Health and social care services across England are currently facing increasing waiting times and unmet demand [1], creating high levels of system pressures that negatively impact both patients and health care staff. What have been defined as avoidable hospital admissions along with prolonged in-patient stays cause patients distress, limit hospital bed capacity, and are costly to the National Health Service (NHS; the publicly funded health care system in the United Kingdom). Older people and those from socioeconomically disadvantaged areas are at a greater risk of emergency admissions [2,3]. The emergence of COVID-19 has brought new urgency to the adoption of strategies that increase hospital bed capacity and reduce the risk of hospital infections.

Virtual wards (VWs) are models of care delivery whereby patients receive the care they need at home rather than in a hospital. This includes either preventing unplanned hospital admissions or supporting people to safely leave the hospital sooner. There is considerable overlap between VW and hospital at home (HaH) models of care, and these terms are sometimes confused or used interchangeably [4]. Frequently used definitions state that HaH services provide face-to-face care at home through community-based multidisciplinary teams [5], whereas VWs are hospital-led and managed to enable the delivery of at-home acute care, monitoring, and treatment using a variable combination of face-to-face care and remote monitoring (eg, through apps, technology platforms, and wearable devices) [6,7]. The scope of VWs as a model of care varies from models using remote monitoring technology to predominantly operate without in-person intervention (eg, telehealth hubs with remote monitoring) to models more heavily reliant on multidisciplinary teams delivering in-person care in patient homes [8]. Therefore, the latter end of this continuum of care overlaps largely with HaH models. VW or HaH models of care have been used across various countries, including the United States [9,10], Australia [11], the Netherlands [12], and Saudi Arabia [13].

During the COVID-19 pandemic, VWs were implemented in the NHS England (NHSE) to manage some patients with COVID-19 infection through monitoring their oxygen levels at home using a portable pulse oximeter, a small device that attaches to a patient’s finger to rapidly measure blood oxygen levels. Patients recorded readings from the pulse oximeter in a diary and were instructed to call the hospital should readings fall below a given level [14]. In some hospitals, oximetry readings were recorded in an app and displayed on a dashboard that the clinical teams could access continuously via a tablet or computer [15]. Since then, VWs have been expanded to different care pathways, primarily for acute respiratory infection and people living with frailty, as per NHSE’s (a public body that oversees budget planning, delivery, and operation of commissioning of the entire NHS in England) instruction [16]. A key criterion of this instruction was that VW models should be enabled by technology, that is, the management of patients via a digital platform managed remotely by a clinical team. VWs have been adopted and introduced in different ways in both primary and secondary care settings across England. As a key part of NHSE’s plans to improve the responsiveness of emergency and urgent care and increase capacity, the national ambition was to have 40 to 50 virtual beds per 100,000 population by December 2023, the equivalent of up to 24,000 beds in total [16]. As of September 2023, the NHS had delivered 10,000 VW beds, with >240,000 patients treated in VWs [17].

Evaluations of VWs are underway. As of July 2022, evidence from reviews of randomized trials indicated that there is consistently low to moderate certainty evidence that clinical outcomes, including mortality and readmission, for patients treated in VWs are as good or better than for those treated as inpatients [4,18]. The evidence on cost-effectiveness is less clear. Although there have been many studies investigating the costs of VWs to health care providers, nearly all of these studies have methodological problems, which may mean they overestimate cost savings. For example, few studies conducted cost-effectiveness analyses, with most having cost-minimization and cost-saving designs, and many studies disregarded informal care costs to patients and carers [4]. There is low-certainty evidence from reviews of randomized trials that patient satisfaction may be improved by VWs compared to inpatient care [4].

On July 1, 2022, 42 integrated care systems (ICSs) were established across England on a statutory basis [19]. These are partnerships of organizations (eg, secondary, community, primary, social care, and mental health services) that come together to plan and deliver joined-up health and care services and to improve the lives of people who live and work in their area [19]. Each ICS has an integrated care board (ICB). The ICBs have strategic and financial responsibility for the planning and delivery of health services. Within the NHS, “commissioning” refers to the process of assessing needs, planning and prioritizing, and purchasing and monitoring health services and is undertaken by those in senior leadership roles [20]. The mandate for VWs was developed centrally by NHSE and detailed the national ambitions and instructions for ICSs to then adopt and implement in their own way according to local needs. ICS commissioners and ICB members have been tasked with developing detailed plans on how they intend to move from current practice to using technology-enabled VWs. These plans are to be based on partnerships between secondary, community, primary, and mental health services. ICS commissioners have an important role in the development of implementation plans and the effective adoption of this new model of care in clinical practice.

Effective adoption and rollout of VWs will depend upon the perceived acceptability and feasibility of this care delivery model to stakeholders. The Medical Research Council framework for developing and evaluating complex interventions [21,22] advocates the use of qualitative research to ascertain what is viewed as acceptable or unacceptable to key stakeholders involved in complex health care interventions. The use of qualitative inquiry in intervention evaluation can provide a contextualized understanding of important experiences, enablers, and constraints regarding the implementation of an intervention across a range of settings. Findings from such research can therefore provide useful insights for decision makers working across different contexts.

This qualitative study drew upon the Consolidated Framework for Implementation Research (CFIR) [23]. The use of theory is critical to understanding the likely processes through which an intervention may work [22]. However, there is an array of often-overlapping implementation theories described in the literature, making the selection of a particular theory for implementation evaluation problematic. The CFIR addresses this issue by providing an overarching, comprehensive framework, comprising common constructs from published implementation theories; the CFIR is therefore well-suited to guide the evaluation of the implementation of health care delivery interventions [24]. The updated version of the CFIR [25] was used in this study. This consists of 5 overarching domains (innovation, outer setting, inner setting, individuals, and implementation process), each with multiple constructs (eg, for the innovation domain, constructs include innovation relative advantage, innovation complexity, and innovation cost).

This study aimed to develop an understanding of the acceptability and feasibility of adopting and implementing VWs across ICSs in England from commissioners’ perspectives, following the issue of the VW mandate by NHSE in the April 2022. Our objectives were to identify commissioners’ views on the potential benefit of VWs within the current health care system, identify the barriers and facilitators to adoption and implementation, and ascertain the ways in which the VW model of care could be optimized in future practice.

We conducted a qualitative study using one-to-one semistructured interviews via an online platform (Microsoft Teams).

The sample comprised individuals working in commissioning roles in ICSs in England. In total, there are 42 ICSs that each provide health care to around 500,000 to 3 million people. Commissioners are staff in senior leadership roles who are involved in the planning of various health care services, including the purchasing of technologies and monitoring implementation of the VWs program. Purposive sampling was used to select staff from ICSs in different geographical regions and those at different stages of the adoption and implementation process to enable comparisons across sites and to identify important contextual factors affecting implementation.

Potential participants were initially identified by our NHSE collaborators as being eligible to participate in the study based on their job role in NHSE. They were approached about the study, and permission to share their contact details with the research team was obtained. The research team invited them by email to participate. Snowball sampling, whereby participants were asked to identify other potentially eligible participants to invite to the study, was subsequently used. Participants were provided with an information sheet that outlined the requirements for participation and data protection measures.

In-depth one-to-one semistructured interviews were conducted via Microsoft Teams by 2 members of the research team (LJM and FG). Only the researcher and participant were present during the interviews. The topic guide (Multimedia Appendix 2) used broad open-ended questions relating to the study aims rather than mapping questions directly to the CFIR. This enabled respondents to discuss issues of importance to them and encouraged a logical conversational flow to the interview, in line with published methodological guidance [13]. The topic guide was developed by the researchers with input from patient and public involvement and engagement (PPIE) representatives and NHSE collaborators. Members of the NHSE team responsible for the VWs program nationally reviewed the topic guide, and modifications were made according to their feedback. Questions were reviewed and revised during the course of data collection to explore emerging issues in more depth. Interviews were video-recorded via Teams and converted to audio-only (MP3) files for storage. Audio files were transcribed verbatim; transcripts were anonymized, ensuring all identifiable names, places, and events were removed. Data collection stopped once the research team agreed data adequacy had been achieved [27,28], which was based on the quality and sufficiency of data in terms of the richness of and variations in the data in relation to the research aims and objectives. To aid this judgment, we applied a color-code rating system to our preliminary themes from the first stage of analysis (refer to Data Analysis section): “red” signified an abundance of data with no new insights gained from subsequent interviews; “yellow” signified data were voluminous but new insights or alternative perspectives were still emerging; and “green” signified that data were minimal and more insight was needed through further data collection. Color codes were updated with each subsequent interview until all preliminary themes were coded red.

The analysis involved 2 separate stages. First, in line with rapid qualitative inquiry methods [29,30], we conducted the preliminary analysis using rapid appraisal procedure sheets. These summarized key findings after each interview based on interview field notes and were then discussed with our policy partners at NHSE. This facilitated the timely dissemination of key findings in a fast-moving policy research area.

The in-depth analysis took place in the second analysis stage. Interview transcripts were analyzed thematically and structured using the framework approach [31,32]. NVivo (version 14; Lumivero) was used to facilitate analysis. The analysis involved both inductive (ie, data driven) and deductive (ie, theory driven) processes and comprised 4 main steps. The first step was reading and rereading transcripts and listening to audio recordings, noting key ideas (to aid familiarization). The second step involved line-by-line coding of transcripts to the updated CFIR [25] (reflecting the deductive element of analysis). CFIR domains were adapted to this study, and a codebook with definitions of domains and subdomains was created. Where it was felt that data did not fit within the framework, new codes were created. In the third step, each CFIR domain was plotted onto a separate thematic matrix, with subdomains (“codes”) presented in separate columns and participants (“cases”) on individual rows. Data were summarized into cells in each matrix according to the respective case and code. In the fourth step, matrices were reviewed and thematically analyzed (reflecting the inductive element of analysis), drawing upon connections made between codes and cases and reflecting on the overall narrative in relation to the research objectives. This facilitated the determination of the final overarching themes and the key issues and meanings within those themes.

The use of a combined inductive and deductive approach in the analysis helped to ensure comprehensive coverage and theoretical consideration while retaining the narrative and nuance of traditional qualitative approaches [33]. The findings presented in this study’s results reflected the final overarching inductive themes derived from the final stage of analysis.

Interviews were conducted by 2 female researchers (both holding PhDs and employed as research associates at the time of data collection) with previous training and experience in qualitative interviewing but little prior knowledge of the VW program and no previous relationship to participants. Participants had no personal knowledge of the researchers. This reduced the likelihood of participants providing ambiguous responses relying upon existing shared understandings. Transcripts were coded and analyzed by multiple members of the research team with a variety of disciplinary backgrounds, including health psychology, sociology, behavioral science, and public health. We also included the perspectives of an NHS clinician as well as a PPIE representative within our analysis discussions to ensure that interpretations were considered from multiple perspectives and that findings were relevant to both practice and public interest.

This study was approved by Newcastle University’s ethics committee on July 28, 2022 (23846/2022). A consent form was emailed to potential participants before the interview, and informed consent was recorded verbally at the start of the interview. Transcripts were deidentified, with any information that could potentially identify an individual removed or replaced with vaguer descriptions to ensure participant anonymity. Participants were given a unique identification number on recruitment to the study, which has been used when reporting participant quotations in the paper. No compensation, financial or otherwise, was provided.

A total of 20 semistructured interviews were conducted between September 2022 and January 2023 (initial implementation plans for VWs were submitted around April of 2022). No participants dropped out once recruited to the study. Interviews lasted between 30 and 80 minutes. Participants’ characteristics are shown in Table 1. There are different models of VW, and their emphasis on technology varies in line with patient needs and preferences. Descriptive information regarding the types of VWs being implemented, as described by commissioners in this study, is available in Multimedia Appendix 3.

Four overarching themes were identified reflecting the acceptability and feasibility of key adoption and implementation processes: (1) assessing the need for VWs, (2) coordinating a system approach, (3) agreement on Program Outcomes: NHSE Versus Organizational Goals, and (4) planning and adapting services.

Overarching themes and subthemes are presented in Textbox 1.

This qualitative study analyzed commissioners’ perspectives on the adoption and implementation of VWs in ICSs in England. Findings demonstrated a strong desire for system-level change in care provision. Though commissioners had an enthusiasm for the VW program as a means to improve patient-centered care, there were concerns about the timing of the mandate and uncertainty over the extent to which hospital beds could be closed to release funds for VWs or whether their purpose is to free bed capacity for those more in need of in-person care. There was also uncertainty over the need to incorporate technology into the VW offer. The VW mandate involved convening a system approach to implementation; however, this was hindered by differing interpretations of the instruction due to a lack of clarity in definitions within the national mandate and the use of hospital-centric language. Other key perceived challenges to implementation included narrow parameters of success, unrealistic timescales to provide evidence of effectiveness the lack of interoperability of technology and time-consuming procurement procedures, and liability concerns. Commissioners felt VWs could offer numerous patient-level benefits but emphasized the need to assess the suitability of patients for the service. Successful implementation was considered contingent upon having motivated and passionate clinical leads.

To our knowledge, no other study has analyzed perspectives on the adoption and implementation of VWs specific to the national 2022 NHSE mandate. However, some of our findings can be understood in the context of research on previous VW and HaH care delivery models. Similar to this study, reviews [4,35] identified patient-level characteristics (eg, technology access, technology literacy, and appropriate home environments) as important issues and emphasized the need for careful consideration of patients’ needs and characteristics before enrollment in VWs. Other studies [36] have highlighted the importance of convening multidisciplinary teams and ensuring integrated involvement of both primary care and hospital staff in VW models. This is in line with our findings on the importance of having multiorganizational steering groups, with particular emphasis on connecting acute care and community care sectors. Furthermore, a recent review on the use of technology in telemedicine [37] identified common barriers to implementation, including resistance to change, technology-challenged staff, interoperability, digital exclusion of patients, clinician perceptions of impersonal care, and issues relating to legal liability. These barriers link to our findings regarding the receptivity and needs of patients and staff (resistance to adapt to new ways of providing care, digital literacy of patients and staff, clinician concern over lack of in-person observation of patients, and liability concerns) and identification of an interoperable technology solution. Therefore, it appears these are common barriers across technology-enabled virtual care that need addressing in future models and iterations of VWs.

Strengths of this study include the involvement of multiple stakeholders in the design and analysis of the research. In addition to academic researchers, we had input from policy, clinical, and PPIE representatives at all stages of the research, including consultation on the topic guide and involvement from the early stages of analysis, to assist with the interpretation of findings and identification of key issues to further probe during data collection. Such involvement helped to ensure the rigor and trustworthiness of the data collected and the diversity in interpretation and relevance of our findings. More broadly, our collaboration with policy makers ensured that this work is policy-relevant, with a direct route to impact decision-making. This was facilitated by our use of rapid appraisal procedure sheets as per the rapid qualitative inquiry approach, ensuring key findings were disseminated quickly in such a fast-moving policy research area.

A further strength of this study is the use of an implementation framework (the CFIR). This helped to guide the analysis and facilitated the comprehensive identification of issues related to implementation. However, our use of inductive analytical techniques following initial coding to the framework ensured that the overall narrative, context, and nuance were retained in our findings [33].

There are also various limitations of this study. First, there was an overrepresentation of ICSs from northern parts of England, particularly the North East and Yorkshire. We attempted to mitigate this by asking our NHSE collaborators to identify commissioners involved in the VW program from southern geographic areas who would be willing to participate; however, our response rate for this was low. This may have been due to the limited capacity of staff in such areas. The overrepresentation of participants from northern ICSs in England may have influenced our findings in terms of the infrastructure and financial circumstances of respective ICSs and the wider social and economic circumstances of areas in northern compared to southern regions. Indeed, there is a worsening divide in terms of health care and health inequality between northern and southern England, with northern areas demonstrating lower life expectancy, higher infant mortality, and worse health and well-being [38]. In light of this, there may well be different challenges to the implementation of VWs in northern England compared to southern regions, such as the impact of the cost of living and patient suitability in terms of access to technology, digital literacy, and having appropriate home environments. A second limitation is that we were unable to characterize systematically the implementation stages of each ICS represented in the study. These data could have helped to identify more nuance in implementation issues across ICSs.

There are several implications of this research. First, if VWs are to be implemented effectively, there is a need to develop clearer communication strategies between NHSE and ICS leads about the purposes and intended goals of the program. In particular, uncertainty about whether VWs are intended as an additional service to increase care delivery capacity or intended solely to improve the quality of patient-centered care needs to be addressed. This may be achieved through the development of a shared logic model providing clear definitions of what does and does not constitute a VW, how VW models could reduce hospital beds, as well as detailing the potential sustainability of such models in terms of continued funding. Such communications should avoid using “hospital-centric” language and terminology to prevent misinterpretation of VWs as an acute care service, with more focus needed on the community-care aspect. Importantly, a shared consensus needs to be reached on the intended outcomes and success measures of the VW program, including patient-level criteria. Clarification and consensus around these issues could help garner support among operational staff and clinicians—strategic leadership and implementers need to push the same agenda to foster a climate where stakeholders are highly motivated and open to changing their way of working.

In terms of the technology enablement of VWs, there should be further clarification over what types of technology would be suitable for particular pathways, with examples given of how such technologies may enhance care within the pathway. This should mitigate the concern over developing pathways around technology solutions rather than assessing the pathway needs first and then identifying how technology may facilitate and enhance these needs. Providing existing and emerging evidence on the benefits and possibilities of using technology in care provision, particularly with respect to digitally enabled remote monitoring, via a national platform would also help alleviate clinician concern and trust issues regarding technologies, and this increased confidence in technology enablement could help to ease clinicians’ resistance to change.

Furthermore, need assessments of both patients and staff should be conducted. Behavioral and skills analysis of different patient populations (eg, respiratory and frailty) should be conducted to develop an understanding of training and support requirements, particularly concerning the use of technology. This could inform the creation of a standardized pro forma to identify suitable patients for a VW. A behavioral and skill analysis of those involved in the implementation of VWs would also be useful to help understand the key barriers and facilitators to implementation tasks and procedures and could inform any required intervention to enable successful implementation. Skill analysis relating to the use of technology and digital literacy would be particularly pertinent and could enable the identification of appropriate training and resources, along with the most suitable type of technology enablement for VW programs. Such behavioral and skill analysis should be grounded in a behavioral framework to identify the most relevant behavior change constructs for each stakeholder group.

In addition, co-design approaches in the design of future iterations of VW models could ensure representation from both acute and community care teams, as well as patient and carer representatives. This could improve the acceptability of VWs to all stakeholders and may help to establish a tolerable level of clinical risk with clear escalation procedures in the case of deterioration in the health of patients.

Further research is also needed to identify and describe the various models of VW currently being implemented across ICSs within the NHS. Such work could enable the development of SOPs that could be distributed nationally and adapted at the local level. Appraisal of the various models could help to identify the most effective model components and could also help to establish the costs involved and predict any potential for future cost savings compared to usual care. Specifically, evidence regarding the effectiveness and cost-effectiveness of the technology components could help to determine the appropriateness and necessity of technology enablement within VW services. Finally, more qualitative research is needed to evaluate the implementation of VWs from staff perspectives as well as patient experiences of being treated in a VW. Such research can feed into future iterations of VW service design and development processes.

This study showed that commissioners involved in the adoption and implementation of VWs in ICSs in England perceived VWs to have the potential to reform some aspects of patient-centered care for some patients. However, for the model to work, there needs to be more clarity over exactly what constitutes a VW, and there should be less use of hospital-centric language in communication strategies. Furthermore, there is a need to include more focus on patient-centered measures of success, for example, clinical safety measures, measures of clinical effectiveness (patient health outcomes), and care quality (patient satisfaction), and sufficient time should be allowed to evidence these. Evidence of patient satisfaction and clinical effectiveness could help to secure future funding, enabling the sustainability of the VW model of care.