This scoping review is based on the widely accepted framework by Arksey and O’Malley [13]. This framework was selected because it allows for the inclusion of various methodological designs across an interdisciplinary field. We searched for papers across 4 databases: PubMed, Scopus, ScienceDirect, and CINAHL. The search terms utilized are presented in Table S1 of Multimedia Appendix 1; they briefly encompassed the following search terms: health, monitor, smart home, ambient assisted living, active assisted living, and AAL. We limited our search to papers published between January 2008 and August 2021. Only peer-reviewed papers published in English were included. Of note, the term “surveillance” was not used in the search query, as its inclusion returned hundreds of results outside of the scope of this research project. A total of 5995 potential papers were identified using the search queries.

Papers were organized into Mendeley and Zotero reference managers. Following the removal of 2159 duplicate papers, 3836 papers remained for title screening (Figure 1). Paper selection was further refined by ensuring that paper titles contained one of our keywords as mentioned above. AO and KSS each reviewed half of the papers. Papers not in English and those not related to humans were excluded: papers related to animal, agricultural, or biology research were excluded. Further, conference papers, book chapters, white papers, reviews, and theses were removed. Following title screening, 1743 papers were selected for abstract review by AO and KSS in Mendeley. AO and KSS screened the abstracts to ensure that the papers focused on remote sensor technology and its application in a home setting. Papers that used synthetic data or described infrastructure architecture or were in hospital or laboratory settings were excluded. The remaining 538 papers proceeded to full-text screening and were transitioned to Zotero for file management due to software issues in Mendeley. Studies using wearables or video-based technologies, theoretical or conceptual papers, and algorithm-based technologies were removed. Both authors independently and unanimously agreed on the inclusion of 29 papers with an additional 97 papers with conflicting votes. These papers were discussed on a case-by-case basis until a unanimous decision was reached. Of the 97 papers that had conflicting votes, 20 papers were included in this review. Thus, 49 papers were found to be eligible for the final scoping review. The selected papers were saved in a database, and a master chart was built by AO and KSS to summarize the key information for subsequent analysis.

To gain an understanding of the types of smart home technologies being used and the information collected, we conducted a literature search across 4 databases (PubMed, Scopus, ScienceDirect, and CINAHL) between January 2008 and August 2021 by using the queries outlined in Table S1 of Multimedia Appendix 1. A total of 49 papers met the inclusion criteria for this scoping review (Table 1). Among the types of studies conducted, 31% (13/49) were pilot studies, 14% (7/49) were proof of concept, 12% (6/49) were algorithm evaluations, 10% (5/49) were proposals, 8% (4/49) were technical validations, 8% (4/49) were case studies, 6% (3/49) were method evaluations, 6% (3/49) were longitudinal studies, 4% (2/49) were platform evaluations, 2% (1/49) were randomized controlled trials, and 2% (1/49) were qualitative studies. When we examined the country of origin for each paper, we found that most of the studies were conducted in western societies, with 47% (23/49) of the papers originating from Europe and 35% (17/49) from North America. Few studies were conducted in Asia (6/49, 12%), Africa (2/49, 4%), and Oceania (1/49, 2%).

We observed an increase in the number of publications in recent years: 71% (35/49) of the papers were published within the last 5 years (2015-2020), while only 29% (14/49) of the papers were published before 2015. All the studies were either directly or indirectly associated with academic institutions. When classified based on a publication’s domain, 64% (31/49) of the selected papers were published primarily in the fields of engineering and computer science, 18% (9/49) were published in biomedical engineering and health informatics journals, and 18% (9/49) were published in health-related journals (Figure 2 and Table S2 of Multimedia Appendix 1).

As it is common practice in computer science or engineering research to use simulated data for platform or algorithm evaluation, we first categorized the studies based on the source of their data. Approximately 78% (38/49) of the papers used data collected from human participants, and the remaining 22% (11/49) of the studies used simulated data (Table 1). The age of the participants ranged from 18 to 93 years. Of the 38 studies that utilized human participants, 63% (31/49) reported participant age, but only 33% (16/49) indicated the gender of the participants. Of those that did report gender, female participants were nearly 3 times more prevalent than male participants (425 females vs 145 males). Volunteer participants were typically students recruited from the researcher’s institution or patients from memory care units and assisted-living residents. Of the papers on human participants, 79% (30/49) reported the health status of the participants.

The 49 papers included in this review can be broadly divided into 2 groups: 41% (20/49) approached the use of IoT for health purposes and 59% (29/49) used IoT for technological validations. The primary research focus was recognizing human mobility patterns (Table 2; complete data in Table S3 of Multimedia Appendix 1). Study length ranged from a single day of data collection to 8 years. Data were collected primarily in real-world settings, including smart apartments or smart workplaces. One of the studies used simulated home environments [39]. If the study took place in an apartment, the number of rooms typically used was between 2 and 3. Typically, there was only a single occupant in the study location.

To determine which smart home technologies were being used, sensors were grouped into 16 main categories (Table 3): utilization of space (bed and chair occupancy, toilet, fridge, kitchen, or GPS), human vitals (blood pressure, electrocardiography, blood glucose, heart rate, or respiratory rate), and environmental sensors (light, air temperature, humidity, sound, airflow, smoke, carbon monoxide, gas, or flooding). Nearly 62% (30/49) of the studies used passive infrared sensors to report on motion detection. As motion detectors and object presence sensors primarily record binary (yes/no) data, it was unsurprising that this data type was the most reported in the studies examined. Quantitative data were reported in many papers. Audiovisual (sound, light), vital indicators (heart rate, respiratory, blood glucose, body temperature), and environmental conditions (room temperature, humidity) typically record quantitative data. Finally, several papers reported spatiotemporal data typical of GPS sensors.

As smart home data collection produces large quantities of data, data management software is frequently employed. Examining the papers, we found SQL [34,35,56,57] and MYSQL [24,25,55] were frequently used to organize the data. MATLAB and Python were used for data analysis and visualization by nearly all the studies. Various statistical methods were used for data analysis, including descriptive statistics, model building, machine learning, and deep learning. Descriptive statistics were primarily used to describe the demographic characteristics of the study participants, whereas multidomain approaches [52], longitudinal linear mixed-effect regression [55], and out-of-sample cross-validation methods [55] were used for statistical models.

As 41% (20/49) of the papers reported the use of machine learning algorithms, we sought to determine which algorithms were more commonly employed. Clustering in 5 studies [1,28,30,34,53] and Hidden Markov Model in 4 studies [23,30,34,39] were the most used in data analysis to identify a regular pattern and predict future patterns. The other algorithms used in the studies were decision tree emerging pattern [11,25,27], clustering conditional random field [37,51], context-aware reasoning [28,42], fuzzy logic [41,49], k-nearest neighbors [10,51], logistic regression classifier [51,55], AdaBoost [10], Bayes network [27], boosting model using ensemble [42], circadian activity rhythms [47], multi-Hidden Markov Model [34], multiple regression model [42], multivariate habits cluster [44], ontological modelling [41], software for automatic measurement of circadian activity deviation [47], and support vector machines [52].

Nearly 14% (7/49) of the papers used deep learning methods, which included artificial neural networks [40], activity recognition using the discontinuous varied-order sequential model [34], latent trajectory models [56], longitudinal linear mixed-effect regression recurrent neural networks [55], open pass neural networks [60], recurrent neural networks [32], and multilayer perceptron [10]. One study used mixed methods and included a thematic analysis of the quantitative data [25]. Another study used the activity discovery method [34], and yet another conducted qualitative data analysis by using a mixed methods approach [25]. Some studies used induction algorithms, behavioral monitoring systems, rapid iterative testing and evaluation [15], or QRS recognition [57] for electrocardiography.

All the studies reported that IoT improved the quality of care, increased participants’ sense of comfort, enabled early detection, and increased participants’ understanding of the impact of health events on overall health. The health indicators specifically measured through smart home technologies included fall detection [24], functional health decline/improvement [10], high-level activities of daily living/instrumental activities of daily living [34,35,48,50,59,61-63], leisure services [59], loneliness [55], medical services [17,21,30,64], patient health status [17,21,30,64], perception [58], physical activity [48], sedentary behaviors [24,62], medication adherence [62], movement patterns [29], sequence of gestures [61], sleep [12,48,56], eating habits [10,24,57,62], situational awareness [30], social engagement [56], time spent outside the home [55], and overall well-being [24].

To gain insight into future research needs in the field of smart home technologies, we extracted information pertaining to the challenges and limitations self-reported by researchers. In the 49 studies, the biggest challenge faced by the researchers was differentiating between multiple participants in a single space. The second challenge identified was the lack of technology interoperability and the ability to scale up. The third challenge identified was linked to data security and privacy. The additional challenges identified by the researchers included calibration of the sensors, cost of technology and data management, data streaming and integration, data velocity, data volume, difficulty differentiating activities, generalization of activities, and demographic discrepancies (data collected from young volunteers, while algorithms were designed for the older population). Heterogeneity, installation of the sensors, lack of patient motivation, large numbers of nodes, limited data bandwidth, limited indoor activities, malfunctioning sensors, privacy, sample size, security, service quality, user acceptance, and varying levels of data accuracy were also noted as challenges.

Existing reviews on the application of smart home technology for health care are limited [2,3]. If at all present, they focus on a very specific specialty within health care, such as geriatric care [65], dementia [66,67], fall prevention [68], or pregnancy [69]. This scoping review aims to address this knowledge gap by elucidating how smart home technologies are being used for health care within the home and community. An extensive database search revealed 49 peer-reviewed publications, which met our inclusion criteria. A wide variety of sensors were used to meet the differing needs in each study. Passive infrared sensors, which report on motion detection, were the most studied smart home technology for health and report primarily binary data. Multiple studies quantified measurable health indicators (eg, heart rate, blood pressure, sleep, physical activity). Reported data were mostly organized using SQL or MYSQL. As expected, diverse data analyses and statistical methods, including machine learning and deep learning, were applied to big data analysis. Of note, although some studies were performed in home settings, none were unobtrusive or zero effort. There were often disruptions to daily routines or participants were required to log activities [70].

We recognize that there are several limitations to our study and that potentially relevant publications may have been overlooked due to the constraints in our search queries and inclusion criteria. As smart home technologies are often developed by the technology industry, not all work is likely published in peer-reviewed journals. Furthermore, our use of the query term “smart home” may have excluded relevant research settings in a community or an institution. For the purposes of this scoping review, database searches were conducted in August 2021. Due to the rapid nature of this field of research, new insights may have emerged since the initial search.

During the preparation of this scoping review, we noted that the health care research papers lacked a concrete definition for a smart home. Based on the available evidence and the identified gaps, we propose the following definition for a smart home for health care.

A smart home for health care can be defined as a home equipped with smart sensors using Bluetooth, Wi-Fi, or similar technology, not restricted to IoT, to automate, regulate, and monitor home occupants’ physical health, mental health, and environments within the home. The focus must be on convenience, safety, and improvement of one’s quality of life, to address the needs of the individual, caregivers, and health professionals.

The studies included in this review were predominantly performed in western societies. This bias could be due to our requirement that studies should be published in English. However, it is known that high-income nations dominate the field of smart home technology. This could be due to several factors. First, western countries are trending toward an aging population, and thus, the interest in assisted living technologies is higher [71]. Second, low- and middle-income countries are focused on reducing mortality and morbidity related to infectious diseases; therefore, their resources are not focused on the needs of an aging population [72-74]. To address global health and knowledge inequalities, researchers and funding bodies must ensure that low- and middle-income countries have the resources to benefit from health technologies. Future research should prioritize including study participants in nonwestern societies.

Computer science or engineering research often use simulated data due to budget, staffing, and time constraints. Traditional technical training does not consider health outcomes and overlooks the social determinants of health. Without health care experts as part of the research team, many are unaware of the importance of reporting the demographic characteristics of human study participants. This was reflected in our scoping review, as many of the included studies failed to report this information. Of those that did report demographics, we found that female participants were more prevalent, being nearly 3 times more likely to have been studied than male participants. This was unexpected, given that research is typically dominated by male participants [75,76]. Some potential reasons for this variance could be that women live longer [77], are more likely to live in assisted care units [78], are more likely to participate in studies [79], or have altruistic considerations [80]. Moreover, the use of simulated data despite the availability of actual data highlights the need for better access to high-quality data.

Smart home technology is a rapidly growing interdisciplinary field at the intersection of health, information technology, and engineering [81]. Yet, our scoping review highlighted a strong bias toward publication within primarily engineering and information technology journals. Many of the papers included in this scoping review contained highly technical language, tools, and databases. However, the primary audience is the health care field. Although we acknowledge that much of the technology is in its early stages, with research focused on technical challenges (data handling, analysis, storage, security, and privacy), this finding highlights a lack of collaboration between health and technical fields. Future work must address this gap—fostering interdisciplinary research teams with a broad spectrum of skills and domain knowledge experts. The involvement of health professionals in smart home technology research will ensure that these tools are relevant and bolster their successful implementation.

Interoperability was a commonly noted challenge faced by researchers. Technology is constantly being upgraded and improved with new products continually hitting the market. As diverse companies compete to create the latest technology, interoperability becomes an issue. Because there are no standardized guidelines, companies develop their own unique protocols and architectures for handling data, which contribute to incompatibility across the IoT landscape. The result is a jungle of systems that are confusing and intimidating to navigate for many non–tech-savvy individuals. One must subscribe to a single system that may not meet all their needs, grapple with the inconvenience of systems that do not communicate seamlessly, or implement third-party software or hardware to bridge the gap. There is a need to continue to develop solutions that allow these systems to integrate and communicate with one another. Similarly, the other 2 challenges faced by the researchers were differentiating individuals within a multiparticipant household and data security and privacy. Health care technology brings a new layer of complexity due to risks associated with personally identified data, health data, privacy, data rights, and ethical considerations [82].

Some of the studies [18,27,55] examined here had insufficient data quality to make their research findings relevant in the health care field. In many cases, the number of study participants was minimal and lacked demographic information. The quality of many of the sensors used in a home setting is lesser than that of the instruments used in a clinical setting, often diminishing the value of the data. Additionally, some of the technologies were not diagnostic tools at all because the health indicators were not quantifiable (video or audio). Other health conditions such as loneliness or mental health cannot be quantified and thus must be measured through the integration of multiple proxy indicators. The challenges of data integration will likely be addressed with continued improvements in artificial intelligence. Here, we have highlighted the existing research on the application of smart home technology to improve health and revealed multiple gaps in our knowledge. The IoT has ushered in a period of ultraconnectivity [83], converting commercial, off-the-shelf sensors like smart Wi-Fi thermostats and wearable devices into vital sources of health data. With the collaborative efforts of technology experts and health care professionals, we have the potential to leverage these data to improve physical and mental health.

Smart home technology has the potential to improve the quality of life by monitoring health indicators in vulnerable persons. Despite their potential, there is still a lack of large-scale utilization of these technologies for health care. A scoping review of the existing literature enabled us to identify the types of sensors and the data being explored. The trends and knowledge gaps identified here will invite new progress in remote patient monitoring in public health. This kind of a care system can support and complement medical interventions to improve population health.