Evidence shows that acute marijuana use impairs psychomotor functions in real-world tasks such as driving [1]. Current research indicates that, in general, marijuana users experience cognitive impairment in areas such as psychomotor speed, attention, and executive functioning [2-6]. Specifically, research has explored the effect of acute marijuana use on psychomotor skills, which are cognitive motor skills imperative for everyday activities such as safe driving. Researchers have assessed psychomotor function on measures of critical tracking, attention, and reaction time, and the findings have indicated that marijuana usage creates deficits in psychomotor function for up to 3.5 hours after use [7,8]. Impairments to psychomotor abilities have practical implications, as these implications impair an individual’s ability to drive [7,9]. While the acute effects of tetrahydrocannabinol use are known, the effect of medical marijuana use on psychomotor function under real-world conditions remains unknown.

Research on psychomotor functioning in older adults generally uses cognitive assessment tools, such as the Trail Making Test, which have limited real-world application [10-15]. The test is performed under artificial, time-pressured conditions that do not resemble the self-paced, contextually rich nature of driving performance. Studies of driving performance using high-fidelity driving simulators do mimic real-world conditions and are valid predictors of on-road driving [16-18]. Driving simulators allow for the systematic representation of events and the manipulation of variables (eg, road type, traffic, intersections), which offers experimental control that is dangerous on the road [16,18]. Driving simulators also offer optimal stimulus presentation through realistic and immersive scenarios, which allows for analysis of both healthy and impaired drivers under similar conditions [17]. Using a high-fidelity driving simulator, we can examine response time, attention, and executive functions by simulating valid real-world driving tasks [18]. Medical marijuana use and how it influences driving performance needs to be examined, and the high-fidelity driving simulator offers a safe, reliable, and valid option to do so [19], anticipating future on-road studies to examine these factors under real-world conditions.

The current study had 2 aims. Aim 1 was to evaluate the feasibility of recruiting and testing adults aged 50 years and older who are newly registered for medical marijuana, along with matched non–marijuana-using controls, in a longitudinal high-fidelity driving simulator protocol [19]. Aim 2 was to explore preliminary associations between medical marijuana initiation and simulated driving performance. Specifically, we examined whether medical marijuana use was associated with changes in divided attention (DA) and reaction time during a driving task among adults aged 50 years and older by assessing participants before and after medical marijuana initiation and comparing them with age-, race-, and sex-matched controls.

The study was approved by the Florida State University Institutional Review Board (IRB; STUDY00001659) and the University of Florida’s IRB (CED000000442). All participants signed an informed consent form prior to enrollment in the study. Study data were coded using participant ID numbers; identifiable data were stored separately from study data on secure, password-protected institutional systems accessible only to authorized study staff. Participants received a US $50 grocery store gift card at each study timepoint. Due to their sensitive nature, data are only available internally to intervention Research Advancing Care Excellence lab staff and related institutional stakeholders (IRB).

This study was a prospective, nonrandomized feasibility cohort study with repeated measures. Adults newly registered for medical marijuana were assessed prior to initiation (T1) and approximately 1 month after initiation (T2). Outcomes were compared with those of age-, race-, and sex-matched non–marijuana-using controls assessed at the same intervals.

Accordingly, the manuscript follows the STROBE (Strengthening the Reporting of Observational Studies in Epidemiology) reporting guidelines for cohort studies. A completed STROBE checklist is provided in Checklist 1.

For aim 1, which evaluated the feasibility of recruiting and testing participants, we summarized participation rates at each stage of recruitment and documented reasons for nonparticipation. Issues encountered during driving simulator testing, including simulator sickness and other technical or participant-related factors, were also summarized descriptively.

For aim 2, which explored preliminary associations between medical marijuana initiation and simulated driving performance, we first reported descriptive statistics for potential confounding variables and the focal driving performance outcomes. Baseline differences between the medical marijuana and control groups were examined using independent samples t tests for continuous variables and chi-square tests for categorical variables.

For the focal driving performance outcomes, we first examined distributional properties within each group versus time condition. Skewness and kurtosis values were inspected to assess departures from normality. Consistent with recommended guidelines (±1.96) [32], most variables demonstrated acceptable normality, with skewness and kurtosis values within this range except for 1 moderate kurtosis value (2.66).

To examine whether medical marijuana initiation was associated with simulated driving performance, we estimated linear mixed effects models [33,34]. These models included fixed effects for group (medical marijuana vs control), time (baseline vs follow-up), sex, and the group × time interaction, with a random intercept for participant to account for repeated observations within individuals. The group × time interaction was used to test whether changes in driving performance differed between groups. Mixed effects models were selected because they can accommodate unequal sample sizes and missing observations across timepoints and are relatively robust to moderate deviations from normality and unequal variances.

Because this study was designed as a feasibility study with a small sample size, the analyses were considered exploratory. Statistical significance was set at P<.05. Effect sizes and 95% CI were reported to aid interpretation of the findings.

A total of 231 individuals were contacted for the medical marijuana cohort (Figure 3) and 189 individuals were contacted for the control cohort (Figure 4). Among them, 87 out of 231 (38%) and 78 out of 189 (41%) participants were successfully contacted, respectively. After screening, 61 medical marijuana candidates (15 not meeting inclusion criteria and 46 declining participation) and 54 control candidates (25 not meeting inclusion criteria and 29 declining participation) were excluded. Subsequently, 26 medical marijuana participants and 24 control participants were scheduled for ROAD ACE (Reactions of Older Adults While Driving After Cannabis Exposure) assessment. Of these, 3 medical marijuana candidates (2 no shows and 1 scheduling conflict) and 3 control candidates (scheduling conflicts) did not complete the baseline visit. Ultimately, 23 medical marijuana participants and 21 control participants enrolled and completed the baseline assessment (T1). Following T1, 4 participants in the medical marijuana cohort and 2 participants in the control cohort were lost to follow-up. As a result, 19 participants in each cohort completed the follow-up assessment (T2).

Table 1 presents the descriptive statistics for self-reported sociodemographic variables assessed at baseline (T1) for the entire sample, the medical marijuana group, and the control group, along with test statistics comparing these variables between the two groups. Across the full sample of 44 participants, the mean age was 62.80 (SD 6.62) years, 25 (57%) were male, 41 (93%) were White, and 43 (98%) were non-Hispanic. The median education level was college (4-y degree or university), and the median annual family income was US $30,001 to $60,000. No significant differences were observed between the medical marijuana and the control groups on any sociodemographic variables, confirming that the groups were successfully matched as planned.

Unfortunately, fewer than half of the enrolled participants (18/44, 41%; 12 in the medical marijuana group and 6 in the control group) completed the driving simulator tests at both assessment timepoints. Table 2 presents the detailed reasons for missing data for each group at each assessment time. The primary reason for missing data was simulator sickness, accounting for 22 of 26 cases of missing data at T1 and 19 of 26 at T2. Simulator sickness occurred in 34.8% (8/23) of participants in the marijuana group and 66.7% (14/21) of participants in the control group. This difference did not reach statistical significance (Fisher exact test, odds ratio=3.63, 95% CI 0.92‐15.77; P=.07). Because simulator sickness can range from mild to severe, participants were allowed to decide whether to continue the simulated driving task if symptoms occurred. As a result, 8 participants who experienced mild simulator sickness at T1 still attempted the driving task at T2.

Consequently, some participants contributed partial driving data. Among the 26 participants with missing data, 12 (46%) had partial data at T1 and 5 (19%) had partial data at T2. The valid sample size for the 6 focal DA outcomes (ie, brake response time, reaction time in 4 DA scenarios, and number of responses to DA events) ranged from 12 to 18 across times among the marijuana group and from 7 to 12 across times among the control group (Table 3).

Little test indicated that missing data for these focal variables were missing completely at random (χ²₆₇=64.23, P=.57). Attrition analyses comparing participants with complete data at T2 (n=18) and those with missing data (n=26) showed no significant differences in sociodemographic or mental health variables at T1, except that participants with missing data were more likely to be female (χ²₁=8.73, P<.01). Thus, in analyses for Aim 2, sex was included as a covariate.

Descriptive statistics of the 6 driving test variables are presented in Table 3. Linear mixed effects models were used to examine whether changes in driving performance differed between the medical marijuana and control groups over time, controlling for sex. The group×time interaction was not statistically significant for any driving performance outcome (Table 5). Effect sizes for the interaction terms were small to moderate (|d|=0.02‐0.66), and all CIs included zero, suggesting no clear evidence that initiation of medical marijuana was associated with changes in simulated driving performance relative to controls over the study period.

This study assessed the feasibility of recruiting patients using medical marijuana prior to the initiation of medical marijuana treatment and age-, race-, and sex-matched controls in a driving simulation study to assess the effects of medical marijuana use on real-world outcomes using driving performance. This feasibility study examined participants under real-world conditions to develop a knowledge base needed for guidelines and best practices that currently do not exist for medical marijuana use in this population. The 7-day marijuana use reflected a spectrum from no use to heavy use, with most participants reporting moderate to heavy use over the past 7 days.

Recruitment efforts produced mixed results; the study successfully recruited the majority of the initial a priori sample size proposed but did not fully meet the goal of age-, race-, and sex-matched controls. Age match was achieved by expanding the criteria to ±5 years of medical marijuana group participant age in matching to control participants, as opposed to ±1 year as initially proposed. We used a multiprong approach to recruit participants in both the medical marijuana and control study conditions. Obstacles that reduced the feasibility of this study included the narrow inclusion criteria for control participants (age, race, and sex matching) and COVID-19 restrictions, which limited the access to potential populations of adults aged 50 years and older, who would qualify as control participants. Although these factors reduced the feasibility of recruiting participants, recruitment efforts were successful for the medical marijuana group participants, particularly because of strong stakeholder engagement and our recruitment partnership with the MMTC.

Current research indicates that marijuana users experience cognitive impairment in areas such as psychomotor speed and DA [35-40]. Our work suggests a more nuanced story; however, all interpretations from this data must be done with great caution due to the limited sample size. Specifically, regarding marijuana initiation, we did not find any evidence that initiation of medical marijuana was associated with changes in simulated driving performance relative to controls over the study period. The data also showed that depressive symptoms significantly decreased in the medical marijuana group. Further work is needed to understand whether medical marijuana use creates a long-term impact on psychomotor functioning that affects real-world tasks, such as response time, attention, and executive functions, all of which are critical for driving.

As expected [41], simulator sickness occurred in numerous cases, leading to the majority of missing data in the study. Data show that older females are more susceptible to simulator sickness than their male counterparts [41]—as such, the mitigation of simulator sickness remains a high priority for any future work with driving simulators. Additional emphasis must be placed on ensuring that the participants understand the importance of implementing the mitigation strategies, given that experiencing simulator sickness symptoms was the primary reason the participants did not complete the driving scenarios. Our findings suggest that when using an immersive, full-body cab driving simulator, it is imperative that it provide both sympathetic and parasympathetic feedback. The limitation of our simulator is that it did not provide kinetic feedback for participants, which led to greater instances of simulator sickness. The substantial proportion of missing data attributable to simulator sickness suggests that simulator-based assessments may present challenges for this population and that more realistic on-road driving assessments in line with the national driving test standard may be a more suitable approach in future studies. An open-road driving task is available and enables a more accurate assessment of older drivers in a heterogeneous group [42].

The present feasibility study demonstrated the ability to prospectively recruit the target population, successfully implement the study protocol, and retain participants in this feasibility study. First, many epidemiological studies on marijuana and driving rely primarily on self-reported marijuana use and driving behaviors [35,43]. Our study was able to prospectively track marijuana use by using multiple data sources and a simulated driving task. Second, we focused on medical marijuana use rather than on recreational marijuana use, which is more prevalent among older adults. Driving studies generally focus on recreational marijuana use [44], which is not comparable to medical marijuana due to the lack of regulatory oversight and creates an additional challenge to measuring and tracking use. Third, our study adds data to the literature above and beyond just the presence or absence of marijuana, as medical marijuana group participants drove within 1 month of starting medical marijuana, and use patterns were clearly classified. Studies that report toxicology at crash sites can only report the presence or absence of tetrahydrocannabinol in the system, but they cannot tell us how proximal marijuana use was to the crash, and if the use pattern, a modifiable risk factor, had any effect. The limitations of the study include difficulty recruiting control participants, which has been particularly challenging during the COVID-19 pandemic. Simulator sickness remains an issue to be managed, as this represented the primary driver of missing data, which may introduce potential bias in this small feasibility study. Therefore, the findings regarding the relationship between medical marijuana initiation and changes in driving performance should be interpreted cautiously and considered preliminary at present. The absence of clinical variables such as sleep quality, psychoactive medication use, and comorbidity burden limits the ability to examine their potential role as unmeasured confounders. Finally, while the study is representative of medical marijuana patients in Florida, the generalizability of the results is limited due to the low enrollment of racial or ethnic minority participants. Although Florida’s Office of Medical Marijuana Use does not publish sociodemographic data on its patient registry, current literature suggests that white patients are overrepresented while patients of racial and ethnic minorities are underrepresented in Florida [45,46] and is consistent with medical marijuana research on a national level [45,46]. Potential barriers to diverse recruitment include financial burden not covered by any insurance associated with acquiring and maintaining a medical marijuana card, historic stigma, and disparities in access to and education about medical marijuana care.

This study is the first to prospectively recruit mid-to-late-life medical marijuana users and examine their driving performance in an immersive full-body cab driving simulator under real-world conditions. The next step in this line of research is a study that systematically assesses driving performance in a rigorous and ecologically valid manner, accounting for long-term medical marijuana use in adults 50 years and older who endorse chronic nonmalignant pain. Fitness-to-drive must be measured using an open-road course under the observation of a Certified Driver Rehabilitation Specialist trained to evaluate impaired driving in a test vehicle with the required evaluator controls (ie, eye-check mirrors, auxiliary brake, and auxiliary gas) to maneuver the car and prevent any adverse events, including near misses or crashes, if any dangerous situations arise. As medical marijuana use increases, data are needed to develop risk screening, best practice guidelines, and relevant interventions for this population.