Assessing Sustainable Autonomous Transportation: Veterans’ Perceptions Prior and After Autonomous Shuttle Exposure

1. Introduction

Veterans frequently encounter various healthcare challenges, with transportation being a critical barrier to accessing necessary services. This issue is particularly prevalent among minority groups, individuals with disabilities, and those residing in rural areas, who often report significant difficulties in securing reliable transportation to medical appointments [1,2]. These transportation limitations contribute to a range of negative health outcomes, including missed appointments, delayed care, and poor management of chronic conditions [1,3]. The consequences of inadequate transportation are far-reaching, leading to worsened health conditions and overall reduced quality of life for many Veterans [4]. This problem is exacerbated in rural areas, where long distances and limited public transit options create additional challenges, making it even harder for Veterans to access timely healthcare [1,2].

In response to these challenges, the Department of Veterans Affairs (VA) has implemented various programs designed to support Veterans’ transportation needs. The Veterans Transportation Program (VTP) includes several sub-programs, such as the Veterans Transportation Service (VTS) and the Beneficiary Travel Program, which aim to assist Veterans in reaching VA medical facilities [5]. Despite these efforts, there remain significant gaps in transportation services, particularly for Veterans living in areas where the VTS is unavailable or where wheelchair-accessible vehicles are limited, posing additional obstacles for those with mobility impairments [6]. Furthermore, while legislative efforts such as the CHOICE Act (Veterans Access, Choice, and Accountability Act) and the MISSION Act (Maintaining Internal Systems and Strengthening Integrated Outside Networks Act) have expanded Veterans’ options for receiving care from non-VA providers, navigating eligibility criteria and coordination of transportation with non-VA providers remains a hurdle [7]. These persistent transportation challenges highlight the need for innovative solutions to ensure that Veterans can access the healthcare services they require.

Autonomous vehicle technology, particularly AS, presents a promising and sustainable solution to address transportation challenges faced by Veterans. With a growing population and increasing urbanization, the demand for efficient and sustainable transportation solutions has become more pressing. The US transportation system grapples with challenges such as aging infrastructure, congestion mitigation, inadequate public transit options, and disparities in access to transportation services, which all point to necessitating innovative solutions to ensure universal access to safe and reliable transportation options [8]. Automated vehicle technology is classified by the Society of Automotive Engineers (SAE) into six levels, ranging from Level 0 (no automation) to Level 5 (full automation) [9]. Levels 0-3 require some degree of human control and Levels 4 and 5 allow for full automation in specific (Level 4) or all environments (Level 5). Level 4, which operates autonomously in controlled environments, holds particular promise for enhancing mobility, reducing human error in driving, and improving road safety [10]. Autonomous shuttles, operating at this level, can act as first- or last-mile service providers—connecting passengers to public transportation hubs or to their final destinations—enhancing mobility options and contributing to a more equitable and sustainable transportation system [11,12]. However, challenges related to infrastructure, regulations, and public acceptance must still be addressed for widespread deployment [13].

Previous studies have primarily focused on the civilian population’s acceptance of AV technology [14], with limited research exploring Veterans’ unique needs and perceptions. Veterans face specific challenges related to their military experience, health conditions, and reliance on specialized transportation services [15]. In a previous study conducted by our research team, Veterans in Gainesville, Florida, were exposed to AS, and their perceptions were assessed [16]. The current study expands upon that work, including additional participants from The Villages, Lake Nona, and Port St. Lucie, Florida.

The Autonomous Vehicle User Perception Survey (AVUPS), designed to assess perceptions of individuals exposed to SAE Levels 4 and 5 AVs, was used in this study to measure Veterans’ perceptions before and after exposure to AS technology across these four locations. The AVUPS has undergone content, face, and construct validity assessments, achieving high internal consistency and reliability [17,18]. The survey generates three subscales—Intention to Use, Perceived Barriers, and Well-Being, along with a Total Acceptance score (Mason et al., 2021). The Intention to Use subscale, comprising the most items (13), assesses concepts including trust, perceived usefulness, intention to use, perceived ease of use, cost, authority, and safety. The Perceived Barriers subscale, containing six items, was developed to measure obstacles related to trust, perceived ease of use, intention to use, driving self-efficacy, and safety. The Well-being subscale, which includes four items, assesses how AVs may influence their ability to stay active, participate in their community, and enhance their quality of life. The Total Acceptance score combines responses from all three subscales. By utilizing both retrospective [16] and prospective AVUPS data, this study evaluated changes in Veterans’ perceptions across these subscales and the total score.

This study aimed to uncover insights into the perceptions influencing AS acceptance among Veterans, a population whose unique needs have been underexplored in the context of AV technology. The study assessed whether exposure to AS changed Veterans’ perceptions, specifically their overall acceptance of the technology (i.e., AVUPS-Total Acceptance). Findings from this research offers guidance for VA decision-makers (e.g., VA medical centers), VA transportation stakeholders (e.g., Veterans Transportation Service), and industry partners (e.g., AS manufacturers and service providers), on the next steps for considering AS deployment as part of transportation solutions for Veterans.

2. Materials and Methods

2.1. Ethics

The retrospective and prospective data were obtained as part of two studies, hereafter referred to as parent studies, both of which were submitted for full board review to the ethics committee and received approval from the University of Florida’s Institutional Review Board, the North Florida/South Georgia Veterans Affairs Human Research Protection Office, and the Office of Rural Health. In both studies, participants provided their informed consent by signing IRB- and VA-approved documents, including the Informed Consent Form (ICF) and Health Insurance Portability and Accountability Act (HIPAA) form, and as such confirmed their agreement to participate voluntarily in the research.

2.2. Study Design

The study used existing data from two parent studies [16]. We conducted a secondary data analysis of quantitative survey responses to assess Veterans’ perceptions of AS before and after their exposure to the technology.

2.3. Study Population

In the parent studies, the inclusion criteria encompassed Veterans who had served or were actively serving in the US military (regardless of combat experience), were 18 years or older, proficient in English, resided in Florida, and were able and willing to travel to one of the study locations (i.e., Gainesville, The Villages, Lake Nona, and Port St. Lucie). Veterans were excluded if they had medical conditions preventing them from completing a 25-minute seated shuttle ride.

As part of the parent studies, Veterans were compensated for their participation. Participants were recruited using convenience and snowball sampling strategies, leveraging various community contacts such as Student Veteran organizations, Wounded Warrior Project sites, clinics, and Department of Veterans Affairs centers. Recruitment efforts were supported by a combination of outreach methods, including flyers, online platforms, and targeted outreach through the VA Informatics and Computing Infrastructure (VINCI) database, ensuring a diverse and representative sample of Veterans.

This current study used the quantitative AVUPS data collected from these participants, with only Veterans who completed both pre- and post-AVUPS surveys being included in the analysis.

2.3.1. Sample Size

In the parent studies, the sample size for survey data was determined through an a priori power analysis using G*Power 3 software [19]. With an alpha level of 0.05, a power of 0.80 for matched pairs with two measures (pre- and post-AS exposure), and a medium effect size of Cohen’s d = 0.53, a sample size of 30 participants was identified. The effect size was estimated based on previous interim findings (i.e., AVUPS-Total Acceptance) comparing older adults’ perceptions via the AVUPS pre- and post-AS exposure [20]. Therefore, this study aimed to include a minimum of 30 participants from the parent studies to maintain 0.80 power for detecting a pre- and post-AVUPS difference in the AVUPS-Total Acceptance score.

2.4. Procedure – Data Source and Description

The research team compiled a comprehensive dataset from the parent studies conducted between 2021 and 2023 [16]. This dataset contains quantitative data from pre- and post-AVUPS assessments. Eligibility criteria were applied to ensure the inclusion of individuals with complete data. In the parent studies, the collection of quantitative data, took 1.5 hours to be completed. Participants completed various study-related documents, including the ICF, HIPAA, payment form, a demographic/medical survey, and additional questionnaires such as the AVUPS and The Motion Sickness Assessment Questionnaire (MSAQ) [21]. The MSAQ was included as a safety precaution to monitor participants’ susceptibility to motion sickness during the study, ensuring that any discomfort experienced while using the AVs could be appropriately managed. Following completion of paperwork, participants experienced the AS ride and again completed the AVUPS and the MSAQ. The shuttle ride duration averaged approximately 25 minutes, except in The Villages, with a 10-minute route in accordance with approvals by the National Highway Traffic Safety Administration. Despite the difference in the duration of the route, The Villages’ route ensured comparable exposure and experience, incorporating busy roads, stop signs, and pedestrian interactions like the other study locations.

The parent studies used SAE Level 4 AS, utilizing cutting-edge technology such as Light Detection and Ranging (LIDAR) sensors, cameras, radar sensors, and a Global Positioning System (GPS) to accurately map the surrounding environment and determine optimal motion behavior in real time. Designed to operate autonomously on specific pre-mapped routes without manual control, these shuttles lack a steering wheel but feature a remote-control interface, such as a joystick, enabling manual operation by a safety operator if necessary. The AS does not exceed 10 miles per hour and may accommodate up to 12 individuals, with six to eight seated and four to six standing, depending on the model (EasyMile or Navya). Wheelchair accessibility makes it suitable for individuals with mobility impairments. In the parent studies, participants were instructed to remain seated during the ride, and, in addition to participants, a safety operator and a research assistant were on board.

2.5. Data Collection

In this study, existing data from the parent studies was used to assess Veterans’ perceptions before and after exposure to an AS. The FY21-22 parent study focused on AS perceptions in Gainesville, FL, while the FY22-23 parent study expanded its scope to three additional locations in Florida—The Villages, Lake Nona, and Port St. Lucie—to better ascertain statewide Veteran AS perceptions. Data from The Villages was obtained as secondary data from an IRB-approved Florida Department of Transportation study (IRB202101677). Both parent studies followed the same research protocol.

In the parent studies, participants completed various questionnaires, including the demographic questionnaire, Montreal Cognitive Assessment (MoCA) [22], Military to Civilian Questionnaire [23], Technology Readiness Index 2.0 [24], Technology Acceptance Model [25], AVUPS [17,18], and the MSAQ [21]. The demographic questionnaire, MoCA, and AVUPS were used for data analysis. The other questionnaires, such as the Military to Civilian Questionnaire, Technology Readiness Index 2.0, and Technology Acceptance Model, were included to gather broader insights but were not analyzed in this context. Additionally, the MSAQ was used as a precautionary measure rather than for analytical purposes.

In this study, the independent variables included MoCA scores and demographics (age, gender, rurality, marital status, military branch). Notably, rurality is determined by the VA using Rural-Urban Commuting Area (RUCA) codes, which consider population density and commuting patterns to provide detailed insights into the distance of rural communities from essential health care services in more populated areas [4]. Participants from The Villages completed a different demographic questionnaire, gathering information on age, gender, rurality, race/ethnicity, education, employment, and health-related impairment. This incongruence in demographic forms occurred because the correct questionnaire was inadvertently omitted from the data collection method for that site, as the data from The Villages was originally collected as part of a separate study. This issue is further addressed in the analysis and results sections of this paper. The dependent variable in this study was Veterans’ perception of AS technology, measured using the AVUPS pre- and post-shuttle exposure.

The MoCA is a widely used cognitive screening tool for detecting mild cognitive impairment (MCI) across various domains, including attention, concentration, memory, language, visuospatial skills, and executive functions [22]. The psychometric properties of MoCA have been extensively studied, and it is a reliable and valid tool for detecting MCI in various populations, including Veterans [26]. Although all participants in the parent studies were included regardless of MoCA scores, this data was collected for transparency and to ensure that the results of the study are not confounded by cognitive deficits.

The AVUPS, a 32-item questionnaire, includes a 28-item visual analog scale and four open-ended questions [17,18]. This study focused on the visual analog scale data. The visual analog scale is a 100 mm horizontal line ranging from disagree to agree. The AVUPS has strong psychometric properties, with excellent internal consistency (α = 0.95), test-retest reliability (ρ = 0.76, ICC = 0.95), and a strong Mokken scale (Hscale = 0.51). While the survey’s visual analog scale consists of 28 items, for the analysis in this paper, and as recommended by the Mokken scale analyses [18], only 20 out of the 28 items are used to assess Total Acceptance and the three subscales. The AVUPS is a reliable and valid tool, with its subscales offering valuable insights into the factors influencing user acceptance of AVs [17,18].

2.6. Data Management

The research team employed strategies to support the security and confidentiality of the data that were obtained from the parent studies. For example, participant responses to questionnaires were securely stored in password-protected systems or locked cabinets within a secure VA research office, adhering to VA and University information security policies. Furthermore, the existing data used for analysis were de-identified to protect the privacy of the participants.

2.7. Data Analysis

The quantitative data, obtained from the parent studies, were analyzed in RStudio using R 4.3.1 [27]. Descriptive statistics were conducted for all study participants on the overlapping demographic variables: age, gender, and rurality. As The Villages used a different demographic questionnaire, and considering the limited count of overlapping variables, a detailed description of demographic variables was additionally provided for each study location. The descriptive statistics by study location included the following demographic variables: age, gender, race/ethnicity, education, rurality, marital status, military branch, employment, health-related impairment, and MoCA scores. Categorical data were presented as count (n) and proportion (%), while continuous data were displayed as mean (M) and standard deviation (SD) or median and interquartile range (IQR) if analysis assumptions were violated.

The current study is powered to assess the pre- and post-AS differences in AVUPS-Total Acceptance which is the primary outcome variable. Secondary analyses were conducted to explore pre- and post-differences in the AVUPS subscales (Intention to Use, Perceived Barriers, and Well-being) for all participants, as well as for rural vs. urban Veterans. Additionally, a subset analysis was conducted on a group of participants (n=30) exhibiting the lowest Total Acceptance scores, assessing pre- and post-AVUPS data differences. Those with the lowest Total Acceptance scores are the most resistant to the technology or least accepting and thus has the most to gain from riding an AS. Lastly, the research team conducted an analysis investigating the specific survey items contributing to each of the three AVUPS subscales. To control for false discovery rate (q < .05), the research team used the “p.adjust” function in R with the Benjamini-Hochberg method, which is a multiple testing correction method that adjusts the p-values for each test based on the number of tests being conducted [28].

The primary outcome variable, Total Acceptance of AS technology among Veterans (pre- and post-AVUPS data), was assessed for normality via visual analysis (i.e., histograms, Q-Q plot, boxplots) and statistical tests (i.e., Levene’s test, skewness and kurtosis indices, and Shapiro-Wilk’s test). Since the normality assumptions were violated for the AVUPS data, a series of Wilcoxon rank-sum tests were used to assess within-group differences for the three AVUPS subscales and the Total Acceptance score (p ≤ 0.05). The research team assessed within-group differences for all study participants and further stratified the analysis to compare rural and urban participants, aiming to explore potential influences of rurality on Veterans’ perceptions of AVs. A subset analysis (n=30) was also performed to investigate within-group differences among participants with the lowest Total Acceptance scores. This analysis specifically aimed to discern pre- and post-AS exposure variations among individuals with lower AS Acceptance scores, excluding those already expressing a high level of acceptance. To control for false discovery rate (q < .05), the research team used the “p.adjust” function in R with the Benjamini-Hochberg method, which is a multiple testing correction method that adjusts the p-values for each test based on the number of tests being conducted [28].

As previously indicated, the AVUPS data were scored into three Mokken subscales (i.e., Intention to Use, Perceived Barriers, and Well-being) and the total score, which represents the overall Acceptance of AS technology score [18]. Specifically, the Mokken subscales were computed using the rowMeans function in R. This function averaged scores across AVUPS items to calculate both the subscale scores and the total score. The Intention to Use mean was derived from 13 items, Perceived Barriers mean from six items, Well-being mean from four items, and the overall Acceptance score mean from 20 items. To identify specific AVUPS items that significantly contributed to a particular subscale, the research team conducted a series of Wilcoxon rank-sum tests, to assess the within-group differences for each AVUPS item within each subscale.

3. Results

3.1. Demographics

Table 1 displays demographic characteristics for all study participants and by study location at baseline. Out of the 113 individuals screened, a total of 77 participants completed Phase I of the study. The AVUPS was administered to Veterans before and after exposure to the AS in The Villages (n=39; 50.7%), Gainesville (n=23; 29.9%), Lake Nona (n=13; 16.9%), and Port St. Lucie (n=2; 2.6%), Florida. The study participants included 66 males and 11 females. The majority of Veterans were between 24 to 64 years of age (n = 41; 53%), with the mean participant age of 60.6 (SD=16) years. In terms of rurality distribution, most participants resided in urban areas (n = 63; 83%). The Villages’ participants completed a different demographic questionnaire collecting data on different descriptive variables. Table 1 provides the demographic characteristics for participants by study location at baseline. Participants from The Villages, an adult retirement community, were older (M=70.9; SD=8.30), while those from Port St. Lucie, recruited from the general population, were younger (M=39.0; SD=4.24). At each study location, the majority were male (>82%) and lived in urban areas (>76%), except for Port St. Lucie, where two participants evenly represented both genders and rurality (50%). In The Villages, the majority were White (92%), highly educated (67% with at least a Bachelor’s degree), retired (74%), and free from impairments impacting their use of transportation (90%). In Gainesville, most reported divorced marital status (39%), army affiliation (44%), and an average MoCA score of 25 ± 2.8. Lake Nona and Port St. Lucie participants reported mostly married status (>61%) and army affiliation (>53%). While Lake Nona participants had an average MoCA score of 25 ± 3.2, those in Port St. Lucie scored an average of 29 ± 0.0, indicating better cognitive functioning than the comparison group.

3.2. The Four AVUPS Scores

Table 2 displays the within-group comparisons of the four AVUPS domains (i.e., Intention to Use, Perceived Barriers, Well-being, and Total Acceptance) between baseline and post-exposure to the AS for all study participants (N=77) and a subset of participants (n=30) with the lowest Total Acceptance scores. At baseline, the majority of participants displayed scores exceeding 65 (out of 100) on the AVUPS for the primary outcome variable, i.e., Total Acceptance. Additionally, most participants scored 69 (out of 100) or higher on the AVUPS in the domains of Intention to Use and Well-being, while the Perceived Barriers domain median score was 35 (out of 100) on the AVUPS. Following exposure to the AS, statistically significant differences (improvements in perceptions) were observed in all domains, except for Well-being. Specifically, Intention to Use and Total Acceptance displayed a statistically significant increase between pre- and post-AS exposure, while Perceived Barriers showed a significant decrease. Descriptively, Well-being showed an increase post-exposure, although this change did not reach statistical significance. In the subset analysis, a similar pattern emerged, with all domains displaying statistically significant differences between pre- and post-AS exposure, except for the Well-being scale. Notably, the subset group exhibited a larger magnitude in the differences between pre- and post-AS exposure, indicating a more substantial increase in Intention to Use and Total Acceptance, as well as a more pronounced decrease in Perceived Barriers among participants with lower AS Acceptance scores at baseline compared to the overall study population.

Table 3 displays the within-group comparisons for the four AVUPS domains between baseline and post-AS exposure for urban (n=63) and rural Veterans (n=13). Descriptively, urban Veterans consistently displayed higher median scores for Intention to Use, Well-being, and Total Acceptance both at baseline and post-AS exposure, compared to their rural counterparts. Conversely, rural Veterans exhibited higher median scores for Perceived Barriers, when compared to urban Veterans at both baseline and post-AS exposure. Urban Veterans demonstrated a statistically significant increase post-AS exposure in Intention to Use and Total Acceptance, and a decrease in Perceived Barriers. For rural Veterans, a statistical significance was observed only for Perceived Barriers, showing a decrease post-AS exposure which indicates more positive perceptions. The Well-being domain did not exhibit statistically significant differences for either the urban or rural Veterans.

Table 4 presents the within-group differences in AVUPS items comprising the three AVUPS subscales—Intention to Use, Perceived Barriers, and Well-being—for all study participants. Within the Intention to Use domain, 4 out of 13 items showed statistically significant differences from pre- to post-AS exposure. Specifically, participants displayed a significant increase in trust (p = .005), willingness to engage in other tasks while in an AV (p = .004), heightened perception of safety for other road users (p = .048), and an elevated sense of safety while riding the AV (p < .001). However, after controlling for multiple comparisons, only the item related to positive feelings of safety while riding in an AV (AVUPS item #27, see Table 4) remained statistically significant (p = .006). In the Perceived Barriers domain, two out of six items showed significant decreases from pre- to post-AS exposure. Participants demonstrated a reduction in perceived barriers post-AS exposure, particularly related to concerns about the potential decline in individual driving abilities (p = .004) and hesitations about AV use (p < .001). These two items within the Perceived Barriers domain retained their significance after adjusting for multiple comparisons. Within the Well-being domain, significant mean differences were not detected for any of the four AVUPS items.

4. Discussion

This study assessed the perceptions of Veterans (N=77) prior to and after exposure to the EasyMile (EZ10) or Navya AS, using existing data from two studies. Specifically, using quantitative methodologies, the research team explored if exposure to an AS changed Veterans’ perception of such technology when comparing pre-exposure to post-exposure.

4.1. Demographics

The study population consisted primarily of middle-aged to older male Veterans, consistent with the broader US Veteran demographic [3]. Participants from The Villages were notably older (M=70.90; SD=8.30) compared to participants in the other study locations. This was not surprising, as The Villages is the largest retirement community in the US and houses older adults (over the age of 55) who are seeking an active lifestyle [29]. In contrast, participants from Lake Nona were younger (M=42.1; SD=9.39). This reflects the area’s focus on productive engagement in industries such as medical research, education, and technology [30]. This age diversity supports the study’s objective to capture varied perspectives on AS from Veterans of various ages.

Although the majority of participants lived in urban areas (83%), about 25% of US Veterans reside in remote (i.e., residing more than 60 minutes from the nearest VA) or highly rural locations (i.e., fewer than seven persons per square mile) [4]. underrepresentation of rural Veterans may be implicated by the associated recruitment challenges. Data were collected from four Florida locations to provide insights into the Veterans’ acceptance of AS. However, the study’s narrow geography limits generalization to communities outside of those that were investigated. As such a need exists for larger studies to examine a greater number of diverse samples, across counties and states, to support knowledge generation for AS development and deployment.

Participants from three of the study locations—Gainesville, Lake Nona, and Port St. Lucie (n=38; 49%)—were primarily Army Veterans, reflecting general US Veteran population trends [3]. Most were married, consistent with national statistics for Veterans [3]. The average MoCA score indicated mild signs of cognitive limitations, consistent with older age and participants’ history of military exposure. In The Villages (n=39; 51%), the majority of participants were White, highly educated, and retired, reflecting the population statistics of The Villages (US Census Bureau, 2022a). Discrepancies in demographic data across locations, particularly in The Villages, arose due to a different demographic questionnaire being used in The Villages.

4.2. The Four AVUPS Scores

The study’s initial assessment of all participants revealed generally favorable perceptions toward AS at baseline, reflected in scores above 65 (out of 100) for the domains of Intention to Use, Well-being, and Total Acceptance. Conversely, scores below 35 (out of 100) in the Perceived Barriers domain indicated the relatively few barriers were perceived. This descriptive data indicate that the majority of participants already held a positive view of the AS even before riding the shuttle. This pre-existing positive orientation might be attributed to self-selection bias, as individuals opposed to the implementation of AVs might have been less likely to participate in the study.

The statistically significant change in AVUPS-Total Acceptance score indicated that exposure to the AS positively influenced participants’ overall acceptance of the technology. Additionally, a significant increase was observed for AVUPS-Intention to Use, particularly in enhanced safety. These findings align with the existing research, emphasizing the role of safety in fostering positive perceptions toward AVs [31]. However, one item under AVUPS-Intention to Use, specifically “I am open to the idea of using automated vehicles” (item #4), did not reach statistical significance, despite the overall domain AVUPS-Intention to Use showing a statistically significant improvement. This suggests the need for caution in interpreting the results. It’s essential to note that acceptance of the technology, as measured by AVUPS-Total Acceptance, doesn’t directly translate to the participants’ intention to use the technology in the future. Acceptance signifies positive attitudes or beliefs towards AVs, but intention to use reflects actual willingness to adopt AS. Therefore, while participants may accept the technology, it doesn’t guarantee their intention to use it in the future. As such, research is necessary to examine the actual adoption vs. just acceptance of the AS technology.

The AS exposure also reduced Perceived Barriers, particularly concerns about the potential decline in individual driving abilities and hesitations about AV use. These findings suggest that the practical experience of interacting with AV technology, through the shuttle ride, positively influenced participants’ perceptions pertaining to barriers and made them more receptive to AS as a viable transportation option. Additionally, these findings are consistent with prior research that explored AS users’ perceptions and attitudes, which also reported an increase in positive perceptions and acceptance of AS after experiencing the technology firsthand [16,32,33]. Although, Intention to Use, Perceived Barriers and Total Acceptance domains showed a significant difference between pre- and post-AS exposure, the actual magnitudes of these differences were relatively small. Specifically, the changes did not exceed six millimeters on the AVUPS scale, which is measured on a 0-100 mm horizontal analog scale ranging from disagree to agree. However, specific items within these domains (AVUPS-Intention to Use items #6, 7, 25, and 27; and AVUPS-Perceived Barriers items #19 and 28), as demonstrated in Table 4, displayed substantial significant differences, ranging from 5 to 21 millimeters. This suggests that these specific items might serve as more sensitive indicators for capturing nuanced changes in perceptions compared to the broader domains.

An exploratory analysis was conducted to gain deeper insights into participants’ responses, specifically by assessing a subset with the lowest Total Acceptance scores. This analysis aimed to uncover whether real-world exposure had a more pronounced effect on individuals initially less accepting of AS technology. In the subset analysis of participants (n=30) with the lowest Total Acceptance scores, a similar pattern emerged, with all domains displaying statistically significant differences post-AS exposure, except for Well-being. Notably, the subset group exhibited a larger magnitude in the differences between pre- and post-AS exposure, indicating a more substantial increase in Intention to Use and Total Acceptance, and a more pronounced decrease in Perceived Barriers. Although the magnitude of these differences remained modest (descriptively, 5 to 10 millimeters), the impact of real-world exposure to AS appears more meaningful among individuals with initially lower acceptance of AVs. Understanding the differential impact of real-world exposure on varying levels of acceptance is key for designing tailored strategies to enhance public acceptance. It highlights the potential of exposure to enhance some perceptions, particularly among individuals with lower initial acceptance levels.

Urban Veterans (compared to rural Veterans) consistently exhibited higher median scores for Intention to Use, Well-being, and Total Acceptance, both before and after AS exposure, indicating a more positive orientation toward AS. Conversely, rural Veterans consistently reported higher median scores for Perceived Barriers, indicating that they perceive more obstacles to the adoption of AS technology compared to their urban counterparts. These findings are consistent with broader literature on rural-urban differences in transportation access, highlighting the unique challenges faced by rural populations, such as limited infrastructure, longer travel distances, and potentially less exposure to emerging technologies [34,35]. The statistically significant increase in Intention to Use and Total Acceptance scores among urban Veterans post-AS exposure is consistent with the overall study population (N=77), indicating a positive impact of direct experience with AS on their perceptions and willingness to accept this technology. However, the actual magnitude of these differences was small, warranting cautious interpretation of their practical implications, as further discussed in the implications section. Interestingly, rural Veterans (n=13) exhibited a statistically significant decrease in Perceived Barriers post-AS exposure, despite the small sample size. However, none of the other domains—Intention to Use, Well-being, and Total Acceptance—reached statistical significance for rural Veterans. The study was underpowered to detect smaller, yet potentially significant differences in secondary analyses, which may have the potential to increase the presence of Type II errors.

The Well-being domain did not exhibit statistical significance for any of the statistical analyses, including the AVUPS items analyses (four items). A previous study also measuring pre- and post-AS perception differences using the AVUPS similarly found no significant changes in Well-being [33]. The type of exposure provided in the study is not consistent to how users would eventually use the technology. For example, if participants were exposed to a real scenario where the shuttle takes them to the VA hospital or somewhere meaningful, their perception might be very different than tested during this trial. The perceived benefit of AS transportation is partially based on need, and if individuals don’t currently have a specific need, they might not see the absolute value in the actual service. Thus, riders may require more exposure to AVs and AS to recognize potential benefits to their well-being—such as incorporating the shuttle into their daily commute or transportation to and from medical appointments. On the other hand, the Well-being scale consistently underperformed in conducted studies where it was utilized [16,33,36]. This may indicate that the Well-being scale is not adequately sensitive to measure perceived differences in well-being as a result of AS exposure. Moreover, the lack of statistical significance in Well-being may also result from the application of the AVUPS in this study’s context. While the AVUPS aims to discern perceptions of AVs as a broader category, this study specifically assesses the effect of AS exposure. Thus, using the AVUPS to measure changes due to AS exposure may lack the specificity needed for nuanced responses related to AS. In fact, post-exposure to AS, only 3 out of 23 AVUPS items (across all domains) exhibited statistically significant differences, after controlling for multiple comparisons (adjusted p-value). Although the study was not specifically powered for this exploratory analysis, it indicates a potential limitation in using AVUPS domains to assess changes in users’ perceptions after AS exposure. Relying on AVUPS domains to assess change after AS-exposure, rather than specific items within each domain, may obscure larger differences and practical implications. Some items may not effectively capture the intended meaning of the lived experience when specifically assessing the effect of AS exposure. In turn, analyzing the AVUPS domains (instead of the individual items) could have resulted in diluted findings. Given the limited number and small magnitude of these differences, additional research specifically focusing on the context of AS transportation, with a measure specifically developed to capture the intended factors influencing (inhibiting or promoting) AS acceptance, is needed.

With a sample size of 77 participants, the study was adequately powered (power of 0.80 for matched pairs with two measures, pre- and post-AS exposure survey, and a medium effect size of Cohen’s d = 0.53) to detect a significant difference in Total Acceptance, the primary outcome variable, among all participants. Although the study did show a statistical significance in Intention to Use and Perceived Barriers, which is consistent with prior research [16,32,33], results need to be interpreted with caution as the study was not specifically powered for these analyses. Additional secondary and exploratory analyses should also be interpreted with caution, even though the Benjamini-Hochberg method was applied to control for multiple comparisons [28].

5. Limitations

The quasi-experimental design and convenience sampling of the parent studies may have led to selection and spectrum bias, where participants with pre-existing positive views of AVs are overrepresented, that may have had the effect of potentially skewing the study’s outcomes. The lack of consistency in demographic data collection across study locations prevented an exploration of how socio-economic indicators, e.g., income variations, may have affected the results. Higher-income areas like Lake Nona and The Villages may have had higher expectations and, consequently, reported more barriers compared to middle- to lower-income areas like Gainesville, which had the highest perceptions of AV benefits and Intention to Use. Additionally, the Lake Nona shuttle had been operational for a longer period compared to other locations, potentially influencing participants’ perceptions due to prolonged exposure. In contrast, areas with shorter exposure periods, such as The Villages, may not have developed the same level of familiarity and acceptance, which could have impacted their responses. Participants in The Villages also had a shorter AS ride (10 minutes versus 25 minutes in the other study locations), further limiting their exposure to the AS. This limited exposure could have negatively influenced their overall perceptions and acceptance.

Additionally, the technology is still in its developmental phases, and the research team in the parent studies encountered numerous field-related challenges during the study execution, well discussed in Report #: BED31-977-26 [37]. For instance, the shuttles were limited by weather conditions, functioning only in light rain, while heavy rain halted operations. During summer, air conditioning drained battery power, requiring temporary suspensions for recharging. Technical issues, such as shuttle reboots, further disrupted operations, leading to participant rescheduling and study delays. Additionally, the shuttles only operated during daytime hours, posing further scheduling limitations. These operational challenges may have negatively influenced participants’ perceptions of AS technology. However, such challenges are inherent in the ongoing development and improvement of AS technology. Although the shuttle offers a first mile–last mile option, it operates on a fixed route and does not fully meet all requirements of the American with Disabilities Act (ADA), a law designed to ensure equal access for individuals with disabilities. For example, the shuttle lacks an automatic ramp for boarding and exiting, which may limit accessibility for Veterans, particularly those who use assistive mobility devices.

Treating AVUPS items as continuous variables without distinguishing between low and high Intention to Use creates artificial hierarchies and limits interpretation. It assumes small differences in scores are meaningful and that all items contribute equally to the overall score, which may not be the case. For example, items like “I am interested in using AVs” and “I am comfortable with AVs operating at night” may reflect different aspects of Intention to Use and treating them as part of a single scale assumes their scores are directly comparable, which can lead to misinterpretations. More nuanced methods, such as item response theory, could provide better insights into these differences and improve the accuracy of findings.

Although the study was adequately powered to detect differences in AVUPS-Total Acceptance, caution is warranted in interpreting the results of the secondary analyses due to the lack of specific power for these analyses. Moreover, the small sample size and underrepresentation of rural Veterans for secondary analyses could lead to Type 2 errors, where an effect may exist but goes undetected due to the limited sample size. Although statistically significant differences were observed in pre- and post-AVUPS Total Acceptance, their practical implications may be limited due to the relatively small magnitude of this difference. Additionally, the lack of statistical significance in the AVUPS-Well-being domain may stem from the survey’s broader focus, underscoring the need for a nuanced assessment of AS-related perceptions.

6. Strengths

Technology-based interventions, such as integrating AS into community mobility options, hold promise for Veterans, particularly those unable or unwilling to drive. This study represents one of the pioneering efforts to explore Veterans’ perceptions of and potential acceptance of AS technology in the United States. Conducted across four different locations in Florida, the study’s geographic breadth enhances its applicability to different regional contexts, strengthening its generalizability. By embracing state-of-the-art technology and fostering collaboration with transportation stakeholders and industry partners, this research offers a unique insight into Veterans’ perspectives on AS.

7. Implications

This study offers significant implications for AS acceptance among Veterans, with applications across research, public health, practice, and sustainability. In research, the modest magnitude of pre- and post-exposure changes in Total Acceptance, Intention to Use, and Perceived Barriers suggests that more sensitive, tailored instruments are needed to better capture nuanced shifts in perception. These improved measures could provide more accurate data for informing the design and deployment of AS programs, particularly for populations that initially exhibit lower acceptance. For public health, this study highlights how AS technology could address significant transportation barriers faced by Veterans, particularly those in rural areas. Improved access to healthcare services and community activities through reliable AS transportation may enhance community integration, reduce isolation, and increase access to healthcare services. Additionally, by decreasing reliance on human drivers, AS technology may reduce transportation-related crashes by minimizing human error and improving overall safety for Veterans and the broader public. In practice, for industry stakeholders, this study indicates the importance of addressing user concerns—especially for older Veterans who expressed more safety concerns. The inclusion of safety operators onboard AS vehicles may help build trust and confidence. The findings also highlight the need for ongoing collaboration with industry partners to address technical and operational challenges, ensuring that AS technology is user-friendly, reliable, and accessible.

Regarding sustainability, one significant benefit is the reduction in greenhouse gas emissions, as these shuttles are often electric powered, reducing dependence on fossil fuels, especially when electricity is sourced from renewable energy. Autonomous shuttles also improve transportation efficiency by operating on optimized routes and schedules, minimizing energy waste associated with idling, congestion, and inefficient driving behaviors. Additionally, they encourage multi-modal transportation by serving as first-mile/last-mile solutions, enhancing access to public transit systems and reducing reliance on private vehicles, which can lower overall vehicle kilometers traveled. Their small size and shared use contribute to decreasing road congestion, further reducing emissions from stop-and-go traffic. Moreover, AS align with sustainable urban planning goals by enabling pedestrian-friendly and low-emission zones, thus reducing the spatial and environmental footprint of transportation. However, to maximize these benefits, challenges such as lifecycle emissions from vehicle production and equitable access must be addressed. Thus, when integrated with policies promoting renewable energy and shared mobility, AS have potential to advance sustainability goals.

8. Conclusion

The study assessed Veterans’ perceptions post-AS exposure across four locations in Florida, leveraging quantitative data from two parent studies. Overall, participants exhibited positive perceptions of AS, with significant increases in Total Acceptance and Intention to Use following AS exposure, particularly related to safety perceptions. Perceived Barriers decreased, especially concerns about the potential decline in their own driving ability and hesitations about using AVs. The Well-being domain did not show significant changes, suggesting that longer exposure or more practical use scenarios may be needed to impact well-being perceptions. The exploratory analysis of participants with the lowest baseline Total Acceptance scores revealed larger improvements post-exposure, particularly in Intention to Use and Total Acceptance, and a more pronounced reduction in Perceived Barriers. These findings suggest that AS exposure may be particularly impactful for individuals with lower initial acceptance levels. While urban Veterans demonstrated more favorable perceptions compared to rural Veterans, both groups benefited from the real-world AS exposure. However, the small magnitude of the observed changes suggests the need for further research. Continuous collaboration with industry partners, further field testing, and overcoming AS technical and operational challenges (e.g., AS weather tolerance) are necessary steps in paving the way for considering AS services for Veterans. VA decision-makers (e.g., VA medical centers), VA transportation stakeholders (e.g., Veterans Transportation Service), and industry partners (e.g., AS manufacturers) can leverage these findings to educate, advocate, and address concerns identified in the study’s findings. Overall, the study findings point to the potential for considering AS as a viable, sustainable, and accessible transportation option for Veterans. However, second-generation shuttles—with increased seating capacity, speed and accessibility—along with more flexible route options and expanded operating hours (day and night, seven days a week) must be considered to provide ubiquitous access and benefits to Veterans as they use these technologies.