Canadian statistics from 2009 indicate that 3.25 million or 74.9 per cent of people aged 65 and over had a driver’s licence, and 200,000 or 40.7 per cent were age 85 years and over (Turcotte, Reference Turcotte2012). It is predicted that approximately 85 per cent of older adults over the age of 65 will have a driver’s licence in the year 2028 in Ontario, but it stands to reason that a similar trend will be seen across Canada as well (Hopkins, Kilik, Day, Rows, & Tseng, Reference Hopkins, Kilik, Day, Rows and Tseng2004). It is also predicted that the discrepancy in licencing rates between male and female older drivers will narrow to 1:1 by as early as 2023 (Dobbs, Reference Dobbs2008). Further to this, Rosenbloom (Reference Rosenbloom2001) noticed an increasing dependency of seniors on the personal automobile to meet their transportation needs, and Turcotte (Reference Turcotte2012) found older adults are living longer in highly car-dependent neighbourhoods.
In addition to the aging driving population, there is the U-shaped curve that indicates per 100 million kilometers (km) driven, older adults have an increased casualty crash rate involving injuries (Langford, Koppel, Charlton, Fildes, & Newstead, Reference Langford, Koppel, Charlton, Fildes and Newstead2006). Additionally, older adults aged 80 and older have the second highest crash rate per km driven, just short of the rate for 15–19 year olds (Langford et al., Reference Langford, Koppel, Charlton, Fildes and Newstead2006; Langford, Koppel, McCarthy, & Srinivasan, Reference Langford, Koppel, McCarthy and Srinivasan2008). There have been two main theories that attempt to explain this increase: The first is the low-mileage bias, which suggests that, regardless of age, drivers who drive relatively low annual mileages are at an increased risk for crashes, due to most of their trips occurring in congested urban areas and intersections where the risk of crashes are increased (Hakamies-Blomqvist, Raitanen, & O’Neill, Reference Hakamies-Blomqvist, Raitanen and O’Neill2002; Janke, Reference Janke1991; Langford et al., Reference Langford, Koppel, Charlton, Fildes and Newstead2006, Reference Langford, Koppel, McCarthy and Srinivasan2008). The second theory is the frailty bias, which suggests that crashes involving injuries are more likely to be reported to the police. In such instances, older drivers are also more likely to be injured in a crash because of age-related fragility (Li, Braver, & Chen, Reference Li, Braver and Chen2003), resulting in more older drivers being included in crash databases compiled from police reports.
In Canada, crash risk for all road users increased with rainfall and snowfall, compared to dry conditions (Andrey, Hambly, Mills, & Afrin, Reference Andrey, Hambly, Mills and Afrin2013), as did risk of injury (rainfall 74%; snowfall 89%; Andrey, Reference Andrey2010). If older adults are at greater risk for crashes and related injury, then these weather-related statistics provide reason for concern. Therefore, it is important to examine the extent to which older adults are driving during these conditions.
As argued by Lindstrom-Forneri, Tuokko, Garrett, and Molnar (Reference Lindstrom-Forneri, Tuokko, Garrett and Molnar2010), it is important to consider both individual and environmental factors as well as interactions between these factors when examining the driving behaviour of older adults. Most frameworks on driving behaviour, including the one by Lindstrom-Forneri et al. (Reference Lindstrom-Forneri, Tuokko, Garrett and Molnar2010), refer to Michon’s model (Michon, Reference Michon, Schwing and Evans1985) that categorizes different levels of compensatory behaviours: (1) strategic, which refers to planning decisions (e.g., using familiar or unfamiliar roads, postponing trips, or shortening trips); (2) tactical, which involves maneuvering such as adjusting one’s speed due to environmental conditions; and (3) operational, which pertains to control of the vehicle (e.g., using turning signal lights).
There is substantial evidence that older drivers, particularly women, appear to modify their driving practices as they age. For example, older drivers tend to drive: less often, closer to home, in the daytime, on weekdays, and in familiar areas (e.g., Collia, Sharp, & Giesbrecht, Reference Collia, Sharp and Giesbrecht2003; D’Ambrosio, Donorfio, Coughlin, Mohyde, & Meyer, Reference D’Ambrosio, Donorfio, Coughlin, Mohyde and Meyer2008; Keall & Frith, Reference Keall and Frith2004). They also tend to avoid: driving at night, driving in bad weather, driving in rush hour, driving on highways, and making complex maneuvers such as left-hand turns (e.g., Baldock, Mathias, McLean & Berndt, Reference Baldock, Mathias, McLean and Berndt2006; Charlton et al., Reference Charlton, Oxley, Fildes, Oxley, Newstead, Koppel and O’Hare2006; D’Ambrosio et al., Reference D’Ambrosio, Donorfio, Coughlin, Mohyde and Meyer2008; Hakamies-Blomqvist & Wahlström, Reference Hakamies-Blomqvist and Wahlström1998; Lyman, McGwin, & Sims, Reference Lyman, McGwin and Sims2001; Oxley, Charlton, Scully, & Koppel, Reference Oxley, Charlton, Scully and Koppel2010). To date, most of the research has used questionnaires or interviews to examine self-reported driving practices (frequency, distance, and types of situations most often avoided).
Evidence is accumulating that self-reported driving practices may not be accurate. Several studies have found that older drivers misestimate driving distance when compared to objectively measured mileage using in-vehicle recording devices (Blanchard, Myers, & Porter, Reference Blanchard, Myers and Porter2010; Crizzle & Myers, Reference Crizzle and Myers2013; Huebner, Porter, & Marshall, Reference Huebner, Porter and Marshall2006; Molnar et al., Reference Molnar, Charlton, Eby, Bogard, Langford and Koppel2013; Porter et al., Reference Porter, Smith, Cull, Myers, Bédard and Gélinas2015; Staplin, Gish, & Joyce, Reference Staplin, Gish and Joyce2008). With regards to weather, Blanchard et al. (Reference Blanchard, Myers and Porter2010) found that in a sample of 61 drivers aged 67–92, 43 per cent of those who reported avoiding driving in bad weather actually drove during “bad weather”.
Related to the present study of seasonal effects on older adults driving, there is a study by Sabback and Mann (Reference Sabback and Mann2005) who specifically addressed the influence of climatic and road conditions on the driving practices of older adults in New York and Florida, albeit using only self-report (telephone surveys) for driving frequency, miles driven, climatic, seasonal, and road conditions. They found older adults (n = 20) in Western New York reported altering their driving patterns due to seasonal variations in weather conditions.
Moreover, Myers, Trang, and Crizzle (Reference Myers, Trang and Crizzle2011) is the only study to date that has looked at the naturalistic driving patterns of older adults specifically during the winter season over a two-week period. Overall, the sample was more likely to drive (69%) than not drive (31%) on days with inclement weather, and 67 per cent drove on days when weather advisories had been issued for the region. This differs from the suggestion by Kilpelainen and Summala (Reference Kilpelainen and Summala2007) that older drivers would not drive during adverse weather conditions as their trips are generally more discretionary.
The primary purposes of the present study were to determine if there were seasonal (winter versus non-winter) and weather-related (inclement versus non-inclement) changes in the driving patterns (specifically trip distances) of Canadian older adults over a full year. Additionally, we examined trip distances at night compared to daylight hours as length of daylight varies by season in northern countries. Our primary hypotheses were that there would be a decrease in trip distance during winter and a decrease in trip distance during inclement weather conditions, regardless of season. We further hypothesized that trip distance would decrease during night-time driving regardless of season or weather conditions, and that increased age, decreased mobility, and poorer health status would result in an even greater decrease in trip distance during both winter and inclement weather conditions.
Methods
This study utilized data from the Canadian Driving Research Initiative for Vehicular Safety in the Elderly (Candrive II) project. Candrive is a longitudinal cohort study assessing the everyday driving patterns of older drivers, as well as the development of a clinical decision rule to stratify the risk of crashes (Marshall et al., Reference Marshall, Man-Son-Hing, Bedard, Charlton, Gagnon and Gelinas2013).
Participants
The participants for Candrive were recruited from Ottawa, Toronto, Montreal, Hamilton, Thunder Bay, Winnipeg, and Victoria. At the time of enrollment, participants (n = 928) had to be aged 70 or older, possess a valid driver’s licence, report driving at least 4 times per week on average, live in their respective city at least 10 months of the year (see Marshall et al., Reference Marshall, Man-Son-Hing, Bedard, Charlton, Gagnon and Gelinas2013), and agree to having an In-Car Recording Device (ICRD) installed in their vehicle for the study duration (see Porter et al., Reference Porter, Smith, Cull, Myers, Bédard and Gélinas2015). Institutional approval for each site was granted by the appropriate ethics board, and all participants provided written informed consent.
Additionally, for the present study, participants must have indicated at the baseline assessment that they were not considering restricting or quitting driving in the next 6 months. As well, participants must have only driven the vehicle that was equipped with the electronic device over the first year of the study (as we could not capture data from other vehicles they drove, which would lead to an underestimation of driving exposure). The latter criterion resulted in 433 people being excluded from the analyses (see Figure 1). Participants (n = 152) with missing, incomplete, and unusable driving data files were also excluded from analysis.
Figure 1: Participant flow chart showing the numbers of participants who were removed from analyses due to various inclusion/exclusion criteria
Apart from changing vehicles, some data was lost due to inadvertent disconnections of the devices (e.g., during vehicle servicing). To allow for short interruptions, participants had to have a minimum of 335 days of driving data (one month short of a year), with at least 90 per cent of usable GPS information collected over the year, to be included in the analyses. Missing GPS data often resulted from trips being very short (unable to establish a satellite connection); however, there could also have been problems with the GPS receiver and/or connection with the ICRD; the latter could have resulted in a systematic loss in trips for analysis.
As illustrated in Figure 1, 279 participants met these criteria for inclusion in the “All trips” dataset. This dataset included all the trips made by the participants during the year. This dataset was used to examine the effects of season but not weather.
Because weather data were obtained from stations within or adjacent to each of the seven research centres, we also created a separate database for “City trips” comprising 248 of the 279 subjects. This dataset allowed us to examine the specific effects of weather (not just season) on trip distance. Participants whose first three digits of their postal code indicated that they lived outside of the 2011 Census population centre of their respective Candrive site were excluded (n = 31) from the weather analysis as shown in Figure 1. Additionally, any trips by these 248 participants that started and ended outside of population centre limits were also removed from this dataset.
Data Collection
Participants in the Candrive study completed a comprehensive assessment at baseline and annually, as described by Marshall et al. (Reference Marshall, Man-Son-Hing, Bedard, Charlton, Gagnon and Gelinas2013). For this study, we examined age, gender, marital status, living arrangements, education, volunteer status, availability of friends and family to drive, self-reported health status, medications, scores on the Expanded Cumulative Illness Rating Scale (CIRS), and time to complete the Rapid Pace Walk and Timed Up and Go tests. The CIRS (Hudon, Fortin, & Soubhi, Reference Hudon, Fortin and Soubhi2007), as modified by Candrive, provided information on 44 medical conditions that participants ranked from no problem (0) to an extremely severe problem (4).
Installed in each participant’s vehicle was the ICRD. The ICRD was the OttoView-CD data acquisition system (Persen Technologies Inc., 2010) which was installed into the vehicles of all participants at the time of enrolment. The GPS antenna receives positional, speed, and Greenwich Mean Time (GMT) time and date information, and the OttoView-CD device collected and stored information on driving time, vehicle speed, distance, and other vehicle engine data; this information was collected once per second (Porter et al., Reference Porter, Smith, Cull, Myers, Bédard and Gélinas2015). For the data collected with this device, a trip is defined as occurring when the ignition is turned on and then turned off again. Therefore, a trip to the grocery store would be one trip, and returning home would be a second trip.
Data Analyses
Winter
Winter in Canada is challenging to define. The meteorological definition of winter is the three months consisting of December, January, and February (MetOffice, 2013), whereas, for the astronomical definition, winter starts at the winter solstice and ends at the spring equinox (MetOffice, 2013). Both of these definitions encompass only three months of the year. In Canada, many people think of the winter season as being longer, beginning with snowfall and/or cold temperatures and ending when the snow melts and the temperature warms up, which varies considerably in different parts of the country (e.g., Winnipeg versus Victoria). For the purposes of this study, six definitions of winter were developed for examination (as shown in Table 1).
Table 1: Winter definitions
The driving trips examined in this study were collected between June 29, 2009, and November 18, 2011. Using each of the definitions shown in Table 1 to classify trips, the proportion of winter days captured for each location ranged from 26 per cent to 58 per cent for Ottawa, 20 per cent to 58 per cent for Toronto, 21 per cent to 62 per cent for Montreal, 25 per cent to 64 per cent for Hamilton, 28 per cent to 57 per cent for Thunder Bay, 30 per cent to 55 per cent for Winnipeg, and 0 per cent to 69 per cent for Victoria.
Night
Trips from both datasets were further categorized according to whether they took place during daylight hours or not (time of day). For brevity purposes, trips that occurred outside of daylight hours are referred to as night. Night driving was defined as any trip where at least a portion of the trip occurred after sunset or before sunrise (Myers et al., Reference Myers, Trang and Crizzle2011). Sunrise and sunset were calculated using the date and the GPS coordinates from the beginning and end of the trip using the method described by Teets (Reference Teets2003).
City
City trips were defined as those which started and ended within the 2011 Census population centre for the respective city or within 10 km of the airport. Trips in proximity to the airport were included for two reasons: (1) weather data were collected from the weather stations at each city’s respective airport, and (2) because the airport in every city except Winnipeg is located outside of the population centre. We used this approach to maximize the number of trips that were included in this dataset, and because trips that were made close to the airport would have valid weather data, even if they were not within the geographic limits of the city border.
Weather
Weather data came from Environment Canada’s hourly historical weather records for each of the seven cities’ airport weather stations. Airport weather stations were used because they consistently collect information on weather conditions (e.g., mainly clear, drizzle). In addition to weather conditions, the hourly weather data provides information on temperature, humidex (humidity index), and wind chill values. Environment Canada usually collected this information once an hour on the hour mark.
The yearly summary file for each participant included in the city driving dataset was merged with the corresponding hourly weather data from Environment Canada. Trips with a start and end time that crossed weather points were averaged. For example, if a trip started at 11:50 a.m. and ended at 12:10 p.m., the temperature at 11 a.m. and 12 p.m. was averaged. If conditions differed between the time points (e.g., “mostly cloudy” at 11:00 a.m. and “snowing” at 12:00 p.m.), both were used. Trips were then coded according to weather condition and categorized as occurring during inclement (four definitions) versus non-inclement weather as shown in Table 2.
Table 2: Inclement weather and precipitation type definitions
Statistical Analyses
To examine the extent to which winter, inclement weather conditions, and night-time driving explained the variation in trip distance exhibited by older drivers over the course of a year, we performed a multi-level random intercept regression. For this study, the outcome of interest was trip distance, which was nested within participants. Because it was not normally distributed, trip distance was transformed using a natural log. To determine whether trip-level variables (i.e., season, weather, time of day) and participant-level characteristics (e.g., age, gender, health status, etc.) had an effect on trip distance, we used a multi-level regression analysis. SAS 9.3M0 software was used for all analyses.
The model building steps outlined by Bell, Ene, Smiley, and Schoeneberger (Reference Bell, Ene, Smiley and Schoeneberger2013) were the ones we used for both the dataset containing All trips (n = 279 participants or 377,464 trips), and for the dataset consisting of City trips (n = 248 participants, or 298,342 trips). First, we ran a sensitivity analysis to determine which definitions of winter (All trips and City trips datasets) and inclement weather (City trips) explained the greatest amount of variation in trip distance. Next, the unconditional model was used to confirm the appropriateness of a multi-level regression and to calculate the intraclass correlation coefficient. Finally, using the definitions emerging from the sensitivity analysis, we added participant-level characteristics into the model using a forward step-wise approach to determine which, if any, were significant. With the change in deviance or the likelihood ratio test, we compared the model containing just the trip-level characteristics with the model containing both the participant- and trip-level characteristics to determine which model best fit the data and therefore the final model (Peugh, Reference Peugh2010).
Results
Participants
Of the 279 participants included in the analysis for the All trips dataset, 25 per cent were from Ottawa (site 1), 14 per cent from Toronto (site 2), 10 per cent from Montreal (site 3), 14 per cent from Hamilton (site 4), 10 per cent from Thunder Bay (site 5), 11 per cent from Winnipeg (site 6), and 15 per cent from Victoria (site 7). The sample ranged in age from 70 to 92 (average age at time of enrollment was 77.5 ± 5.2 years), with a relatively equal gender distribution (51% male). 54 per cent reported still living in a house, 44 per cent were married, and 39 per cent had some university education. Only 7.9 per cent were working full or part-time, while 53 per cent reported volunteering full- or part-time. Sixty-six per cent reported having family or friends that were willing to drive them.
In terms of health, 76 per cent reported their health status as either excellent or very good. The average number of reported medications was 4.7 ± 2.9, and the average CIRS score was 10.9 ± 5.0. The average times to complete the Rapid Pace Walk and Timed Up and Go tests were 6.8 ± 1.7 seconds and 10.4 ± 2.4 seconds, respectively. The sample characteristics for the City trips dataset were essentially the same.
Inclement versus Non-inclement Weather
Of the 107,674 hours of weather data examined, 23.8 per cent met the criteria for inclement weather using the first definition, 26.9 per cent using the second definition, and 30.6 per cent using the third definition (see Table 2 for definitions). No precipitation was recorded for 78 per cent of the hourly data, while 9.8 per cent had some form of rain, 8.3 per cent had some form of snow or frozen precipitation (e.g., freezing drizzle or snow), while 3.2 per cent indicated vision obstructing precipitation (e.g., blowing snow or fog).
Multi-level Regression
Sensitivity Analysis
All Trips Dataset
We ran a model with each definition of winter to determine which definition best explained the variation in trip distance when trips were treated as independent observations. The fourth definition of winter (i.e., starting on the first date of the first four consecutive days of lows below 0°C and ending on the first date of the last four consecutive days of lows below 0°C), explained the most variation at the trip level when controlling for time of day and site.
City Trips Dataset
Similar to the model we ran with the All trips dataset, we ran a model with each combination of winter and inclement weather to determine which combination explained the most variation in trip distance when each trip was treated as an independent observation. The sixth definition of winter (i.e., starting on the first day of temperatures less than or equal to halfway between the minimum and maximum highs for the year and ending on the last day of temperatures less than or equal to halfway between the minimum and maximum highs for the year), and the fourth definition of inclement weather (i.e., weather conditions indicating precipitation), explained the most variation in trip distance when controlling for time of day and site.
Regardless of which definition of winter we used, there was a small but statistically significant decrease in trip distance during winter when compared to non-winter driving when controlling for weather (City trips dataset only), time of day, and site.
Final Model
Table 3 presents the final models for the All trips and City trips datasets. Given that trips are nested within participants for both, the multi-level model was an appropriate method for analysis (All trips: trip level variance estimate = 1.53 [Z = 434.27, p < 0.001], and participant level variance estimate = 0.11 [Z = 11.65, p < 0.001]; City trips: trip level variance estimate = 1.15 [Z = 349.74, p < 0.001], and participant level variance estimate = 0.08 [Z = 9.97, p < 0.001]). From the level 1 and level 2 variance, we calculated the ICC (All trips: 0.11/[0.11 + 1.53] = 0.07, City trips: 0.08/[0.08 + 1.15] = 0.07), indicating that only 7 per cent of the variation in trip distance could be explained by the participant characteristics (e.g., age, gender). Most of the variation in trip distance occurred at the trip level for both datasets.
Table 3: Final model for All trips and City trips datasets
Variance values are the variance estimate (all variances were significant at p < 0.0001), and regression values are the parameter estimate (standard error). *p < 0.1, **p < 0.05, ***p < 0.01
AIC = Akaike information criterion
–2LL = Negative of twice the restricted log likelihood
Participant-level characteristics were then added to the unconditional model to determine which, if any, were significant predictors of trips distance. It was determined for the All trips dataset that marital status – and, for the City trips dataset, living arrangements – were the only participant-level characteristics that were significant. When we added these characteristics to their respective model containing season, weather (City trips dataset only), time of day, and site, they did not improve the fit of the models and were therefore excluded from the final models.
The final models can be interpreted as:
-
• All Trips Dataset:
-
◦ In the winter (compared to non-winter), there was a 7 per cent decrease in trip distance when controlling for time of day and site.
-
◦ For night-time driving, there was a 33 per cent increase in trip distance when controlling for site and winter.
-
-
• City Trips Dataset:
-
◦ In the winter, there was a 1 per cent decrease in trip distance when controlling for weather, time of day, and site.
-
◦ There was a 5 per cent increase during rain, and no change during snow or vision obstructing precipitation when controlling for season, time of day, and site.
-
◦ There was a 29 per cent increase in trip distance during night-time driving when controlling for weather, site, and season.
-
Discussion
As we noted at the outset, only a few studies to date have objectively examined the naturalistic driving practices of older adults, and fewer still have examined weather conditions. Although the findings were informative, monitoring periods were short (one to two weeks) and samples were limited. The current study is the first to examine a full year of driving for a sample of older adults from seven locations across Canada, to determine if older drivers reduce their trip distance based on season or weather. The study employed strict exclusion criteria such as driving only one vehicle and examined trip-level rather than aggregate data.
As expected, there was a decrease in trip distance in the winter (compared to non-winter), when all trips were examined (7%) or even for trips within their city (1%), although the decrease was relatively small. As noted by other researchers, in harsh climates like Finland (Kilpelainen & Summala, Reference Kilpelainen and Summala2007) and Canada (Myers et al., Reference Myers, Trang and Crizzle2011), people must deal with winter driving conditions if they wish to maintain their activities. This might explain why there was no difference in city trip distance when there was snow or vision obstructing precipitation. In addition, it was surprising that participants actually drove farther when it was raining. It is possible that people chose to take longer trips for indoor activities (e.g., to the mall across town) when it was raining as opposed to doing outdoor activities such as walking or gardening. In the only other study to objectively examine weather-related driving practices of older drivers, Myers et al. (Reference Myers, Trang and Crizzle2011) found that 70% drove on days with inclement weather (snow or rain) and poor road conditions.
One of the most interesting and unexpected findings of this study was the increase in trip distance (about 30%) at night compared to daytime driving. Prior naturalistic studies using shorter monitoring periods have found that older drivers make proportionately fewer trips at night and drive shorter distances than in the daytime (Myers et al., Reference Myers, Trang and Crizzle2011; Crizzle & Myers, Reference Crizzle and Myers2013). However, these studies used shorter monitoring periods (two weeks) and did not compare seasons. Crizzle and Myers (Reference Crizzle and Myers2013) found that the number of night trips and duration were significantly correlated with leisure trips in their group of healthy older drivers. Although we did not assess trip purposes in this study, it is possible that night driving trips might be more discretionary in nature, and may take someone farther from home than they would normally travel (e.g., to watch a grandchild’s hockey game).
Together these results suggest that older Canadian drivers may alter their trip distances based on weather, season, and time of day, although not always in the expected direction. As Lindstrom-Forneri et al. (Reference Lindstrom-Forneri, Tuokko, Garrett and Molnar2010) proposed, “Decisions made at the strategic, tactical, and operational levels must be viewed within the social/physical environmental context …” (p. 283). For older North Americans, transportation and hence community mobility are “synonymous with being able to operate an automobile” and is “inextricably linked to independence, autonomy and quality of life” (Dickerson et al., Reference Dickerson, Molnar, Eby, Adler, Bedard and Berg-Weger2007, p. 579). Therefore, an older driver might choose to make longer trips even if the weather were poor or if it were night, in order to maintain their community mobility. City planners, transportation engineers, health practitioners, and those offering programs and events (e.g., concerts) that older adults attend need to consider the travel patterns of older drivers. For example, those who are responsible for roadways could make accommodations to make night or poor-weather trips more safe and comfortable for older drivers. This could include, for example, better signage; road maintenance; intersection design; and lighting (Boot, Stothart, & Charness, Reference Boot, Stothart and Charness2014). Automotive engineers could also ensure that vehicles are designed with older adults in mind (e.g., to accommodate their ability to use vehicle temperature controls at night). Event planners, meanwhile, may need to address issues of clear signage, and sufficient lighting in entry ways and parking areas.
Although this study provided an extremely detailed analysis of trip distances of older drivers, there were several limitations. We observed very low R 2 values, indicating that when each trip was treated as an independent observation, the covariates chosen did not explain much of the variation and therefore other factors may be more predictive. Although studies of shorter duration (two weeks) have examined driver perceptions of weather and driving trip purposes using logs (e.g., Crizzle & Myers, Reference Crizzle and Myers2013; Myers et al., Reference Myers, Trang and Crizzle2011), it was unrealistic to ask our participants to record such information for every trip, 365 days a year, over several years. Driving diaries can quickly become burdensome and thus compliance becomes an issue (Marshall et al., Reference Marshall, Wilson, Molnar, Man-Son-Hing, Stiell and Porter2007). Another major challenge was defining winter and inclement weather in a consistent manner across sites. Individuals likely have different perceptions about what constitutes “winter” and “inclement weather” based on where they live, and these perceptions in turn could affect changes in driving patterns.
Road conditions were also not considered in this analysis as there are no databases that contain historical records on road conditions, especially within cities. This is important because road conditions can be affected not only by current precipitation, but also by past precipitation, temperature, and wind-related factors. For example, in Winnipeg the weather could be clear, but the roads could be extremely slippery due to low temperatures and wind “polishing” the roads. Finally, we collected weather data from only one location for each of seven different sites. Thus, the weather data were likely more accurate for trips that occurred near the respective weather stations, and less accurate for trips made farther from these stations. Furthermore, although the weather data was reported once per hour, weather can change quickly and it is possible that conditions were not always accurately captured (e.g., a short downpour).
Trip distance (measured once people got into their vehicle; i.e., chose to drive) was the variable of interest in this study examining winter and weather. Older drivers may adjust to adverse driving conditions by choosing not to drive (i.e., cancelling or postponing trips) or adopt other strategies (e.g., taking alternate routes or reducing their speed) rather than reducing their trip distance. Participant characteristics such as health or mobility did not seem to explain much of the variation in trip distance. If the number of trips had been examined, then participant characteristics may have played more of a role. In future, self-regulation can be explored further by examining numbers of trips taken, distance from home, and speed, using Candrive data. Additionally, objectively measured driving data could be combined with information on older drivers’ attitudes towards driving and perceived comfort level to try to better understand the mechanisms of self-regulation.
Conclusion
This study provides a detailed examination of seasonal/weather-related driving trip distances by a large sample of older adults at seven Canadian sites over a full year. Although older adults drove slightly shorter trip distances during winter conditions, trip distance was slightly longer during both rain and vision-obstructing precipitation types (as compared to no precipitation), and substantially longer at night. Additional research is clearly needed to further understand the driving patterns of older drivers, including reasons for trip cancellation/postponement and route changes, especially in relation to season and weather in different geographical areas.
