Preliminary validation of the Candrive Older Driver Risk Stratification Tool
Author(s): Marshall, Bedard, Vrkljan, Tuokko, Porter ,Naglie, Rapoport, Mazer, Glinas, Gagnon, Charlton, Koppel, MacLeay, Myers, Mallick, Ramsay, Stiell, Wells, Man-Son-Hing
Slidedeck Presentation:
Candrive RST preliminary validation (20220617)_Marshall
Abstract:
Background:
While the Candrive risk stratification tool (RST) has been developed using a rigourous process, further validation of the tool is essential to ensure its applicability in the clinical setting.
Aims:
Ozcandrive at-fault crashes: We recruited active Australian drivers aged 75+ (the “Ozcandrive” cohort). Participants underwent the same Candrive assessments and were assigned a risk category as per the RST classification process. The primary outcome measure was self-reported at-fault crashes, per person-years. Because of the limited number of crashes in the two highest risk categories they were combined into one category. We then calculated the incidence rate ratio of an at-fault crash compared to the lower risk category. Candrive traffic violations: The main outcome measure was a police-reported traffic violation. Because not all traffic violations may represent poor fitness-to-drive, they were divided a priori into speeding and “cognitive violations” (e.g., failure to obey traffic signals). Because of the limited number of violations, we combined the two lower-risk categories (low-risk), and the two higher-risk categories (high-risk). We generated odds ratios (OR) comparing the high-risk group to the low-risk group, using a generalized linear model to account for the clustered nature of the data, and adjusting for age, gender, and annual distance travelled.
Methods:
Ozcandrive at-fault crashes: We recruited 257 older drivers (71% male). The average age at enrolment was 79.7 years (SD = 3.5). The participants combined for a total of 990 person-years of driving and were involved in 248 crashes of which 140 (56%) were at-fault crashes. Based on the Candrive RST, the majority of the person-years (74.4%) fell within the lowest risk category, whereas 10.1% fell in the low-medium category, and 15.5% fell within the combined medium-high and high-risk categories. The incidence rate ratio for an at-fault collision was 1.73 (95% CI = 1.20, 2.50) for the highest risk category compared to the lowest. Candrive traffic violations: A total of 163 speeding violations and 87 cognitive violations were reported over the seven-year study period. The OR for speeding violations did not differ between groups (OR = 0.97, 95% CI = 0.64, 1.45). However, for cognitive violations, it was 1.62 (95% CI = 1.02, 2.57) for the high-risk group compared to the low-risk group. For both analyses, age and gender were not associated with violations, but greater annual distances traveled were associated with speeding violations.
Results:
A higher-risk RST classification was associated with at-fault crashes in the Ozcandrive sample. Similarly, Candrive participants in the higher-risk category had higher odds of having a police-reported cognitive violation. While there are limitations in using self-reported crashes and traffic violations as outcome measures, both types of adverse driving outcomes support the validity of the RST.
Discussion:
Preliminary validation of the RST supports further study of the tool. Implementation trials in clinical settings will provide the evidence required to support the wider use of the RST.