Identifying models of care to improve outcomes for older people with urgent care needs: a mixed methods approach to develop a system dynamics model

Work package 1: identifying best practice

• WP1.1 – review of reviews of UEC interventions for older people, their outcomes and costs and any implementation factors identified

  • Research question (RQ) 1.1.1 – what is the evidence base for UEC interventions for older people, the outcomes of these interventions and the costs associated with these interventions?

  • RQ 1.1.2 – what factors have been described in the evidence base to date that influence implementation of UEC interventions for older people?

  • Output 1.1 – a taxonomy of UEC interventions with outcome effect sizes (where available) and descriptors of costs and implementation factors.

• WP1.2 – patient and carer preferences.

  • RQ 1.2.1 – what elements of care are most important to older people and their carers with UEC needs?

  • RQ 1.2.2 – how could UEC interventions be configured to best meet the needs of older people?

  • Output 1.2 – description of patient and carer priorities for UEC.

• WP1.3 – staff perspectives.

  • RQ 1.3.1 – what other interventions, not yet reported in the literature, offer promising models for improving outcomes for older people in the UEC pathway?

  • Output 1.3 – staff perspectives on the ‘state of the art’ and factors that will facilitate implementation.

Work package 2: qualitative study of delivery of exemplar urgent and emergency care pathways

• RQ 2.1 – what aspects of interventions, context and approaches to implementation facilitate and hinder delivery of UEC interventions for older people?

• Output 2.1 – context-, implementation- and intervention-related influences on delivery of interventions for older people in UEC settings.

Work package 3: routine patient-level data analysis

• RQ 3.1 – are some UEC pathways associated with better patient outcomes than others?

• RQ 3.2 – have UEC pathways improved or got worse over time?

• RQ 3.3 – over and above the UEC pathway, what patient characteristics, demand and supply factors explain differences in outcomes from place to place and over time?

• RQ 3.4 – what is the relationship between outcomes and the costs of the UEC pathway and are some pathways more cost-effective than others?

• Output 3.1 – estimates of what drives better outcomes and lower costs for older people with UEC needs.

Work package 4: improving emergency care pathways

• WP4.1 – baseline simulation model development.

  • RQ 4.1 – what is the best way to build an ‘archetypal’ or baseline model of patient flow into and out of one specific UEC pathway, ensuring that the model is valid and captures all the relevant factors?

  • Output 4.1 – a validated quantitative (stock-flow) system dynamics model of patient flow, describing the status quo situation in one specific setting.

• WP4.2 – ‘what if’ analyses.

  • RQ 4.2 – what changes can be made to existing UEC pathways that will lead to greatest improvements, and what might the consequences of such changes be for the wider health-care system?

  • Output 4.2 – a family of system dynamics models based on output 4.1 describing the whole-system impact of evidence based, patient centred interventions applied to UEC pathways.

Search methods for identification of reviews

Criteria for selecting reviews for inclusion

Screening of references at title and abstract level was done by one reviewer (LP or JvO), with 50% also screened by SA. All reviews that met the title and abstract screening criteria (Table 1) were screened at full text by LP and SC. Any reviews that were excluded at title stage were listed and reasons given for their exclusion. Agreement between reviewers was assessed using kappa scores.

Application of reporting standard criteria

As part of the screening process, reviews were identified which were relevant (by inclusion criteria; Table 2) but did not meet the methodological standards for a systematic review. In the protocol, we stated that we would assess reviews according to A MeaSurement Tool to Assess systematic Reviews version 2 (AMSTAR2) to assess systematic reviews. However, we adapted this approach and added an additional stage where we screened reviews to assess their methods and reporting prior to the use of AMSTAR2. This prior assessment allowed the exclusion of evidence described as a review but where data extraction and quality assessment were not possible. The reporting standards criteria developed were based on the Cochrane Handbook definition of a systematic review and criteria developed by Brunton. 16 Reviews were included in the review of reviews if they met three or more of the criteria. Conference abstracts were not assessed according to these criteria.

Data extraction and synthesis

Data were single extracted by one reviewer (LP, SA or JvO). The full data extraction tables are included in Appendix 2. These tables were all checked by LP and a random sample of 10% were also checked by SA. Data were extracted on review description, review methods, description of included studies, all reported outcomes (including whether they had been synthesised or reported as individual studies) and a headline message/conclusion.

Assessment of the evidence base

For the review of reviews, we used two different tools: AMSTAR2 (for reviews reporting interventions) and the Joanna Briggs Institute Critical Appraisal Checklist for Systematic Reviews and Research Syntheses (for non-intervention papers). AMSTAR2 was chosen because it allows the appraisal of reviews that include non-randomised studies of interventions, not just randomised controlled trials (RCTs). 17 We undertook a narrative assessment of the applicability of the results to the UK setting.

Size and scope of the evidence base

To examine the overall evidence base, a citation matrix18 was drawn up using EndNote 9 and Microsoft Word, mapping primary studies against the reviews in which they were cited.

Taxonomy development and validation

One of the review objectives was to develop a taxonomy of models of care delivery to older people in the ED; however, the evidence was not organised according to models of care. Three alternative frameworks (McCusker et al. 2018,53 American College of Emergency Physicians and the original framework described in the project protocol), were tested on three selected reviews by SC, JvO and LP. A process of discussion and consensus led to the choice of the Elder-Friendly Emergency Department Quality Assessment Tool. 19 This tool was used to identify whether the reviews included in our review of reviews included evidence of interventions related to any of the 13 subscales, which are as follows:

• two‐step screening

• standardised assessment tools

• clinical care protocols

• geriatric team

• multidisciplinary staff

• discharge planning

• family‐centred discharge

• linkages between ed and homecare services

• physical environment and design

• furniture and equipment

• educational sessions

• quality improvement

• administrative data monitoring.

Again, three reviewers classified the reviews and then we discussed and agreed via consensus which of the subscales were represented in the reviews.

Characteristics of included reviews

A total of 18 reviews and three conference abstracts were included in the review of reviews. The date range was 2004 to 2019, with primary papers included in the reviews dating from 1984 to 2018. None of the reviews limited the papers by geographical setting. Not all the papers in all the reviews were included; where evidence related to screening for frailty or frailty related syndromes, these papers were excluded. The range of (relevant) papers included in reviews was 2–28 and there were 128 unique papers included in the 18 reviews. None of the conference abstracts reported bibliographic details of included studies.

Six supplementary reviews were identified in the 2021 searches.

Overview of included reviews

Outcomes

Figure 2 lists the reported outcome; the patient outcomes map closely to those described by Akpan et al.,38 although they were inconsistently reported. The reported range of follow up across all the reviews was from 0 days to 18 months.

FIGURE 2.

Overview of included reviews – outcomes

Overview of included conference abstracts

Three conference abstracts were included in this review. Cherian et al. 39 reported interventions to enhance EDs to be more suitable for older patients. Gupta et al. 40 described a review of geriatric trauma protocols and interventions. Tran et al. 41 examined interventions for patients at high risk following discharge from the ED.

Synthesis

There was a high degree of heterogeneity in the review methods and reporting across the 18 reviews included in the review of reviews, so meta-analysis was not possible. Instead, data were reported as:

• summary tables, reporting intervention descriptions and outcomes reported (where a statistically significant positive effect was measured)

• narratively within evidence clusters designed to gather evidence by intervention type.

Summary tables

These tables report intervention descriptions (by review) and any statistically significant outcomes reported in the review, with the number of studies in which the outcome occurs. Table 4 contains data from reviews where results were reported narratively for patient outcomes. Table 5 contains from reviews where results were reported narratively for health service outcomes and Table 6 contains data from reviews where meta-analysis was undertaken. It is important to note that this table does not account for overlap across the reviews in terms of the primary studies included.

Table 4 shows that reduced functional decline was the most frequently reported outcome, although in Table 5 it appears that most of the positive intervention effects were for health service outcomes.

Reviews which undertook meta-analysis24,32,33 reported on a variety of patient and health service outcomes but did not identify any statistically significant effects on outcomes as a result of interventions with the exception of Malik et al.,33 who reported reduced readmission rates.

Narrative synthesis

Table 7 describes the five evidence clusters that were developed from the review findings. These were developed iteratively and grouped thematically;42 they are presented narratively, and are summarised according to their reported effectiveness. This does not reflect the primary studies but on the strength of the findings as reported at review level.

Study assessment

Applicability of evidence to the UK setting

Although much of the primary research summarised in the reviews emanated from North America, it is likely that the review findings will apply to UK settings. Older patients with urgent care needs will present similarly in North America compared with the UK (e.g. high prevalence of non-specific presentations). The interventions were broadly holistic in nature, consistent with the international literature supporting CGA as an intervention to improve outcomes in older people with acute care needs. Finally, the outcomes described as important are consistent with the outcomes valued by older people in the UK, as well as the service outcomes being valued by providers and commissioners of care and consistent with the International Consortium for Health Outcomes Measurement report. 38 It is therefore likely that the review findings will be relevant to the delivery of urgent care for older people in the UK.

Taxonomy

To address the research objective of developing a taxonomy of models of care, as reported in the introduction, the McCusker et al. 19 Elder Friendly Emergency Department Quality Assessment Tool was used to identify whether the reviews included in our review of reviews included evidence of interventions related to any of the 13 subscales of the tool. Use of this tool to classify the evidence allows us to describe which areas are more likely to have interventions relating to them (Table 9). The most frequently occurring subscales were interventions relating to multidisciplinary staff, with standardised assessment tools, two‐step screening, geriatric team, discharge planning and linkages between ED and homecare services also occurring in either 11 or 12 reviews. The remaining subscales were mentioned in few of the reviews, although there was often mention of issues relating to these in the background or discussion sections of reviews as being important, particularly to patients.

Introduction

Little is known about the experiences and preferences of older people for emergency care, even less for older people living with frailty. Reviews of existing evidence highlight waiting times, physical and psychological support alongside good communication and information provision in shaping the experiences of older patients in the ED. 20,22,48 Although existing studies included people with stereotypical markers of frailty, none used a validated frailty assessment. As such there is currently limited evidence about whether or how the presence of frailty impacts upon older people’s experiences and preferences for emergency care. 48 These needs must be understood and addressed to provide person-centred care.

Methods

Results

Discussion

People wished to be treated by caring, responsive staff and with respect, which was understood in terms of dignity and appropriate care. The provision of timely information communicated clearly and honestly to both patients and relatives was also highlighted, not least in enabling patients to be involved in decision-making about their own care. Participants wanted to experience emergency care in a calm, quiet environment where basic needs for privacy, comfort and food/drink are met, and where they can be supported by relatives and friends. The desire for shorter waiting times was also prominent. Study participants described very mixed and sometimes negative experiences in terms of whether these preferences were met. This explains the reluctance expressed by many about attending hospital in the first place.

The findings reported from this study are consistent with previous research on the experiences of emergency care for older people, which emphasise the importance of information and communication, staff attitudes, waiting times and physical/environmental needs in shaping those experiences. 20,22,48,53,54 These aspects of care have also been identified as instrumental in determining the experiences of adult patients in the ED. 55–57 In many respects, the needs and priorities of older people living with frailty in the ED mirror those of patients of all ages, not least in the import attached to relational aspects of care (communication, compassion and empathy) over technical competence. 55–57 Our findings suggest, however, that these needs and preferences are particularly pressing for older people living with frailty as they are perhaps more vulnerable in the ED environment both physically and emotionally when these needs are not met. 48

For older people living with frailty, interactions with staff are inextricably linked to their own sense of self and dignity. Being treated with respect and kindness provides validation and legitimacy, whereas being ignored or dismissed (e.g. when seeking help with toileting) can have a long lasting negative impact upon self-esteem. 20 Similarly with regard to information and communication, older people living with frailty may need particular support to be involved in decision-making,58 such as timely information communicated clearly, and for staff to check that patients have understood. 59 Communication barriers relating to hearing/visual problems, language or cognitive impairment need to be identified and overcome. 22 The presence, involvement and advocacy of relatives is imperative for both older people living with frailty and their carers. 59,60

Our findings show that the noisy, busy ED environment poses a particular challenge to older people living with frailty who prefer to receive care in a calm and quiet setting. They are particularly sensitive to and distressed by the physical privations they often experience in ED in the form of uncomfortable trolleys and a lack of food and drink. While previous studies remark on the ‘resilience’ or ‘tolerance’ demonstrated by older people encountering long waits in ED,61,62 this may be more challenging for older people with frailty. Participants reflected very negatively on this element of their experience; they had no choice but to ‘tolerate’ long and/or uncomfortable waits which took a considerable toll upon them at the time and appeared to make them more reluctant to attend the ED when faced with future emergency episodes.

Participants included those living with mild to severe frailty and, as such, we report the views of a group of patients who have previously been excluded from research in this area yet represent a sizeable and growing number of those accessing emergency care. Although views on desirable outcomes from ED attendance are not explored here, these data were collected during the study and are presented separately.

Despite the efforts of recruiting staff, the lack of ethnic diversity within our sample is a limitation. We did not recruit any patients from the very severe frailty categories (CFS 8 and 9). Further, we were unable to explore whether there were any relationships or associations between distinct levels of frailty (i.e. mild/moderate/severe) to experiences and preferences for emergency care. This would require a larger mixed methods study including more patients within each frailty classification. Almost one in five participants had cognitive impairment; the inability to gain consent (or consultee consent) or the inability to participate in an interview may have influenced greater representation of people with cognitive impartment. Although most interviews were conducted within 30 days of the ED attendance there is always the possibility of inaccurate recall. Moreover, the focus on relational aspects of care may reflect the purview of both patients and interviewers who, having little knowledge of medical processes and procedures, focused upon interpersonal elements of the ED experience. 55

This study provides important evidence about the extent to which frailty impacts upon older people’s experiences of emergency care. Our research suggests that frailty can result in a particular vulnerability in ED if physical (ED environment, personal comfort, waiting) and emotional (sense of dignity, communication, involvement, family support) needs are not met.

At the level of practice, we make several recommendations which might potentially be delivered in clinical practice (Figure 3). Recommendations relating to staff care and attitudes and information and communication could make a significant positive difference to the experiences of older people living with frailty in ED. While the ED environment and waiting times may be harder to change, health care professionals can help older people living with frailty by being mindful of their comfort, physical needs, the flow of information and the importance of patient/carer involvement.

FIGURE 3.

Meeting the needs of frail older people in ED: recommendations for practice.

More broadly and given the challenges of more fundamental changes to the fabric of the ED and the pressures on this part of the health-care system, policy-makers and practitioners need to consider service development changes when responding to the needs of older people living with frailty requiring UEC. These may include the adoption of new approaches closer to the hospital ‘front door’, interventions which minimise or even bypass the time these patients need to spend in ED, as well as frailty-friendly design in ED. 63,64 Any new models of care must be based on robust research evidence, which should relate to clinical and cost effectiveness but also to patient and carer perspectives. 65

Conclusion

Frailty reflects a vulnerability and a need for support in basic activities of daily living, which EDs in this study, and perhaps more widely, are not set up to provide. Changes at the levels of clinical practice and service design are required to deliver even the most basic care for older people living with frailty in the ED environment.

Methods

Using the same study design, participants and data but by analysing with a focus on outcomes rather than preferences, we report here on patients’ preferred outcomes from emergency care. The methods and recruited participants data are not repeated here, but the reader is referred to the Methods section above.

Results

Findings are presented as goals of care and attainment of those goals (expressed verbatim or illustrated through experiences).

Discussion

This analysis explored the goals of care of older people living with frailty and their families/carers during an urgent care episode. The key finding was the discrepancy between what people sought and what was attained (Table 11).

This research adds to existing evidence by expanding the issues identified by older people, those living with frailty, who are poorly represented in previous work. 48,53 Existing outcomes measures for older people38 do not address all of the goals of urgent care identified by our participants from their perspective; for example: addressing the medical problem and feeling better; receiving information about diagnosis and treatment, medication changes and involving relatives and carers; involvement in discharge planning and appropriate discharge; referral, communication and coordination with primary and social care; returning to usual activities; not being readmitted or needing to reattend ED. It is somewhat surprising and disappointing that some of these goals of care were really very modest yet were often not attained.

Our participants did not particularly discuss mental health as desired goals of care other than perceiving these as medical needs requiring attention. They did not expect urgent health care episodes to tackle their experiences of loneliness and isolation. Carer burden was evident from our interviews.

To our knowledge, this is the first paper discussing goals of care and their (non)attainment for older people living with frailty and emergency care needs. Participants were interviewed by expert, non-clinical researchers with subsequent clinical validation, facilitating free and open discourse about health care. In this cohort of older people living with frailty, we were able to identify multiple domains of meaningful goals of care not apparently addressed in previous reviews,48 adding to the understanding of how these might be better elicited and addressed. This study took place in three English hospitals with quite different emergency care processes, yet the goals of care identified were consistent across all three. There was a lack of ethnic diversity in our sample, despite the efforts of the recruitment staff. We also failed to recruit participants with high frailty scores, who may have a different focus in terms of goals of care.

Many older people living with frailty have poor service-centred outcomes from urgent care. 9 These interviews describe limited attainment of person-centred goals of care, with many participants reporting unattained goals or deteriorating while in hospital; this qualitative description adds depth to previous studies reporting on poor outcomes from emergency66,67 and acute care68–70 in older people living with frailty.

Emergency care research and quality improvement usually focuses upon service metrics such as length of stay and readmission rates, which do not necessarily align with patient perspectives. 23,71 Such service metrics are easy to generate but may not be meaningful to older people living with frailty. 48,72 In particular, holistic, person-centred care which can best serve people living with frailty, is poorly captured using such metrics. 48

There is a need for improved communication across the whole pathway of care, with patients being able to feel involved at every stage. Readmissions are a concern within increasingly pressurised systems. The data presented here fit with other reports that have identified that many reasons for readmission might reasonably be identified during the index acute admission. 48,73 It might be that a focus on delivering patient-centred care, driven by eliciting and addressing patient care goals, might lead not only to higher quality care but also to more efficient care. For example, if patients report that their goal is to improve function, and this is the focus of the acute admission then readmissions might be avoided.

Future research could explore if capturing goals of care at the beginning and throughout an urgent care episode might influence clinical care and pathways, and thence the achievement of said goals. Engagement in research of people of minority-ethnic backgrounds is needed, as is engagement of people with high levels of frailty.

Most participants in our study did not attain their goals of care and many deteriorated or experienced adverse effects in hospital and difficulties after discharge. Attainment of care goals from first contact to post discharge for older people living with frailty and urgent care needs should be improved. Further work is needed to develop goal-oriented care for older people living with frailty which is personalised to their specific needs.

Conclusion

Older people living with frailty have heterogeneous urgent care goals which require individual ascertainment. Identifying these goals of care early could result in improved attainment through person-centred care.

Approaches to urgent and emergency care for older people

Participants in interviews and focus groups described a wide range of models of care for older people seeking UEC, with various labels and some variation in terms of staffing profile, referral criteria, objectives and care provided. However, they tended to fall under three broad headings, with alternative models of provision under each: community-based teams focused on reducing attendances and admissions; interventions delivered within the ED providing specialised care and support; and interventions offered elsewhere in the acute hospital, either as an alternative to care in the main ED or an alternative to admission to a ward from the ED.

Community-based teams tended to offer a mixture of reactive and proactive interventions. Proactive interventions sought to pre-empt crises and associated emergency attendances and admissions, involving for example community frailty teams, multidisciplinary teams in general practices and community matrons. They tended to include a wide range of multidisciplinary staff, collaborating closely with GPs to identify patients at risk of crisis, assess needs through CGA and plan care accordingly. Reactive interventions responded to immediate needs for emergency care among older people with a view to avoiding an unnecessary attendance, by providing an assessment service equivalent to that offered in the ED in a patient’s own home or in a separate medical assessment unit, together with referral routes to appropriate diagnostic, treatment and care services. Services of this kind typically took referrals from GPs, call centres (emergency hubs and NHS 111) and ambulance crews and had rapid response times (e.g. under two hours), although they tended not to be available 24/7. Some included ‘in-reach’ functions, to ‘pull’ patients out from an ED following attendance and provide assessment and care in a community setting. Typically, they were nurse or therapist led with access to geriatrician input as necessary; some also included input from pharmacists and other specialist practitioners.

Services within the ED included dedicated teams and/or dedicated areas of the department. Dedicated teams included those led by practitioners who were usually community-based, such as community matrons, and those led by staff based in the ED or wider hospital. Their remits overlapped, although the focus of community-led teams tended to be on identifying and assessing patients who could potentially return home swiftly and safely once appropriate support was put in place. These teams would look to identify such patients early after arrival and take advantage of strong connections with community-based providers of health and social care to set up support packages rapidly, ideally within a few hours and usually the same day. Teams led by ED and wider hospital staff would also look to discharge patients rapidly where appropriate, but also had links to provision within the hospital, such as frailty units (see below), which could offer the right multidisciplinary input to support safe and prompt discharge. Typically, these teams were geriatrician led, with input from nurses, therapists and other relevant clinicians. Whether formed of community or hospital based practitioners, dedicated teams did not generally offer 24/7 input. Dedicated areas of the ED comprised beds or chairs specifically set aside for older people living with frailty, with a view to providing a calmer environment, and were served by dedicated teams. Some could be accessed directly by GPs or ambulance crews, while others relied on triage at the ED’s ‘front door’. Again, a key aim was to rapidly assess the potential for discharge to usual place of residence, with support as required, and to secure admission to an appropriate unit within the hospital if necessary.

The third main form of intervention identified by participants was services elsewhere in the hospital. These fell into two main categories: services offered as an alternative to the ED, such as frailty assessment units, and those offering an alternative to admission to a regular ward, with the expectation of a short (< 72-hour) stay, such as frailty units. Alternatives to the ED were located in physically distinct areas of the hospital and generally led by geriatricians, with multidisciplinary staff, and tended to operate during ‘office hours’, typically between 8 and 12 hours per day on weekdays only. Access arrangements varied but, in some cases, they could be accessed directly by ambulance crews, rather than requiring triage in the ED. Short-stay services were provided for individuals for whom admission was necessary but expected to be brief, though in practice, pressures on bed availability meant that patients perhaps more suitable for other wards were often accommodated, sometimes for longer periods. Some units included fairly substantial numbers of beds (20–30); others were much smaller. In some cases, short-stay frailty units and frailty assessment units were integrated, providing a ‘one-stop’ service for frail older patients that avoided both the ED and the main hospital wards.

Key components of urgent and emergency care interventions

Participants identified a range of components that they saw as important in the delivery of UEC interventions, which were necessary regardless of the specific approach taken. To a large extent, these mapped well to the McCusker framework used to organise the components of care described in the literature reviewed by WP1.11 There were some omissions in what participants described; for example, there was relatively little reference to the use of routine data for audit and evaluation purposes or to the use of quality improvement methods. Participants’ focus was primarily on components of care delivery rather than on continuous monitoring or improvement.

Common to many participants was the view that the physical environment of the ED was ill suited to the needs of older people living with frailty, and that the noise and bustle could even lead to deterioration. These physical features were partly what drove efforts to develop dedicated areas within or outside the ED, such as assessment units. Where dedicated areas had been set up, much care had often gone into making them ‘frailty friendly’. However, other than short-stay frailty units that provided an alternative to admission to a ward, few services in and around the ED were available at all hours and, indeed, many were accessible only during ‘office hours’ on weekdays. The general view among participants was that the ebbs and flows of patients meant that 24/7 community or ED-based services might be excessive, but that opening 12 hours a day, 7 days a week would help to ensure availability to patients who could most benefit from such provision.

While the physical limitations of the ED environment were one key reason for setting up separate dedicated areas for older people, a more pressing driver was their distinctive needs in terms of care, assessment and care planning and the implications of these needs for skill mix, organisation of care pathway and timescales. Prompt and thorough assessment of older people, ideally covering all of the domains of CGA, was seen as a vital component of good care, but the ED itself offered a poor setting in which to deliver it. Assessments led by non-specialised ED staff were often patchy and rushed and tended not to be prioritised given the need to make decisions within the window of the four-hour standard. Delivering a high-quality, multidisciplinary assessment required either a highly coordinated and focused process or a more decompressed timescale, offered respectively by dedicated teams and spaces outside the ED. Comprehensive assessment offered by community teams, for example, could offer a key part of durable plans to avoid unnecessary attendance, while assessments by dedicated teams or in in dedicated areas ensured rapid and accurate identification of needs for the duration of any admission, and to secure a safe and timely discharge.

[Describing the approach taken in a frailty unit] The way the patients were being managed in the pitstop, everybody had [CGA] so by the time they leave the pitstop, they had a medical plan, they had an OT plan, a physio plan, pharmacy review, everything done. So there was not a lot that needed to be done once they actually went into the ward except just follow the plan that was made by them, they didn’t have to chase physios, chase OT because it was all done.

(P02a AMD, service improvement)

Though some EDs did seek to offer CGA to appropriate patients routinely, without the input of specialist teams, there was scepticism among participants about the adequacy and accuracy of such approaches. Where assessments were begun in the ED, they were often incomplete at the point of discharge or admission and left to teams in the hospital or community to complete—an inherently fallible process.

With the proper CGA it takes a lot of time and resources. And unfortunately you know they don’t have that in ED with the volume of patients. So what happens is that this is just an eyeballing sort of tool to categorise patients not necessarily with a proper CGA (P19 Pharmacist, Lead care of older people and stroke)

There’s a real time pressure in A&E to get it all done. […] Because obviously you don’t necessarily complete the whole of the comprehensive geriatric assessment in the time, if there are outstanding bits, so we need to make sure that we can then hand that back onto the community to follow up.

(P05 consultant physician and geriatrician)

Closely related to timely and accurate assessment was the need for discharge planning right from the start of contact. Again, the coordination of multidisciplinary input across sectors was central to this process, and a fundamental part of many of the models of care put forward as alternatives to the standard pathway through the ED. There was a sense among some participants that proactive efforts put into discharge planning by frailty units and other alternative models meant that admissions were more likely to be avoided in such settings (even accounting for case mix). Participants argued that this was due to better assessment and better links with community service providers, as well as to staff who were more confident and less risk-averse than decision-makers in EDs, and able to make decisions based on a wide range of disciplinary inputs.

So it might be that when the consultants are seeing the patients, some of them will spend a lot more time than an acute medical consultant would spend with them. But that is the point of having us there because we’re trying to deliver the patient a more in-depth care, if you like, because if we’re going to discharge them we want to make sure they don’t come bouncing back or that we tie them into the right service.

(P20, advanced nurse practitioner)

Correspondingly, strong links across sectors and services were seen as vital, particularly with primary care and other community-based teams, whose input would be vital to ensuring that care plans were delivered and arrangements for avoiding admissions were more than short-term patch-ups. As well as intervening to prevent unnecessarily attendances, some community-based teams also had close links with ED and other hospital-based staff, providing additional expert input and broadening the range of options available for assessment and care. Similarly, links from community-based teams to hospitals also offered streamlined access to diagnostic facilities for patients in the community, reducing the need for admission. Establishing consistent links across a hospital’s catchment area, though, could be challenging when it spanned multiple primary care networks, community health-care providers and local authorities, and this could create uncertainty and inequality in discharge decisions.

I would say the biggest negative is around readmission. So, it really highlighted how poor we are as an integrated care system. So as we were balancing risk and taking more risk around discharge it almost feels often that your discharging is a bit of a black hole. And the next real step change I think that needs to happen is around that integrated pathway. And proper discharge from hospital and trying to create a fluid team really that works across into the community.

(P09 consultant geriatrician)

Good-quality communication was at the heart of creating such links, and participants reported that information could sometimes go missing in discussions across teams and sectors–for example in relation to patients’ preferences regarding admission, treatment, and resuscitation. Participants also noted inconsistencies in the quality of communication with care homes and with patients’ families, both important to ensuring durable returns to living in the community.

Other critical dependencies with other services included palliative care and ambulance services. Patients nearing the end of their lives were seen as frequent attenders at EDs, sometimes due to poor symptom control, and at risk of inappropriate admission and intervention if coordination with palliative care teams was lacking. Participants saw the decisions of ambulance crews as representing a potentially crucial point in selecting the right emergency care pathway for older people but reported mixed experiences of the extent to which ambulance staff were skilled and empowered to diverge from the default of conveyance to the ED. In some cases, protocols had been established for ambulance crews to refer patients to community teams or hospital-based frailty teams, based on initial screening and assessment undertaken by crews. In others, ambulance crews could contact frailty teams or geriatricians by telephone for advice about how to manage patients on a case-by-case basis. Elsewhere, however, ambulance staff were seen as lacking in knowledge about frailty and unwilling to take risks, even if patients themselves expressed a preference to avoid hospital.

Yeah, but because of the area I’m in, the older people—because as well one of the things that we’ve tried to do is say actually older people going to hospital is not the safest option, because there’s always been this kind of mentality in the ambulance service, ‘Well you can’t harm them if you take them to hospital, you can’t get in trouble if you take them in’, and they’ve always had this issue about leaving people at home.

(P07)

Similarly, interfaces with other groups who might channel patients towards frail-friendlier services were also critical, but inconsistent in quality. One issue was responsibility for frailty assessment and ensuring that it identified the patients most likely to benefit from available services in a timely way. Staff in the ED, and in other ‘gatekeeping’ roles such as ambulance crews, were seen as well placed to assess for frailty, but ensuring reliable and accurate screening could be challenging. What frailty teams saw as a vital and relatively straightforward process was not always perceived that way by other teams expected to implement it. Despite the apparent simplicity of frailty scoring systems, some participants reported systematic differences between their own scores and those derived by colleagues involved in triaging in the ED. Some efforts had been undertaken to provide training and to routinise frailty screening in the workflows of colleagues, but with mixed success, in part due to disagreements about where responsibility lay.

We were told – this was in discussion recently – that [ambulance service] can’t implement anything new to the paramedics that isn’t in their core training, even though it’s just a pictorial.

(Participant, focus group 2)

Participants also identified other opportunities to undertake frailty assessment to better anticipate, manage and avert demand on emergency care, for example in primary care and in care homes, with a view to early detection of deterioration.

Staffing urgent and emergency care interventions

Across service models, participants perceived multidisciplinary team working as vital. Multidisciplinary input into assessment was seen to help to expedite decision-making and improve the quality of care and strengthened the breadth and quality of connections with other services available to teams. A breadth of expertise and experiences could also help to ensure that all pathways available to patients were given due consideration and avoid defaulting to seemingly lower-risk options such as admission.

We are sharing to resolve the problem, so there is a cross boundary in terms of practices, in terms of opinion, in terms of clinical decision. I don’t know, I may be right, I may be wrong, but I feel like in frailty it has to be like interdisciplinary so we kind of help each other rather than a strong opinion like that is not a safe discharge.’

(P03 physiotherapist, frailty lead)

‘I think there’s always that risk of over-treating, but we talk about that a lot. We have a weekly MDT where we talk about patients and we talk about that a lot and reflect a lot on scenarios where perhaps there may have been some—not conflicts, but slight differences of opinions in the team about whether we should escalate to IVs or not for example.

(P13, advanced nurse practitioner)

The three overarching types of intervention varied in terms of the staff available but typically either comprised or had agreed arrangements for access to key forms of specialist input, including from geriatricians, pharmacists, and therapists. Leads included physicians (including geriatricians, less commonly ED doctors, and occasionally GPs for community-based services), nurses and therapists.

Community-based teams mainly consisted of nurses, therapists and geriatricians, with others included in or readily accessible to team members as necessary. They tended to have close relationships with community nursing teams, integrated care teams, care homes, general practices and mental health teams. Medical cover was provided variously by community geriatricians and GPs, with geriatricians usually on hand to provide advice to GPs when needed. Senior nurses, including community matrons and advanced nurse practitioners, were often central to the activity of community-based teams, leading assessments, decisions about prescriptions, and day-to-day patient care.

Teams in the ED varied in scale and disciplinary involvement. In some cases, large, multidisciplinary teams of frailty specialists were available, including geriatricians, advanced nurse practitioners, therapists, pharmacists and social workers. These teams were largely geriatrician led, although in some cases emergency care clinicians had been involved in developing, staffing and leading them. While some teams had dedicated time from geriatricians, including in some cases geriatricians who devoted most of their time to working with the ED, others relied on ‘on call’ geriatrician input, and some participants reported that geriatrician colleagues could be reticent to attract geriatricians with an appropriate interest.

They couldn’t appoint a geriatrician; it was like an hour or two hours for one of our geriatricians. So it happened that our medical director during the time is actually a geriatrician, so she dedicated two hours of her time every afternoon, so from one to three in the afternoon, that she will see patients who have been identified as needing frailty assessment.

(P03 physiotherapist, frailty lead)

Teams and units outside the ED but within the hospital also tended to be led by geriatricians, with dedicated specialised nurses leading on day-to-day care, and in some cases input from therapists, social workers and pharmacists.

The services described by participants faced various recruitment and retention challenges. Particularly notable were the issues some faced in securing consistent geriatrician input. For services where medical oversight was provided by physicians from a different specialty, such as general practice or emergency medicine, obtaining consultative input from geriatricians was not always straightforward and could make informed and patient-centred decision-making more difficult. Particularly for services that operated across sectors, clear agreements about service level and commitment were crucial to underpin high-quality care and decision-making.

So most units – and my previous unit from a different trust had the GOD phone, whatever you want to call it, geriatrician for the day, silver phone after the Silver Book, there’s lots of different name for it, where the GPs can call any geriatrician. So attached to that is advice, very quick assessments.

(Focus group 1)

We’ve got […] a phone-based service that GPs can use to get advice. But the commitment to respond to those phone calls is quite low. Part of the thing in [trust] is the consultants, there isn’t paid time to answer those calls. They don’t necessarily see it culturally as something that they absolutely should be doing. […] I think if you went back 20 years there was much more of a sense that that was absolutely a core part of a consultant’s job. So we’ve lost a little bit of that. And I think pressures on consultant time doesn’t help.

(P18 commissioner)

Also seen as vital was the input of pharmacists, but here too securing consistent access could be challenging. Pharmacists had a distinctive knowledge base that could feed into both immediate decision-making and assessment for longer-term care planning, but while some participants reported that pharmacists were a core part of the multidisciplinary team, others had patchier access to their expertise. Notably, pharmacists’ contributions at the point of ED attendance were often not consistently built into care pathways, due to capacity constraints and the time needed for a full review of medications management, with many services only able to obtain pharmacist input post admission to hospital wards or to short-term frailty units.

We don’t have pharmacists in our ED unfortunately. Within our trust two of the hospitals have got emergency departments but only one of them has got a specialist pharmacist in that ED. But unfortunately that pharmacist sees only the highly complex patients and they’re absolutely swamped because the resources are not there. […] So once they’re admitted then the service then has a greater impact because then they will be seen by OTs and pharmacists within the hospital, the ward admission structure. [… But] they may get discharged so quickly that they can possibly something put in place. It’s only the patients who are then admitted that may require a greater input of service.’

(P19 pharmacist, lead care of older people and stroke)

Community-based services and dedicated teams within the hospital but outside the ED reported more consistent involvement of pharmacists in assessment and decision-making, and continuing connections across sectors following a return to the community were again seen as important, not least to ensure that any changes to medications regimens made by pharmacists to reduce polypharmacy and risk of admission were not reversed by GPs following discharge.

We’re also now starting to have care home pharmacists, so pharmacists that will review patients in hospital and then follow them up in the community to make sure that the GPs are following the medication plans. We’re finding a lot of patients are getting readmitted because GPs don’t follow the plan.

(Focus group 1)

Overall, key features of approaches to staffing included the importance of dedicated teams, comprising experienced and confident staff, who were comfortable in working with each and in taking an approach to multidisciplinary decision-making that was not always common in other parts of the health-care system. Where the responsibility of teams was well demarcated and working relationships and practices well established, participants felt that they had better ownership of decisions and were able make good decisions more quickly.

I’ve got key members of the team. We’ve had very little change over in our staff which has been key. Because they all feel like they’re making an impact. (P17 consultant geriatrician)

It’s not just about their presentation or their age, it’s about their level of dependence and, you know, what’s going to happen to them if we do bring them in? Where is the best place for them to be looked after? Often more junior doctors, specifically in the emergency department, they don’t have the confidence or the expertise to make that decision. It’s far easier [to say], ‘This is really complicated, let’s just get them in. There’s too much going on and I can’t tell which is the most important thing to deal with today’. Whereas a senior clinician can often get to the bottom of it quite quickly.

(Focus group 1)

Being comfortable with a degree of managed risk-taking was seen as crucial to effective service provision, and again was underwritten by experience, multidisciplinary input, and trust in the competence and judgement of colleagues. This was true for those with overall accountability for decision-making.

There is an element of what is an acceptable risk. And I think that’s different for a lot of people, you know, and certainly even with our therapists and nursing, they will absolutely stop clinicians where somebody is saying, ‘Well they can’t possibly go home’, and actually our therapists and nurses are saying, ‘This is what they are like all the time, they manage at home like this, we can let this person go home’. And actually I think that’s having knowledge, you know, because we often see a lot of patients again and again, we know them relatively well, but there is something about that sort of level of risk. And also knowing the backup that you have to support that level of risk going out into community.

(P10 consultant geriatrician)

One concomitant risk of building teams with strong experience and clearly bounded remits, however, was a depletion of skills and confidence among colleagues in the ED and elsewhere. As noted above, effective service delivery depended in part on accurate triage to identify the patients who could benefit most from the input of dedicate teams, but the willingness and ability of ED and ambulance staff to assess and route patients was mixed. Equally, there was a sense among some participants that enhancing the skills and profile of frailty-focused teams could result in a deskilling elsewhere or promote a sense that care of older people was the responsibility of those teams alone, rather than a core part of emergency health-care delivery. Conversely, from the perspective of those outside specialised teams, there was sometimes a sense that the mainstream service was being deprived of capacity that it badly needed.

Using the therapists, the OT and physio in a frailty practitioner role, I think some of the other therapists were a little bit, sort of, territorial and didn’t really know what the frailty OTs were doing and then I think the OT has now found it a little bit difficult because she was a bit like, ‘Well should I be getting the equipment out of the cupboard or should I be filling out the CGA?’, you know, there was a little bit of, you know, because obviously these are kind of new roles so there was a little bit of finding feet.

(P05 consultant geriatrician, frailty lead)

The emergency department side of it, it can sometimes because they have that early intervention team, that maybe the medics can be a little bit hands off you know, in sort of identifying people. ‘Oh yeah, they’re for early intervention team’, and then they’re not getting the other aspects of what the patients require so that’s taken a bit of time to get that.

(P12)

In places where hospitals were setting these units up, they would then say, ‘Well we have to provide this service, so we’re now not going to participate in the acute medicine rota.’ And then the acute medicine rota falls over because all of a sudden if you’re a small hospital and you’ve only got, I don’t know, 12 consultants and all of a sudden three of them are not playing, it’s a quarter of your consultants gone. […] The next thing you know you’ve got a massive artificial workforce shortage. […] So actually it heavily dichotomises aspects of the patient’s care depending on how you set your unit up, you know, if you set up a unit, you can only be there for two days and everything’s brilliant and then you go to another ward that can’t deliver anything because someone’s hogged all the resource, actually it makes the team resentful and it’s not good for the patient.

(P11 senior clinical fellow and acute physician)

Given demographic changes, the limited capacity and opening hours of frailty-oriented services, and the need for risk stratification and involvement in the care of patients with less complex needs, broad-based knowledge of the needs of older people and how best to care for them was seen as vital by participants, as was continued input from geriatricians and other specialists in the management of frailty into mainstream care provided in the ED and elsewhere in the hospital.

Influences on implementation and consolidation of services

More broadly, the dispositions of staff at various levels towards the value that frailty-oriented services could add to a health economy’s urgent and emergency system services appeared crucial to the extent to which such services were initiated and sustained. In most cases, it was needs and opportunities identified by frontline clinicians that had led to services being set up, often on a pilot basis and with ‘soft’, non-recurrent, funding. While some had become increasingly mainstreamed, others remained precarious; often, participants felt, they were the first services to be curtailed when there were staffing shortages or surges in demand in the short term, while in the long-term arrangements for their funding were uncertain.

It has been challenging. I think we were able to fund the service initially for the early supported discharge aspect. The funding for our nursing staff, which was part of the move to another ward and because of the budget that was already there for that ward, we were actually able to argue for two frailty nurses to develop the service and say that, you know, ‘If you want us to do more, this is what we need.’ So we were able to negotiate that, but that was quite difficult and that took a while to get through. I think we have been struggling to expand the service to seven days and that has been a challenge in trying to get funding for that despite knowing that that would be a helpful move, so that’s just within the financial environment that our trust is in.

(P10, consultant geriatrician).

In practice, frailty-oriented services were often held together by just a single individual or a small team, leaving them vulnerable to changes in personnel or changes in priorities. The need for experienced staff, comfortable with risk and confident in multidisciplinary team working, compounded the fragility of such teams, especially when such highly skilled staff were in strong demand in other parts of the system.

I’ve had comments from a lot of my colleagues who are I suppose a bit surprised that I’m just doing it all myself and it seems that I don’t have a huge team to sort of help me get this off the ground or help this be sustainable, so there is a sort of…I do have a worry that it’s not going to be a sustainable thing forever, but I hope so.

(P21, lead pharmacist, elderly care).

The problem is that there’s still just one person in the hospital that is running the whole show basically. So, the resources are still quite thin. The aim is that every pharmacist should be trained to be able to offer the service. But it comes down to resources, it comes down to time. Those are your biggest challenges. And also retaining staff. That’s a big thing. You train staff and then they move unfortunately.

(P19, pharmacist, lead care of older people and stroke)

In contrast, participants reported that some organisations displayed more progressive, facilitative approaches to provision for older people. Personal buy-in and engagement from senior figures within a trust could help to ensure that the importance of UEC for older people was recognised, and that interdependencies with other aspects of provision were understood; recognition of the issue as a whole-system challenge could also help to ensure co-ordinated approaches to addressing it across integrated care systems (ICS), overcoming sometimes divergent priorities across sectors.

So one of the key things is having people like [colleague], so [they’ve] been very proactive in terms of supporting and promoting it in the right channels, which if we don’t have that support then it’s very difficult to really—you can make progress, but you’re never going to make the progress that you need.

(P02b, consultant geriatrician)

We’re an integrated care system, so we’re very much taking what we would describe as a system approach to frailty, so we are looking at the kind of proactive, preventative stuff that we can do in the community using population screening tools basically to identify people at high risk and then obviously once we’ve looked, making sure that we’re then supporting the community to try and keep them at home and obviously, if it is appropriate, then to come back to hospital, that’s fine, but if it’s not appropriate for them to come back to hospital then we can obviously give a bit of support to the community team about what we can do.

(P05, consultant physician and geriatrician)

In contrast, a lack of joined-up thinking in local health economies could frustrate the efforts of individual organisations to try to improve emergency and urgent care pathways.

Another issue which is kind of generic, but it does impact on our frail elderly when we’re trying to get them home, is our availability for transport. That’s a huge problem sometimes because you can’t get the ambulance because they’re too busy bringing people into hospital.

(P13, advanced nurse practitioner, rapid response and treatment service).

At the level of the organisation of care on a day-to-day basis, too, there was a need for well-developed relationships and protocols with adjacent teams in the ED and elsewhere to ensure that frailty-oriented services could work effectively. As noted above, a careful line needed to be followed between upskilling dedicated teams and ensuring that skill and responsibility remained distributed across the UEC pathway, particularly given the volume of patients and variation in the complexity of their needs. Some colleagues were sceptical about the value of frailty-specific teams, and time and care were needed in agreeing the role of frailty-oriented services and the distribution of responsibilities across teams.

I think ED would just prefer the patient to be taken away and it took us quite a long time to establish that actually no, the frailty team work with you, we’re not going to take the patient off your hands, you know, we’re working as part of your team, not as a separate team that’s going to do everything else, we’re going to be part of the team, we’re going to work together, so that provided some of the more difficult conversations at the beginning.

(P10, consultant geriatrician)

So they became ours, well we don’t have a team for that, so even things like, ‘Well your patient wants to go to the toilet,’ ‘Well I can’t provide all the nursing care for all the older people down here on my own and do my job’. People were really awful, I mean, it was really difficult.

(P06, nurse).

Lack of clarity regarding the boundaries of responsibility or how frailty-oriented services were intended to complement ‘regular’ emergency and urgent care pathways could result not just in poor working relationships but in reluctance to take responsibility for the high-quality frailty screening necessary for the identification of patients who could most benefit from such services.

Conversely, overlapping jurisdictions between professionals and teams could also pose problems. Participants highlighted a need for clear governance arrangements for frailty-oriented teams, including how multidisciplinary team arrangements for decision-making interfaced with intraprofessional norms of accountability.

My colleague, who’s a physio, she struggled in the early days because she was trying to be really, really proactive which was part of our role and some of the therapists were very junior and very risk averse, and also the processes were quite constrained. So for example people have to be signed off by a therapist before certain things can happen, but it has to be a therapist who does it. […] They weren’t really sure whether [colleague] was able to say that to them because she wasn’t in their hierarchy, so she was a separate physio.

(P16, Advanced clinical practitioner in frailty, nurse)

Training and awareness-raising regarding the needs of frail patients, the process of frailty screening and the roles of different teams were identified as helping to address these challenges, but again were time consuming and vulnerable to staff turnover.

Several participants identified the importance of extra-organisational networks in supporting their work to develop and sustain frailty services, including the Acute Frailty Network (AFN), Health Improvement Scotland and various professional associations. Such networks were seen as offering structured approaches to develop services using improvement methodologies, as well as frameworks for measuring, benchmarking, and demonstrating impact.

Being a member of the Frailty Network […] gave me a lot of resources in terms of different hospitals, I can quote them or I can use their data to show them that look this is what I’ve done, this is what happened and kind of putting it more personalised according to their staff group. So it gives me a better confidence actually, the confidence to talk about it’s not just me, I’m not creating it, it happens everywhere.

(P03, physiotherapist, frailty lead)

Staff described a range of approaches to seeking to demonstrate the impact of their services, with mixed outcomes. Evidencing impact on measures such as admission rates or length of stay could help to secure longer-term funding for services; some also sought to collect data on patient-oriented outcomes. In cases where participants felt that their services were comparatively well established, some form of data-based case had been crucial, even if only derived from audits or before-and-after comparisons.

That just sold it to the chief executive and everybody when we presented this, that it just was sold and they didn’t even need a proper business case, they said ‘just go and recruit as many people as you need to get this eight til eight working.’ So even that recruitment has been an issue, but definitely the support was there and they could see that you would not need the beds that they predicted that they would need if we did this.

(P02a AMD, service improvement)

I think now we’ve proven our worth. We’re established because we’re able to demonstrate reduced length of stay that has then been sited at unscheduled care meetings. This is what [name of service] are doing. So we’re kind of proven now.

(P17, consultant in medicine for the elderly).

Complementing generalisable evidence about the likely value of frailty-oriented services with data demonstrating local impact could, in some cases, provide a sufficient foundation for a sustained service, though ongoing challenges of staffing and competing priorities meant that relatively few participants felt confident that their services were truly embedded and normalised.

Intervention-related implementation factors

Context-related implementation influences

Approaches to implementation

Data

We conducted a data analysis of linked routine health-care data for the entire YH region of the UK (population 5.5 million). Emergency care in the region over the period was provided by one ambulance service, the Yorkshire Ambulance Service (YAS), with a single emergency phone number (999), an NHS telephone triage service (NHS 111), 18 type 1 EDs (24-hour consultant-led services with resuscitation facilities) and 13 acute hospital trusts. We extracted data from the CUREd database, which collates routine NHS data from NHS 111, the YAS computer-aided dispatch data and the ED and inpatient administration systems. The database makes it possible to track each patient from their initial contact, through transit to and from the ED, and during and after their hospital stay. The full dataset runs from 1 April 2011 until 31 March 2017, but we ran the analyses from 1 April 2013 so that the hospital frailty risk score (HFRS) was always constructed consistently using a full two years’ worth of historic data for all patients admitted to hospital. 77

The full CUREd dataset contains data on over 5.5 million individuals who accessed urgent care in the region. Of these, we identified 359,945 people aged 75 years or older who had an ED attendance or emergency hospital admission between 1 April 2013 and 31 March 2017. Many patients attended the ED or hospital on multiple occasions (e.g. 1% of patients attended more than 12 times over the full period), generating 1,035,045 separate observations.

Outcome variables

We analysed three outcomes for patients attending ED:

• Whether they were seen and discharged from the ED within four hours of arrival. This was calculated by comparing time of arrival in ED and departure time from ED.

• Whether they were transferred from the ED and admitted to hospital.

• Whether they reattended the ED within 30 days of discharge either from the ED or hospital, this being readily ascertained from the linked data. This analysis took account of whether the patient died.

For patients admitted to hospital, we analysed three sets of outcomes:

• Whether they had a short, long or excessive length of stay. These were defined according to Getting it Right First Time (GIRFT: https://gettingitrightfirsttime.co.uk) standards for geriatric medicine which define short length of stay as 2 or fewer days, long length of stay as more than 7 days and excessive length of stay as more than 21 days.

• Whether they died while in hospital. In-hospital death is indicated by the discharge status of the patient.

• Whether, conditional on the patient surviving the hospital admission, they were readmitted to hospital as an emergency within 30 days of discharge, and whether the patient had more than three emergency admissions within 90 and 365 days, these also being GIRFT standards.

For all patients, we analysed the full pathway costs of their use of emergency care service, calculated by applying national tariffs to each element of care delivery from each patient’s first contract with emergency services through to their discharge either from the ED or hospital (but not including subsequent reattendances or admissions). We also analysed hospital costs for those admitted to hospital. The methods by which costs were calculated are detailed in Appendix 4.

Patient characteristics

All analyses controlled for the patients’ age, categorised into five-year age bands, sex and the socioeconomic conditions of where they lived using the deciles of index of multiple deprivation (IMD) with IMD = 1 indicating the worst-off communities,78 number of emergency admissions in the past year, whether patients were care home residents, as recorded in the CUREd dataset, and for the estimated travel time by road between the patient´s residence and the hospital. 79

For those admitted to hospital, we controlled for whether the patient was admitted to hospital through the ED, and for frailty risk using the HFRS. 77 The Charlson comorbidity index80 uses age and International Classification of Diseases 10th revision indicators of comorbidity to estimate mortality risk and takes values from 0 to 17 but was categorised for the analysis into four groups (0, 1, 2 and 3+). 81 The analyses also accounted for the number of emergency admissions in the past year, counts of the number of operation codes, and whether or not they had ambulatory care sensitive conditions. 82

In the analyses of hospital length of stay, we accounted for whether the patient died in hospital. In the analyses of in-hospital death and 30-day readmission, we accounted for the patient’s length of stay during the index admission. The analysis of ED reattendance accounted for whether the patient died in ED or in hospital while the analysis of 30-day readmission is conditioned on the patient surviving the ED or hospital admission.

Pathway characteristics

All analyses included variables capturing the patient’s UEC journey prior to ED attendance or hospital admission; namely, the number and length of the individual’s emergency (NHS 111 and 999) calls, time of the ambulance on scene (arrival to departure), time taken between calling the ambulance and arrival at the ED, and the urgency with which the ambulance was dispatched, assigned by the NHS Pathways triage system based on answers from the caller to scripted questions asked by the call-handler. 83

Site and time variables

In the analyses of ED outcomes, ED size was measured using the number of attendances (in 1000s) during the year that the patient attended. Differences in staffing ratios across EDs were captured by the number of attendances divided by the number of senior doctors. We accounted for out-of-hours ED attendances occurring on weekends, public holidays and weekdays from 6.30 p.m. to 8 a.m. 84

Although suppressed in the tables, we also included variables accounting for the day of ED attendance or hospital admission and whether this was on a public holiday, and month and year variables to capture seasonal effects and annual trends.

Descriptive statistics

The dataset comprise a total of 1,035,045 observations, of which 867,902 were ED attendances and 167,143 were emergency admissions direct to hospital. Table 10 reports summary statistics across all those observations with non-missing values for each of the outcome and explanatory variables.

Attendances lasting more than four hours totalled 256,215 (29.5%) of the 867,902 ED attendances. Following receipt of care, 363,526 (41.9%) were discharged from the ED (of whom 3901 (0.5%) died) while the remaining 504,376 (58.1%) were admitted to hospital for further treatment; 178,553 (20.6%) reattended the ED within 30 days of discharge.

Of the 671,519 patients admitted to hospital, 504,376 (75.1%) had been transferred from the ED while 167,143 (24.9%) were admitted directly; 199,506 (29.7%) had a short stay, 233,600 (34.8%) stayed more than 7 days and 73,665 (11%) stayed more than 21 days; 51,323 (7.6%) of patients died in hospital, and 129,971 (19.4%) had an emergency readmission within 30 days of being discharged, with 21,579 (3.2%) having three subsequent emergency admissions within 90 days and 90,639 (13.5%) within 365 days.

The average cost of hospital care for those admitted was £2760 while the average cost of the entire UEC pathway amounted to £2007.

In terms of patient characteristics, the mean age was 83.5 years, 42% were male, 17% lived in the most deprived areas (IMD 1) with 7% in the least deprived areas (IMD 10); 16.5% of attendances were by care home residents and the average travel time was 14 minutes from the patient’s place of residence to the ED or hospital.

For those admitted to hospital, 39% were classified to the low frailty risk category, 39.1% to the intermediate risk category and 21.8% to the high-risk category. 20.9% were classified to the 0 Charlson comorbidity category and 31.4% to the 3+ category. On average, patients had 1.28 operations and 19.4% were diagnosed with an ambulatory care sensitive condition.

Figure 4 summarises the possible routes to the ED and hospital, with summary statistics about the pathway reported in the bottom part of Table 13. Of the total, 94,178 (9.1%) first phoned NHS 111, 6.7% of whom more than one call that same day and the average call lasting 15 minutes, and they were either advised to go directly to the ED (10,024, 1%) or were taken by ambulance (84,154, 8.1%). A further 451,030 (43.6%) made a 999 call to the ambulance service, 4.4% calling more than once that same day, and were transferred to the ED by ambulance. Of all the 535,184 conveyances by ambulance, 50.1% were designated by the call handlers as less urgent (category 4), 18.8% as urgent (category 3), 30.2% as an emergency (category 2) and 0.9% as in a life-threatening condition (category 1). 83 The average time between the call for an ambulance and arrival at the ED was 85 minutes, with the average time on scene being 40 minutes. The remaining 489,837 (47.3%) went directly to the ED or hospital, without prior NHS 111 or 999 phone contact and no ambulance journey.

FIGURE 4.

The emergency care pathway.

Some 54.2% attended the ED out of hours, 44.5% had not attended an ED in the preceding 12 months, but 18.73% had attended three times or more. Of those admitted to hospital, 47.5% had not been admitted previously in the preceding year while 15.1% had been admitted three or more times.

ED pathway outcomes

Figures 5–7 and Table 14 present the results from the analyses of the three ED outcomes. Patients in older age categories were more likely to experience poor outcomes. For example, compared with those aged 75–79 years, those aged ≥ 95 years were more likely to stay more than 4 hours in the ED (OR 1.146), to be admitted to hospital (OR 1.345) and to reattend the ED within 30 days (OR 1.091). Men were more likely to be admitted to hospital (OR 1.139) and had a higher likelihood of ED reattendance (OR 1.123). Compared with those living in the most deprived areas (IMD 1), those in better-off areas (IMD 10) were less likely to spend more than four hours in the ED (OR 0.921), to be admitted to hospital (OR 0.882) or to reattend (OR 0.862).

FIGURE 5.

ED pathway outcomes: ED duration (more than four hours).

FIGURE 6.

ED pathway outcomes: 30-day reattendance.

FIGURE 7.

ED pathway outcomes: hospital admission from ED.

Care home residents were more likely to spend more than four hours in the ED (OR 1.095) and less likely to be admitted to hospital (OR 0.795). People living closer to hospital were more likely to be admitted to hospital (OR. 1.006) and were less likely to reattend (OR 0.993).

Patients were more likely to spend more than four hours in the ED or to reattend if they made more than one NHS 111 or 999 call on the day of ED attendance. The call handler designation of urgency for those conveyed by ambulance was associated with all three outcomes. Compared with those who made their own way to the ED, the probability of waiting more than four hours was higher for those designated as less urgent (OR 1.238), urgent (OR 1.359), emergency (OR 1.35) and with a life-threatening condition (OR 1.448). There is also clear gradient across urgency categories in the likelihood of hospital admission, increasing from OR 2.155 for those designated less urgent to OR 3.603 for those with life-threatening conditions. This pattern is not evident and the effects are smaller or insignificant when considering reattendance. The longer the ambulance was on the scene, the higher the likelihood that the patient stayed more than four hours in the ED (OR 1.003) or was admitted to hospital (OR 1.009).

Patients who attended the ED out of hours spent longer in the ED (OR 1.419) were more likely to be admitted to hospital (OR 1.261) and were more likely to reattend (OR 1.071). Previous attendances were significantly associated with all outcomes, particularly for frequent attenders. Compared with those who had not previously attended the ED in the past year, those who had attended three or more times were more likely to spend more than four hours in ED (OR 1.087), to be admitted to hospital (OR 1.146) and to reattend the ED within 30 days (OR 3.133). The larger the ED (expressed as the number of attendances), the higher the likelihood of waiting more than four hours and of reattendance but the lower the likelihood of hospital admission. The effect sizes associated with number of attendances per senior doctor are very close to OR 1 for all three outcomes.

Ambulance pathway outcomes

Table 15 and Figures 8–10 report the analyses of the ED outcomes for those who were or were not conveyed to the ED by ambulance, revealing clear differences between these two groups in what factors were associated with their outcomes.

FIGURE 8.

Ambulance pathway outcomes: 30-day reattendance.

FIGURE 9.

Ambulance pathway outcomes: hospital admission from ED.

FIGURE 10.

Ambulance pathway outcomes: ED duration (more than four hours).

In terms of the likelihood of waiting more than four hours (Table 15 columns 1 and 2 and Figure 5), for those conveyed by ambulance, few of the patient characteristics are particularly influential, suggesting that people who arrived by ambulance were afforded priority based primarily upon their clinical urgency rather than any other characteristic. For this group, the most influential factor was whether they attended out of hours (OR 1.295).

Out-of-hours attendance was also the most influential factor (OR 1.622) in explaining the probability of waiting more than four hours for those who did not arrive at the ED by ambulance, but there is also an age gradient, older people facing longer ED waits (e.g. OR 1.374 for those ≥ 95 years) as did those from care homes (OR 1.184).

In terms of admission to hospital (Table 15 columns 3 and 4 and Figure 7), for those arriving by ambulance the most important factors were the call-handlers designation of urgency, those considered to have urgent (OR 1.183), emergency (OR 1.467) or life-threatening conditions (OR 1.732) being more likely to be admitted than those designated less-urgent. Older people were also more likely to be admitted (e.g. OR 1.104 for those ≥ 95 years).

Age was also an important predictor of admission for those who did not arrive by ambulance (e.g. OR 1.59 for those ≥ 95 years) and the likelihood of admission increased the more times they had previously attended the ED in the past year (e.g. OR 1.371 if three times) and if they attended out of hours (OR 1.531).

The factors associated with 30-day reattendance were little different for those who did and did not arrive at the ED by ambulance, previous attendances being the dominant factor.

Hospital pathway outcomes

Figures 11–13 and Table 16 report the analyses of the GIRFT measures of hospital length of stay. This reveals that frailty risk is more important at explaining length of stay than any of the other patient or pathway characteristics considered. Those with intermediate and high frailty risk were significantly less likely to have a short stay (OR 0.476 and OR 0.293, respectively) and significantly more likely to have a long stay (OR 2.909 and OR 5.872, respectively) and excessive stay (OR 4.25 and OR 11.78, respectively). When considering length of stay in excess of either 7 or 21 days, none of the other variables has an OR of more than 2. Of these other variables, the Charlson index (e.g. OR 1.891 for Charlson of 3+) and presence of an ambulatory care sensitive conditions (OR 1.384) were most important in explaining short length of stay and long stay, with similar effects for excessive length of stay.

FIGURE 11.

Hospital outcomes – length of stay (> 21 days).

FIGURE 12.

Hospital outcomes – length of stay (> 7 days).

FIGURE 13.

Hospital outcomes – length of stay (< 2 days).

Column (1) of Table 17 and Figure 16 report the results pertaining to the analysis of in-hospital death. Frailty risk is associated with greater likelihood of in-hospital death (HR 1.872 for intermediate risk and 2.042 for high risk). The likelihood also increases for older people (e.g. HR 1.832 for those ≥ 95 years), in line with the Charlson measures (e.g. HR 3.239 for Charlson of 3+), for those with an ambulatory care sensitive condition (HR 1.483) and according to the call-handler’s designation of urgency (e.g. HR 1.838 for those designated as having a life-threatening condition).

The remaining columns of Table 17 report the results for the three GIRFT measures of emergency readmission to hospital, which are also summarised in Figures 14 and 15. The number of past admissions in the previous year stand out as by far the most important explanatory variables. For example, those who had three or more admissions in the past year are 15% more likely to be readmitted within 30 days (average marginal effect 0.15) and to have more than three admissions within 90 days (OR 7.525) and 365 days (OR 5.156; Figure 14).

FIGURE 14.

Hospital pathway outcomes: > 3 admissions in 365 or 90 days.

FIGURE 15.

Hospital pathway outcomes: 30-day readmissions.

FIGURE 16.

Hospital pathway outcomes – in-hospital death.

Pathway and hospital costs

In Table 18 and Figures 17 and 18, the results of the cost analyses are presented. The most important factor influencing the costs of the overall UEC pathway is whether there is a hospital admission, costs being £4262 higher for those admitted to hospital compared with those who were not. Costs were £1328 and £2209 higher for patients designated by call-handlers to have an emergency or life-threatening condition and £1236 higher for those who died, whether in the ED or hospital.

FIGURE 17.

Pathway and hospital costs: total pathway cost (in 1000 pounds).

FIGURE 18.

Pathway and hospital costs: healthcare resource group tariff inpatient (in £1000s).

The analysis of hospital costs is reported in Table 18 column (2) and allows for the inclusion of a wider set of explanatory variables derived from the hospital data. Compared with those with low frailty risk, hospital costs were £269 higher for those of intermediate frailty risk and £288 higher for those of high frailty risk. Cost was £333 higher for those who died and increased by £87 for each day in hospital. Costs increased according to the Charlson categories, being £459 higher for those categorised as 3+ and by £179 for each operation conducted. Costs were also £137 higher for those who made more than one call for an ambulance on the day. Compared with those not conveyed to hospital by ambulance, costs were higher for those designated by the call-handlers as less urgent (£117), urgent (£41) and life-threatening (£202), but lower for those designated as an emergency (£92).

Key results

To our knowledge, this study is the first to analyse a set of ED and hospital outcomes and costs, contextualised as a function of the UEC pathway. This allowed us to identify a number of factors associated with adverse outcomes using readily available system-level data. Our analyses of the three ED outcomes identified some factors that are already well known, notably that those of higher age, from more deprived areas and who attend out-of-hours had a greater likelihood of a longer ED wait, hospital admission and ED reattendance. But we offer novel insights, notably in recognising the importance of call-handler designation for those conveyed to the ED by ambulance.

Turing to the hospital outcomes, we find that frailty risk is a strong predictor of length of stay and in-hospital death, according with other studies that have applied the HFRS to older people. 87,88 We found that being categorised as having intermediate or high frailty risk was a more important predictor of length of stay than any of the other rich set of control variables included in our analyses. These categories also proved important explanators of in-hospital mortality, only the Charlson comorbidity index offering greater explanatory power. The fact that both the HFRS and Charlson categories were highly significant predictors when jointly considered demonstrates that they capture different risk factors associated with in-hospital death.

In contrast, frailty risk was not a strong predictor of 30-day readmission, as with McAlister87,89 but unlike Gilbert88 and Mitsutake. 90 The inability of the HFRS to predict readmission may be because of the care that frail patients receive: if post-discharge care packages are put in place to support frail individuals when they return to their usual place of residence, their risk of readmission may be reduced.

Strengths and limitations

The strengths of this study include the large longitudinal dataset which captured information from a populous region with diverse EDs. We benefited from a large sample size, thereby generating more precise estimates than for small samples. We controlled for a wide range of patient characteristics, included measures of the urgent care pathway not considered in previous studies, and were able to perform analyses of different UEC journeys.

The study has limitations. Although not a national study, the data are from large, representative region of the UK. Our analysis was limited to routinely available data, which prioritise service and system-level over patient-centred outcomes. 91 The data only reflect the period 2011–2017 and patterns of care may have changed in the interim. The explanatory variables can be improved, for instance by accounting for ethnicity, known to be poorly coded in English hospital data, and by using better measures of socioeconomic status. Even though diagnoses are likely to be important predictors of ED outcomes, unfortunately the ED data in the CUREd database do not capture diagnoses codes accurately or in a consistent fashion across EDs, with 61.5% of attendances having no diagnostic information recorded at all. While frailty identification is now commonplace in EDs,64 no measure of frailty could be constructed from our data for those attending the ED, only for those admitted to hospital. 88 For those admitted to hospital, the positive association between the HFRS and longer stay may be upwardly biased since patients with longer stays tend to have more extensive documentation of frailty-related diagnoses,92 though use of three-character rather than four-character ICD-10 codes offers some protection against this bias.

Implications for practice

Call handler designation of urgency could be relatively easily embedded in EDs that wish to focus efforts on those at the highest risk of a long ED wait, hospital admission and ED reattendance upon arrival at ED. Systems might test the use of some simple urgent care risk stratification based on call-handler designation to identify cohorts of people who might especially benefit from alternative care pathways, for instance with the use of a GP or geriatric emergency medicine-led telephone advice line for paramedics, or routine notation of the call-handler designation during ambulance handovers to ED triage practitioners. Call-handler designations are routinely captured in electronic patient report forms but often not in handwritten ones, and may not be routinely considered by professionals working in the ED. Given the characteristics of those prioritised urgent by call-handlers, and the emerging evidence base for frailty-attuned interventions in the ED,13,93 it might be that senior decision-makers delivering initial assessment could use the designation to inform early intervention discussions or involvement of professionals with frailty expertise.

Call handler designation of urgency could be relatively easily embedded in EDs that wish to focus efforts on those at the highest risk of adverse events upon arrival at ED. Systems might test the use of some simple urgent care risk stratification based on call-handler designation to identify cohorts of people who might especially benefit from alternative care pathways, such as early senior decision-making that can facilitate early intervention of transfer to a bed. Given the characteristics of those prioritised as requiring urgent attention by call handlers, and the emerging evidence base for frailty-attuned interventions in the ED,13,93 it might be that frailty expertise in combination with emergency medicine expertise could be usefully deployed to support these individuals.

Our analyses of hospital outcomes indicated the importance of frailty risk in explaining length of stay and in-hospital mortality. While no frailty tool discriminates well enough to direct clinical care at the individual patient level, tools can be used to identify cohorts at greater risk of poor outcomes. 94 If embedded into hospital electronic health records, the HFRS could be used to target frailty interventions, thereby avoiding the burden of implementing manual scores such as the CFS and improving the standardisation of frailty assessment. The combination of the electronic Frailty Index (eFI) for primary care95 and the HFRS for secondary care risk stratification would advance frailty-friendly care, as called for in various policy documents. 65,96 The eFI and HFRS identify different at-risk cohorts,97 so their combination offers opportunity to identify the whole population at risk. Further research could examine how well the HFRS performs in other health care systems, refine the HFRS through the incorporation of real-time data, and develop frailty-attuned clinical decision support systems. 98

Strengths and limitations of the model

Key strengths of the system dynamics model are that it uses robust published evidence from WP 1.1 to create the scenarios, an integrated dataset reflecting the whole of the UEC pathway (WP3), and extensive stakeholder engagement to ensure that it is both user friendly and a realistic representation of the system.

Another strength of both versions of the model is in the ability of being able to visualise the impact of the intervention through the comprehensive set of graphs which compare the baseline and intervention results for several key hospital metrics over a simulated year of operation. The user has the option to view the model as it updates or run the model and view the output results at the end of the simulated year. The involvement of interdisciplinary research team, coupled with stakeholder meetings have provided very useful feedback on the model throughout its development, especially in how it looks and feels to the end user.

One final strength of the tool is that the underlying system dynamics model includes feedback loops which enable the user to see how various parts of the system connect, mirroring the real world. For example, in the acute frailty unit intervention scenario, the evidence suggests that with the intervention increasing a patient’s length of stay by 0.5 days, there should be more patients in hospital. However, the reduction in patients readmitted leads to an overall reduction in hospital numbers.

However, there are limitations to the model. First, we were not able to include frailty measures into the model, as these were not routinely embedded into the CUREd dataset. While it would have been possible to capture HFRS for the admitted cohorts, this would not be available to include in the whole system model, which also captures outcomes on people attending but discharged from EDs. We have only used interventions that have been reported in peer-reviewed papers; emerging care models, such as prehospital frailty services, which offer promise but the effect sizes are uncertain. Future iterations of the system-dynamics model could allow for these novel interventions to be incorporated. We were unable to model jointly delivered interventions, such as front-door frailty in combination with acute frailty units, as only separate effects have been reported, so we do not know the combined effect sizes. A final limitation is that the model does not include an estimate of the staffing resources needed to provide a service and would need to be considered separately by those seeking to implement a new or different service for older people accessing emergency care in their region.

Participant representation

WP 1.2 attempted to recruit a diverse range of patient and carer participants, using frailty data as well as ethnicity from the local systems to inform the sampling frame. Despite which, people with severe frailty and those from ethnic minorities were under-represented. This is in part related to the complexities associated with recruiting people living with severe frailty, who typically have barriers to providing informed consent (e.g. sensory or cognitive impairment). Although our team have considerable experience in recruiting people without capacity, we worked with the Clinical Research Networks, who may be less comfortable using consultee consent processes. We are aware that they are wider issues beyond appearance and language – for example, trust in the NHS, which might have had an impact on the relative under-representation of people from ethnic minorities. In contrast, the econometric analysis and the stakeholder engagement broadly reflected the distribution of ethnicity found in the community and workforce, respectively.

Reflections on the research team and wider involvement

We are aware that our research team was predominantly white (although we had PPI representation from ethnic minutes on our EMT meetings), and this may have had some unconscious influence on our ability to recruit people from ethnic minorities. We did include early career researchers in our EMT meetings – James van Oppen, undertaking a NIHR doctoral research fellowship, who contributed actively, not least in developing the PROM for older people with emergency care needs. We also included Abdullah Alshibani, a University of Leicester PhD student who was working on pre-hospital frailty, although more from a trauma perspective.