Stroke has become a global problem [1]. One new case is reported every 2 seconds, and the number of stroke patients is predicted to increase by 59% over the next 20 years [2]. In the United Kingdom alone, more than 100,000 stroke cases are reported annually [1], with impairment or disability affecting two-thirds of the 1.2 million stroke survivors [1]. In the United Kingdom, only 77% of stroke survivors are taken directly to the stroke unit. Due to the high number of patients, in England, for example, the social care costs are almost £1.7 billion per annum. The social care cost varies with the age of the patient: the older the patient, the higher the cost. The cost for a person who has had a stroke was reported in 2017 to be around £22,000 per annum. Thus, cost is one of the main drives for service delivery practices. In that respect, early discharge units have been used due to better outcomes and greater success on rehabilitation. Early discharge units consist of specialized personnel who offer an intensive rehabilitation program to the patient. However, after this intensive program of relatively short duration, the patient is discharged and continues the rehabilitation at home. This is expected to reduce costs by £1600 over 5 years for every patient, according to a 2017 report [1].

Due to increasing pressure to discharge patients early from hospital [3], they rely increasingly on home rehabilitation to improve their condition after discharge. As a result, the need has been increasing for home rehabilitation systems that are not dependent on specialist or clinician operators [1,4,5] while providing service similar to a clinical environment. Technological advances in home rehabilitation have been mainly focused on motor control impairments due to their prevalence in the patient population (85% worldwide [1]).

Rehabilitation in a home environment can prove more efficient than that in a clinical environment, as the home environment supports patient empowerment through self-efficacy [6,7]. The presence of supportive family members and a familiarity with the space are significant contributors to motivation. Additionally, rehabilitation in cooperation or in competition with family members demonstrates higher level of engagement [8].

Though rehabilitation in the comfort of a patient’s home seems an attractive option, home environments have limitations that can affect the use of clinical devices. The most prevalent limitations are related to space and the lack of qualified personnel to operate devices. The number of occupants; the patient’s mobility, individual personality, and mood disorders following stroke; and sound insulation, home modification requirements, and cost [9,10] also contribute to limitations of home rehabilitation. Finally, different age groups react differently to technology and devices; for example, elderly survivors often do not engage with wearable devices or video games [11]. As a result, stroke rehabilitation requires a person-centric approach that is suitable for the home environment and that does not require infrastructure change in the home.

The success of stroke rehabilitation depends heavily on personal commitment and effort. Recent studies, for example, on applied psychology in behavior change theories for stroke rehabilitation [12-14], do support that the self-esteem of the patient is limited after stroke. In addition, there is an extended sedentary period due to disability and, thus, different programs of activities are set to motivate the patients. Thus, the patient’s motivation and engagement have a critical impact on the success of any routine that is to be encouraged [15]. This is especially critical for devices used at home, since patients are usually interacting with them alone without frequent checks. Indeed, if a device does not provide a high level of engagement or motivation enhancement, it is more likely to be abandoned within 90 days [16]. Motivation levels depend on the individual, their achievements, and their needs at each given point in time. For example, once the patients achieve their physiotherapy exercise targets, they lose motivation for further practice. There are 3 main approaches to enhancing patients’ motivation: (1) goal-setting theory, (2) self-efficacy improvement theory, and (3) possible selves theory.

Regardless of the approach implemented, several factors affect motivation positively (Figure 3) or negatively (Figure 4). The main contributor to a positive motivation effect is the information available to the patient. This includes acknowledgment of the condition, control of one’s actions, achievement of goals, individualized care, overcoming an uncertain psychological condition, and receiving timely feedback [15,23,26]. Motivational feedback can be oral or visual. Also, receiving performance feedback is instrumental in maintaining motivation and engagement. Finally, self-selecting goals and having personal control is a major contributor. Negative factors usually arise from the patient’s environment and, thus, care needs to be adapted to these environmental aspects to minimize their impact [27,28].

Additionally, constructive, supportive, and competitive motivational activities, such as specially designed games, can further enhance motivation and engagement with the required rehabilitation activity [8,22].

Based on the above approaches, motivation levels can be increased and engagement maintained at a high level. This is particularly important for home rehabilitation, as it reduces the requirement for caregiver engagement and provides greater control and independence to the patient. Thus, home rehabilitation has a direct impact on the cost of care and the requirement for physiotherapist visits.

Moreover, patients who have an increased capacity to perform daily activities are less dependent on other family members or care providers. This personal improvement in daily activities turns the home environment into a positive contributor to rehabilitation and recovery.

We aimed to examine the state-of-the-art in home rehabilitation systems and to assess their suitability and functionality from a patient engagement perspective. Although several review (narrative and systematic) articles have been published on rehabilitation technologies focused on particular areas of the taxonomy (eg, wearable sensor systems review [21] and robotic systems review [29]), to our knowledge, no extensive narrative review of existing home-based rehabilitation technologies to identify criteria for designing future solutions has been done.

Our goal was to make the following contributions: (1) extend the state-of-the-art in assessment of home-based rehabilitation by combining research from 3 research domains: motivation enhancement as part of patient psychology, home rehabilitation technologies, and monitoring technologies through an interdisciplinary approach; (2) provide an in-depth narrative review of home rehabilitation systems that addresses both information and communication technologies and mechanical engineering solutions; (3) develop a patient motivation and engagement analysis of the reviewed technologies; and (4) identify a list of comparative criteria and successful device requirements to address patient motivation and engagement designed based on research findings from all 3 research domains.

The literature search returned a total of 307,550 results after the inclusion criteria were applied as presented in Table 1.

We used the following exclusion criteria to identify the most relevant sources and reduce the number of literature search results: (1) no relevance to stroke rehabilitation in the home environment, (2) trained personnel required to operate the technology; (3) medication or other clinical intervention required, (4) no report of engagement or motivation as a result of using the technology or other form of patient feedback, (5) no description of the technology, (6) no report of usability especially for older people, and (7) no additional contribution to the review findings compared with the previously reviewed articles.

Overall, we read 420 sources, as we excluded the majority by reading the abstracts. A total of 96 sources remained for analysis after meeting the inclusion criteria and having not been eliminated through the exclusion process.

Some rehabilitation technologies required renovation and other modifications in the patient’s home [9,10,16]. Other challenges for the successful deployment and engagement with home assistive rehabilitation technologies were design and technical limitations [43], and many systems did not meet the acceptance and motivation criteria as reviewed above. Indeed, retrofitting infrastructure in existing homes can be significantly more challenging than designing a smart home that will already be equipped with embedded technology. To avoid these issues, research has mostly focused on smart home environment or monitoring devices that stand alone and do not require redesign of the home [94]. Such systems mostly focus on monitoring generic parameters and provide individualization through pattern recognition algorithms, but do not contribute to rehabilitation activities. Hence, monitoring systems can be tailored to the individual home environment [95-101] and to study individual patterns. To support rehabilitation, their scope would need to be altered to encompass rehabilitation goals, and patient motivation and engagement, while at the same time being transferable (supporting different application domains).

Systems for smart home environments have been proposed for various health-related applications [11,94-101]. They usually require extensive installation of sensors in several locations such as doors, windows, electrical appliances, and furniture. Monitoring devices can provide increasing understanding of the home environment, and of the patient and their condition; they can even provide a diagnosis. Such systems might be more appropriate to support rehabilitation based on performance of daily activities [21,22].

Research in this area has focused on machine learning for information extraction based on recorded data streams [11,94-101]. However, other challenges are introduced when such devices are used, including data security, data correctness, and device operational efficacy [22,101]. Cavallo et al [102] presented an example of such a system targeting rehabilitation. The system provided monitoring and remote supervision, but did not actively support rehabilitation. Some monitoring systems focused on specialist support [102] and behavior analysis [22]. Thus, rehabilitation, motivation, and engagement are outside the scope of these systems (leading to low scores in motivation).

According to Fell et al [22], there is a need for new monitoring devices that incorporate different sensors or input data streams. Figure 6 presents the additional information that can be used to support patient rehabilitation at home. The environment must be monitored and any deviation from the “normal” behavior must be identified. Changes in Figure 6 refer to identification of unexpected behavior, events, or abnormalities identified in the recorded data streams. “Other sensors” refers to the need for data fusion, acquisition of information from multiple sources, and a higher level of information extraction. For example, energy patterns can be identified through active power (smart meter) measurements; appliance use, though, might require other information such as time, temperature, location, and motion. Similarly, a health event may be recorded by a single monitoring parameter (eg, glucose level dropped below a threshold). However, changes in behavior and competence in daily activities require a series of additional measurements (eg, sound, motion, temperature, humidity, gases).

We found a gap in applying monitoring technologies (supported, for example, by machine learning, fusion, or pattern recognition) to rehabilitation that, at the same time, support patients in performing daily activities and enhance motivation. We suggest that, by using these monitoring technologies, individualization can be achieved for rehabilitation purposes, via providing appropriate feedback, applying machine learning to the individual patient, and focusing on daily activities, thus meeting the acceptance, motivation, and engagement requirements reviewed above (see Enhancing Motivation).

Tables 2-4 summarize the criteria we selected for the comparative analysis. The methods for selecting the criteria were as follows. For Table 2, we selected these criteria based on the narrative review of motivation and engagement aspects we analyzed (see Enhancing Motivation). For Table 3, we selected these criteria according to commonly used and evaluated metrics in the majority of the reviewed articles. This was additionally informed by the conclusions outlined in the Enhancing Motivation and Home Rehabilitation Systems sections. For Table 4, we selected the criteria to meet other acceptability and economic aspects (including long term usability and transferability), as well as a separate category for the application area presented in the reviewed articles.

We used the extracted information from the reviewed articles to establish the criteria and to identify whether the criteria were met by the proposed systems. For the engagement and motivation criteria, as well as acceptance, all of the reviewed articles reported results on a common basis; thus, we needed no additional steps to cross-validate the reported results.

Tables 2-4 present a detailed comparative analysis of all the aforementioned technologies that were applicable for use in the home environment. We selected the technologies as representative examples of each category we analyzed.

Our analysis identified 3 aspects of technologies that we used for comparison: (1) motivation, (2) acceptance of technology, and (3) technological aspects. We selected these aspects for their importance in supporting patients’ motivation and engagement (motivation) and in being incorporated into patients’ rehabilitation routines (acceptance, technology).

For each aspect, we identified several comparison criteria. Regarding motivation, the criteria are (1) the motivation method used, (2) the patient’s engagement with the technology, and (3) whether the technology supports daily activities as an additional measure of motivation. There are 3 motivation methods: cooperative, supportive, and constructive. When the method used in a technology was not specified, we characterized it as general. With respect to acceptance, the criteria are (1) individualization of the device to meet patients’ needs, (2) suitability of the device for elderly patients, (3) the use of wearable components, and (4) the use of intrusive monitoring methods (eg, wearable sensors, on-body sensors, cameras, microphones). Wearable and intrusive methods have a negative impact on acceptance. Technological aspects are (1) intended use for the technology (monitoring, rehabilitation, diagnosis), (2) cost, (3) complexity, and (4) transferability to other domains.

Regarding intended use, besides our focus of rehabilitation, we also included 2 systems that perform monitoring and diagnostics.

“Yes” in the table highlights that the system met the criteria, which was demonstrated in the published work, subject to our interpretation. “No” indicates that the system did not meet the criteria.

We assessed a system to be individualized or personalized or person centric when it learned or adapted to the needs of a particular patient by incorporating some type of feedback loop mechanism where the device adjusted the requested task(s) to the ability of the patient. Examples of such mechanisms are machine learning approaches and increasing task difficulty. We assessed suitability for the elderly based on Debes et al [11]. We classified nonwearable (on-body sensors, wearable components) and nonintrusive (cameras, microphones) systems according to the system inputs that were used. The intended use of the system can be for rehabilitation, smart home monitoring, or smart home diagnosis of a health condition.

In the analysis, we considered systems that could be purchased by an average household in the United Kingdom to be cost-effective. We considered systems that would require a high capital investment, and thus reimbursement from the health care provider, to be not cost-effective. We considered technologically complex systems to be those that had a significant number of components, required significant training before use, or required extensive installation to be usable in a household. Finally, transferable systems were those that could be used for other rehabilitation purposes and were not restricted to stroke rehabilitation.

As the tables show, no technology met all the selected criteria. Most of the technologies were suitable for the elderly and were nonintrusive. However, most technologies lacked motivation and engagement enhancement through the use of a variety of motivation methods. The developed approaches were technology centric, whereas a person-centric approach is necessary to keep patients engaged and motivated in achieving their rehabilitation goals. Several devices claimed to enhance motivation but produced little or no evidence of patient engagement [5,8,70,71,74,75,81,86]. None of the devices intended for rehabilitation were transferable to other uses. Devices intended for monitoring or diagnosis had the desired transferability features [22,102]. Only 1 of the reviewed technologies proposed for rehabilitation supported individualization [71]; however, it did not meet the requirements for elderly patients and it used wearable components. On the other hand, individualization was supported by monitoring devices that were not intended for rehabilitation use [22,100]. Several technologies we reviewed were inappropriate for home rehabilitation, as they were technologically complex and expensive.

The first rows of Tables 2-4 list all the selected criteria, drawn from our extensive literature review, that must be met for a home rehabilitation system to be engaging and enable stroke recovery patients to meet their progressively ambitious goals or targets. Based on the above analysis, an ideal home rehabilitation device should meet all the identified criteria and requirements. The device needs to avoid wearable or intrusive components. It needs to support enhanced motivation and engagement by being incorporated into the daily activity routine. It must be cost-effective and not complex to install, maintain, and use. It needs to support the needs of all patients, regardless of age and background. Moreover, it needs to be portable and transferable to other domains.

The successful design of an assistive technology or rehabilitation device should take under consideration what the individual should and can achieve during rehabilitation [16]. Quantification and further analysis of the present and future conditions of the patient could overcome difficulties and unforeseen circumstances and could result in better assistive technology design.

Data and patterns from electronic databases are quite important to tailor rehabilitation, as the device can learn patients requirements and goals, adapt to their individual needs, and provide suitable challenges, for example, through machine learning. Individual choice and personal control are mandatory for success (see Monitoring and Home Rehabilitation). Technology design has to follow a person-centric approach considering technology ability levels. Given the developments in smart devices, algorithms, and information extraction, devices can adopt a person-centric approach while meeting the requirements for cost and complexity.

Contributions from the patient’s environment can be used to enhance motivation and engagement with the activity. Members of the family or others can provide a competitive or cooperative stimulation. Additionally, rehabilitation incorporated into the completion of daily activities could enhance motivation. Finally, the continuous adjustment of the technology or device to the patient’s changing requirements has a more beneficial effect. The device should adapt to increased levels of difficulty to provide stimulation for achieving higher targets.

Thus, a system catering to every occasion, individualized and adapted to support the patient’s daily activities in their home environment, has a higher potential for successful acceptance and engagement. This system should incorporate a device, hardware, and additional software. However, developing such a system for the full range of impairments and rehabilitation tasks is an unrealistic goal. This is due to the requirement for different types of inputs for each condition, the range of rehabilitation goals, the differences between patients, and the differences between home environments. Hence, the successful system should focus on supporting specific daily activities that have measurable outcomes specified in recognized health care rehabilitation tests (see Comparative Analysis).

We reviewed rehabilitation devices for stroke patients in detail. The focus was on systems that are intended for use within the home environment for self-rehabilitation routines. We reviewed several technology domains under the criteria of motivation and engagement enhancement for continued use without the need for clinical or specialist involvement. We demonstrated that the existing approaches do not meet all the criteria in the motivation, acceptance, and technological categories. However, there is evidence that some devices proposed for monitoring instead of rehabilitation might provide solutions to individualization and thus wider engagement and acceptance. We identified the criteria for a device and system that will provide the required level of self-rehabilitation commitment as nonintrusive, nonwearable, motivation and engagement enhancing through a list of motivation methods, individualized, supporting daily activities, suitable for the elderly, cost-effective, simple, transferable, and intended for use in rehabilitation.