Improving a Mobile Telepresence Robot for People With Alzheimer Disease and Related Dementias: Semistructured Interviews With Stakeholders

Background By 2050, nearly 13 million Americans will have Alzheimer disease and related dementias (ADRD), with most of those with ADRD or mild cognitive impairment (MCI) receiving home care. Mobile telepresence robots may allow persons with MCI or ADRD to remain living independently at home and ease the burden of caregiving. The goal of this study was to identify how an existing mobile telepresence robot can be enhanced to support at-home care of people with MCI or ADRD through key stakeholder input. Objective The specific aims were to assess what applications should be integrated into the robot to further support the independence of individuals with MCI or ADRD and understand stakeholders’ overall opinions about the robot. Methods We conducted in-person interviews with 21 stakeholders, including 6 people aged >50 years with MCI or ADRD living in the community, 9 family caregivers of people with MCI or ADRD, and 6 clinicians who work with the ADRD population. Interview questions about the robot focused on technology use, design and functionality, future applications to incorporate, and overall opinions. We conducted a thematic analysis of the data obtained and assessed the patterns within and across stakeholder groups using a matrix analysis technique. Results Overall, most stakeholders across groups felt positively about the robot’s ability to support individuals with MCI or ADRD and decrease caregiver burden. Most ADRD stakeholders felt that the greatest benefits would be receiving help in emergency cases and having fewer in-person visits to the doctor’s office. Caregivers and clinicians also noted that remote video communication with their family members using the robot was valuable. Adding voice commands and 1-touch lifesaving or help buttons to the robot were the top suggestions offered by the stakeholders. The 4 types of applications that were suggested included health-related alerts; reminders; smart-home–related applications; and social, entertainment, or well-being applications. Stakeholders across groups liked the robot’s mobility, size, interactive connection, and communication abilities. However, stakeholders raised concerns about their physical stability and size for individuals living in smaller, cluttered spaces; screen quality for those with visual impairments; and privacy or data security. Conclusions Although stakeholders generally expressed positive opinions about the robot, additional adaptations were suggested to strengthen functionality. Adding applications and making improvements to the design may help mitigate concerns and better support individuals with ADRD to live independently in the community. As the number of individuals living with ADRD in the United States increases, mobile telepresence robots are a promising way to support them and their caregivers. Engaging all 3 stakeholder groups in the development of these robots is a critical first step in ensuring that the technology matches their needs. Integrating the feedback obtained from our stakeholders and evaluating their effectiveness will be important next steps in adapting telepresence robots.


Introduction
Background More than 6 million Americans aged ≥65 years are currently live with Alzheimer disease and related dementias (ADRD), and it is projected that this number will increase to nearly 13 million Americans aged ≥65 by 2050 [1]. Approximately 80% of people with ADRD and most people with mild cognitive impairment (MCI) receive care in their homes [2]. In the United States, 83% of care provided to older adults is from family members, friends, or other unpaid caregivers; 48% of these caregivers provide help to older adults with ADRD [1]. Caregivers of people with ADRD can often be burdened emotionally and financially; they experience emotional, financial, and physical challenges due to care responsibilities at twice the rate of caregivers of older adults without dementia [1]. The primary reason family members or friends act as caregivers for people with MCI or ADRD is to allow them to remain living in the community rather than in a long-term care facility, but this may become challenging, especially for caregivers supporting people with ADRD who live alone. ADRD caregivers report worsening of their own health due to care responsibilities and are required to take a leave of absence from work or quit work [3].
Various types of robots can help people with MCI or ADRD. Mobile telepresence robots, which are the focus of this study, can be used to mediate communication and social exchange between family members, friends, and others who may not be colocated. Such contact can alleviate loneliness, support medication compliance and other important daily activities, and alert caregivers and clinicians in a timely way to address changing patient needs. Studies have established the feasibility of mobile telepresence robots for supporting social interactions between caregivers and people with MCI or ADRD [4]. If telepresence is complemented by autonomous robotics, the potential range of benefits expands even further through innovative approaches to care that can allow people with MCI or ADRD to remain living independently in the community, which may help to alleviate loneliness and enhance the quality of life of people with ADRD, as well as ease the burden of caregiving [5][6][7][8][9][10][11].
Although several types of robots have been developed to assist people with MCI or ADRD, and the feasibility of telepresence for supporting social interactions between caregivers and people with dementia has been established [4], research on dementia-specific adaptations and acceptance of these robots that incorporate perspectives across 3 key end-user stakeholder groups (ie, individuals with MCI or ADRD, caregivers, and clinicians) has been limited in scope [4,5,[11][12][13][14]. For example, there is limited research on how autonomous robotics features can assist a mobile telepresence robot in supporting at-home care of people with MCI or mild to moderate ADRD. Robots have often been developed in an ad hoc manner by technology companies with limited understanding of the needs, preferences, and feedback from key stakeholders [15][16][17]. The ways in which robots can assist people with MCI or ADRD, their caregivers, and clinicians need to be explored further.

Objectives
In this study, we engaged these 3 key stakeholder groups in the design of dementia-specific adaptations to an existing, commercially available mobile telepresence robot that is already being used in several other settings, such as in schools for homebound students and for home health care by remote clinicians. The overall purpose of this qualitative study was to obtain stakeholder feedback and identify how an existing, commercially available mobile telepresence robot can be enhanced to support at-home care of people with MCI or mild to moderate ADRD, with the ultimate goal of developing an autonomous mobile telepresence robot. The specific aims of the study were to (1) assess what applications should be integrated into the robot to support the independence of individuals with MCI or ADRD and (2) understand stakeholders' overall opinions about the robot.

Overview
The study consisted of qualitative interviews with 3 key stakeholder groups (people with MCI or ADRD, caregivers, and clinicians) to obtain feedback on how to adapt an available mobile telepresence robot and further understand stakeholders' opinions about the potential utility of the robot.

Ethics Approval
The study protocol was approved by the Institutional Review Board at the Veterans Affairs (VA) Bedford Healthcare System (approval number 110818). Participation in the study was voluntary, and informed consent was obtained from each participant.

Study Population
Study inclusion criteria consisted of the following for each of the three stakeholder groups, respectively: (1) male or female Veteran aged ≥50 years with MCI or early ADRD living in the community (ie, not a nursing home or assisted living facility): people with MCI or ADRD were categorized as having MCI versus early ADRD based on the most recent clinic visit note at the time of recruitment and had to be alone at home for at least 4 hours, and there had to be no indication of incompetence in the Veteran's medical record; (2) male or female family caregivers aged ≥18 years who had a family member with an established diagnosis of memory loss, MCI, or any cause of dementia: the caregiver needed to have a family member with MCI or early ADRD who frequently spent at least 4-hour stretches alone at home apart from any caregiver in the last 12 months; and (3) male or female clinician with experience in managing patients with MCI or ADRD: clinician participants must have provided care to patients with cognitive impairment and/or supported families of such patients for at least one year.

Recruitment
Our goal was to recruit at least six participants from each of the 3 key stakeholder groups: people with MCI or ADRD, caregivers, and clinicians. To recruit people with dementia, we identified patients with MCI or early ADRD from the current VA Bedford outpatient primary care and dementia clinics using electronic medical records. Once identified, the VA Bedford study coordinator (JM) reviewed the electronic medical records of the patients to determine whether they met the inclusion criteria. Those who met the inclusion criteria were mailed an invitation letter for participation in the study; the letter included a phone number to call should they wish to opt out of the study. The study coordinator made a recruitment phone call to patients who did not call back to opt out.
Family caregivers were recruited through family relationships with people with dementia who were recruited for this study or through inputs from VA clinical providers. Caregivers also received a letter of invitation, and the study coordinator followed up with telephone calls or email.
VA Bedford providers with expertise and experience in managing patients with MCI or ADRD were recruited by the VA study principal investigator (LM) and/or study coordinator via email. Using an opt-in approach, the recruitment email instructed the invited clinician to contact the study coordinator if they were willing to participate. Up to 3 emails were sent to each clinician.

Interview Guide
The interview guide contained semistructured questions to elicit open responses from participants, as well as structured questions that asked participants to rate their responses on a Likert scale as well as to categorize options (eg, from most useful to least useful) and then provide a rationale for their response. Questions related to perceived usefulness and perceived ease of use were used to assist in understanding stakeholders' preferences regarding technology and identify adaptations that can be made to the robot (ie, the mobile robot, as both concepts may affect stakeholders' attitude toward using the mobile robot, which can then influence their behavioral intention to use the mobile robot and thereafter, affect actual use). We conducted a literature review to understand the types of questions and qualitative themes found in previous research studies relating to mobile telepresence robots as well as other types of robots that have been developed to assist people with ADRD [18][19][20][21][22][23]. On the basis of this literature review, we developed interview questions that focused on domains that we established a priori to help guide our understanding of how stakeholders perceive the mobile robot's usefulness and ease of use: (1) technology use, (2) design and functionality of the robot, (3) future applications to be incorporated into the robot, and (4) overall opinions about the robot. We established these as key domains because they would allow us to elicit information from stakeholders to accomplish this study's overall purpose and specific aims. We asked participants similar questions across all stakeholder groups. Table 1 provides relevant questions for these domains. Of note, for the question "Please let me know how you would rank these applications that could be used or built into the robot (from most useful to least useful)," the 6 categories of applications were chosen by subject-matter expert team members with extensive clinical knowledge (LM) and technical knowledge (AR)-based on their professional experience in working with people with MCI or ADRD, conducting research on robots, and/or developing robots. The subject-matter experts met to brainstorm and discuss what types of applications would enhance the mobile telepresence robot for this population and chose the 6 applications that would be the most feasible to add to the robot.

Data Collection
From July to August 2019, a total of 3 health service researchers (MS, JM, and JS) who are experienced in qualitative methods, conducted the semistructured interviews. Each stakeholder participated in 1 in-person individual or group interview, which lasted approximately 60 minutes. Interviews were conducted either at the VA Bedford main facility or at the participant's home. People with MCI or ADRD and family caregiver participants received a US $50 gift card for their participation in the interview and an additional US $30 gift card to compensate for travel time if they participated in person at VA Bedford. To thank them for their time and insight, the VA clinician participants received a modest meal, which did not exceed a value of US $20 per person. Before the interviews, all participants provided informed consent and agreed to be audio-recorded for transcription purposes.
After obtaining consent, the interviews began by asking participants to complete a stakeholder group-specific demographics questionnaire. As part of the interview process, we also provided participants with information about the existing, commercially available mobile telepresence robot, which included a photo ( Figure 1) and a brief explanation.
This robot is a wheeled, upright (approximately 122 cm tall), self-propelled device that weighs 8.6 kilograms with a 14.5-cm screen; it moves about under remote control and provides 2-way video communication between a remote user who pilots the robot through an app accessed via a computer, tablet, or smartphone and a person who is physically present with the robot. In addition, we presented a video (television news segment) to visually display the robot's current capabilities. The video featured a student who was immunocompromised and could not attend school in person. However, the student was able to attend school and remain in the classroom virtually through the mobile telepresence robot. Through this video, the participants were able to see how the robot displayed the student's face on the screen and enabled her to talk to and socialize with her classmates and teacher in the classroom and how she could control the robot, allowing it to move through the classroom and down the hallways. We then asked the participants questions from the interview guide.

Data Analysis
Interview transcripts were the primary data source for the directed content analysis. Coding and data analysis focused on four a priori domains: (1) technology use, (2) design and functionality of the robot, (3) future applications to be incorporated into the robot, and (4) overall opinions about the robot. Using a code book that contained domains with definitions, the core analytic team (MS and JS) jointly coded 1 interview transcript and examined the transcript for evidence of the established domains. The 2 researchers discussed the coding process, resolved any discrepancies, and reached a consensus on definitions and coding classification. Thereafter, the transcripts were divided and coded separately by the 2 researchers using NVivo (version 12; QSR International) qualitative software.
Upon completion of coding, we generated coding reports for each domain and stakeholder group. The data in the coding reports were reviewed and then synthesized into a matrix, which allowed us to organize the data to facilitate the final interpretation. Next, we used a matrix analysis technique to assess patterns within and across stakeholder groups [24,25]. Thereafter, for each of the 4 domains, we identified salient themes while noting similarities and variations across stakeholder groups. In addition, we compared patterns in the Likert ratings and responses across stakeholder groups. To ensure wider perspectives, we included study team members (JM, LM, and AR) beyond the core analytic team to review and comment on the study findings.

Participant Characteristics
Overall, a majority of participants (20/21, 95%) were White across all 3 stakeholder groups. The 6 people with MCI or ADRD in our study sample were men (mean age 73 years, range 69-80 years), with 4 (67%) of these participants living with a family caregiver and 2 (33%) were living alone. All 9 caregiver participants in this study were women (mean age 60 years, range 42-80 years). Of the 9 participants, caregivers' relationships with family members with MCI or ADRD included 6 (67%) participants who were the adult child or child-in-law of the family member with MCI or ADRD and 3 (33%) participants who were the spouse or partner. Of the 6 clinician participants, 4 (67%) were women and 2 (33%) were men. All had advanced degrees ranging from master's to doctoral, with an average of 20 years of experience in their respective fields. They estimated that, on average, 89% of their patients were aged ≥60 years, and 76% of their patients had cognitive impairment.

Person With Dementia and Family Caregiver Technology Use
Although most individuals with MCI or ADRD reported that they were able to use their mobile phones for basic functions (eg, phone calls or SMS text messages), overall, they reported lower use of technology and the internet than caregivers, who reported higher use: Most individuals with MCI or ADRD had never used the internet or applications for health-related reasons or to get answers to health-related questions; similarly, caregivers also reported lower use for health-related reasons or questions. However, most caregivers reported using at least one app for health purposes.

Design and Functionality of the Robot
Overall, stakeholders across all 3 groups expressed that they liked the mobility and size of the robot (approximately 122 cm tall), as well as the ability to connect and communicate in an interactive way. However, they expressed concerns related to the size of the robot for individuals living in smaller apartments and trailers, as well as the stability of the robot, questioning whether it could tip over easily. In addition, stakeholders queried whether the screen size and quality (14.5 cm and 320×240-pixel display) would provide enough support for those with visual impairments who would need a screen with higher-quality contrast and that could display larger text or image. Across all stakeholder groups, voice commands and 1-touch lifesaving or help buttons were the 2 top suggestions offered: As presented in Table 2, stakeholders offered several suggestions on what features could be added to the robot and provided input on how those features could help improve the design, functionality, and use of the robot in the context of people with MCI or ADRD. The features listed in Table 2 are presented from most often suggested to least often suggested across all stakeholder groups. Adding a voice command feature to the robot was suggested most often, whereas adding lights to the robot was the least often suggested feature.  Figure 2 presents the mean scores of the 6 applications ranked across each stakeholder group with 1 being the lowest and 6 being the highest. In addition to ranking the options of possible applications provided by the study team, stakeholders across the 3 groups offered their own specific suggestions on the types of applications they perceived as useful to include in the mobile robot to enhance the opportunity for individuals with MCI or ADRD to live at home independently. We categorized these suggestions across the 3 stakeholder groups into 4 types of applications: health-related alerts, reminder prompts, smart-home-related applications, and social, entertainment, or well-being applications.

Health-Related Alerts
Suggested applications include health-related alerts such as those that could inform providers of a change in patients' blood sugar or blood pressure levels, as well as those that could be triggered if a person falls. Similar to a bracelet or necklace alert, activation and connection to a previously imputed emergency response system such as family members, ambulance, police, emergency medical technicians, and/or health care providers, could occur.

Reminders
Suggested applications include those offering real-time, step-by-step prompts on how to complete multistep activities such as getting dressed, making a microwave meal, or turning the oven on/off; those about medication, doctors' appointments, other scheduled activities, and wake-up or bedtime prompts, especially for those who have sleep problems; and those that can prompt a person with MCI or ADRD to perform and help with activities of daily living or instrumental activities of daily living (eg, prompt for eating, bathing, or bathroom use).

Smart-Home-Related Applications
Suggested applications include smart-home technology integration, in which the robot can detect motion, connect to a home security alert, and recognize home hazards (eg, smoke and carbon monoxide detection, or tripping hazards). In addition, the robot can work similarly to or in conjunction with a voice-activated smart-home device (eg, the robot turns on or off appliances in the home).

Social, Entertainment, or Well-being Applications
Suggested applications that could enhance the well-being and comfort of a person with MCI or ADRD include community or social connection applications, such as ride share applications that could assist transportation arrangement and allow individuals with MCI or ADRD to go out into the community independently (eg, attending religious groups for social connection); those that stream music, movies, and television shows either directly on the robot's screen or projected onto a wall, which could also be turned on virtually by a caregiver; and games to help with cognitive engagement and exercises to help with balance and mobility. In addition, an app related to identifying family members or friends to assist with recall could enhance social connection and well-being, such as integrating contact numbers of specific family members or friends into the robot, uploading photos and voice recordings of their family members or friends, and inputting specific memories with family members or friends and important dates (eg, birthdays and anniversaries) into the robot.

Overall Opinions About the Robot
Of the 3 stakeholder groups, caregivers reported the highest overall rating for the robot (mean 8.1; rated on a scale of 1 to 10). All caregivers (9/9, 100%) positively perceived the robot and liked that they could virtually check in and communicate with their family members, especially if their family members were to be left alone for an extended period. A caregiver described how her family member may trust the robot more than humans to help around the house: However, of the 9 caregiver stakeholders, 7 (78%) noted that additional features and applications would have to be added to the robot for it to be more useful; thus, the reason for not rating the robot a 10.
Overall, ratings from the clinician and people with MCI or ADRD stakeholder groups (mean 7.4 and 6.4, respectively) were slightly lower than the caregiver group, noting that additional features and applications had to be added to the robot for it to be more useful. Similar to the caregiver group, all clinicians (6/6, 100%) expressed positive perceptions about the robot and its potential to increase a person's independence and ability to stay home alone: What we worry about with people at that stage is that they can continue to do as much as they can, but they need support. Although the robot was perceived to be beneficial in terms of living independently if additional features and applications were added, 50% (3/6) of the people with MCI or ADRD rated the robot lower on the scale for two main reasons: (1) they perceived that they did not need the robot as they had family members who cared for them and (2) they did not know if they could use the robot in their homes because of challenges with size and mobility: We

Concerns
Although there were many positive responses to the mobile telepresence robot, all 3 stakeholder groups expressed concerns about using the robot, including technology usability, privacy, and security. In all, 50% (3/6) of people with MCI or ADRD were concerned about the technology needed to operate the robot because they did not own a computer/smartphone and lacked knowledge of how to use this technology: In all, 44% (4/9) of the caregivers and 67% (4/6) of the clinicians also expressed similar concerns about the ability to use the robot by people with MCI or ADRD (eg, ability to fix the robot if presented with technological challenges, lack of Wi-Fi, or equipment needed to operate the robot): Of the 6 people, although 1 (17%) with MCI or ADRD was concerned about privacy or security, the other 5 (83%) were not. However, caregivers and clinicians expressed more concerns about privacy and security. They discussed the tension between caregivers wanting to check in with their family members but feeling as though they may be infringing on their family members' privacy when controlling the robot. Concerns regarding privacy and patient dignity (eg, when a patient is changing or going to the bathroom), especially when there are multiple permissible remote users, were also noted: To help alleviate concerns, caregiver and clinician stakeholders noted that privacy and security terms would need to be spelled out clearly, such as those related to who would have main control over the robot, if any data are recorded and the security associated with it, and whether someone could easily access the robot and obtain private information.

Principal Findings
We conducted this qualitative study to obtain feedback from the 3 groups of key stakeholders (people with MCI or ADRD, family caregivers, and clinicians) on how to adapt an existing, commercially available mobile telepresence robot to specifically support individuals with MCI or early ADRD so that they can continue to live at home independently. Through these interviews, we received consistent feedback across groups, which could enhance the robot's usability. Suggestions that differed by stakeholder group provided us with a complete understanding of the adaptations that need to be made to strengthen the utility of the robot for all stakeholder groups.
People with MCI or ADRD, family caregivers, and clinicians all described multiple possible updates to the design and functionality as well as the applications of the remotely navigable telepresence robot. Family caregivers and clinicians clearly perceived the need for additional support for people with MCI or early stage ADRD to allow them to live alone or be left alone for long periods. Both stakeholder groups felt that the augmented version of the mobile telepresence robot in our study could play an important role. Compared with clinicians and family caregivers who were the most enthusiastic of the 3 stakeholder groups, people with MCI or ADRD reported lower ratings for the robot. Despite these lower ratings from people with MCI or ADRD, many stakeholders across all groups perceived that the robot had the potential to increase a person's independence and ability to stay at home. For example, stakeholders perceived the videoconferencing function (already available in the robot) to be useful in facilitating communication with friends or family members and for video telehealth visits with providers, which helps strengthen relationships by bridging the distance between individuals [26]. However, similar to previous research that highlights the importance of developing robots based on stakeholder feedback [18,19,27], incorporating several adaptations regarding the robot's design, functionality, and applications would be critical to enhance use for their needs, such as additional development of voice command and help button functions as well as applications related to medication reminders or dispensers and emergency response access [20]. These were perceived as critical features or applications to help people with MCI or ADRD maintain independent living at home; stakeholders raised concerns that these may be barriers to adoption if not incorporated. Among the 3 stakeholder groups, medication reminders and emergency help access ranked the highest, with reminders regarding daily schedules, rounding out the top 3 for those with MCI or ADRD. Although all 3 functions were beyond what the robot could do at the time, the fact that 2 of the 3 related specifically to help with memory decline highlights the importance of engaging disease-specific stakeholders in such studies. Similarly, concerns about the ease of use of the technology with strong recommendations from all 3 groups regarding the integration of voice commands also underline the benefits of including people with MCI or ADRD in the development of products or interventions for which they are the primary target as well as family caregivers and clinicians.
Our results underscore the importance of engaging and obtaining end-user input from different groups of stakeholders in technology development-the individual with MCI or ADRD, the caregiver, and the clinician-which provides for the opportunity to tailor according to the needs and interests of all who are involved in the care of people with MCI or ADRD. Although this is an important component of technology development, to the best of our knowledge, only a few previous studies have interviewed all 3 groups [12,18]. In addition, engaging community-dwelling people with MCI or ADRD is a critical and feasible component of technology development; however, studies have usually lacked the involvement of people with MCI or ADRD in technology development [16]. This lack of involvement may lead to the implementation of technology that is not tailored or suitable to the individuals who the technology intends to serve [16]. Family caregivers are also important stakeholders and are quite likely one of the most critical given how this type of technology (mobile telepresence robots) may help alleviate caregiver burden. Caregivers of people with MCI or ADRD are their primary advocates; the primary people with whom they communicate with; and may usually be the primary decision-makers in the household. By involving caregivers as a stakeholder group in our study, we were able to obtain input on what types of features and applications of the mobile telepresence robot can help reduce caregiver burden.
The feedback provided on the adaptations to the mobile telepresence robot and possible applications for the inclusion of all 3 stakeholder groups aligned with concepts such as perceived ease of use and usefulness as well as trust [32]. These concepts are critical to the adaptation of mobile telepresence robots in health care communication settings and can lead to barriers in adoption if not resolved. Previous studies have noted that barriers to the acceptance of mobile telepresence robots included challenges in using technology and concerns about the ability of older adults to operate the robots [18,33,34]. In our study, compared with caregivers, people with MCI or ADRD reported overall lower use of technology and the internet, which may have affected their perceptions about the usefulness and ease of use of mobile telepresence robots. People with MCI or ADRD and caregivers expressed concerns regarding the technology use required by people with MCI or ADRD to operate the robot. One of the most often suggested design features that stakeholders across the 3 groups wanted (particularly people with MCI or ADRD) to have incorporated into the robot was voice command technology. This simple design improvement has the potential to increase the perceived ease of use and usefulness of robots. In addition, stakeholder feedback can assist in understanding preferences that may potentially enhance trust, thereby leading to a higher adoption of technology. Similar to previous research [19], people with MCI or ADRD in our study did not seem to be concerned with privacy and security. However, in contrast to previous research on caregivers who were more likely to perceive no ethical dilemma when balancing this with the safety of people with dementia [35], privacy and security were the main concerns expressed by both caregivers and clinicians. Caregivers and clinicians were unable to offer solutions that they felt would alleviate their concerns but noted that these challenges would need to be resolved for end users, such as themselves, to trust the robot. Given the dearth of studies focusing on privacy and security with robots for people with MCI or ADRD, additional exploration is warranted and should be incorporated into future research [13].
Although applications related to social stimulation activities (eg, games or reading the news) ranked lowest across all 3 stakeholder groups, stakeholders offered suggestions in terms of what types of social, entertainment, or well-being applications they would want to see in the robot; for example, ride share applications that could assist transportation arrangements and allow individuals with MCI or ADRD to go out into the community independently and applications that could assist with recall to help identify family members or friends as well as specific memories and important dates (eg, birthdays and anniversaries). These suggested applications to enhance the robot are important because they may enable people with MCI or ADRD to be even more socially engaged and connected to their family and friends-a key component of what a telepresence robot is supposed to be doing. In particular, as highlighted during the COVID-19 pandemic, social interactions and connections are critical. The pandemic heightened feelings of loneliness as the number of older adults who were socially isolated grew and they were unable to participate in social activities, thus significantly affecting their mental health [36]. Before the COVID-19 pandemic, about one-fourth of the older adults living in the community in the United States were thought to be socially isolated, with approximately 40% of these older adults reporting feeling lonely [37]. Compared with well-established risk factors such as smoking, high blood pressure, and obesity [37][38][39], social isolation and loneliness can increase the risk of depression, poorer cognitive function, and dementia leading to an increased risk of mortality and morbidity [39]. In the United States, deaths caused by Alzheimer disease and dementia have increased by 16% during the COVID-19 pandemic [1]. The pandemic has shown the importance of developing innovative technology that can improve social connections and support for older adults [36,40].
Our study results echo findings from another study of community-dwelling older adults, highlighting that one of the roles of a mobile robot was to provide friendship or companionship or to provide help [27]. In addition, our study also underscores stakeholders' perceptions that one of the greatest values of a mobile telepresence robot is the connection that it can offer to people with MCI or ADRD through remote video communication with family members and friends. Although the robot, as is, provides access and social connections, integrating stakeholders' suggestions on applications into the robot would only further enhance engagement and social connections.
Through this qualitative study, we were able to obtain feedback from key stakeholders and identify the types of features and adaptations they prefer on an existing, commercially available mobile telepresence robot to enhance the support of at-home care for people with MCI or mild to moderate ADRD. As discussed earlier, stakeholders offer several suggestions regarding their desires. Some of these features and adaptations may be more feasible than others. For example, participants noted voice command as a highly desired feature, which is becoming more common in robots (eg, Alexa and Siri). A technology company that focuses on designing and engineering robots may find this feature more feasible to include in a mobile telepresence robot because of the advancements in this technology compared with other desired features such as attachments (eg, arms, handles, or cupholders), which may be challenging because of the robot's center of gravity, or a robot that cleans the house, which is a function that is beyond current capabilities. Moving forward, an important next step for companies that design and engineer robots is to assess the feasibility of the desires from an engineering perspective and balance the challenges in fulfilling the desired features and functions of the stakeholders while also ensuring the robot's usability for the target population.

Limitations
There are a few limitations to our study that should be acknowledged. Similar to the challenges faced by most qualitative studies, participants volunteered to be a part of the study, and interviews elicited those particular perspectives; thus, interviews may have been subject to selection bias. In addition, stakeholders only viewed the robot through a video and did not see the robot in person; this may have limited their ability to visualize and understand the robot's full capacity of what it could offer. Our study participants were all from the Greater Boston area; thus, we only had a sample of participants from 1 geographic area. In addition, because of the small sample size, particularly within each stakeholder group, we were unable to compare similarities and differences in feedback within each stakeholder group, such as whether caregivers of varying education levels made similar or different suggestions on features and applications to incorporate into a mobile telepresence robot. Because this was a qualitative research study, we did not collect large quantitative data sets that would allow comparisons with the general or larger Veteran patient population. These comparisons are interesting areas for future research. However, a major strength of our work is that, to our knowledge, this is one of the few studies to elicit stakeholder feedback about adaptations that can be made to a telepresence robot from all 3 user groups: people with MCI or ADRD, family caregivers, and clinicians. In addition, our findings are