Adopting and embedding home sensors in social care: findings from a mixed methods, rapid evaluation

Learnings and gaps in the current literature on the use of home sensors for proactive care in England

For our targeted scoping review of existing literature, we focused on the deployment and evaluation of ‘connected care’ platforms (i.e. data from a range of non-intrusive sensors around the home are aggregated and analyzed for changes in service users’ day-to-day living patterns) within English social care services. This review of both academic and grey literature builds on our previous horizon scanning and stakeholder engagement to outline the current national terrain on the use of telecare to support proactive models of social care delivery. Further details on the scoping review method and findings can be found on the project website (www.phc.ox.ac.uk/research/decide).

Overall, we found that the evidence-base on the implementation and evaluation of home sensors for proactive care was extremely limited. The five academic articles we identified described three evaluations, each of which considered evidence from multiple stakeholders and settings^3,10–13. A general review of proactive telecare services using smart home technology in the UK^10,11 drew on primary interviews with stakeholders (mainly the technology designers/suppliers and social care housing providers) and a limited number of interviews with older people. An evaluation of Phase 2 of the Technology for our Aging Population: Panel for Innovation (TAPPI) project described how six testbed sites had implemented a range of care technologies, including three sites that had piloted proactive remote monitoring¹². A recent rapid evaluation of proactive remote monitoring within adult social care was conducted by the Birmingham, RAND, and Cambridge Evaluation (BRACE) Centre, which included two local authorities and a national charity provider of social care services as case sites^3,13.

The 16 grey literature articles identified reflected a variety of stakeholder perspectives, including sector leaders in adult social care^14–17, local government¹⁸, and technology-enabled care^16,19. Initiatives to implement proactive remote monitoring within specific local government settings were identified^20–25, with some evaluation evidence available for Hertfordshire²⁴ and Kent²⁰. Within the context of integrated care systems, implementation efforts were also identified in a home care agency²² and in collaboration with NHS partners²³. One technology supplier has produced or sponsored multiple reports that highlight a range of case studies and emerging issues within the sector^17,26–28, which are highly informative but also at greater risk of bias for a favorable review of proactive remote monitoring technologies and the supplier’s specific products.

Across the reviewed articles, there were few examples of robust evidence from independent evaluation. Many of the grey literature reports provided some information on outputs, outcomes, and/or projected impacts, but they did not contain clear evidence in support of the claims (i.e., the methods that supported the findings were not described). The academic literature provides some evaluation evidence through more robustly described methods, but most data are in the form of qualitative stakeholder interviews, with more representation from suppliers and providers of the technologies rather than service users.

All of the reported case studies were for the initial implementation stages, with a small number of service users (usually 30 or less). We found no examples where proactive remote monitoring had already been scaled up to the level of a standard service offer within local government organizations, and only one reference to scale-up in progress¹⁵. Most evidence of evaluation was smaller-scale and qualitative rather than capturing system-wide quantitative data.

The design and analysis of our rapid evaluation informed by three key insights from the literature. First, there was a consensus on how proactive telecare within adult social care is framed to address system-level challenges. Policy makers, strategic decision makers within social care organizations, and technology suppliers were in alignment with the potential outcomes that these technologies could support: discharging patients from hospital as quickly as possible without compromising on safety, tailoring care packages to real-time needs in order to maximize efficiency in the context of chronic social care workforce pressures, and avoiding or delaying escalation to more intensive forms of care (e.g., moving into residential care) through earlier intervention. However, there remained tension in prioritizing system-level outcomes within an ethos of personalized care. If service users and the people in their care networks do not benefit from using the technology at a personal level, then uptake could remain low and aspirational system-level outcomes would not be realized.

Second, there was a mismatch between the articulation of system-level goals and current approaches for evaluating preventative telecare service models. The demonstration of system-level efficiencies, cost savings, and maintained/improved quality of life for target user groups requires large-scale data that capture a range of health and care outcomes over time. However, we found no examples of quantitative data on the long-term outcomes used to evaluate these technologies. Independent evaluation was limited to the few available academic sources and a small number of grey literature reports, providing mostly qualitative data on strategic decisions and early experiences of trialling technology. Most evidence of the impact was presented within the grey literature as individual case examples and unverified projections of cost savings. To effectively evaluate proactive remote monitoring technologies within social care, a wider range of data must be captured.

Third, the literature most strongly reflects the perspectives of sector leaders, organizational decision-makers, and technology suppliers rather than end users. It provided some emerging evidence of benefits to service users and their families, notably a sense of reassurance when additional care is not needed but confidence that carers will be alerted quickly if the situation changes and needs emerge. However, the evidence on end user experience of the technology remains sparse, with some indication of organizational difficulties in sustaining use. The technology prompts new roles, responsibilities, and (often hidden) workflows for a range of end users, particularly care staff, who are expected to engage with the data produced by remote monitoring systems. The values and experiences of all end users – the social care service users, their family carers, front-line care professionals, and any other staff responsible for data monitoring and response – must therefore be understood.

In summary, the shift to preventative telecare models involving proactive remote monitoring technologies has gained considerable momentum since the Covid-19 pandemic. The value proposition of the technology’s potential for addressing priority system-level challenges has been driven strongly by technology suppliers and strategic decision-makers, in line with overarching policy aims. There is emerging evidence of implementation outcomes from numerous pilot studies, but currently no evidence of local authorities in England (in their role as statutory providers of adult social care) that have demonstrated the scale-up and sustained use of these technologies beyond the pilot. There is a notable gap in the evidence of on-the-ground experiences of using these technologies, particularly regarding how different stakeholders respond to the information presented on data dashboards for each system. There was also a lack of available quantitative data to support the robust evaluation of longer-term, system-level outcomes and the impacts of implementing preventative telecare models. These gaps informed the mixed-methods design for collecting and analyzing new primary evidence within the three case study sites selected for this rapid evaluation.

Aims, objectives and evaluation questions

Against this background, we sought to study the use of home sensors for proactive care within adult social care services in England, in order to define good practices in the implementation and use of such technology, and draw transferable lessons that can inform sustained adoption at scale. The evaluation was guided by the following questions.

Origins of study and evaluation approach

The evaluation was conducted as part of the Digitally Enabled Care in Diverse Environments (DECIDE) program, a partnership between the University of Oxford and RAND Europe, undertaking rapid evaluation of technology-enabled remote monitoring across health and social care²⁹. The topic of focus was identified and refined through initial scoping meetings with national stakeholders and guidance from the DECIDE steering committee (a cross-sector group bringing perspectives from across the four UK nations). This study was informed by the tradition of developmental evaluation involving an emergent approach that captures data that can inform ongoing developments and focus³⁰. Working in partnership with three organisational case sites in England (local authorities and collaborating organisations using/supporting in-home monitoring within social care), we explored the multiple system-wide influences on implementing and embedding home sensors for proactive care within routine practice.

Case sites

We initially engaged with six local authorities where in-home sensing was being implemented and used as part of service transformation. From these, we selected three local authorities as collaborating case sites to ensure maximum variation with regard to the technology platforms and approaches to proactive monitoring, geographical contexts, and populations served (with inequalities being a key theme of interest). Data collection sources, protocols, and the associated governance requirements were established for each site.

The evaluation was conducted across three case sites (pseudonyms for the sites and technologies were used to maintain anonymity). All three sites use one or more technology platforms consisting of a range of in-home sensors (e.g., motion, doors, fridge, bed/mat, smart plugs), an activity dashboard accessible via computing or mobile devices, and machine learning capabilities and/or customizable parameters to detect changes in usual routines and predefined activity/event notifications. Table 1 summarizes the case sites.

The Riverbourne County Council has multiple districts and boroughs with a mix of relatively affluent towns and rural villages, 73% of which took part in a pilot service that started in 2021, with the service subsequently made available across the whole county. The county council’s adult social care team estimates over 700 use cases of home sensor services in the county during 2021–2024, although there are typically around 100 service users at one time. They use two different technology platforms with similar functionality: Pulse and Responda (though predominately using the former). Both of these platforms are used across different pathways and are monitored by a ‘proactive monitoring team’ alongside the traditional alarm receiving center, who reviews activity data and contacts family/informal networks regarding ‘actionable’ insights and ‘urgent’ alerts, based on changes in activity.

The Eastvale County Council is a predominantly rural county with market towns, rural areas, and pockets of high deprivation. The use of home sensors forms part of a wider TEC service, which was established in 2019, with home sensors introduced into the service model in 2020. More than 600 home sensor devices are used at any time. The technology provider, KinLink, provides all other TEC as well as home sensors, and conducts installation and monitoring. KinLink monitors urgent alerts (e.g., falls); however, continuous proactive monitoring is undertaken by the family/informal care network.

The Steelgate City Council is an urban area with a diverse population and high levels of deprivation in some inner-city areas. Home sensing has been introduced as part of a wider digital care transformation program. At the time of the study, they piloted two different technology platforms for two distinct pathways. The Responda system was used for long-term monitoring, while SafeNest was used for short-term reablement following hospital discharge.

Qualitative data collection and analysis

The data were collected between August 2024 and April 2025. Table 2 provides an overview of the qualitative data (supplementary case summaries are included in the data repository).

A total of 51 interviews were conducted with social care staff and system stakeholders (professionals from other related organizations, including technology providers, service delivery partners, and domiciliary care services). Participants were recruited through combined purposive and snowball sampling, asking the main contacts at each site and successive interviewees to connect with potential participants. This provided a diverse sample of staff and stakeholders with clinical/care and technical, managerial, strategic, and administrative roles in the delivery and use of home sensors. The staff interviews were semi-structured, conducted online, and lasted up to one hour, covering their role and use of home sensors within their organizations, implementation and delivery, perceived impact, and the value of the technology and service model.

In addition, we conducted system mapping workshops with 4–5 interviewed staff members at each case site. Workshops allowed us to develop and refine visual maps of the service user journey (to explore key steps in the delivery and use of home sensors), data flows (to explore how sensor data are shared and used across different systems and organizations), and funding (to explore different sources of funding for the service). To this end, we invited representatives with different roles and from different organizational partners (e.g., social care TEC strategy, management and administration, installation and technical support, and frontline care practice).

Service users and caregivers were identified by the staff at each case site. All were provided (or were caring for someone) using home sensors. In total, we conducted 16 family/informal caregiver interviews and three service user interviews. Most interviews were conducted with family/informal caregivers due to the service users’ lack of capacity to consent to the interview. Nearly all interviews were conducted over the telephone and lasted up to 60 minutes (two interviews were conducted at the person’s home). Participants were identified and initially approached by care practitioners on the basis that the service users were provided with home sensors.

All interviews were written using RAP (Rapid Assessment Procedure) sheets, along with selected audio transcripts where needed, using a guiding template to cover content related to the technology, use of the home sensor data, organizational processes and structures, perceived impact and experience with the technology, and challenges and enablers to implementation and use. The templates were used adaptively, depending on the background and role of the participants.

Data from staff and carer/user interviews at each site were synthesized using a structured narrative document. Narrative documents were used to support cross-case comparisons to develop a final theorization of the sustained adoption of home sensors at scale.

Emerging findings were explored and tested through stakeholder workshops with 25 attendees from across policy, health and social care, TEC and social care advisory bodies, the technology industry, third sector, and PPIE representatives. Participants were identified and invited with the support of our policy partners, project advisory, and PPIE groups. Informal interviews were also conducted with six national stakeholders: three were unable to attend the workshop and three followed up on points raised during the workshop discussion. In addition, a cross-site learning workshop with 1–2 representatives from each case site was conducted to discuss emerging findings and facilitate shared learning.

The analysis was guided by the NASSS (non-adoption, abandonment, and challenges to scale-up spread and sustainability) framework³¹. The NASSS framework was used adaptively as an analytical tool to surface and explain the challenges and complexities in technology-supported service changes. It includes seven interacting domains: the condition or illness, technology, value proposition, adopter system (intended users), organization(s), wider system (especially regulatory, legal, and policy issues), and emergence over time (see Figure 1 below). Alongside the NASSS framework as the overarching analytical framework, we applied three main theoretical lenses to further explore the influences on the sustained adoption of home sensors in routine care practice. First, we drew on diffusion of innovation theory^32,33 to further explore the organizational conditions and the work of introducing, embedding, and scaling up technology. Second, we drew on the notion of organizational routines (defined as recognizable, repetitive patterns of interdependent action carried out by multiple actors) as a source for stability and change³⁴, with a particular focus on how day-to-day work and coordination is achieved across organizational boundaries. Thirdly, we applied the theoretical conceptualization ‘information infrastructure’ (defined as the backgrounded things that other things ‘run on’)³⁵, in order to gain further insight into the invisible and taken-for-granted aspects of infrastructure (which includes technology and IT systems, but also material, social and regulatory aspects) that shape and constrain the pace and trajectory of innovation and service development.

Figure 1. NASSS framework³¹.

Economic component

We initially planned to undertake a full cost-consequence analysis. However, this was not possible because of the unavailability of the required data, coupled with the rapid evaluation timeframe. However, an analysis was conducted to explore the conditions under which a fully powered economic evaluation can be conducted in the context of proactive telecare. We sought to assess and inform approaches for monitoring implementation costs, service use, and economic and service user outcomes across sites. To this end, we worked with collaborating case-site teams to assess local data quality and availability as part of a feasibility analysis.

Drawing on the initial scoping work and emerging findings from the qualitative case studies, we sought to identify measures related to: (i) provision and uptake of the technology (demographics and variability across different population groups and with different care needs); (ii) service user outcomes (e.g., clinical, social, and quality of life outcomes associated with the implementation of proactive telecare); and (iii) service use and economic data (e.g., health and social care utilization, including rates of hospitalization, A&E attendances, inpatient days, outpatient hospital visits, prescription costs, and social care package cost).

We struggled to obtain high-quality data covering all the areas we wanted to consider. For example, quality of life data were not collected by care providers. Additionally, key data were held outside of social care systems and control, for example, by healthcare providers, Occupational Health settings, and technology providers that run the services. We were unable to obtain data from these sources.

There is no unique model for the management of social care provision, nor do social care organizations use the same data systems. Therefore, it was difficult to navigate the system data specifically for our evaluation question on the economic impact of home sensors. We also faced challenges regarding information governance for data-sharing. These included uncertainties around consent, confidentiality, and legal definitions required to facilitate data sharing for the purposes of service evaluation. Our case site partners worked collaboratively to address such challenges but also faced resource limitations. In addition, when analyzing the direct impact of specific interventions, establishing a comparator and including a robust-sized sample is important, but proved difficult in our study. All of these factors impacted the quality of data we were able to access, and the timeliness of access. Despite these challenges, we were able to analyze some service users’ demographic and social care provision data from all three case sites, which came from social care management systems. This enabled us to undertake a feasibility analysis to establish what would be needed for future economic analyses.

Eastvale provided the richest set of data, which included 6399 service users – 491 sensor customers and 5908 service users who did not have sensors, dating back to 2022. The data included rich demographic data, information on the health conditions for which service users received social care support, information about the care packages service users received, dates of major events (such as time they entered the care system, changes in care packages, and dates at which they stopped receiving care), and coded reasons for any changes in care. The site had access to some data on health outcomes, but was unable to share it with us because the information was controlled by the service users’ integrated care board (ICB). The costs of specific sensor devices and their maintenance and monitoring in users’ homes were incomplete. These data were provided by technology providers, but they did not maintain full historic records for all devices allocated to users. This meant that we were unsure about the full suite of items provided to any given individual. This dataset allowed us to perform a descriptive analysis but did not facilitate cost/benefit analysis.

Riverbourne provided data on 226 users with devices, and a matched sample of 226 users who did not have home sensors. They were unable to provide data on their health outcomes or health conditions.

Steelgate provided a large dataset of 6325 customers. Of these, 1233 had a TEC device and 5092 had never used any technology but it was not possible to separate home sensor users from those who were provided with passive devices such as fall sensors (coded broadly as ‘sensors’). Additionally, as with the two other sites, it was not possible to obtain data regarding the health outcomes of service users, but they were able to provide data on the number of hospital admissions for each service user.

None of the sites routinely collected information related to the quality of life of service users. Steelgate performed a trial of home sensors with a small group of participants (up to 30); therefore, we were able to conduct a survey using a questionnaire to collect primary data. This was designed to capture demographic data, quality of life measured using the EQ-5D and the Adult Social Care Outcomes Toolkit (ASCOT), health status, and interactions with health/social care providers so that we could fully cost their care and identify any savings that could accrue as a result of their use of home sensors.

We received 18 responses to the baseline questionnaire and were able to follow up with 12 participants who responded to a second questionnaire administered approximately 12 weeks later. Sections regarding interactions with healthcare and social care providers were not completed by the majority of users, and the quality of life questions did not provide consistent results, so we were unable to draw any meaningful conclusions from this component of the study. However, our work with the staff at Steelgate produced a questionnaire that could be used in subsequent studies, and we learned about the resource requirements for a fuller study.

Ethics and governance

The Research Governance Ethics and Assurance (RGEA) team at the University of Oxford (sponsor) classified the project as a service evaluation, and therefore did not require research ethics approval.

All participants provided informed consent. All participants were provided with an information sheet and consent form detailing the study aims, design, contacts, and data privacy. The information sheet also clarified that participants had the right to withdraw from the study at any point, without needing to give a reason. All interviews were audio-recorded and written using summary notes, with key quotes included. To protect identities, we used pseudonyms for sites and the technologies they used.

Patient and public engagement and involvement

A dedicated project PPIE advisory group with six members was established, meeting periodically for three meetings during the project period of 12 months. Four of the six members of the group included individuals with prior experience of living with and caring for someone with TEC. The other two members were representatives from Carers UK. The group advised on study design, data collection, and analysis, as well as guiding focus on issues of equity of access and ethical considerations. Members of the group also joined the cross-stakeholder workshop and advised on written outputs and plans for dissemination.

Origins and use of home sensors in participating case sites

The use of home sensors within the three sites aligned with the established classifications of proactive telecare pathways, as described above, including assessment, reablement, hospital discharge, and long-term monitoring. However, there was wide variation across sites in terms of how pathways were operationalized and the extent to which home sensors were implemented and used within them.

Box 1 summarizes the diverse origins, evolution, and adaptations of the use of home sensors in the case sites.

Details of the origin and development of home sensing across case sites illuminate the convoluted work of introducing the use of home sensors within social care settings and developing associated roles and processes. The approach and focus for service delivery were shaped by both new and legacy systems, cross-sectoral partnerships, varied contracting models, and incremental additions funded through a combination of local and national sources. Social care systems are known to be diverse, fragmented, and path-dependent, shaped by historic strategic decisions that set organizations on a particular path. Consequently, the technology could not simply be ‘installed’; it had to build on, and interface with the existing infrastructure or ‘installed base’³⁵. This led to gaps in planned and actual use and shifts in focus across care pathways, depending on technical, regulatory, social, financial, and logistical contingencies.

Across all three sites, home sensors were predominantly used for older adults with mild cognitive impairment (e.g., dementia), frailty, acquired brain injury, or learning disabilities. Decisions for use were based on the need to gain insight into levels of independence and safety that would be difficult to establish via a one-off assessment (e.g., memory, communication difficulties). This tendency was also reflected in the descriptive quantitative data captured as part of the economic evaluation component. For example, in Eastvale, 77% of service users with home sensors were aged 75 years or over and 36% were identified as having dementia. Similarly, in Riverbourne, 81% were aged 75 years or older, with over half (47%) identified as having dementia or mild cognitive impairment. It should be noted, however, that while the quantitative data may be indicative of trends, affirming what staff have told us, caution should be taken in interpreting the figures due to methodological challenges, as described in the methods section.

All service users lived at home as opposed to residential care. However, the living arrangements of service users varied depending on the affordances, functions, and configurability of the technology. In Riverbourne, most deployments were for people living alone, because the platforms could not differentiate individual activities in shared settings. In contrast, the SafeNest system in Steelgate was designed to analyze activity patterns in multi-occupancy households, constructing a picture of typical activity patterns across the home as a whole rather than per person. In Eastvale the sensors could, in some cases, be configured for shared settings using targeted sensors like bed or ‘usual chair’ mats to infer individual activity, as well as targeted night-time (bedroom) activity in shared supported accommodation for mental health and learning disabilities.

Integrating the use of home sensors into existing workflows required significant and distributed effort across all sites, and the technology operated within a nexus of interconnected and interdependent practices across organizational boundaries and systems. Although roles and workflows varied within and across sites, system mapping revealed consistent stages and considerations in the service delivery process summarized in Figure 2, including identification, assessment, installation, maintenance/monitoring, and review/removal.

Figure 2. Overview of service delivery process based on system mapping across case sites.

Perceived impact of home sensors in social care

Supporting independence and safety. It was highlighted that home sensors can support independent living in a safe environment in a way that helps maintain service user privacy. However, this relied on good knowledge of the individual’s existing routines, and in some cases, additional insight from the activity data introduced new concerns and perceptions of risk.

Staff, family/informal carers, and service users described how home sensors helped identify inactivity periods and reducing 'long lies’ after a fall, as well as unsafe ‘wandering’ outdoors. Sensors have also been reported to support activities of daily living, such as medication adherence, home temperature management, nutrition, and hydration, thereby reducing the need for frequent supervision.

Relatedly, many of the staff, family/informal carers, and service users interviewed felt this supported service user autonomy and dignity at home, reducing the need for formal and family/informal carer visits. One informal carer talked about how the sensors enabled her father, who refused all domiciliary support, to live alone despite his rapidly deteriorating dementia:

However, this did require careful and individualized trade-offs between support for autonomy and protection from harm. As illustrated in the below extract from the son of one service user, in which access to the activity data raised additional concerns about opening the front door, which he felt a need to restrict:

Family/Informal carer reassurance. The use of home sensors was generally seen to reassure family/informal caregivers and reduce the burden of frequent visits or contact. However, this still required efforts among informal networks to monitor and respond to activity, effectively reshaping rather than merely replacing or reducing their caring roles and responsibilities.

Staff and family members across sites highlighted the value of being able to see the activity data, especially when the service user was not communicating (e.g. not answering the phone):

Within Riverbourne and Steelgate, service users were further reassured that there was a service monitoring their family members when they were unable to and would provide alerts of any problems identified by the sensors. For some families, a professional monitoring service was crucial, as they were already under strain to support the service user and were unable to constantly monitor the sensors.

However, this idea was queried by one family member who felt that no care provider or monitoring hub could know her mum like she could, and gave an example of a time her mum did something ‘normal’ (left the house during the afternoon) which was in fact highly unusual in its context and necessitated an emergency escalation:

Some family/informal carers noted that home sensors did not necessarily lead to less informal caring work; rather, the nature of the caring work changes.

Another aspect of reassurance for family/informal carers related to monitoring the activity of others visiting the person in their home (e.g. that domiciliary carers were visiting within the expected timeframes):

Identifying health care needs. Staff and family/informal carers provided examples of when the activity data informed health-related decisions, either from insights picked up by the formal monitoring services or based on their own interpretation of the activity data. However, such use of home sensors relied heavily on informal networks to take these insights forward with the relevant healthcare service. The extent to which these insights contributed to meaningful health outcomes and decisions was also unclear.

There were examples of home sensors being used to inform healthcare decisions. Common aspects of activity included increased bathroom usage (as an indication of UTI), nighttime agitation and wake-up patterns (as an indication of potential sleep disorders), and reduction in eating activities (as an indication of ill health or low mood). Within Riverbourne and Steelgate, such monitoring was undertaken as part of the short- and long-term care pathways. If changes were identified that indicated a need for action (referred broadly as ‘actionable insights’), then the relevant contact (e.g., family/informal carer, social worker) would be notified to take forward. For example, in Riverbourne, one family caregiver described how such data were used to request a sleep study from her son’s specialist health professional to investigate potential sleep apnea:

However, as highlighted by one Steelgate staff member involved in remote monitoring, not only does such use of the technology require capacity to monitor the data, it depends on there being an individual able and willing to take forward:

More holistic and objective assessment for personalised care. It was highlighted that an analysis of service users’ activity patterns can be used to provide a richer and more objective account of their lives, capabilities, and support needs. However, this also raised challenges and uncertainties regarding how and the extent to which such quantitative representations should be relied upon against more direct meetings and encounters.

Social care practitioners discussed how activity data helped gain a fuller picture of an individual's life, especially for those unable to communicate their needs, and identify behavioral patterns and exceptions that might otherwise be unknown. Furthermore, it was reported that the data helped increase practitioner confidence in their assessments:

Additionally, home sensors were described as helping identify ‘points of escalation’ that, if acted upon, could help avoid unnecessary care placements, need for higher levels of care, and hospital admissions:

Examples were also provided as to how home sensor data were used to adapt and tailor a person’s care provision (e.g., decreased assistance with meal times, as sensor data showed that the service user could prepare their own food if encouraged and prompted earlier in the day).

However, it was also acknowledged by practitioners that “you still have to sort of unpick it and do more sort of fact finding - it just gives you a start really to find out what is happening with people” [Staff, Steelgate_S_5]. Many practitioners cautioned against placing too much weight on the sensor data, seeing it as something to be used in conjunction with existing assessment practices.

Similarly, some family/informal caregivers raised concerns regarding the extent to which the data would inform an assessment. The following extract is from one case in which the family felt that the home sensor data misleadingly downplayed one practitioner’s assessment of the safety concerns of their grandmother’s wandering:

Supporting dialogue and communication. Some incidents were described in which the activity data appeared to work as an effective communication and ‘evidencing’ tool when considering and agreeing to an appropriate level of care for a particular service user. This appeared to be used on an ad hoc, case-by-case basis as opposed to routine practice.

In some cases, TEC leads and care practitioners highlighted the value of using activity data as a communication tool with commissioned care providers regarding the provision of care packages. For example, a staff member in Riverbourne described a situation in which a paid care provider advocated for a service user’s care package to increase because they regularly did not go to bed until 2am. However, the sensor data showed the service user in bed at 11pm and did not move until the morning. As highlighted in the following extract from a TEC practitioner in Eastvale, such uses of the data or facilitation tool or ‘boundary object’³⁶ supported dialogue and shared understanding:

Some practitioners also described how the data was used to ‘evidence’ additional care needs for internal reviews and decision making, such as special funding panels:

An additional, and perhaps unintended, use of the technology by social care staff has also been to monitor care staff activity, cross-referencing care logs with sensor data where there have been care discrepancies or differing accounts from family members.

While some staff members considered the use of the technology to enable quality and safeguarding, others highlighted concerns and unease among some home care staff about their privacy and how the data would be used.

Influences on the embedding and use of home sensors

Materiality and dependability of the technology. Participants considered the technology generally reliable as long as there was adequate maintenance and connectivity, which could depend on both technological factors and user behaviors. Materiality (material properties, design features, and affordances) also influenced how and the extent to which it would fit within people’s lives and home environment. Effective use and sustained adoption required adaptive and practical efforts to ensure that the technology remained in good working order in people’s homes.

Across all sites, Wi-Fi connectivity was provided to service users who did not have broadband in place (at the cost of the council). Because of the typical user populations, this was a common requirement. In Steelgate, over 60% of service users required Internet access, which had not been factored into the initial phase of service provision. In some regions, particularly the rural areas of Riverbourne and Eastvale, the installers could not establish an internet connection, and thus the technology offered had to be withdrawn. Similarly, connectivity was sometimes influenced by building architecture (e.g., size, layout, and thickness of walls), resulting in poor internet signals that required installers to find workarounds:

Staff and family/informal carers also talked about the practicalities of replacing batteries that had run out-of-charge. Across all sites, family/informal carers were shown or provided instructions on how to replace batteries during installation, and most said they were confident in doing so. However, maintenance support was sometimes required to replace batteries. In Riverbourne, the service delivery team incorporated routine maintenance visits (every six months) to people’s homes because the battery levels could not be monitored remotely.

Informal/family caregivers also mentioned situations in which sensors were displaced or fell from the walls, requiring either them or someone else to reattach them. As noted by the domiciliary carer below, engaging in such repair and maintenance work brought new risks and accountabilities:

The technology might also be disconnected by the service users themselves or other people in their homes (e.g., unplugging the hub unit, removing sensor devices), requiring family members to recover and reconnect the devices, contact others to help them, or request a maintenance visit from the service provider.

Compatibility of technology with service users’ lives and care networks. The use of home sensors works well and was considered useful to the extent that it aligned with the service user’s wider care network, priority concerns, and values. Staff and family/informal caregivers reported that service users’ initial responses to technology were mixed. Often, it was considered that the service user had forgotten about or did not notice the technology because of their cognitive impairment. In some cases, families needed to provide reassurance and familiarization, such as relating the technology with something that they knew the service user would understand or would be more palatable to them. Therefore, the introduction and explanations were a situated judgement based on the knowledge of the service user:

The balance between autonomy, safety, and dignity also necessitates various forms of ‘emotional labor’³⁷ by members of informal and formal care networks:

However, there were also occasions in which service users considered the technology too intrusive and had to be withdrawn. It was reported that the physical resemblance of some sensors to cameras increased such concern, and a sense of being ‘watched’. Some practitioners also reported that, for some service users, particularly within mental health services, the sensors were not conducive to their well-being, as it would heighten anxieties and paranoia.

The effective use of home sensors relied heavily on the involvement and engagement of informal carers to monitor and respond to data and alerts. For example, in Riverbourne, the proactive monitoring team would alert family members responsible for checking in or arranging for someone else (e.g., a paid carer or neighbor) to do so. In Eastvale, the long-term use of home sensors depended entirely on informal support. Similar concerns were raised regarding loss of engagement due to caregiver breakdown and difficulties maintaining the level of support and involvement required:

Across all sites, staff emphasized the importance of ‘wrap around’ support for the family/informal network, to align it with particular setup, capabilities, and priority concerns. The installers, for example, would often spend time with the family to help download the app on their phone, demonstrate how the sensor technology worked, and configure the devices and settings accordingly. In some cases, family/informal networks would adapt the technology, or request additions or adaptations to be made, once they become accustomed to the technology (and limitations) while in use. For example, there were a number of cases in which family members used home sensor data alongside video cameras to help contextualize and interpret the graphical visualization on the dashboard. This was explained by family members as a step towards comprising a more ‘tailored’ suite of technology, which provided a fuller picture of what was going on in the home; cameras were usually not in bedrooms or bathrooms. However, it was also acknowledged by some that the service users’ understanding of the placement and usage of cameras might be limited:

Workforce knowledge, capabilities and capacity. Strategic decisions about the use of home sensors were generally made by TEC leads within the councils. However, the attitudes, capabilities, and actions of frontline social care practitioners greatly influenced the assimilation, embedding, and ongoing use of the technology. There were various aspects to this, including identifying service users who may benefit from the technology, tacit (know-how) knowledge of the ways in which the technology can be configured to meet particular needs, and interpretation of the sensor data. It was reported that social care practitioners varied both in their attitudes towards the use of home sensors and in their confidence and capabilities in applying the technology and interpreting activity data.

Training and upskilling among care practitioners was generally done on the job as part of a team, as opposed to formal teaching or study. This was supported through multiple means across sites, including the establishment of local champions to advocate and support the use of the technology and provide a conduit between strategic decision-making and frontline practice. Members of the TEC teams or service delivery partners would also work closely with such champions to offer ‘drop-in’ sessions, informal training, and the development of internal guidance materials. Such interactions were reported as important for building workforce knowledge and confidence, as well as gaining insights into common problems or misunderstandings on the ground for wider learning and system change.

However, given their informal nature, such networks were also considered difficult to establish and maintain universally across counties, with varying levels of engagement. Some strategic leads within the TEC teams felt this was reflected in a higher proportion of referrals from ‘usual’ practitioners and early adopting regions and teams, as opposed to service user need:

Variations in levels of engagement were linked to differences in practitioner attitudes, professional values, and styles of practice. The importance of focusing on positive messaging and cultural changes within the workforce has been highlighted by many strategic leads and champions. However, engagement among frontline practitioners remains a challenge in the face of limited time and capacity. Concerns about the extra logistical and administrative work that may be created by the use of home sensors further accentuated this challenge, particularly if the immediate benefits to staff or service users were not clearly evidenced or understood.

In addition to social care practitioners, domiciliary care workers were considered by TEC leads to be important for the success of sensor-based technology because of the frequent contact they have with the service user in their home; for instance, ensuring that the sensors remain online. This aspect of the care workforce tended not to be involved in the onboarding process and could be unaware of the purpose of the sensors and how they work, or concerned that the sensors may be monitoring their own activities and delivery on responsibilities. The high volume and heterogeneity of care, as well as variations in the level of domiciliary care provided to service users, made it difficult to operationalize enrolment and capacity building within this aspect of the care workforce.

Inter-organisational routines and coordination. The sustained adoption and embedding of home sensors within the social care ecosystem required close attention to the co-adaptation of interdependent roles and workflows within and across organizations.

The staff emphasized the importance of clearly defined processes for home-sensor-enabled pathway delivery and differentiation in roles (advanced division of labor with different specialist functions allocated to different teams). For example, in Eastvale and Riverbourne, practitioners noted the value of having distinct responsibilities between social care practitioners (for assessments and referrals) and installers (for set up and continued support):

The coordination of interdependent routines required the mutual adaptation of social and technical sub-stems across professional and organizational boundaries, supported by the joint and iterative development of 'structuring devices'³⁸ that guides and interconnects routine workflows. For example, in Eastvale, the TEC assessment form (used for ordering home sensors) was eventually integrated into the pre-existing social care assessment form hosted on the social care electronic record system, in order to effectively embed it within existing practitioner routines. This was subsequently linked to the technology providers CPR (Customer Reporting Management) system, enabling referrals to be sent directly on for installation, and in turn, keep referring practitioners updated as to when the technology had been installed, facilitating ‘mutual awareness’ (the sense of what the other collaborators are doing in order to provide a context for your own activity)³⁹.

In cases where technical integration was not possible (e.g., due to governance or interoperability constraints), staff would be required to bridge such systems as part of routine practice. For instance, in the case of Riverbourne, the home sensor dashboard and service delivery team IT systems were not integrated with social care IT systems; thus, referrals and PDFs of the dashboard activity data were securely emailed to the social care practitioners to manually save onto the social care record system. In Eastvale, lack of integration between NHS and social care systems meant the creation of dedicated staff within the hospitals to relay referrals onto the social care IT systems as part of the reablement and hospital discharge pathway.

The effective and continual co-adaptation of organizational routines and structuring devices over disparate systems relied on good collaborative relationships between the service delivery team, commissioning team, and technology provider. Establishing these processes was an ongoing accomplishment and remained a challenge in some key aspects of service delivery. In particular, structuring clear thresholds and protocols for proactive responses to activity data remained a persistent challenge. Even within Riverbourne, where a proactive monitoring team had been established, uncertainties remained with regard to organizational roles and accountabilities in operationalizing ‘actionable insights’ on a continual basis:

Strategic alignment, evidence and funding. Readiness and capacity for the introduction and expanded use of home sensors were strongly influenced by strategic alignment across organizations, in which such initiatives had been assessed by respective organizations (either as business plans or informally) and considered favorable with strong senior-level buy-in. A key challenge was establishing shared goals and evidencing outcomes related to different institutional logic and strategic priorities.

Strategic alignment and collaboration were important for extending, adapting, and experimenting with different technological options, service processes, and team roles. For example, Eastvale initially formed the hospital discharge pathway through temporary funding of a digital practice lead post to facilitate referrals as well as the dedicated installer reserved by the technology provider to ensure rapid deployment. Once the feasibility and outcomes of the service model had been shown, the temporary post was made permanent through NHS funding, and the technology provider decided to continue and expand the enrolment of installers specifically for the hospital discharge pathway [see Box 1]. As highlighted below, strategic collaboration and alignment for this initial (temporary, experimental) arrangement were contingent on both formal and informal relationships, trust, and shared strategic interests:

In addition to the importance of long-standing and trusting relationships across organizational boundaries, this example illuminates the importance of systems and appropriate skills in place to monitor and evaluate the impact of the innovation. Accordingly, strategic leads across sites also talked about difficulties aligning mutual interests and demonstrating benefits with other key parts of the social care and healthcare systems:

Additionally, some TEC leads felt that strategic alignment with private care services remained a challenge, in part because reduced in-person care provision contradicted the domiciliary care business model:

Strong leadership and senior-level strategic support appeared to be positively associated with securing funding and slack resources³² to expand and sustain the use of home sensors across social care pathways. A challenge in this regard has been to make the case to invest in new approaches to proactive care, in the face of wider financial constraints across health and social care. As a technology supplier noted in the following:

Economic feasibility analysis. Some staff talked to us about how the sensors saved costs for health and social care services by reducing care packages, delaying residential care placements, reducing ambulance call-outs, and avoiding hospital admissions. Economic data for this study was lacking. However, it was possible to obtain descriptive data on the characteristics of home sensor users (ie service users provided with home sensor technology) compared to home care users (ie service users not provided with home sensors).

In the dataset provided by Eastvale, sensor users tended to be slightly older, although almost half of the sensor and home care users were 85 or older. There was slightly less ethnic diversity in home sensor users than in those without sensors, although a large part of this could be due to the much lower proportion of sensor users for whom this information had not been provided or obtained. The average IMD score of the sensor users was slightly lower, indicating that they were less deprived, although the difference was small. Sensor users were more likely than home-care users to have autism, dementia, other neurological health conditions, and learning disability support.

One interesting observation from the data provided by Eastvale indicated that sensors were being used by service users with more costly care packages, which persisted even when controlling for the features of the care user in a Propensity Score Matching (PSM) model. This technique first creates a probability (the “propensity score”) that each care user would be given home sensors based on their observable characteristics, such as age, gender, and health condition, and then selects individuals to identify groups with and without home sensors whose needs are more equal than in the equivalent groups observed in the overall sample.

In the example below, the control variables explain £365.80 of the total cost of care in the full sample, but this increases to £724.66 when looking at a sample defined using PSM. This means that home sensor users were significantly more expensive than care users who were not provided with home sensors and implies that there is another variable, such as the complexity of care needs, that we were not able to identify from the data we were using in this estimation. See Table 3 for descriptive data and Table 4 for the PSM model.

The data provided by Riverbourne presented a demographic picture similar to that of Eastvale, in which more than half were over 85 years old and the majority (79.6%) lived alone. However, the classification system used for primary conditions was much more limited than that used in Eastvale. The three main reasons for installing motion sensors were a “decline in memory” (47.1%), “decline in health” (21.8%) and “decline in physical mobility” (19.1%).

Our experiences demonstrate that it is possible to use a combination of primary and secondary data to analyze the cost consequences of home sensors, but significant time and resources are required in a full study to overcome the challenges that we faced in this rapid evaluation context.

Summary of key findings

This evaluation of the use of home sensors for proactive care generated several key findings. Across the three sites, home sensors were used across multiple care pathways and contexts, including assessment, reablement (to varying degrees) and long-term care pathways. The perceived value and impact of home sensors in social care were numerous: Home sensor activity data were widely considered beneficial for service user independence and safety; family/informal carer reassurance; identifying health and social care needs; providing more holistic and objective assessments for personalised care; and communication and ‘evidencing’ need. However, the findings also highlight the tensions, uncertainties, and conflicting views on the use of home sensors and how they can shape the quality of care. In particular, there were some concerns regarding the reliability and suitability of quantitative activity data for care assessments and how this could be appropriately balanced against more direct and contextual information about the service user and their support needs. Additionally, while technology can provide reassurance and peace of mind for family/informal carers, there is potential for unintended consequences of increasing caregiver anxieties and demands, especially if the technology is not supported or compatible with people’s needs, priorities, and capabilities.

Implementation and use varied across sites and were shaped and constrained by numerous factors, including technical, regulatory, social, financial, and logistical contingencies. The adoption and embedding of home sensors in routine practice was an ongoing pursuit across case sites involving incremental and distributed changes across a nexus of interconnected cross-sectoral systems and practices. Key areas of focus in this regard included the material properties and dependability of the technology (involving maintenance work and supporting infrastructures); compatibility with service users and their care networks (aligning use with needs and capabilities of service users and their care networks); workforce knowledge and engagement (e.g., confidence, skills, and capacity); inter-organizational routines and coordination (co-adapting routines and supporting structures across organizational boundaries); and strategic alignment and funding (including senior-level leadership and buy-in, collaborative partnerships, and slack resources for continual adaptation and experimentation).

Despite some promising insights into the potential benefits to service users, family/informal care, and health and social care services, there remains a lack of compelling evidence on system-level impacts and cost-effectiveness. This has been underpinned by challenges in collecting and accessing relevant data. Further work is needed to support local evaluation capabilities to monitor and assess the impact and cost-effectiveness in ways that continually inform strategic and operational decisions on the implementation and scaling up of home sensors for proactive care.

Comparisons with previous literature

This study extends existing literature on the use of home sensors in social care. As reflected in our initial scoping literature review, participants described the multiple roles of such technology, including assessing care needs, supporting timely interventions, and providing reassurance. However, there is little published detail in relation to the on-the-ground experience of implementing, embedding, and using such technology within routine care practice, and implications for sustained adoption at scale. Our findings, therefore, provide further insight into experiential accounts with the technology and services around it, and the work of introducing and embedding such technology within social care.

Our findings illuminate how the effective use of home sensors emerges through situated practices and relational negotiations, ‘reshaping’ (as opposed to replacing) care roles and practices. For example, such technology introduces new and extended roles for family/informal care networks to monitor data and respond to associated insights. The emergent use of home sensor data by practitioners (e.g., to assess and evidence need or monitor care provision) also raises practical, interpersonal, and ethical implications regarding how such data are used. This reflects Pol’s⁴⁰ emphasis on the need to understand and accommodate the human and relational dimensions to technology-enabled care if it is to become successfully adopted and embedded within frontline care practice.

Additionally, following our initial scoping literature review, we specifically sought to shed light on the practical and logistical challenges of incorporating technology within social care workflows and infrastructure. Our findings in this regard resonate with wider socio-technical systems theory, which has long emphasized the need for mutual adaptation of technologies and processes to extend and improve the fit between them^41,42. In particular, the case sites draw attention to the importance of attending to ‘organizational routines,’ defined as ‘recognizable, repetitive patterns of interdependent action carried out by multiple actors’³⁴ across organizational and sectoral boundaries. Routines help reduce uncertainty, support complex collaborative work, and maintain ‘mutual awareness’³⁹ of distributed roles and actions. Attending to routines is therefore important in establishing the foundations for implementation (through alignment with established routines and minimizing disruption to existing practice), scale-up (through stability and replicability), and sustainability (through consistency in process and retaining organizational memory). Organizational routines are also situated within a socio-material context; in that, actions are structured around time, physical spaces, material, and technological artifacts. Within our data, we observed the importance of co-adapting routines and associated ‘structuring devices’ (e.g. online forms, electronic record workflows, protocols)³⁸, in order to maintain clear processes and roles and accountabilities between staff. This required close collaborative work across the social care, service delivery, and technology provider teams, which has been key in this regard, requiring time and long-term strategic planning to build relationships and trust for ongoing adaptation and system learning. However, defining and negotiating such routines and accountabilities remained an ongoing challenge in many aspects of service delivery. In particular, uncertainties remained in the formalizing of decision-making protocols and accountability with regard to proactive responses to home sensor activity data.

Our focus on organizational routines further highlights the need to consider various types of ‘articulation work’ being performed by different actors involved in the provision, maintenance, and day-to-day management of home sensors. Articulation work can be described as a ‘supra type of work’ that connects and integrates tasks and ensures activities ‘mesh’ across divisions of labor^43,44. In our study, we observed the role of articulation work in the coordination of activities (e.g. between referring practitioners and technology installers), data management (e.g. sharing and information across disparate IT systems), keeping the technology in working order (e.g. maintenance visits, check-in calls and monitoring technology status in use) and helping manage relationships (e.g. ‘emotional labor’³⁷) with service users and their families. We also observed the contribution of articulation work undertaken by service users’ families and informal networks as they attempted to appropriate and ‘domesticate’⁴⁵ the technology within the home, and take on extended roles in monitoring, responding, and interacting with call center operators, for the system to work as a whole. Previous research has shown how a mismatch between formal procedures and what is actually necessary in practice to ‘get the job done’ may compromise the scalability of the service at the organizational level^44,46, stressing the importance of having mechanisms to help surface such work and establish how it may be supported and resourced efficiently, and hence more sustainable and scalable.

Our findings on workforce influence reflect wider debates on how ‘digital skills’ are defined and understood in health and social care⁴⁷. With regard to home sensing, social care practitioners’ knowledge was not limited to how the technology worked in general, but involved situated knowledge and skill, which extends beyond general awareness and understanding of the technology. There has been much policy interest in ensuring that the health and social care workforce has the right skills to apply and utilize digital technologies⁴⁸. However, as highlighted by Whitfield et al.⁴⁹, current policy discourse on skilling the workforce underplays the diversity of roles across the care system and overlooks the need to manage the heterogeneity and shortcomings of digital devices and systems adaptively. The adoption and use of home sensors across social care relied heavily on the development and sharing of tacit (know-how) knowledge through a combination of formal training (e.g. demonstrations with technology by providers), informal support (e.g. drop in sessions on activity data interpretation) and ‘learning by doing’ (e.g. how aspects of the technology can be configured for particular needs, troubleshooting issues) across multiple actors in the system. However, opportunities for learning and support were often limited to an already overstretched, under-resourced workforce and often constituted additional ‘hidden work’ that was difficult to manage and sustain.

In addition to understanding the process of implementing and embedding the use of home sensors in social care, we also sought to assess service-level outcomes through economic analysis. As highlighted in the initial scoping review, while there are widely held perceptions and assumptions on the cost benefits for health and social care, there is currently a lack of methodologically robust evidence. Our collaborative efforts on this component revealed how the national absence of such evidence is not merely a methodological issue but is underpinned by wider infrastructural (technical and regulatory) challenges faced widely across the social care sector. Similarly, Mendes et al.⁵⁰ recently highlighted how system fragmentation significantly limits the scope of comprehensive economic analysis across health and social care, linked to data access and governance constraints. Improved integration is needed to gain a deeper understanding of how and the extent to which home sensors can advance proactive care and mobilize resources and investments to scale up and spread such service models. This calls for closer engagement at the national policy level, in collaboration with regional stakeholders across health and social care. We have found that domains to focus on include service set-up and running costs (including technology, supporting infrastructure, service delivery and ‘wrap around’ support), social care provision (including other technology as well as in-person care), health outcomes (including hospital admission and treatment), as well as social and quality of life data (e.g. using short standardized measures within assessments and reviews).

Finally, our analysis affirms Glasby et al.’s³ previous study of similar technology, highlighting that existing pressures on health and social care constrain the scope of pursuing a wider strategic interest in the long-term benefits of proactive and preventative approaches to technology-enabled care. Our findings are in line with the wider literature on the diffusion of innovations that emphasizes the importance of establishing shared (and evolving) meanings and values in relation to the innovation in use, and a coherent and shared understanding or ‘organizing vision’⁵¹ for describing the role and impact.

Recommendations

The findings offer key insights and recommendations relevant to different stakeholder groups.

For national policymakers:

For local authority decision makers and commissioners:

For social care practitioners

For technology providers

In summary, sustained adoption of home sensors ‘beyond the pilot’ demands a coordinated, multi-level approach across all sectors and levels of care. This should include a focus on: consolidating emerging guidance and expertise; investment in workforce capacity and skills; evidence-based commissioning for incremental and iterative service development; support and guiding frameworks for local evaluation capabilities; and cross-sectoral partnerships to continually co-evolve the social and technical subsystems that constitute service delivery.

Strengths and limitations

This rapid evaluation is small-scale and was conducted over a period of 13 months, during which data availability and access needed to be established and agreed upon. Available data for the economic component were limited; therefore, a comprehensive analysis could not be conducted. However, the process of assessing data availability in collaboration with our site partners provides useful insights for methodological, practical, and policy recommendations for future economic analyses.

While the study generated rich qualitative data, a larger study with more in-depth and longitudinal data could have provided further insight into contextual influences and complexities and how these change over time. The findings are drawn on three diverse sites that are not generalizable to the wider social care sector. However, through a theoretically informed analysis alongside wider stakeholder engagement, we sought to draw transferable lessons to inform future policy, practice, and research. Through system mapping and strategic sampling, we interviewed a diverse sample of staff and system stakeholders as well as service users and family/informal carers. We did not gather data from people who had not been offered home sensors or who were unable to access the technology. Furthermore, we were unable to interview service users who lacked the capacity to consent due to governance restrictions for the service evaluation. Given the high proportion of service users with cognitive impairment, a longer study with relevant ethical approvals with MCA (Mental Capacity Act) consultee procedures could provide insight into the lived experience of this user group. However, our study significantly extends the existing body of literature with regard to lived experience and care practices related to the use of home sensors for proactive care.