pediatrics, 120, 330-339.

15.

Groll

D. L.

Bombardier

Wright

J. G.

(2005). The development of a comorbidity index with physical function as the outcome. Journal of Clinical Epidemiology, 58, 595-602. doi:10.1016/j.jclinepi.2004.10.018

16.

Hawe

Shiell

Riley

(2004). Complex interventions: How “out of control” can a randomised controlled trial be? British Medical Journal, 328, 1561-1563.

17.

Kay

L. G.

Bundy

A. C.

Clemson

L. M.

(2009a). Predicting fitness to drive in people with cognitive impairments by using DriveSafe and DriveAware. Archives of Physical Medicine and Rehabilitation, 90, 1514-1522. doi:10.1016/j.apmr.2009.03.011

18.

Kay

L. G.

Bundy

A. C.

Clemson

L. M.

(2009b). Awareness of driving ability in senior drivers with neurological conditions. American Journal of Occupational Therapy, 63, 146-150.

19.

Keay

Coxon

Chevalier

Brown

Rogers

Clarke

Ivers

R. Q.

(2018). Sex differences evident in self-reported but not objective measures of driving. Accident Analysis and Prevention, 111, 155-160. doi: 10.1016/j.aap.2017.11.011

20.

Korner-Bitensky

Kua

von Zweck

Van Benthem

(2009). Older driver retraining: An updated systematic review of evidence of effectiveness. Journal of Safety Research, 40, 105-111. doi:10.1016/j.jsr.2009.02.002

21.

Kostyniuk

L. P.

Molnar

L. J.

(2008). Self-regulatory driving practices among older adults: Health, age and sex effects. Accident Analysis & Prevention, 40, 1576-1580. doi:10.1016/j.aap.2008.04.005

22.

Linnan

Steckler

(2002). Process evaluation for public health interventions and research: An overview. In Steckler

Linnan

(Eds.), Process evaluation for public health interventions and research (pp. 1-24). San Francisco, CA: Jossey-Bass.

23.

McLaughlin

J. A.

Jordan

G. B.

(1999). Logic models: A tool for telling your programs performance story. Evaluation and Program Planning, 22(1), 65-72

24.

Molnar

L. J.

Eby

D. W.

(2008). The relationship between self-regulation and driving-related abilities in older drivers: An exploratory study. Traffic Injury Prevention, 9, 314-319. doi:10.1080/15389580801895319

25.

Molnar

L. J.

Eby

D. W.

Langford

Charlton

J. L.

St. Louis

R. M.

Roberts

J. S.

(2013). Tactical, strategic, and life-goal self-regulation of driving by older adults: Development and testing of a questionnaire. Journal of Safety Research, 46, 107-117. doi:10.1016/j.jsr.2013.05.001

26.

Musselwhite

C. A.

Shergold

(2013). Examining the process of driving cessation in later life. European Journal of Ageing, 10, 89-100. doi:10.1007/s10433-012-0252-6

27.

Oakley

Strange

Bonell

Allen

Stephenson

(2006). Process evaluation in randomised controlled trials of complex interventions. British Medical Journal, 332, 413-416. doi:10.1136/bmj.332.7538.413

28.

Owsley

McGwin

Jr. Phillips

J. M.

McNeal

S. F.

Stalvey

B. T.

(2004). Impact of an educational program on the safety of high-risk, visually impaired, older drivers. American Journal of Preventive Medicine, 26, 222-229. doi:10.1016/j.amepre.2003.12.005

29.

Owsley

Stalvey

B. T.

Phillips

J. M.

(2003). The efficacy of an educational intervention in promoting self-regulation among high-risk older drivers. Accident Analysis and Prevention, 35, 393-400.

30.

Owsley

Stalvey

B. T.

Wells

Sloane

M. E.

(1999). Older drivers and cataract: Driving habits and crash risk. The Journals of Gerontology Series A: Biological Sciences and Medical Sciences, 54, M203-M211.

31.

Oxley

Charlton

(2009). Attitudes to and mobility impacts of driving cessation: Differences between current and former drivers. Topics in Geriatric Rehabilitation, 25(1), 43-54.

32.

Ross

L. A.

Anstey

K. J.

Kiely

K. M.

Windsor

T. D.

Byles

J. E.

Luszcz

M. A.

Mitchell

(2009). Older drivers in Australia: Trends in driving status and cognitive and visual impairment. Journal of the American Geriatrics Society, 57, 1868-1873. doi:10.1111/j.1532-5415.2009.02439.x

33.

Rudman

D. L.

Friedland

Chipman

Sciortino

(2006). Holding on and letting go: The perspectives of pre-seniors and seniors on driving self-regulation in later life. Canadian Journal of Aging, 25(1), 65-76.

34.

Saunders

R. P.

Evans

M. H.

Joshi

(2005). Developing a process-evaluation plan for assessing health promotion program implementation: A how-to guide. Health Promotion Practice, 6, 134-147. doi:10.1177/1524839904273387

35.

Stalvey

B. T.

Owsley

(2003). The development and efficacy of a theory-based educational curriculum to promote self-regulation among high-risk older drivers. Health Promotion Practice, 4, 109-119. doi:10.1177/1524839902250757

36.

Taylor-Powell

Henert

(2008). Developing a logic model: Teaching and training guide. Benefits, 3, 1-38.

37.

Vance

D. E.

Roenker

D. L.

Cissell

G. M.

Edwards

J. D.

Wadley

V. G.

Ball

K. K.

(2006). Predictors of driving exposure and avoidance in a field study of older drivers from the state of Maryland. Accident Analysis & Prevention, 38, 823-831. doi:10.1016/j.aap.2006.02.008

38.

Weinstein

N. D.

Lyon

J. E.

Sandman

P. M.

Cuite

C. L.

(1998). Experimental evidence for stages of health behavior change: The precaution adoption process model applied to home radon testing. Health Psychology, 17, 445-453.

39.

Weinstein

N. D.

Sandman

P. M.

(1992). A model of the precaution adoption process: Evidence from home radon testing. Health Psychology, 11, 170-180.

40.

Weinstein

N. D.

Sandman

P. M.

Blalock

S. J.

(2008). The precaution adoption process model. In Glanz

Rimer

B. K.

Viswanath

(Eds.), Health behavior and health education: Theory, research, and practice (4th ed., pp. 123-148). San Francisco, CA: Jossey-Bass.

41.

Wong

I. Y.

Smith

S. S.

Sullivan

K. A.

Allan

A. C.

(2016). Toward the multilevel older person’s transportation and road safety model: A new perspective on the role of demographic, functional, and psychosocial factors. The Journals of Gerontology, Series B: Psychological Sciences and Social Sciences, 71, 71-86. doi:10.1093/geronb/gbu099