The UCD process begins with identifying end users and understanding their characteristics, abilities, needs, and preferences [9]. Although the primary users of remote assessment technology are the participants who generate the digital activity data by living in the midst of these sensors, secondary users include anyone who installs the sensors and maintains their integrity, and those who access the remote sensor data streams to analyze activity or health outcomes. Depending on the type of research study, secondary users may include clinicians, family members, or other individuals who are involved in the maintenance of, or interaction with, the sensors or the data derived from them.

In general, UCD strategies aim to make technology more appealing to encourage uptake by users. Paradoxically, UCD strategies for passive sensing generally aim to decrease the need for the user’s interaction with the system and instead to make it as unobtrusive and as “low-touch” as possible. The remote passive sensing issues described in this paper largely concern the assumptions that users make about the sensors, rather than their interaction with them. Thus, the strategies that we have used to mitigate these issues among participants in our research groups must address the users understanding and knowledge about the purpose and function of passive sensing. In research studies, this process begins with informing and educating participants about the system prior to installation and then continues during beta testing and use in research studies.

At the National Institute on Aging–funded Roybal Center ORCASTRAIT (Oregon Roybal Center for Care Support Translational Research Advantaged by Integrating Technology), education about the study technologies begins during recruitment and continues throughout study participation. The language used in recruitment materials and during informed consent sets the tone and expectations for participants. Consent forms contain plain language about the function of remote sensing, such as “Please note: the study technologies are not a security system. There is no video, photograph, or audio recording. The technology will not detect a break-in, a fall, or other emergencies. We will not be able to notify anyone in the event of an emergency.”

Educating and re-educating participants continues during studies about the capabilities (or lack thereof) of technology. One way that the ORCASTRAIT teams achieve this is by showing participants examples of their data to illustrate the deidentified nature of the data as it appears to the research team (Figure 1). While showing participants the data depicted in Figure 1, an ORCASTRAIT technician will describe it as follows: “The presence of a shape on the graph indicates that the sensor was activated on that day. This is the 30,000 feet view of the data, zoomed out, giving us an overall picture of the status of the sensors in the homes.”

Users bring their preconceived notions of technology, though, and often retain them even after the study team’s educational efforts. Clearly, the fit between the participant’s goals and the goals of the research will not always be possible and should be taken into account. Yet, a concerted effort must be applied to ensure that we address participants’ concerns and misconceptions when possible, thereby improving readiness. These efforts must be robust in order to include the most representative sample possible since participants from diverse backgrounds vary in terms of digital literacy or misconceptions about remote sensing technology.

UCD principles are frequently considered in the initial development of digital devices and technology interventions; however, UCD is less often considered during the translation and implementation processes of intervention testing, refinement, and expansion [10]. Mental and behavioral health interventions in particular can benefit from applying UCD principles [17]. Recently, the remote monitoring system was developed by the Oregon Center for Aging and Technology (ORCATECH) and used in 2 pilot studies of behavioral interventions funded by the National Institute on Aging–Roybal Center ORCASTRAIT at Oregon Health and Science University (OHSU). These studies were conducted among persons living with dementia and their care partners to detect changes in daily routines such as sleep, physical activity, and time out of the home: READyR (Remote Assessment and Dynamic Response Intervention to Support Co-Residing Care Dyads in Identifying Dementia-Related Care Needs; ClinicalTrials.gov NCT04542109) and STELLA (Support via Technology: Living and Learning with Advancing Alzheimer’s disease and related dementias; ClinicalTrials.gov NCT04335110).

The READyR Program is a behavioral intervention designed to assess and target everyday routines of care dyads (older adults with cognitive impairment and their cohabiting partners) using in-home sensors and videoconference sessions in order to understand future care needs and promote greater well-being [18]. In the pilot study of READyR (OHSU Institutional Review Board #20397), ORCATECH infrastructure and a platform of sensors were deployed to collect objective data from remote passive assessments of daily routines. The technology platform used in the READyR study included PIR sensors that captured home exits and activity within or transitions between rooms (NYCE, wireless PIR sensors using a ZigBee network), an actigraphy watch that detected step count (Withings Activité/Steel Watch), an electronic pillbox that captured medication-taking behavior, (TimerCap, wirelessly transmits data about when and which day of the week the pillbox doors were open or closed), and an under-the-mattress bed mat that captured a variety of sleep metrics (eg, bed exits at night and total sleep time) and physiologic data such as heart rate and respirations (Emfit, ballistocardiography sensor).

One of the goals of the READyR intervention was to better assess the alignment of participants’ daily routines with their care values such as autonomy and safety. To help participants visualize their daily activity data in intervention sessions, a tailored web-based tool was created. Data specific to each participant’s home were extracted from the ORCATECH central server. These data were processed and analyzed to generate plots and figures of activity related to participants’ daily routines, which were incorporated into an HTML template to create a static web page that could be displayed for participants.

User experiences with having passive sensors in the home, and with viewing their daily activity data, were video recorded in intervention sessions and acceptability surveys. Of the sample of 24 participant dyads (half of which had cognitive impairment, and the other half were their spouses) who completed the acceptability survey, 23 participants rated the sensor data that were presented to them in web-based intervention sessions as “very easy” or “easy” to understand, and only 1 participant rated it as difficult. The most common user issue was that when presented with data about their daily routines, over half of the participants expressed doubt that the sensors accurately captured at least one of their routines such as time out of home or sleep disturbances. A few participants gave concrete reasons for their doubt, for example:

The ORCATECH platform includes multiple ways of assessing some functions, such as sleep, and participants were more reassured that the data represented their patterns when we could show them 2 sources of data that were in agreement.

The issue of therapeutic misconception arose several times during the READyR pilot study. One participant living with Alzheimer disease was prone to somnambulism, and her partner expressed hope that the sensors could offer a way to detect it and alert them, especially when she opened the front door at night. Another participant with mild cognitive impairment went for long bike rides and did not bring his phone. His spouse expressed disappointment that there was no capability to track his whereabouts with the wearable in case he got lost or hurt. A third participant living with Alzheimer disease had a tendency to fall, and his spouse was concerned that there were no sensors in their garage to detect a fall. In all cases, re-educating the participants about the functionality of the sensors, and the lack thereof, was our main strategy for addressing this UCD issue. In a future iteration and testing of the READyR Program, we plan to produce a sensor guide for families to refer to after installation and examine other ways of providing intermittent information to curb misperceptions.

The STELLA study enrolled 13 care partners and their family members with moderate to late dementia. The study tested an intervention designed to help care partners recognize and address behavioral symptoms of dementia, with the primary aim being to reduce care partner burden (OHSU Institutional Review Board #19306). The STELLA study used bed mats to assess sleep duration and quality (under-the-mattress, ballistocardiography-based sensor, Emfit), PIR wall sensors (NYCE) to assess motion, and wearables to assess step counts (Withings Activité/Steel Watch).

After completing the intervention, 4 STELLA care partners participated in a focus group that addressed their perceptions of STELLA and their experiences with technology. Several technology-related user experience themes emerged from the focus groups: (1) a mismatch between study goals and the participants’ needs; (2) trust of the system, or privacy concerns; and (3) and lack of aesthetic and minimalist design.

Care partners found that their needs were not met with the technology in that they had to manage the interaction between their family member with dementia and the sensors: “I should have put them (the sensors) up higher.” His family member with dementia saw “these little things on the wall, and pulls them off. And I find them in her pocket later.” Two STELLA care partners had difficulty convincing their family members to wear the watch. The technology on the watch did not update to adjust to time changes, also leading to frustration, as this care partner noted:

Some STELLA participants also expressed worries about privacy and the “undertrust” discussed above. They were concerned that the team could monitor their activity while they were in bed:

Some of the focus group participants did not like the appearance of the sensors, indicating a lack of aesthetic acceptability:

Another participant commented that he was somewhat mystified by the wall sensor, indicating the lack of aesthetic acceptability and minimalist design:

Despite the study team’s efforts to use unobtrusive technology, the STELLA care partners found the devices far from “passive.” The existence of the sensors in the home, even if passive, led some participants to worry about their privacy, struggle to help their family members use devices, like watches, and grapple with malfunctions:

It should be noted that all STELLA participants had experience using technological devices such as phones, computers, and email. They all had consistent support from the ORCASTRAIT team:

The STELLA focus group revealed that as we tested this intervention that addressed the care partner burden, it is possible we increased the care partners’ sense of daily hassle with the technology.

Overall, the user experience feedback from focus groups may reflect the fact that family caregivers of older adults with later-stage dementia are a distinct type of user that had not been considered in the original design of the passive sensing platform. The additional needs and preferences of a new set of users should be considered [17], and after identifying those distinct needs, the UCD process involves ideation and exploration of prototypes and testing [9]. In this case, the new user group of family caregivers of persons with later-stage dementia will prompt a redesign of the passive sensing platform before testing again with this user group.

More commonly accepted passive technologies, such as a long-term wearable device, can also highlight UCD issues in behavior intervention studies. One example is from a recent study evaluating a digitally delivered physical activity and social support intervention among community-dwelling older adults (ClinicalTrials.gov NCT03538158), which included the use of wearables (similar to a wristwatch) to objectively assess activity engagement or step count. The participants ranged in age from 65 to 98 years and some had minimal experience with technology. Although a standard commercially available wearable was selected for the study, an issue that arose was a lack of adherence among the participants with respect to the use of wearable. Some participants were not used to wearing a watch, and others were not accustomed to wearing a watch at bedtime. These aspects of comfort and acceptability resulted in gaps in the data when participants forgot to put on the wearable or chose not to wear it at night (SJ Czaja, 2024, unpublished data). This reinforces the need for reminding study participants about the function of the passive technology, and the importance of adherence with respect to study outcomes.

Long-term use of a remote sensing platform is often the intended duration. One of the advantages of a remote assessment system is the ability to monitor daily life in the home setting and detect changes over time. However, new UCD issues arise during long-term monitoring that cannot be taken into account during short-term testing of the system.

MODERATE is a longitudinal study for which the target population is couples living at their residences, with one having dementia and one being the care partner. MODERATE focuses on identifying digital biomarkers of agitated behaviors exhibited by participants living with dementia and identifying environmental precipitants of such behaviors. Passive motion sensors and contact sensors (NYCE, wireless PIR sensors using a ZigBee network), were installed in participants’ homes to continuously assess in-home activity.

Algorithms developed at ORCATECH extract time spent out of home based on whether there was motion detected between doors opening or closing. In single-resident homes, the metric, time spent out of home, can be relatively easy to interpret. However, in double-resident homes, time spent out of home refers to the time both residents are out of home and that does not include the time when only one of them is out of the home. In addition, in double-resident homes, it is difficult to attribute motion detected in a specific room to a particular person, and innovative algorithms are required to synthesize the motion sensor data. We have also developed an algorithm that extracts the duration of independent life-space activities within the home. This metric refers to the duration when there were activities simultaneously detected in two or more life spaces within the home. Such a metric can be used to infer a basic level of physical function and independence of the participants. Yet, the primary users’ unanticipated activity patterns impact the error rate of these algorithms. Long sedentary periods of sitting on the couch may look identical to an unoccupied room. Exiting and re-entering a room quickly may not be recorded as a room transition. Testing these and other behaviors using scripted trials in the home was necessary to identify and recover from these types of errors in the algorithms.

In MODERATE, we also deployed actigraphy devices, research-grade Actiwatch Spectrum, to be worn by people living with dementia, and Withings Activité/Steel Watch, to be worn by caregivers, to track their daily level of activity. Although we have demonstrated powerful examples of behavioral symptoms being detected by actigraphy devices among persons living with dementia [19], we have also encountered “errors” of the actigraphy devices being lost or damaged. One of the wearables was found in the participant’s clothes dryer. In addition, one of the caregivers’ actigraphs was struck by a person living with dementia during a bout of agitation. As a result, the glass surface broke. We learned that there are distinct UCD issues with deploying digital technology in the homes of people living with later-stage dementia as their cognition and behavioral symptoms worsen. In addition, in our research settings, the Actiwatch Spectrums (activity monitoring watch with solid-state “piezo-electric” accelerometer and color-sensitive photodiodes) only have around 4 weeks of battery life. As a result, we continuously need to swap out devices via mailing which is labor-intensive for both our staff and participants.

The Collaborative Aging Research Using Technology (CART) initiative was an interagency (National Institutes of Health and Department of Veterans Affairs) program of research funded to spur wider, evidenced-based use of digital technologies in aging research. CART developed and validated protocols and infrastructure in a longitudinal demonstration study, enrolling 301 participants across 4 diverse cohorts (rural residing residents, older adults living in low-income housing, African American older adults living in Chicago, and Latino older adults living in Miami). The project demonstrated that the system could enable digital health monitoring and intervention delivery that is technology agnostic and extensible to many use cases and research scenarios [20]. The scalable system is being used by multiple research groups in North America and Europe [21]; versions of this platform were used in the READyR and STELLA studies described above. In developing the system, priority was placed on ensuring that the methods were incorporated into the daily lives of diverse populations of older adults. User experience feedback was crucial and remains key to the continuous improvement of the experience and engagement of older populations in research; this is crucial to ensuring long-term adherence to research protocols, reducing missing data and maintaining ecological validity.

To highlight UCD principles further, we note some observations from the CART study. For example, the low-income housing cohort (participants living in US Department of Housing and Urban Development, Section 202 subsidized housing) and the rural-residing veterans cohort were found to have overall excellent daily adherence to wearing a wrist-worn watch (Withings Activité/Steel Watch) that recorded step counts and sleep times, with an overall daily wear compliance of 86.5% [22]. However, it was observed that despite using a device chosen for its typical wrist-watch appearance (ie, higher aesthetic acceptability for a research-device wearable), which required battery replacement only every 8-9 months, and for which there was clear participant agreement at the study entry that the watch was intended to be worn throughout the day and night (except for bathing), the wear compliance at night was found to be lower after a mean of 4.7 (SD 1.8) months of study (69.5%) [20]. When asked about the lower wear compliance at bedtimes, participants relayed that they found they were not comfortable wearing any wrist-worn device every night, even though they thought that this would be feasible for them at the beginning of the study. This experience highlights, as noted here and for the READyR and STELLA intervention studies described above, that user-experience needs to be studied under the actual conditions of use. CART participants using the watch in short-term (2 weeks in duration) device validation pilot studies [23] did not voice this potential longer-term compliance issue.

Although the above examples illustrate how participants may choose for understandable reasons to not comply with a protocol because of usability factors and personal life circumstances, we also note that many dedicated participants will invent or devise “workarounds” to enable them to continue in a study. These must be identified during the study so that, for example, if a protocol states that a wearable should be worn all the time (as in our example) or worn on the nondominant wrist, and a participant decides they are more comfortable wearing the device on their other wrist, at variable times, this can add inaccuracies to the data. In some cases, we have discovered that participants devise an alternative that does not affect the data (Figure 2). Participants’ ideas and solutions are welcome, and in fact, the research team may gain valuable knowledge from their relating to that experience. Finding ways to intentionally incorporate the user experience into the design process of remote sensing research can mitigate many UCD challenges.