With the development of artificial intelligence (AI), physical robots with intelligent capabilities based on AI (hereinafter intelligent physical robots) have been applied in the health care context to expand the digitization of health care work processes and increase the use, fairness, and cost-effectiveness of health care services, such as in smart health care services [1,2], including telemedicine [3], ambient-assisted living [4], intelligent health management [5], psychotherapy [6], and companionship [7].

The use of intelligent physical robots has attracted the attention of scholars, and various studies have examined the use of intelligent physical robots in health care from different angles. For instance, some studies have investigated the use of intelligent physical robots from the perspective of anthropomorphic design and features [8], social interaction [9], personality [10], and intelligence function [11] in various health care contexts, such as nursing homes [12], hospitals [13], psychiatric clinics [14], and patients’ homes [15,16]. Another research stream has mainly investigated how robot use affects individuals, such as users’ mood and behavior [17], user attitudes toward robots [18], and health promotion [19,20]. Although prior studies provide an understanding of intelligent physical robot use in health care, each study only examines the topic from a specific point of view. A couple of studies have attempted to provide an overview of robots in the health care context, but these studies have focused on either specific health care contexts or specific robot devices [21]. For instance, Sarker et al [21] reviewed the literature to identify how intelligent robots can help health care professionals fight the COVID-19 pandemic, whereas Vélez-Guerrero et al [22] focused on AI-based wearable robotic exoskeletons for rehabilitation by reviewing relevant articles. These studies failed to provide an overview of the use of intelligent physical robots in the general health care context.

In addition, prior literature has stated that it is important to understand the antecedents and consequences of innovative IT use to improve both user acceptance and IT performance [23,24]. Thus, it is imperative to obtain an overview of the antecedents and consequences of intelligent physical robot use in health care based on a literature review, which will provide state-of-the-art knowledge for both scholars and practitioners.

To fill this research gap, we aimed to provide an overview of the research on the use of intelligent physical robots in health care through a systematic literature review, especially to identify its antecedents and consequences. In addition, we aimed to identify the potential future research agendas in the field to guide scholars’ future research in the field.

A widely used definition of a robot is provided by Nejat et al [25]: autonomous or semiautonomous artificial objects and devices programmed to act and perform tasks in their environment. Depending on the nature of the embodiment, robots can be divided into physical robots with visually observable bodies (eg, the humanoid robot Pepper and the animal-like robot Paro) and internet-based robots generated through computer algorithms that respond to users in natural language (eg, chatbots and animation robots). Many studies have shown that physical robots are more expressive than internet-based robots in terms of interactive functions and physical embodiment [26,27]. From a functional point of view, the intelligent capabilities of physical robots are mainly reflected in three aspects: (1) perceiving surrounding informational and environmental changes, (2) thinking and learning, and (3) handling various complex tasks autonomously and proactively [28].

Some research has focused on the use of intelligent physical robots at the individual level, such as among patients and health care professionals. Fasola and Matarić [29] investigated the intrinsic motivations for older adults to engage in physical exercise using socially assistive physical robots and found that users’ perceptions of the enjoyment, usefulness, helpfulness, social attraction, and social presence in robot use motivated their use. Kuo et al [30] identified gender differences in patients’ attitudes and reactions toward intelligent physical robots in home care centers. Chang et al [31] examined how intelligent physical robots can reduce nurses’ workloads and turnover intentions in hospitals. Mettler et al [32] found that different professionals’ acceptance of and resistance to physical service robots are determined by their shared beliefs and concerns as well as their perceived affordance of physical service robots. Some studies have also found that robot design is closely linked to individuals’ use of robots in health care, such as anthropomorphism, social capabilities, and intellectual capabilities [33,34]. Meanwhile, some studies have investigated the use of intelligent physical robots from an organizational point of view. For example, Lee et al [35] found that management support facilitated the adoption of intelligent physical robots in an organization. Intelligent physical robots have also been found to be an effective means for hospitals to improve cost-effectiveness and health care service delivery [36,37].

Although prior studies offer important insights into intelligent physical robot use in health care, these studies have mainly provided knowledge of its antecedents and consequences from different perspectives and cannot yet provide an overview of its use in health care, which gives rise to the need for this study.

We conducted a systematic literature review following the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) guidelines. The details of the PRISMA guidelines are presented in Multimedia Appendix 1 [38]. The literature review comprised the following steps: (1) database search; (2) eligibility criteria; (3) study selection and screening; (4) quality appraisal of studies; and (5) data extraction, analysis, and synthesis.

We conducted database searches in the PubMed, Scopus, PsycINFO, Embase, and CINAHL electronic databases in May 2021. These 5 electronic databases in health, nursing, biomedicine, and psychology were selected to search for eligible studies as widely as possible. The following search terms were used to search titles and abstracts of articles and to find subject-specific articles for the literature review: “healthcare,” “health care,” “nursing, robot*,” and “bot.” The search string had wide coverage to avoid missing any research of interest. In our search, there were no restrictions on publication time or study design. Studies published in languages other than English were excluded. The search strategy for each database is shown in Textbox 1, and the detailed explanation of these search strategies is provided in Multimedia Appendix 2.

Studies were selected for this literature review according to the following inclusion criteria: (1) studies examining at least 1 physical robot with intelligent capabilities to assist users in completing tasks; (2) studies examining robots for health care purposes, which means they aim to use robots to promote or monitor health, to assist in tasks that are difficult to perform because of health problems, or to prevent further health decline [39]; (3) peer-reviewed, full-length articles published in journals, conferences, and books; and (4) studies published in English.

Some studies were excluded from this literature review because of the following reasons: (1) studies were not published in English; (2) studies were incomplete or non–peer-reviewed; (3) studies examining robots without embodied physical appearance (eg, conversational agents, robotic process automation, and robotic software); (4) studies examining robots without intelligent capabilities, including robots mainly for automation that allow users to choose from predefined options or robots that cannot adapt to dynamic and uncertain environments; (5) studies examining the design and development of robots without the actual use implementation and evaluation of robots; and (6) studies in which robots were not implemented for the purpose of health care (such as using robots to promote or monitor health, to assist in tasks that are difficult to perform owing to health problems or to prevent further health decline).

All relevant studies identified by database searches were downloaded and stored in the reference management software EndNote (version X9; Clarivate), which automatically eliminated duplicates. Initial selection of the studies was performed independently by the first and second authors by screening the titles and abstracts of the identified articles. All disagreements were resolved through discussion. In a second screening step, the full texts of the relevant articles were independently examined by the first and second authors according to the inclusion and exclusion criteria. All disagreements were resolved with joint discussion and final agreement between the 2 authors.

To evaluate the methodological quality of the studies, the first and second authors appraised the quality of the selected articles independently according to the Mixed Methods Appraisal Tool (MMAT; version 2018) [40], which can be applied to evaluate empirical studies using 5 different research methods: qualitative methods, quantitative randomized controlled trials, quantitative nonrandomized trials, quantitative descriptive methods, and mixed methods. Each category is assessed by 5 different quality parameters, with final scores of 1 to 2=low quality, 3=moderate quality, or 4 to 5=high quality. All disagreements were resolved through discussions between the first and second authors.

The following information was extracted for each eligible article by the first and second authors: publication type, titles, authors, publication year, research method, theoretical base, robotic platform, context, and main findings. Any disagreement was resolved through a discussion with the entire research team.

We aimed to provide an overview of the antecedents and consequences of intelligent physical robot use in health care. As there was no existing theory to serve as a framework for our study, we performed an exploratory conventional content analysis [41] to analyze the included articles. We captured a list of terms or phrases regarding the antecedents or the consequences of intelligent physical robot use based on our reading of the included articles and coded them. We first conducted the coding based on our reading of 15 articles and set up the preliminary codes. Next, we coded the remaining articles using these codes. We added new codes when we found new antecedents or consequences that were not in the existing codes.

The codes of the antecedents and the consequences of intelligent physical robot use were sorted into categories based on their characteristics, and a hierarchical structure of the antecedents and the consequences was established to guide the synthesis of the findings from the included articles. Specifically, the antecedents were categorized into individual-, organization-, and robot-related factors, and the consequences consisted of both non–health-related consequences and consequences for health promotion. The non–health-related consequences include emotional outcomes, attitude and evaluation outcomes, and behavioral outcomes of technology use, whereas the consequences for health promotion include physical health promotion, mental health promotion, and social health promotion. Finally, we discussed and finalized the results of the data synthesis.

This section includes a description of the articles included in this study, the quality assessment of these included articles, the publication year, and the publication sources of these included studies. We have summarized the applied research methods and theories, research contexts, robotic devices, and target users in these studies. Finally, we synthesized the antecedents and consequences of intelligent physical robot use in health care based on the included studies and proposed an integrative framework for the antecedents and consequences of physical robot use in health care to provide an overview of physical intelligent robot use in health care based on the findings of the included studies.

Figure 1 shows the PRISMA flowchart for the study selection process. Initially, the database search identified 8059 articles. Some articles were duplicated in different databases. After removing the duplicate articles retrieved from different databases, 5365 entries were left. Of the remaining 5365 articles, 5224 (97.37%) articles were excluded from the literature review after title and abstract screening as they did not meet the eligibility criteria to be included in this study. After full-text screening of the 141 remaining articles, 47 more were excluded. Thus, 94 articles were included in this systematic review. The details of the 94 selected articles are provided in Multimedia Appendix 3 [3-6,10-18,20,30-37,42-113].

We performed a methodological quality assessment of the 94 included studies using the MMAT. Among these 94 studies, 62 (66%) were rated as high quality, 24 (26%) as moderate quality, and 8 (8%) as low quality. Although these 8 studies have methodological limitations, they provided some new insights in some specific research contexts regarding intelligent physical robot use, such as in cognitive interventions and physical rehabilitation, which have been rarely covered by other studies. Thus, we included the 8 articles with low methodological quality in this systematic literature review. The details of the quality assessment are provided in Multimedia Appendix 4 [3-6,10-18,20,30-37,42-113].

The 94 articles included in the literature review were published between 2009 and May 2021 (Table 1). More than half of these articles were published after 2017. Among them, 66% (64/94) were published in journals and the rest in proceedings of international conferences.

With regard to the methods used in the included studies, the experiment was the dominant research method; 55% (52/94) of studies opted for an experimental approach. Surveys (16/94, 17%), interviews (10/94, 11%), and mixed methods (10/94, 11%) were also used to gauge the representativeness of individual views and experiences, accounting for 38% (36/94) of the included studies. Among the studies that applied mixed methods, 30% (3/10) used focus groups and questionnaires, 60% (6/10) used interviews and questionnaires, and 10% (1/10) used Q-methodology, which combines qualitative and quantitative data analysis methods. Some studies applied case studies (3/94, 3%), observational studies (2/94, 2%), and ethnographic methods (1/94, 1%).

Different theories have been applied in these included studies to investigate robot use in the health care context. Two popular technology use models, the technology acceptance model and the unified theory of acceptance and use of technology, have been widely applied in some studies to investigate the use of robots in the health care context [15,42-45]. In addition, other theories regarding user behavior, such as applied behavior analysis [46], the theory of planned behavior [47,48], IT affordance [32], and activity engagement theory [49], have been used to explain user behavior regarding robot use. In mental health contexts, some psychology-related theories have been applied to understand robot use from a psychological viewpoint, such as the theory of mind [44,45], emotional well-being [50], emotional appraisal [51], and the capability approach [52]. Moreover, some studies have applied nursing-related theories to investigate the use of robots from a health care professional perspective, such as the transactive relationship theory of nursing [53], professional task engagement [31], job satisfaction [31], and person-centered care [14].

Robots have been widely studied in care facilities, including older care facilities [12,54,55], long-term care facilities [13,50], and mental care facilities [56,57]. The use of robots in professional medical contexts has also been studied, such as in hospitals [31], outpatient clinics [58], and simulation laboratories [49]. Health care robots serving residents and their caregivers in retirement villages [20] and communities [59] are popular contexts in the included studies. Several studies have investigated the use of health care robots in participants’ homes [30,60] and universities [61]. Multimedia Appendix 5 [3-6,10-18,20,30-37,42-95,97-113] describes the contexts of the included studies.

As shown in Multimedia Appendix 6 [3,5,6,10,12, 14,16,17,20,30,33,34,36,37,42,45,46,49-54,56-59,61-64,66-71,73, 74,77,78,83,85-88,90-101,103,105-107,109,111], a total of 33 different robotic platforms were investigated in the included studies. Among them, the most popular robotic platforms are Pepper, Paro, Nao, iRobiQ, and Healthbot. These robots are usually equipped with a variety of sensors to support their performance of different functions, such as sight, sound, balance, and touch [65,66,114,115]. In terms of their physical characteristics, these robotic devices can be divided into mechanoid, humanoid, android, and animalistic.

Twelve of the robotic devices mentioned in 18% (17/94) of the included studies were mechanoid robots, which were described as having a mechanical appearance and no overtly human-like features, such as a kiosk with a screen [116]. For example, Lio looks like a robotic arm placed on top of a mobile platform and can provide humans with daily life assistance such as navigation, grasping, and monitoring [88].

In addition, 21% (20/94) of studies investigated the use of animalistic robots. These animal-like robots have a cute appearance and are mainly used for psychological and emotional interventions for older patients with mental illnesses. For example, a robotic baby seal named Paro has been widely used to accompany and help older people with cognitive and psychological impairments [56,57].

Among the included studies, 28% (26/94) of studies investigated humanoid robots, which usually have physical structures and movement patterns similar to those of humans [62,63]. Furthermore, 4% (4/94) of studies examined the use of android robots, which have a more realistic human-like appearance than humanoid robots [10,16,51,64].

The target users studied in the included articles can be classified into 2 types: end customers and professionals. Specifically, regarding end customers, 40% (38/94) of studies investigated robot use among older adults, such as the impact of robots on their physical health [12], cognitive health [55], and quality of life [3]. In addition, 21% (20/94) of studies emphasized patients as end customers, such as hospitalized patients (4/94, 4%) [10,13,34,36], people with cognitive impairment (10/94, 11%) [54], children with chronic diseases (4/94, 4%) [46,59,65,105], and people with limited arm or leg mobility (2/94, 2%) [66,100]. Furthermore, 14% (13/94) of studies investigated the general public’s awareness of health care robots, including healthy adults (5/94, 5%) [5,12,16,30,49,80] and university students and employees (8/94, 9%) [10,48,51,61,64,68,71,104]. Overall, 4% (4/94) of studies focused on patients’ relatives [75,97,105,111]. Health care professionals include nurses (20/94, 21%) [31], personal care and home care workers (16/94, 17%) [43], medical doctors (5/94, 5%) [6,15,18,32,44], management personnel (11/94, 12%) [32], and other staff (8/94, 9%) [11,15,18,44,67,70,72,76]. Multimedia Appendix 7 [3-6,10-18,30-37,42-59,61-87,89-113] presents details of the target users in the included studies.

Various antecedents and consequences related to intelligent physical robot use in health care were identified based on reviewing the included articles. The factors related to individuals, organizations, and robots provide a multidimensional understanding of the factors determining intelligent physical robot use in health care, and the use of intelligent physical robots in health care can lead to both non–health-related (emotional, attitude and evaluation, and behavioral) outcomes and consequences for (physical, mental, and social) health promotion. A theoretical framework to present a holistic view of these factors is developed (Figure 2), which could guide future research in the field.

In this systematic review, we examined 94 studies that focused on intelligent physical robot use (eg, mechanoid, humanoid, android, and animalistic) in various health care contexts (eg, hospitals, personal care facilities, laboratories, and patients’ homes). The target users of intelligent physical robots in the health care context include end customers and health care professionals. We identified the antecedents of intelligent physical robot use at the individual level (eg, social and psychological factors), organizational level (environment and resources), and robot level (eg, robot design and functions). We also synthesized the consequences of intelligent physical robot use for individuals into non–health-related consequences (emotional outcomes, attitude and evaluation outcomes, and behavioral outcomes regarding technology use) and consequences (physical, mental, and social) for health promotion. We proposed an integrative framework of these antecedents and consequences to obtain a holistic overview of the intelligent physical robot use in the health care field.

In the health care context, social and psychological factors affect the use of intelligent physical robots at the individual level, including end customers and health care professionals [106,107]. These findings are consistent with prior research findings, which also found that social and psychological factors affect nurses’ and other health care and social care workers’ use of physical robots and internet-based robots [27,117]. Although only 2% (2/94) of studies examined cultural differences, these studies do provide evidence of cultural differences in user perception [43,75]. In addition to user characteristics, the design and functions of intelligent physical robots are also important factors affecting their use in the health care context, which is consistent with the findings on intelligent physical robot use in other contexts. For example, the anthropomorphic features and designs of intelligent physical robots have been argued to play important roles in explaining robot use among end customers in the context of hospitality and education [19]. Some organizational factors, such as the physical environment and relevant resources of health care organizations, have also been found to be important in intelligent physical robot use [44,83]. Among the included 94 studies, there are fewer studies examining the use of intelligent physical robots at the organizational level compared with the number of studies focusing on individual users [35,67].

Various non–health-related consequences of intelligent physical robot use in health care were identified in this study, including emotional, attitudinal, and behavioral outcomes. These findings provide a rich understanding of the non–health-related consequences of intelligent physical robot use in health care among different user groups with various robot use purposes across different contexts in health care [108,114,118]. The use of intelligent physical robots in health care could, to some degree, improve the physical, mental, and social health of individual users [3,74,90]. These findings are consistent with the results of a literature review focusing on intelligent physical robots in care delivery during the COVID-19 pandemic from an organizational perspective. This review also found that the use of intelligent physical robots could promote health during the COVID-19 pandemic, such as diagnosis, risk assessment, monitoring, telehealth care, disinfection, and service automation [21]. The findings of this study complement prior research by providing a view on the potential physical, mental, and social consequences of intelligent physical robot use among various groups across different health care scenarios, such as older care, emotional health care, cognitive improvement, social participation, quality of life, and well-being [88,110].

Some studies have mentioned the following challenges and barriers in intelligent physical robot use in health care: perceived fear, distrust, and uncomfortable feelings caused by robots’ anthropomorphic design, technical barriers, and limited intelligence capabilities [111]. As Vélez-Guerrero et al [22] proposed, there is a need to improve the weight, operation mode, and control systems of robotic exoskeletons from the view of robot design to address certain challenges and barriers related to robot design. Clearly, more research is needed to address different challenges and barriers in intelligent physical robot use in health care and to make intelligent physical robots serve society in a health care setting.

It is important to acknowledge the limitations of this study. First, we searched the articles from 5 databases using some keywords, which may have limited the selection of articles and excluded some relevant articles that did not contain these applied keywords in their abstracts and titles or were published in other databases. Future work could expand the scope of the database and search terms to include more relevant articles in literature reviews in the field. Second, 8 studies with low-quality research methodologies according to the MMAT were included in the review. Thus, there might be a potential risk of bias in our findings based on the literature review. Future research could consider including only articles with high quality in research methodologies in literature reviews in the field. Third, some studies had a small sample size or included research participants from a specific geographical area, which may have potentially affected the data representativeness in our findings. Future research could consider including articles with a broader geographical and population diversity. Fourth, this study focuses on identifying the antecedents and consequences of intelligent physical robot use in the health care context, and other aspects related to intelligent physical robots, such as human-robot interaction and collaboration, are worthy of further research.

This study has made several contributions to the literature. First, it provides an overview of the antecedents and consequences of intelligent physical robot use in the health care context by synthesizing and analyzing the current literature on intelligent physical robots in the health care context. Specifically, this study contributes to the literature by identifying individual-related factors (eg, social and psychological factors), organization-related factors (environment and resource view), and robot-related factors (eg, robot design and function) as antecedents of intelligent physical robot use and non–health-related consequences (emotional outcomes, attitude and evaluation outcomes, and behavioral outcomes regarding technology use), and consequences for (physical, mental, and social) health promotion. Second, this study proposed an integrative framework to synthesize the antecedents and consequences of intelligent physical robot use in the health care context, which provides scholars with an integrated overview of the antecedents and consequences of intelligent physical robot use in the health care context. Finally, the literature review helps identify some gaps in the research on intelligent physical robots in the health care context and provides scholars with some future research directions from the perspectives of concepts, methodologies, research themes, and robot technologies. Thus, this study also provides guidance to scholars for identifying future research topics regarding intelligent physical robot use in the health care context, which could potentially address existing research gaps and improve research in the field.