A cross-sectional study using a convenience sample of adult participants was used to survey the use and perceptions of DVAs from patients at Michigan Medicine, the academic medical center affiliated with the University of Michigan. Patients from 3 purposely sampled clinics in a single city were recruited. The clinics included a family medicine (Briarwood Family Medicine Center), low vision (Kellogg Eye Center Low Vision and Rehabilitation Clinic), and physical medicine & rehabilitation (PM&R) clinic at Michigan Medicine. The family medicine clinic was selected to reflect a typical primary care setting with a range of patients. The low vision clinic was selected for its access to patients with visual disabilities. Finally, the PM&R clinic was selected to enrich the sample for individuals who had physical-related disabilities. Potential participants younger than 18 years were excluded from participation.

Potential participants at the Kellogg and PM&R clinics were informed of the study by clinic staff members during check in. If interested, participants were given a paper survey to complete. Completed surveys were returned to clinic staff members. In addition, participants were recruited at the monthly low-vision support group offered at the Kellogg Eye Center. At the family medicine clinic, study staff members were posted in the lobby and informed potential participants of the study. Those interested in participating were handed a survey, which was returned to study staff at completion. Recruitment occurred over 19 months from September 2019 to March 2021, with recruitment delays experienced due to the COVID-19 pandemic.

At each of the 3 clinic sites, an information sheet explaining the purpose of the survey, as well as a self-administered paper copy of the survey, were given to potential participants by clinic or study staff members. Participants then had the option to fill out the survey at any point during their clinic visit and return the paper survey to the clinic or study staff members upon completion. The assessment survey was self-administered by the respondent; however, in cases where assistance was needed for survey completion, participants could select a friend or caregiver accompanying them for assistance. In the case of participants recruited from the low vision clinic, clinic staff members were available to help respondents complete the survey.

The study survey used was a 46-item investigator-developed paper-based instrument that surveyed participants about their knowledge, use, and perceptions of DVAs, their interest to participate in related future studies, and demographic questions (Multimedia Appendix 1). Questions regarding knowledge and use of DVAs were presented in 6 closed, multiple-choice format questions with some including an “other” option for further elaboration. There was one additional open-ended question that asked participants to give examples of potential use cases of DVAs that aligned with their interests. Participants were also asked to rate their concerns about 5 specific issues related to the use of DVAs, and one opportunity to voice their own concern. These ratings used a 5-point Likert scale (0=not concerned at all; 4=highly concerned) and were used to measure concerns relating to privacy, security, confidentiality, accuracy, and reliability of the DVAs. The questions about privacy, security, and confidentiality included definitions of the respective terms to clearly define the specific concern. “Privacy” referred to freedom to opt out of monitoring and to determine how personal data is used. “Security” referred to protection of personal data from loss or theft. “Confidentiality” was defined as the obligation of trusted persons to keep data safe and held in confidence. To assess participants’ interest in participating in future studies relating to DVAs, 3 closed multiple choice questions were asked.

The next set of questions presented assessed whether participants had a visual disability. If participants endorsed having a visual disability, 9 additional questions were presented asking about the nature of their disability, access to care, and access to resources. These questions were presented in mixed format. Participants who did not have visual disabilities were asked to skip to the next section, which was focused on physical disability.

The physical disability portion of the survey asked questions generically about disability, which allowed participants to self-identify and describe the nature of their disability. Those endorsing no physical disability were asked to skip to the final set of questions on demographics; whereas those who did endorse a physical disability were asked mixed format questions about the nature of their disability, access to care, and access to resources. As participants were allowed to self-identify with having a physical disability, this study included a wide range of physical disabilities and documented disability within the electronic health record was not required by any participants; rather, participants were asked to endorse whether they had any kind of physical disability and if so, questions probed into the nature of their self-reported disability.

Given that participants with visual disabilities were expected to have more difficulty taking the survey and would require assistance, a shortened 34-item survey presented in large print was administered to participants recruited from the Kellogg low vision clinic (Multimedia Appendix 2). The 34-item survey contained the same questions as the aforementioned 46-item instrument, except it excluded the question set focused on physical disabilities. Both versions of the survey were designed to be completely anonymous, with no protected health information or HIPAA (Health Insurance Portability and Accountability Act)-protected data collected. See Multimedia Appendices 1 and 2 for the survey instrument.

Participants gave their completed surveys to study or clinic staff members, who then placed the surveys into a box clearly labeled with the study title. After the collection of the completed surveys, the data were entered into a secure web application REDCap (Research Electronic Data Capture; Vanderbilt University) by approved study personnel. The entered data were verified through a 10% random sampling for entry errors. Incomplete survey responses were also retained and collected during this study.

The survey data were analyzed using descriptive statistics including frequencies, means, and SD. To determine differences in responses between subgroups, bivariate associations between subgroups and demographics, DVA use, and DVA concerns were performed using chi-square tests or Fisher exact tests as appropriate depending on the numbers in each group. Differences were examined by previous DVA use, clinic site, visual disability status, physical disability status, and willingness to participate in future research studies. All analyses were performed using SAS Software (version 9.4; SAS Institute Inc).

Our study was registered under a self-exemption process at the University of Michigan, classified under EXEMPTION 2 (i) and 2 (ii) as described in 45 CFR 46.104(d)(IRBMED HUM00156811). Participants were handed an information sheet outlining the risks and benefits of the study before completing the survey. All study materials were stored in locked file cabinets in the locked office of the principal investigator. All data were collected anonymously and reported in the aggregate, and participant responses cannot be linked to individual participants. No financial or nonfinancial incentives were offered for study participation.

A total of 218 eligible participants completed the survey. The majority of the sample was White (71.6%, 156/218) and female (61.7%, 124/201) with an average age of 52.6 (SD 18.9) years. The education levels of the participants varied, with the largest portion having a graduate degree (35.0%, 71/203), followed by a split between those who have completed a bachelor’s degree (20.2%, 41/203) and those who have completed some college education (19.7%, 40/203). Participants’ employment status was primarily full-time (40.8%, 82/201) followed by retired (29.9%, 60/201). Roughly half of the participant population earned an annual income of US $75,000 or higher (46.3%, 88/190) and was insured through their employer (54.1%, 118/218). Almost half of the participant population (49.5%, 101/204) indicated interest in participating in an additional research study that provides DVAs for use (Table 1).

The results of the survey showed that DVAs are being used for various tasks and there is a willingness to use them in the future. The majority of all participants had previously used a DVA (69.7%, 152/218) and were or might be willing to use DVAs in the future (84.0%, 178/212) including 57.1% (36/63) of participants who had never used a DVA before (Table 1). Among the 152 participants who had used a DVA, Siri was the most highly used DVA (67.8%, 103/152). Other commonly used DVAs were Alexa and Google Assistant which were both used similarly by about nearly half (70/152 and 71/152, respectively) of participants. Of those who did report DVA use, the majority accessed the tool from their personal cell phone (88.8%, 135/152), and the main task was to answer questions (75.0% 114/152), with entertainment (57.2%, 87/152), accessing information online (55.9%, 85/152), and setting timers (51.3%, 78/152) also being frequent tasks done with a DVA. While many of the participants have used DVAs, the frequency of this use was split between those who use it rarely or sometimes (57.4%, 85/148) and those who use it often or always (42.6%, 63/148). Although the participants use DVAs, the majority did not rely on the DVA for everyday needs (57.0%, 85/149; Table 2).

Among respondents, 36.5% (77/211) reported having visual disabilities. Of the participants who reported having any visual disability, 81.6% (62/76) found it mildly to extremely difficult to perform activities because of their vision impairment and 41.9% (31/74) reported having vision-related needs not being met. Only 38.2% (26/68) of participants with visual disabilities indicated being very or highly satisfied with current services and devices, revealing a crucial need for improvement.

A greater proportion of participants with visual disabilities reported having never used DVA (39.0%, 30/77) than those without visual disabilities (24.6%, 33/134; P=.03). Most participants with visual disabilities who had used a DVA did so on their personal cell phone (53.2%, 41/77) but this was significantly less than the proportion of those without visual disabilities who use DVAs on cell phones (68.7%, 92/134) (P=.03). As in the entire sample, the most common DVA task was answering questions regardless of visual disability status, but this task was done by fewer people with visual disabilities (40.3%, 31/77) than those without visual disabilities (60.4%, 81/134; P=.005). Those with visual disabilities indicated relying on DVA sometimes or always (63.8%, 30/47), more often than those without visual disabilities (32.7%, 33/101; P=.001). In addition, participants with visual disabilities were equally as willing to potentially use DVAs in the future (85.3%, 64/75) as those without visual disabilities (85.0%, 113/133; P=.78). Within the group of participants who had never used a DVA, those with and those without visual disabilities had similar willingness to potentially use DVAs in the future (62.1%, 18/29 vs 56.3%, 18/32, respectively; P=.64). Those with visual disabilities were also more willing to grant access to past data for DVA studies than those without visual disabilities (22.5%, 16/71 vs 13.1%, 17/130; P=.01; Table 3).

Among respondents, 47 (30.1%, 47/156) reported having a physical disability. Those with physical disabilities were more likely to report also having visual disabilities (29.8%, 14/47) than those who did not report having any physical disabilities (9.2%, 10/109, respectively; P=.001). Among the participants who possessed physical disabilities, 63.0% (29/46) found it moderately, very, or extremely difficult to perform everyday activities. The most common activities affected by the physical disability were walking or climbing stairs (87.2%, 41/47) and holding objects in one’s hands (31.9%, 15/47). Most participants (80.4%, 37/46) reported that their physical needs are being met. Still, only 55.6% (10/18) of participants with physical disabilities report being very or highly satisfied with the services or devices meeting their needs.

Participants with physical disabilities reported having never used DVA (34.0%, 16/47) more often than those without a physical disability (21.1%, 23/109), although the difference is not statistically significant (P=.09). The proportion of participants that used DVAs for certain tasks tended to be lower for those with physical disabilities than those without physical disabilities but again these differences were not statistically significant. There was a statistically significant difference (P=.01) in the use of Alexa or Amazon Echo for those with physical disabilities vs those without physical disabilities (21.3%, 10/47 vs 42.2%, 46/109, respectively). Similarly to those with visual disabilities, the most common task participants with physical disabilities used DVAs for was answering questions (59.6%, 28/47). There was no significant difference in the participants’ reliance on DVA for everyday needs by physical disability status. Most participants with physical disabilities were willing, or possibly willing, to use DVA in the future (86.7%, 39/45). There is a significant difference of reliance on DVA with those without either disability being less reliant and those with visual or both visual and physical disabilities being more reliant (P=.01; Table 3).

Overall, a large proportion of participants had at least a slight concern about privacy (86.2%, 175/203), security (89.1%, 180/202), confidentiality (86.7% 176/203), accuracy of information (81.7%, 165/202), and reliability of information (85.1%, 171/201). More specifically, roughly half of the participants had moderate to high concerns about the privacy (47.8%, 97/203), security (54.5%, 110/202), and confidentiality (51.7%, 105/203) of DVAs. The degree of concern about privacy, security, and confidentiality had no significant differences between the participants with and those without visual disabilities. We found a significant difference between concerns about accuracy (P=.04) and reliability (P=.048) of information from DVAs based on visual disability status with participants with visual disabilities tending to be moderately or highly concerned more often. The degree of concern about privacy, security, confidentiality, accuracy, and reliability had no significant differences between the participants with and those without physical disabilities. Despite these concerns, we generally did not see a significant decrease in DVA use, frequency, or reliance for those with higher levels of concern. The one exception was in those with high levels of concern about security who tended to use DVAs less frequently than those with slight or no concern about security (P=.04; Table 4).

In this study, we found the majority of participants had previous experience with DVAs and that there was interest in using them in the future. Looking specifically at those with self-reported disabilities, many adults with physical disabilities and visual disabilities do not regularly use DVAs, however, those with visual disabilities have a greater reliance on the technology. Overall, participants were moderately concerned with the issues of privacy, security, confidentiality, accuracy, and reliability of DVAs. Roughly half of those with visual disabilities had moderate or high concern for privacy, security, and confidentiality, but these numbers do not drastically differ from those without visual disabilities, or with or without physical disabilities, nor does it seem to affect their use and reliance on the technology. Previous literature has discussed smartphone-based low vision aids where DVAs are briefly examined among other smartphone tools for low vision usability [17,18]. The use of DVAs for people with disabilities, especially for the purposes of living independently, is a rapidly expanding field of study [15]. Many recent studies have focused on the accessibility, perceptions, and feasibility of the applications of DVAs, but few have evaluated the actual use cases and frequency of DVA use among individuals with visual or physical disabilities. This is an important step toward understanding and developing technology that better meets the needs of these individuals to improve health outcomes and quality of life.

Overall, it was observed that more than half of the participants used DVAs only “sometimes” or “rarely.” Despite the lack of current use when the survey instrument was completed, more than half of the participants were willing to use DVAs in the future. This was perhaps a function of age, seeing that the average age of participants was older at 52.6 years old. From the survey, it was observed that the dominant use of DVAs was found in smartphones. Smartphone use among older age groups differs from those younger as a 2024 study from the Pew Research Center reported that 91% of adults aged between 50‐64 years and 79% of adults aged 65 years and older own smartphones in the United States, compared with 99% of adults aged 18‐29 years [1]. Using smart speakers as DVAs in populations of older adults in low socioeconomic positions has also proved to be promising, though more research should be conducted to promote the answering of health-related questions [19]. The additional smartphone functions and tools available for consumers are not necessarily used by older populations. While lower DVA use is more common among older adults, this could be related to a number of factors including lesser comfort with the technology, lower digital literacy, usability factors, technology learning curves, and other personal hesitations.

The perceptions of DVAs were quite similar across those with and those without physical and visual disabilities. Overall, about half of participants expressed a moderate to high concern with the security and reliability of DVAs. It was interesting to note that those with visual disabilities were less concerned with privacy and confidentiality issues than those without visual disabilities. Concerns regarding online data privacy, security, and reliability are not novel, as literature spanning decades has documented these concerns. With specific regards to data privacy, DVAs represent an interesting use case as large data giants such as Facebook (Meta) and Amazon are increasing their use of personal user data for advertising, as well as individual targeting and tailoring of products and services. DVA companies have a reputation for lacking transparency with users about privacy protection which can lead to a lack of trust among users [20]. Many DVAs rely on the “activation keyword approach,” where devices listen for a specific keyword before they start recording, which means that these technologies are always listening, unless muted. It also means that DVAs are susceptible to false activation, which means that recording could be unintentionally turned on at a time when users are not aware. Overall, there needs to be improved transparency and additional user controls for personal recording for DVAs and their devices. However, despite the concern over data privacy, 23% of participants reporting visual disabilities and 15% of participants reporting physical disabilities in our study were willing to grant access to their past existing data or share future data for future DVA studies.

The landscape of DVAs along with the integration of generative artificial intelligence (AI) is everchanging, which could lead to evolving data privacy and data quality concerns regarding DVAs. Since the release of ChatGPT by OpenAI in 2022, the interest and number of publications regarding “trust” in conversational agents has risen rapidly. Before 2024, computer science held the vast majority of publications regarding trust in conversational agents; however, the field of medicine has shown greater growth in interdisciplinary research with conversational agents over the last 8 years [21]. Applications of AI in health systems will require close monitoring and transparency for patients to feel comfortable granting access [22]. Recent research has shown that many individuals feel relatively comfortable sharing their data with national health organizations and universities, but much less willing to share with private commercial organizations such as insurance and technological companies [23]. This further emphasizes the importance of exploring utility and applications of DVAs and AI in health care and research contexts.

The survey showed that those with visual disabilities had a greater reliance on DVAs than those without visual disabilities. This is likely due to those with visual disabilities relying on voice commands to read and assist them with tasks they otherwise would not be able to do. In terms of specific activities, participants with visual disabilities reported using DVAs more frequently for answering questions than those without visual disabilities.

The survey showed that the DVA use cases of those with physical disabilities did not differ significantly from those without physical disabilities. Due to the nature of the relatively small convenience sample, there was not a large degree of heterogeneity in terms of types of physical disabilities. Those with physical disabilities were shown to rely more on DVAs than those without. The finding of increased reliance is supported by previous interview-based research that noted DVAs promote autonomy by increasing the ability to perform daily activities and tasks that were normally obstructed by mobility or other barriers [24].

To our knowledge, this is one of the first studies to assess the utility and perceptions of DVAs for adults with specific focus on those with visual and physical disabilities. Despite the novelty of the study, there are multiple limitations to consider that can guide future exploration. It is crucial to note that due to the COVID-19 pandemic, data collection was disrupted with delays in analysis. Despite the lag experienced, the study begins to fill a literature gap on the use of DVAs, particularly from the perspective of those living with visual and physical disabilities.

This study is not without limitations. The study findings relied on participant self-reported information, which may introduce response or recall bias, such as participants not remembering, or choosing not to share information related to disabilities or DVA use. Moreover, this study used a convenience sample, which is a limitation to our approach. In addition, participants were recruited from one health system, so results may not be generalizable to other populations. To reduce participant burden among those with visual disabilities, the survey was shortened. This may have overlooked potential dual disability concerns (physical disabilities among those with visual disabilities). In addition, the study survey did not capture the specifics related to the reported disability, such as etiology, onset, severity and overall quality of life. In addition, the survey instrument used was created specifically for this study and not validated. Despite these limitations, our study was intended to be exploratory in nature and serve as a foundation for future work in this field.

Future work should examine DVA use through device use data recordings and compare this across different disabilities and functional levels. Additional areas of investigation should incorporate qualitative approaches to more fully elucidate the ways that DVAs have been used, or could be used, to support the health of older adults and those individuals living with disabilities. Not only does this have great implications for the more than 25% of individuals in the United States living with a disability [25], but it also has great implications for the number of aging individuals who may benefit from DVA use (eg, social isolation and improved independence).

Self-reported use of DVA varied across the surveyed participants. Tasks such as answering questions and seeking more information were the most used functions of DVAs by individuals with physical and visual disabilities. Participant perceptions of DVAs, regardless of disability status, suggest that concerns with data privacy, security, confidentiality, reliability, and accuracy exist.