Technology options to help people with dementia or acquired cognitive impairment perform multistep daily tasks: a scoping review

Purpose – The purpose of this paper is to review studies that evaluated technology-based prompting systems for supporting participants with dementia or acquired cognitive impairment in their performance ofmultistep dailytasks. Design/methodology/approach – A scoping review was conducted to identify eligible studies through a search of four electronic databases, that is, PubMed, PsycINFO, Web of Science and Institute of ElectricalandElectronicsEngineers. Findings – Thesearch,which coveredthe2010 – 2020period, ledtothe identification of1,311articles, 30 of which were included in the review. These articles evaluated six different types of prompting systems: context-aware, automatic computer prompting, context-aware, mediated computer prompting, teleoperated robot prompting, self-operated augmented reality prompting, self-operated computer or tablet prompting andtime-based(preset)computer,tabletorsmartphoneprompting. Originality/value – Technology-aided prompting to help people with dementia or acquired cognitive impairment perform relevant multistep daily tasks is considered increasingly important. This review provides a picture of the different prompting options available and of their level of readiness for application indailycontexts.


Introduction
People with dementia or acquired cognitive impairment (i.e. forms of the functional decline occurring in the early stages of neurodegenerative diseases or subsequent to neurological damage such as traumatic brain injury) represent a vast and increasing part of the world population and call for medical/pharmacological interventions, as well as behavioral interventions (Gagnon-Roy et al., 2020;. Medical/pharmacological interventions are generally focused on identifying and implementing treatments to halt further deterioration of people's social and cognitive functions and possibly improve those functions (Liu et al., 2020). Behavioral interventions are typically directed at setting up forms of environmental support to help the people remain functionally active, and thus maintain a level of practical integration within their daily context and a positive social image and succeed in slowing down any further decline .
One of the main goals of behavioral interventions for people with mild or moderate cognitive impairment and mild or moderate dementia is to support their performance of functional multistep daily tasks (e.g. preparing meals and drinks, washing hands or going through the (Information about the authors can be found at the end of this article.) morning routine) (Braley et al., 2019). Such support is important because these people are often unable to carry out those tasks independently largely due to memory problems that interfere with their recalling of the task steps and their organization of those steps in an orderly fashion (Lancioni et al., 2017;Pereyda et al., 2019).
Supporting these people's functional task engagement through direct caregiver supervision does not seem a very realistic or desirable option. In a daily context, in fact, caregivers may not necessarily have the time and energy to provide extensive supervision. Moreover, performance under supervision can make people with dementia or cognitive impairment feel dependent and inadequate and develop/consolidate a poor social image of themselves (Burleson et al., 2018;Mahoney et al., 2015).
An alternate to increased caregiver supervision may be the use of technology systems that provide prompts (instructions) for the single steps of those tasks. Such prompts are intended to compensate for the aforementioned memory and executive problems, and thus enable people to manage the tasks in a largely correct and independent manner (Rohrbach et al., 2019;Wolf et al., 2018). Those systems vary widely in terms of technological complexity (e.g. from systems relying on robots to systems relying on tablets or smartphones), prompt conditions (e.g. from automatic prompt delivery to self-prompting) and prompt characteristics (e.g. from simple verbal or pictorial prompts to combinations of verbal and video prompts) Pinard et al., 2019;Wang et al., 2017). This paper is aimed at providing a picture of the different systems and their use with people with dementia or acquired cognitive impairment. Specifically, this paper is directed at reviewing studies carried out between 2010 and 2020 that evaluated those systems so as to determine the level of development (readiness) reached by the systems and the systems' possible/expected contribution in supporting people's performance of multistep tasks in daily contexts. Providing such a general picture to professionals working in the area may be considered useful to orient their choice of technology options in daily practice and stimulate their research initiatives to add essential evidence about those options.

Search strategy
A systematic search was conducted following the Preferred Reporting Items for Systematic Reviews and Meta-Analyses Extension for Scoping Reviewes (PRISMA-ScR) (Tricco et al., 2018) to identify studies that relied on technology arrangements to enable persons with dementia or acquired cognitive impairment to perform multistep tasks. A scoping review approach was used, as our primary aim was to examine the range of technology options available and identifying knowledge gaps (Munn et al., 2018).
Studies were identified through a search of four electronic databases, that is, PubMed, PsycINFO, Web of Science and Institute of Electrical and Electronics Engineers (IEEE). The search process included the following free-text terms for the first three databases: (dementia OR Alzheimer Ã OR mild cognitive impairment OR acquired cognitive impairment OR brain impairment OR brain injur Ã ) AND (assistive technology OR technology) AND (instruction Ã OR prompt OR prompting OR micro-prompting OR step by step OR multistep). For the IEEE database, the terms technology and assistive technology were omitted, as they were deemed redundant in that the IEEE database includes primarily technical content.
The search, which covered the period between January 2010 and August 2020, resulted in a total of 1,311 papers. The number of papers was reduced to 1,048 after duplicates were removed. Figure 1 illustrates the search process and outcome. Initially, titles and abstracts of the 1,048 papers were screened. When the titles and abstracts were judged to be in line with the inclusion criteria (see below), the corresponding full-text articles were downloaded.
Following this process, 73 full-text articles were downloaded. Those full-text articles were then read by the first and second authors and 29 of them were found eligible for the review. Subsequently, the references of these 29 articles were inspected to possibly identify other relevant articles. In addition, a Google Scholar "cited by" search was conducted using the aforementioned 29 articles. This search led to the finding of one extra article, with the consequence that 30 articles were finally included in the review.

Inclusion and exclusion criteria
Two basic inclusion criteria were used in selecting the studies for the review. First, the studies involved the participation of adults (i.e. aged > 18 years) with dementia or acquired cognitive impairment (i.e. forms of the functional decline occurring in the early stages of neurodegenerative diseases or subsequent to neurological damage such as traumatic brain injury). Second, the studies used technology systems aimed at helping the participants to perform one or more multistep daily tasks. Studies aimed at developing technology systems to support multistep tasks were excluded if the evaluation of the systems occurred via focus groups or the involvement of participants other than those on whom this review was concentrated (studies such as those reported by Bouchard et al., 2020;Gagnon-Roy et al., 2020;Pereyda et al., 2019;Wilson et al., 2019;Yaddaden et al., 2020).

Interrater agreement
Interrater agreement was checked between the first two authors on scoring the eligibility of the 73 full-text articles, which were downloaded after screening titles and abstracts and on reporting the data extracted from the articles reviewed (see Results and Table 1). Interrater agreement on this latter measure was checked over 10 articles. The percentage of agreement on the 73 full-text articles was 97%. That is, the authors provided the same label (i.e. "included" or "excluded") for 71 of the 73 articles. Consensus between authors on the two articles with initial disagreement was then achieved after a brief discussion. Interrater agreement on reporting the data extracted from 10 of the articles included in the review was 100%.

Data extraction and coding
Data were extracted in terms of country in which the study was conducted, number and functioning characteristics (i.e. dementia or cognitive impairment) of the participants included, tasks targeted, setting where the study was implemented (i.e. with the use of the technology), the method used to assess the impact of the prompting system and data reported on task performance (Table 1). Moreover, following a consensus-based approach between the first two authors, codes were created to group the studies included in the review (results) according to the forms of technology support they used to foster multistep task performance.

Results
As indicated above, 30 studies were identified that relied on technology arrangements to enable persons with dementia or acquired cognitive impairment to perform multistep tasks (see Table 1 for an overview). Those studies were conducted in Canada (n = 8), UK (n = 8), Italy (n = 7), the USA (n = 3), Taiwan (n = 2) and Germany (n = 2). In total, 302 participants were involved.
The various forms of technology reported by the studies show clear differences in terms of the level of support afforded to the participants to enable their successful task performance. Forms of technology providing context-aware, automatic prompting, for example, are designed to ensure extensive support (i.e. require minimal human input/oversight). Conversely, forms of technology entailing self-operated prompting ensure relatively moderate support. Accordingly, the complexity of the former types of technology is much higher than the complexity of the latter types.
The studies show large variability in terms of the methodology used to assess the effects of prompting. A concise description of the methodological approach adopted in each study is provided in Table 1 (see "Assessment methods" column). In total, 14 studies used a clearly recognizable experimental approach (i.e. ABAB, multiple baselines, multi-element, multiple probes, alternating treatments and cross-over designs or a randomized controlled trial; Table 1). The remaining studies reported simple AB (baseline-prompting) sequences, Prompting led the participants to complete the task more than 90% of the times comparisons of various prompting strategies or of a prompting strategy with other conditions without specifying the design used and video recordings of prompting trials, interviews or self-rating. The following paragraphs summarize studies that clearly illustrate the different types of technology systems available and some of their applications.

Context-aware, automatic computer prompting
The eight studies using this type of technology have set up systems capable of monitoring the participants' responses and determining the kind of prompting to provide to help the participants' successful task performance. For example, Hoey et al. (2010) and Czarnuch et al. (2013) used a system known as Cognitive Orthosis for Assisting aCtivities in the Home (COACH; see Mihailidis et al., 2008) to help participants with mild to severe dementia to wash their hands successfully. The system involves a video camera mounted above a sink, which feeds a tracking device (following the position of hands and towel), an evaluation unit that estimates the progress of the participant in the task and an intervention unit that translates the information on progress difficulties into action. Action may involve an audio or video prompt or a call for human assistance. Hoey et al. (2010) adopted an ABAB (in which baseline and system use were alternated) but did not report specific data as to the level of improvement in handwashing the six participants of their study achieved with the help of the system. Czarnuch et al. (2013) reported that the 20 participants included in their study completed 206 of the 246 steps available in the trials recorded (83.7%). The system correctly identified 96 of those steps (39.0%) as completed, while it failed to correctly identify the other 110 steps completed (44.7%).
O'Neill et al. (2018) used a system called Guide to support the morning routine of 10 people with mild to moderate cognitive impairment due to brain damage. The Guide system relies on a computer with a voice tracker, speech recognition software, activity protocols and activity protocol player. Specifically, the computer entails audio-verbal interactive forms of prompting emulating the verbal prompts and questions that are characteristically used by staff. During the morning routine, the system was set up to present the participants with a variety of step-related checks (questions) and instructions/prompts. Staff intervention was used when the participants failed to make progress in their performance. Data indicate that the system had a beneficial impact. In fact, the 10 participants using the system were able to complete the routine with a significantly smaller number of staff interventions than counterparts randomly assigned to a control group not using the system.

Context-aware, mediated computer prompting
The four studies using this technology model, like those of the previous group, seek to monitor the participants during task performance and to provide them with the type of prompts that are matching their performance needs. In contrast to the studies of the previous group, however, these studies do not deliver the prompts automatically. Essentially, they rely on sensors technology to monitor the participants' task behavior and need of prompts and human supervisors to approve or activate the prompts to be delivered. For example, Wang et al. (2019) carried out a study aimed at supporting 16 participants with mild to moderate cognitive impairment (secondary to traumatic brain injury) in performing two cooking tasks. The experimental kitchen included an integrated sensor network and prompting elements, which allowed monitoring the participants' behavior and to deliver verbal and visual prompts subject to confirmation/approval from a human supervisor. The authors compared the effectiveness of this prompting model with the effectiveness of an iPad Mini, which required the participants to activate prompts on their own, according to a randomized cross-over design. Task performance data indicate that the participants required a significantly lower amount of assistance from the investigator with the context-aware system than with the self-operated iPad. The system proved relatively accurate, thus requiring mere supervisor's approval, regarding prompt decisions based on contact switch sensors (e.g. decisions based on the participants' opening/closing cabinets and drawers). Decisions based on power consumption and Kinect sensors were less accurate (i.e. only about 60% of them were correct) asking for the supervisor's specific/corrective input.

Tele-operated robot prompting
The four studies carried out to ensure this form of prompting rely on human supervisors to operate the robot's prompts in relation to the participants' needs as determined by the monitoring process. For example,  carried out a study involving 10 participants with mild to severe dementia who were informed that the robot's movements, speech and orientation/pointing were controlled by a researcher (teleoperator). The participants were asked to interact with the robot (and use its help) while making a cup of tea in the kitchen of a simulated home. The teleoperator continuously monitored the participants' task progress and their overall mood conditions in a video stream sent by the robot and made the robot initiate social conversation, ask task-related questions, provide confirmations and deliver prompts to guide the participants toward successful completion of the tea-making task. The robot enabled the participants to start the task steps but delivered the appropriate prompts if the participants looked around or asked for directions. The prompts could vary from a suggestion to a direct verbal instruction plus a video display. If a participant asked a question about the location of an item needed for a specific step, the robot indicated the place by orienting to it. Data on task performance are not specifically reported. The authors' analysis of task video recordings and post-intervention interviews focused on a number of participants' behaviors such as interaction, natural dialogue and emotion. Such an analysis was extended by Rudzicz et al. (2015), who concentrated particularly on the communication between participants and robots to identify difficulties that need to be addressed.

Self-operated augmented reality prompting
The two studies using this type of technology follow the view that augmented reality instructions/prompts might be advantageous over paper-based and screen-based instructions (Lin et al., 2016). They used a head-mounted display (i.e. Microsoft HoloLens) to visually present the instructions/prompts and guide the participants through the task steps (Rohrbach et al., 2019;Wolf et al., 2019). For example, Rohrbach et al. (2019) developed a prompting application installed on a HoloLens to enable participants with mild and moderate dementia to prepare a cup of tea. A prompt consisted of a holographic simulation of the corresponding step projected on the head-mounted display that the participants wore during the tasks. The holographic simulation was supplemented by a young female voice instructing the participant about the step illustrated by the hologram (i.e. to be performed) and the appearance of the corresponding written instruction. The participant could proceed to the next step by uttering the word "Next." In total, 10 participants were involved in carrying out the task via the augmented reality application and a standard condition (i.e. without such application) according to a cross-over design. Data show that seven participants managed to carry out the task with the augmented reality application. This performance was not statistically different (not more satisfactory) than that observed in the standard/control condition.

Self-operated computer or tablet prompting
The four studies carried out to set up and evaluate this form of prompting stress the importance of simple technology that can be used in daily (home) contexts. Early work by Boyd et al. (2017b) was essentially directed at developing a simple prompting product that could be adjusted by a caregiver to fit the functioning level of participants with mild or moderate dementia so as to help them carry out multistep tasks independently at home. Their study compared four types of prompting, that is, written text, verbal prompts, picture prompts and video prompts. The prompts were displayed on the touchscreen of a tablet. The nine participants were to try each of the prompts on different tasks. Together with the evaluation of the various types of prompting, the authors also assessed the possibility of using a touch/push area on the screen to enable the participants to move to the next prompt for the next task step. With regard to the prompts, the authors reported that written text and verbal instructions were more effective with one of the two tasks directly evaluated while no differences among prompts were found with the second task. The touch/push area on the screen was a viable means to allow access to prompts. Task performance data suggest that participants could still require some caregiver support to complete the tasks.  extended the work just described by assessing whether caregivers and people with dementia could manage to use a prompting package (i.e. a tablet and a manual) without any previous training on it. Their preliminary data seem quite encouraging in that eight of 11 participants with dementia had some improvement in their task performance .

Time-based computer, tablet or smartphone prompting
The eight studies using this approach are based on the view that providing participants with task step prompts occurring at preset intervals (i.e. at intervals deemed appropriate to enable the participants to complete the steps being prompted) may be advantageous as compared to asking the participants to operate their own prompting (Lancioni et al., 2011). Indeed, a system-based prompt delivery spares the participants from the need of operating their prompts (i.e. from an extra demand on their weakening memory function). Programming the intervals between prompts based on the participants' performance pace is critically important to secure a timely occurrence of the prompts and enhance the prompts' efficacy. For example, Lancioni et al. (2017) evaluated such an approach with eight participants with mild to moderate dementia using a non-concurrent multiple baseline design across participants. For each participant, 12 or 14 daily tasks of practical relevance were selected (e.g. preparing coffee, setting the table and watering plants). The technology included a tablet device with a Talking Alarm Clock application and a wireless Bluetooth earpiece through which the participants received the tablet's verbal outputs. Specifically, the tablet was programmed to remind the participants of any specific task at the time when the task was due, provide prompts for the task steps and deliver encouragement and praise in between prompts to foster their motivation to remain active and accurate. The task performance data indicate that the participants started (virtually) all tasks independently in relation to the tablet's reminders and carried out nearly or more than 90% of the task steps correctly following the tablet's prompts.

Discussion
Supporting the independence of people with dementia or acquired cognitive impairment in performing functional tasks is one of the main goals of behavioral interventions. The aim of this paper was to provide an overview of technology-based prompting systems to promote independence in daily tasks that require the execution of a sequence of steps (i.e. multistep tasks). In light of the findings of this scoping review, it is relevant to discuss the prompting systems reported in terms of complexity and readiness for use within applied contexts. Regarding the complexity aspect, one could divide the systems into at least two groups. The first group would involve the more sophisticated systems, that is, contextaware, automatic computer prompting, context-aware, mediated computer prompting, teleoperated robot prompting and self-operated augmented reality prompting. The second group would include the remaining two systems, which are much simpler.
j JOURNAL OF ENABLING TECHNOLOGIES j

Complex systems
If one examines the COACH system, for example, the first consideration is that it is aimed at providing appropriate prompting to the participants in an automatic and independent manner. Such an objective is based on the use of monitoring devices that inform the prompting system as to the participant's progress or difficulties and cause such system to provide the prompting the participant requires at any particular point of the task sequence. While this approach has been proven promising in earlier studies (Mihailidis et al., 2008), the accuracy of the monitoring process (i.e. the dependability of the intelligent judgment of the system) may not yet be viewed as satisfactory. The data provided by Czarnuch et al. (2013) with regard to the COACH's accuracy in identifying task steps completed are an indication of the problem and pose questions as to the system's readiness for general use.
Whether the problem noticed with COACH (i.e. a system based on visual monitoring) might be less serious in a system such as Guide (i.e. based on verbal monitoring) is not known. It should also be pointed out that the data reported on task performance need to be taken with caution due to methodological issues such as the involvement of small numbers of participants or the use of observational data.
Context-aware, mediated computer or robot prompting systems might be viewed as technology packages, which have achieved different levels of development. For example, the computer prompting system described by Wang et al. (2019) appears to be at an advanced stage of development compared to other systems. In fact, the authors' evaluation goal was to determine how much human supervision the system still required and what obstacles needed to be overcome to make it function independently. The robot prompting system described by  and Rudzicz et al. (2015) appears to be at a lower stage of development. The experimental work was focused on determining how functional the robot might be in guiding the participants through the tasks, but the robot was still operated by a human supervisor. That means no evidence exists as to whether the robot can move and provide accurate prompting based on environmental information (i.e. ambient sensors and participants' verbal and physical behavior). Data on task performance (reported by all studies using mediated computer prompting and one of the robot-based studies) may need to be viewed with caution, as only Wang et al. The self-operated augmented reality prompting systems appear less complex than the previous systems. Yet, their readiness for use is difficult to judge due to the exploratory nature of the studies conducted and the uncertainty as to the willingness and ability of people with dementia or cognitive impairments to adapt to the unnatural interaction with the holographic system (Rohrbach et al., 2019;Wolf et al., 2019).
In light of the comments made above, one might argue that the complex systems have significant potential but need to undergo further developments and/or assessments before they can be considered suitable and ready for use in daily contexts. Those developments would need to make the systems function independent of human inputs, have a level of accuracy sufficient to improve the participants' task performance significantly, have an application cost that is affordable for daily contexts and be acceptable to the participants and their caregivers (Wang et al., 2017).

Simple systems
Self-operated computer or tablet prompting appears the simplest system of all. Indeed, such a system is particularly straightforward for caregivers to set up and seemingly helpful for supporting the participants' task performance. One more advantage of this system is that it is highly affordable in terms of cost and suitable for home environments . A question one might raise about this system is that the need to operate the prompting may be a progressively significant burden on the participants' memory function.
This function, which grows weaker in people with more severe forms of cognitive impairment and dementia, may become insufficient to guarantee that the participants operate the prompts reliably, and thus maintain reasonably high levels of correct task performance (Chang et al., 2013;Lancioni et al., 2017).
Time-based computer, tablet or smartphone prompting may be viewed as a system complementary to the self-operated prompting mentioned above. Time-based prompt delivery provides the participants with a regular stream of prompts related to the different steps of the tasks to be performed. This programmed prompt availability spares the participants from the need of operating the prompts and allows them to fully concentrate on the task-step performance those prompts are meant to support. A difficulty inherent to this system concerns the identification of adequate intervals between prompts. To tackle this difficulty, one would need to ensure that the intervals are set up for each participant following preliminary observations of the time the participant requires for the single task steps and the intervals are re-adjusted over time if the participant's performance speed changes. Although these measures cannot be a guarantee of errorless performance, they can increase the likelihood of success and heighten the participant's level of satisfaction (Lancioni et al., 2018).
In light of the aforementioned comments and the performance data reported, one could argue that those systems represent a viable option for daily contexts. Indeed, they may constitute the only realistic option at this point in time given the fact that the more complex/ sophisticated systems may not be ready or accessible for daily use. It may also be noted that the simple systems would be largely affordable and easy to transport, while the more sophisticated systems would reasonably be rather expensive, as well as difficult to move across settings.

Limitations
Two limitations of this paper may be underlined. First, the focus on articles written in English may have prevented the inclusion of pertinent studies reported in other languages. Second, although the search strategy we used has been refined over a series of pilot searches to ensure comprehensiveness, the exclusion of relevant databases such as the Association for Computing Machinery (ACM) Digital Library might have reduced the number of articles eventually identified. Notwithstanding this latter limitation, the combination of academic databases used for this review is generally considered adequate to ensure an acceptable outcome (Bramer et al., 2017).
It might also be stressed here that the studies reviewed do not allow to formulate any definite statements about the effectiveness of prompting technology to support independent task performance in adults with dementia or acquired cognitive impairment. This is due mainly to the absence of clear experimental designs in many of the studies and to the rather small sample sizes involved.

Conclusion
Six technology systems to support multistep task performance in people with dementia or cognitive impairment were analyzed: 1. context-aware, automatic computer prompting; 2. context-aware, mediated computer prompting; 3. teleoperated robot prompting; 4. self-operated augmented reality prompting; j JOURNAL OF ENABLING TECHNOLOGIES j 5. self-operated computer or tablet prompting; and 6. time-based computer, tablet or smartphone prompting.
The first four systems involve sophisticated technology arrangements while the last two are based on rather simple technology solutions. Application of the complex systems in daily contexts would require new developments of such systems capable of increasing their accuracy, ensuring their functioning independent of staff and making them affordable, as well as acceptable to participants and caregivers. The relatively positive task performance data reported for the simple systems suggest that they may represent the only realistic option for daily use at this point in time. In fact, notwithstanding their limits, they may provide a meaningful level of support at a very modest cost and with minor operational difficulty. New research will need to advance the development of the complex systems (with the aim of making them suitable for use in non-experimental settings) and extend the evaluation of the simple systems and possibly upgrade them to increase their applicability in daily contexts and their impact on the lives of people with cognitive impairment or dementia, as well as on the lives of their caregivers.