This study used a qualitative descriptive design. Qualitative description is a naturalistic and pragmatic approach that seeks to provide clear, comprehensive, and data-near accounts of participants’ experiences and perspectives in everyday terms [35,36]. This design was appropriate for describing nursing assistants’ preimplementation expectations of smart care robots without seeking to generate a formal theory or test hypotheses. The interview data were analyzed using framework analysis, as described below. The study was reported in accordance with the Consolidated Criteria for Reporting Qualitative Research (COREQ; Checklist 1) [37].

The TAM and the UTAUT primarily directed attention to individual-level factors associated with technology acceptance [28,38], whereas the Consolidated Framework for Implementation Research (CFIR) further sensitized the research team to organizational, external, and implementation-related contextual influences [39,40]. These frameworks were used as sensitizing concepts rather than coding templates, and the data were not formally mapped onto their predefined constructs. Consistent with framework analysis, the analysis combined deductive and inductive elements: the research questions and interview guide informed 3 initial overarching analytical domains, whereas the specific codes, analytical categories, and their boundaries were developed and iteratively refined through engagement with participants’ accounts.

The target participants were nursing assistants working in long-term care facilities. In this study, nursing assistants were defined as paid workers who provided direct care in long-term care facilities but were not licensed nurses. Their responsibilities could include assistance with activities of daily living, basic care, rehabilitation support, psychosocial support, and care-related observation. The inclusion criteria were: (1) direct involvement in frontline care, (2) provision of informed consent and voluntary agreement to participate, and (3) adequate ability to communicate and express their views. The exclusion criteria were: (1) managerial, administrative, or support staff not directly involved in care; and (2) interns, trainees, or short-term part-time workers.

Between September and November 2025, purposive sampling was used to recruit participants from long-term care facilities in Chengdu, Sichuan Province, China. Participating facilities were identified from the official list of registered long-term care facilities published by the Chengdu Civil Affairs Bureau. To include facilities with different organizational backgrounds, the research team purposively selected 18 facilities from this list based on facility size, operating model (public, private, or publicly built and privately operated), and geographical location (urban or suburban), and sent each facility an invitation describing the study purpose, participant eligibility criteria, and data collection procedures. Five facilities responded and agreed to participate, and all were included in subsequent participant recruitment. The participating facilities had 65‐720 beds and employed 8‐102 nursing assistants. Further facility characteristics are presented in Table S1 in Multimedia Appendix 1.

The 5 participating facilities employed a total of 201 nursing assistants. With assistance from facility administrators, 41 potentially eligible nursing assistants were identified for formal screening, with consideration given to the eligibility criteria and variation in sex, age, educational background, professional title, and years of care experience. The research team screened all 41 individuals and confirmed that they met the eligibility criteria. Of these, 34 agreed to participate, and 7 declined. Interviews with consenting participants were scheduled sequentially according to their work schedules and availability. Three of the 34 consenting participants subsequently did not complete an interview because of scheduling conflicts or personal reasons. The remaining 31 nursing assistants completed interviews and were included in the analysis.

The final sample of 31 participants was pragmatically determined by operational constraints during the planned data collection period, including the number of eligible nursing assistants identified through the participating facilities, their willingness and availability to participate, and scheduling constraints related to their care work. Participant characteristics, including sex, age, educational background, professional title, years of care experience, daily working hours, monthly days off, and night-shift frequency, are presented in Table 1. Participants’ previous exposure to care robots was also recorded to clarify whether their accounts were based on direct use, observation in real-world settings, or indirect exposure.

Data were collected through semistructured, face-to-face interviews conducted between September and November 2025. All formal interviews were conducted by the first author (YA), a female graduate student who had received training in qualitative research methods. The interviewer had no prior relationship with the participants. The interview guide was developed from the study aim and research questions, informed by previous literature on care robots, health technology adoption, and technology implementation in long-term care facilities, and finalized through research team discussion.

Before formal data collection, 3 nursing assistants were recruited separately to participate in pilot interviews assessing the clarity, relevance, and comprehensibility of the interview questions. Based on their feedback, the wording and order of several questions were revised, resulting in the final interview guide presented below:

The 3 pilot participants were recruited separately from the same 5 facilities before formal participant screening began and were not drawn from the 41 individuals identified as eligible for formal recruitment. They were therefore not counted among the 41 eligible individuals and were excluded from Table 1 and all other formal study sample counts. The pilot interview data were used only to refine the interview guide and were not included in the formal analysis.

Interviews were scheduled at times convenient for participants and conducted in relatively quiet, private spaces within the facilities, such as activity rooms or nursing assistants’ break rooms, to minimize interruptions and allow participants to express their views privately. Before each interview, the researcher (YA) introduced her role and research background and explained the study purpose, confidentiality arrangements, and voluntary nature of participation. Written informed consent was obtained before the interview, and interviews were audio-recorded with participants’ permission. All interviews were conducted face-to-face and one-to-one and lasted approximately 20‐40 minutes. The researcher recorded relevant nonverbal cues, including facial expressions, body posture, and changes in tone. Probing, clarification, and reflective summarizing were used to explore the meanings and contexts of participants’ accounts. The interviewer sought to minimize leading questions and used follow-up prompts primarily to clarify or confirm participants’ intended meanings.

Within 24 hours of each interview, the first author (YA) listened repeatedly to the audio recording and transcribed it verbatim. The second author (FZ) checked each transcript against the recording for accuracy. Unclear content was discussed by the research team and, where necessary, clarified with the relevant participant. Verified transcripts were deidentified and imported into NVivo 15 (Lumivero), and the recordings were securely deleted after their accuracy had been confirmed. The data were analyzed using framework analysis, which was appropriate for addressing focused, applied research questions and enabled systematic comparison within and across cases and analytical categories while retaining openness to unanticipated, context-specific findings [41]. Following established guidance, the analysis involved 6 iterative stages after transcription: familiarization, initial coding, development of a working analytical framework, application of the framework, charting, and interpretation [42].

YA and FZ independently read and line-by-line coded all 31 transcripts. The research questions and interview guide provided 3 initial overarching analytical domains: (1) expected care functions of smart care robots, (2) conditions influencing their integration into routine care, and (3) preparations required before implementation. Facilitating and hindering factors, organizational support, training, and external support were acknowledged as a priori areas of interest, while data-driven codes were generated within and beyond these domains to capture unanticipated care needs, implementation concerns, and contextual factors. Illustrative coding examples are provided in Table S2 in Multimedia Appendix 1. Based on the initial coding, conceptually related codes were grouped into working analytical categories. Although the overarching domains were informed by the research questions and interview guide, the specific codes, category labels, definitions, and boundaries were developed and refined through engagement with the data. In particular, accounts of existing enabling conditions were differentiated from substantive implementation barriers, dividing the initial domain of influencing conditions into “limited facilitating conditions” and “major anticipated barriers.” Examples of this analytical progression are presented in Table S3 in Multimedia Appendix 1.

The working framework was iteratively revised by adding, merging, dividing, renaming, or redefining categories when necessary. Previously coded transcripts were revisited after substantive revisions to ensure consistent application. The revised framework was then applied to all transcripts for indexing, and a matrix was constructed with participants in rows and analytical categories in columns. Each cell contained a concise summary of the participant’s relevant account, with the participant identifier retained to support comparison with the original transcript. An excerpt from the matrix is provided in Table S4 in Multimedia Appendix 1. Interpretation involved comparing similarities, differences, contrasting accounts, and relationships across cases and categories. The final framework comprised 4 analytical domains and 15 analytical categories, as presented in Table S5 in Multimedia Appendix 1. YA and FZ jointly developed and revised the working analytical framework and interpreted the matrix. They resolved differences in coding or category boundaries through discussion. When additional adjudication was required, the third researcher (XX) reviewed the relevant transcript excerpts, codes under discussion, category definitions, and related analytical records. Final decisions were reached through team discussion and documented in the analytic audit log. Successive framework versions, matrices, analytic notes, and discussion records were retained as an audit trail.

This study was conducted in accordance with the Declaration of Helsinki and was approved by the Biomedical Ethics Review Committee of West China Hospital, Sichuan University (No. 20252018). Written informed consent was obtained from all participants. To protect privacy, participants’ names and other identifiable information were removed during the research process and replaced with study codes.

Trustworthiness was supported in accordance with the criteria of credibility, transferability, dependability, and confirmability [43]. All 31 transcripts were independently coded by both YA and FZ. Coding disagreements were resolved through discussion and, when necessary, referred to XX for additional adjudication. To enhance credibility, member checking was conducted in February 2026 by the first author (YA), who had conducted the interviews, after initial coding and development of the working analytical framework but before its finalization. Using the contact details provided during recruitment, the first author attempted to contact all 31 participants by telephone to explain the purpose and procedure. Participants reached by telephone were subsequently sent via WeChat (Tencent) a summary of the preliminary findings relevant to their accounts, including the key points from their interviews and the research team’s preliminary interpretation. They were informed that they could assess whether the summary accurately reflected their intended meanings and main views and provide any corrections, clarifications, or additions within 7 days. For participants who could not be reached by telephone or did not respond via WeChat within the specified timeframe, the relevant summary was presented and discussed verbally during a face-to-face review. All 31 participants completed member checking within 10 days, with 14 completing it via WeChat and 17 through face-to-face follow-up. Participants generally considered the summaries consistent with their views, and no participant requested substantive corrections, although some reiterated points they considered important. Based on this feedback, no substantive changes were made to the working analytical framework. Following each completed review, the first author (YA) promptly documented the review process and feedback in an anonymized member-checking summary log, as detailed in Table S6 in Multimedia Appendix 1. Transferability was supported through detailed reporting of the study setting, participant characteristics, and research procedures. To support dependability and confirmability, the research team maintained an audit trail documenting data management, coding decisions, revisions to the analytical framework, and analytic discussions. All transcripts were prepared and analyzed in Chinese, the participants’ primary language. The findings and selected quotations were translated into English only after the analysis had been completed to preserve the meanings and contexts of participants’ original accounts as far as possible.

Most participants had no direct hands-on experience with care robots. Their accounts were based primarily on information from news or social media and observations of robots in care-related or public service settings. Only a small number reported direct experience with robots. Participants’ types of exposure and the robots involved are presented in Table 2. Framework analysis of the interviews with 31 nursing assistants produced 4 analytical domains and 15 analytical categories (Figure 1; Table S5 in Multimedia Appendix 1). The first domain described participants’ expectations regarding the care functions of smart care robots. The second and third domains captured the limited facilitating conditions and major anticipated barriers associated with integrating robots into routine care practice. The fourth domain outlined the preparations participants considered necessary before formal implementation. To enhance transparency, Table 3 summarizes the number of participants whose accounts contributed to each analytical domain and category [44,45]. These numbers are provided only to describe the breadth of participant accounts informing each analytic category within this sample and should not be interpreted as statistical prevalence, inferential results, or indicators of the relative importance of analytical domains or categories [46,47].

This study examined nursing assistants’ expectations of the potential roles of smart care robots, existing facilitating conditions, anticipated barriers, and preparations that may be required before their introduction into long-term care facilities. As most participants lacked direct or sustained experience with care robots, their accounts reflected anticipatory judgments shaped by care experience, news, social media, and everyday observations and therefore require further examination in practice. Participants perceived potential value in safety monitoring, standardized management, repetitive physical tasks, and basic interaction. Openness to new technologies, a supportive organizational climate, and positive external signals provided preliminary favorable conditions, but appeared insufficient to support integration into routine care. Participants anticipated implementation challenges related to unclear human-robot collaboration, limited organizational capacity, potential care-related risks, and insufficient external standards and accountability arrangements. Organizational assessment, contextual adaptation, real-world pilot testing, and human-robot collaboration training were identified as areas warranting attention during preimplementation preparation.

Participants’ expectations of smart care robots extended beyond standardized tasks and repetitive physical work, reflecting the need for continuous observation, timely response, and contextual adjustment in long-term care [48,49]. They expected robots to complement existing care through functions such as dynamic following [50], low-disturbance monitoring [51,52], and immediate response, potentially supporting safety, greater resident autonomy, sleep continuity, and timely care. For emotional support, participants generally considered robots suitable only for basic interaction and entertainment, consistent with previous research suggesting that human care relationships cannot be fully replaced [53]. Notably, the perceived value of robots appeared to lie less in replacing deeper emotional support for residents than in sharing some of nursing assistants’ emotional labor. Long-term care work requires sustained attention, patience, and emotional regulation [54,55], particularly when caring for older adults with cognitive impairment [56]. Participants therefore envisaged robots providing stable, repetitive, and less emotionally complex responses, allowing nursing assistants to focus more on interactions requiring empathy, reassurance, and professional judgment. Previous research has focused mainly on reducing physical workload or improving older adults’ well-being [11,57], with less attention to caregivers’ emotional experiences [58,59]. The current findings therefore suggest that nursing assistants’ emotional labor and work experiences may be useful exploratory dimensions in future real-world evaluations of care robots.

This study identified preliminary facilitating conditions at the individual, organizational, and external levels, but these mainly reflected general staff resources, supportive climates, and positive signals rather than concrete implementation capacity. At the individual level, care experience and technological confidence may provide an initial basis for discussing, testing, and adapting robots. More experienced participants were better able to identify potential uses in specific care tasks, whereas those with limited technological experience focused more on learnability, operability, and controllability. Previous research similarly suggests that perceived controllability, transparency, and reliability may influence robot acceptance [60,61]. A small number of younger participants with junior college education appeared to view robots more in terms of workflow integration and human-robot collaboration, broadly consistent with previous findings that age and education may shape technology perceptions [14,62]. However, as this observation was based on only 3 participants (P27, P28, and P30), it should be considered exploratory. Future research could further examine how individual differences relate to implementation readiness and support needs.

A similar gap between openness and implementation capacity was evident at the organizational and external levels. Supportive management and teamwork may facilitate discussion and adaptation [63,64], while social and policy attention to smart eldercare provides a positive context [65]. However, these conditions have not yet translated into specific training, technical support, workflows, or accountability arrangements. Fragmented or exaggerated external information may also make it difficult for participants without direct experience to judge robots’ real-world capabilities. Overall, these facilitating conditions remain limited and may require further development before they can support routine implementation.

Participants anticipated uncertainties concerning organizational capacity, human-robot collaboration, contextual fit, and accountability. Embodied robots generally require continued testing, iteration, and contextual adaptation before stable workflows and responses to unexpected situations can be established [12,66], potentially increasing demands on staffing, time, and technical support in resource-constrained facilities. Human-robot task allocation was a particular concern because, unlike industrial settings with relatively stable task boundaries [67], long-term care requires continuous adjustment to residents’ functional status, cognition, and immediate needs [48,68]. Task allocation must therefore consider care context and the respective capabilities of staff and robots. Moreover, real-world robot use may require staff supervision, technical coordination, and team communication [22,69], potentially creating new work related to coordination and troubleshooting rather than simply reducing workload. Participants consequently questioned how changes in staff responsibilities and workload would be assessed.

Contextual fit and accountability introduced additional uncertainty. Functional limitations, cognitive impairment, and communication difficulties among long-term care residents [70,71] may restrict the applicability of robots designed primarily for users who can clearly express their needs and reliably interact with the technology, potentially raising equity concerns. Responsibility for robot-related incidents may also extend across developers, manufacturers, facilities, and operators [72]. In the absence of adequate standards, reliable information, technical support, and clear accountability arrangements, facilities may bear greater responsibility for managing these risks. Participants therefore identified organizational assessment, rule preparation, contextual adaptation, real-world pilot testing, and human-robot collaboration training as relevant preparatory needs. However, these suggestions reflected anticipatory judgments grounded in care experience and limited technology exposure and require further examination in practice.

The findings can be interpreted alongside, rather than as a formal test of, existing technology acceptance and implementation frameworks. Nursing assistants’ accounts of robot usefulness extended beyond efficiency and task support to resident autonomy, timeliness of care, and caregivers’ emotional labor, illustrating how perceptions of usefulness are embedded in care relationships and work contexts. The identified importance of organizational support, external context, human-robot collaboration, and accountability also resonates with the multilevel contextual emphasis of the CFIR [39]. Thus, the analysis provides a context-specific account of individual, organizational, workflow-related, and external conditions surrounding the potential integration of smart care robots from nursing assistants’ perspectives.

Transferability should be considered in relation to the organizational and social context of long-term care facilities in Sichuan Province. Expectations regarding safety monitoring, risk warning, repetitive physical tasks, and basic interaction may reflect broader long-term care needs, including continuous observation, physical workload, and timely responses to residents, which have also been reported internationally [14]. In contrast, concerns about family blame, institutional responsibility, human takeover, and limited transitional support may be more closely tied to local care organization and accountability structures. In this setting, nursing assistants undertake broad direct care responsibilities, while role boundaries and technology-related accountability arrangements remain under development [73]. Similar concerns may therefore take different forms or carry different weight in settings with clearer staffing roles, regulatory standards, and accountability mechanisms. Transferability should accordingly be judged against local staffing, workflows, organizational resources, and governance contexts.

This study was based on interviews from 5 long-term care facilities in Sichuan Province, China. Transferability to other regions and facility types should therefore be considered in relation to local workforce characteristics, organizational resources, and care practices. The sample was predominantly female, older, less educated, and lower in professional rank, which may partly reflect the local nursing assistant workforce but may also limit the range of perspectives represented. Most participants also lacked direct or sustained experience with smart care robots; therefore, their concerns about human-robot collaboration, care risks, and accountability were anticipatory. These views were grounded in care experience but may also have been influenced by news, social media, and everyday observations. Because the sample size was pragmatically determined by operational constraints, the scope of recruitment may have shaped the coverage of the dataset. The breadth and depth of the data varied across analytical categories, with some categories informed by relatively few participants or limited contextual detail. The findings should therefore be interpreted in relation to the scope of the available data and may not encompass the full range of nursing assistants’ perspectives in other settings. Future studies could recruit nursing assistants with more diverse backgrounds from a wider range of facilities to explore perspectives that may be less represented in the current dataset. Future research could also combine real-world pilot testing with longitudinal follow-up to compare preimplementation expectations with actual use experiences and examine how smart care robots are gradually integrated into long-term care practice.

This study identified nursing assistants’ expected roles for smart care robots and the conditions they considered relevant to their future integration into long-term care practice. Openness to technology, supportive organizational climates, and positive policy and social signals provided an initial foundation, but uncertainties remained around human-robot collaboration, risk management, organizational capacity, and concrete external support. Organizational assessment, contextual adaptation, real-world pilot testing, and practical training may help facilities prepare for implementation, although their feasibility and value require further evaluation.