Summary of findings
Key findings
Comparison of outcomes at 6 months between patients attended in areas where cross-service case management approaches were available (n = 550) and control areas using within-service management (n = 633) did not show any consistent differences in mortality, emergency admissions to hospital, ED attendances or further calls to the emergency ambulance service. Within this follow-up period, mortality was higher than expected (10.5% in intervention sites; 14.1% in control sites); approximately two-thirds of patients experienced an emergency admission (68% in intervention sites; 67% in control sites); approximately three-quarters of patients attended ED (77% in intervention sites; 74% in control sites); and most patients called the emergency ambulance service at least once (88% in intervention sites; 86% in control sites), with an average of 17 calls in intervention sites and 15 calls in control sites. Overall, a high proportion (~ 95%) of patients in both arms were found to have experienced at least one of these outcomes.
Nearly all patients in both arms ceased to meet the criteria for inclusion on a ‘frequent callers’ list at some point in the 6 months following their first classification (98% in intervention sites; 99% in control sites), although about a third of these subsequently met the criteria once more during this 6-month period. Patients had an average of nine (in intervention sites) and eight (in control sites) emergency ambulance attendances, but fewer than half of these attendances resulted in conveyance to hospital (on average, four in both study arms).
Implementation costs of case management varied between different intervention sites, reflecting differences in staff mixes and in intervention management. Differences in costs per patient of healthcare resource utilisation linked to the primary (composite) clinical outcome were not statistically significant between study arms in both unadjusted and adjusted analyses; the latter also showed that costs per patient associated with secondary clinical outcomes were also not statistically significantly different.
Consistent with findings on implementation costs, models of case management were highly varied in terms of provision, resourcing and leadership, and most were subject to change over time. Service providers recognised the range of possible drivers for frequent calling, with some categories of need much more amenable to intervention than others. Many service users have deep-seated, complex needs, and contact emergency services because of a lack of more appropriate support.
Logic model
Following completion of all analyses, we revisited the logic model developed through a stakeholder event at the outset of the study; Figure 3 shows this revised logic model.
FIGURE 3 Revised STRETCHED logic model. a, Outcomes included in STRETCHED study protocol. (Figure not reproduced here; see the original article.)
Our epidemiological data showed high mortality (12.6% overall) and healthcare needs (68.2% overall with an emergency hospital admission) within 6-month follow-up. These outcomes do not appear to align with the predicted mechanisms of change in this patient group – in particular, the notion of case management achieving a change in patient help-seeking behaviour. This apparent responsibility on patients to reduce emergency ambulance service use thus seems to be an unreasonable expectation.
Natural experiment
Gender, ethnicity and proportion of patients with health service contacts in the 6 months before patients became eligible for inclusion were similar between arms. The patient cohort tended to be elderly (median age 65, upper quartile 81) and living in more socioeconomically deprived areas. However, patients in case management areas tended to be younger than in standard care areas (median age 60 vs. 69), and more likely to live in the areas classed in the most deprived quintile (47% vs. 31%).
All eligible patients met the 5+ calls within 1 month eligibility criterion; some patients also met the 12+ calls criterion, but this was always in conjunction with (or subsequent to) the former.
Data on referral for case management and harms were generally unavailable. We were only able to ascertain the proportion of patients referred for or receiving case management (40%) in one ambulance service, with partial data available in a second. Data on police arrests and convictions were generally not held by ambulance services or were only collected if a member of ambulance service staff was directly involved. Due to low numbers of events, the available data were grouped with other recorded adverse events, including requests for serious case review. The routine data set available to the study did not provide complete information on the use of primary care, social care or other agency provided care.
Health economics
Due to the complexity and variability of intervention delivery and the fact that action plans are tailored to individual patients’ needs, it was difficult to describe fully the intervention components for implementation costing. Furthermore, the pause of the study during the COVID-19 pandemic meant that, by the time data collection could start for the implementation costs, key members of staff had moved on and were unavailable and intervention delivery models had evolved from the models during the study period. As such, not all information provided referred to 2018–9 case management delivery and, where no information was available for 2018–9, information from 2020 onwards needed to be used to fill the gaps. Moreover, increased pressures on ambulance services post pandemic meant that the response from frequent caller teams to our questionnaire was low with information often incomplete which will cause uncertainty in the implementation cost, despite all efforts having been made to retrieve the required information. We encountered similar problems in obtaining case management set-up costs incurred by police and the wider care system.
Similarly, the healthcare resource use and related cost for this patient group are likely to have been underestimated – for instance, while most frequent callers are assumed to have had numerous contacts with (at least) a GP and social worker, once patients were referred to these other services, the composition and duration of contact with these services were reported to range massively depending on the support each patient requires with no data available to quantify this. It was therefore not possible to cost case management beyond referral to relevant services and the opportunity cost of managing these patients beyond referral from the frequent caller team could not be taken into account.
The cost–benefit approach originally planned was not pursued, as preliminary analyses showed no overall downstream effect of the intervention in aggregated form presented as net benefit. It was therefore decided to use a disaggregated cost–consequences approach to allow a more detailed overview of implementation costs and resource consequences, which was considered the best possible analysis approach for the case management intervention.
Qualitative work
Service providers
Interviews revealed a highly variable range of models of intervention for frequent callers. Although some key components of case management were in place in each site – such as the MDT – the interventions delivered varied. In some places primary care providers took a leading role in terms of ‘ownership’ of support for frequent callers, while in others the lead was taken by mental health providers. In all areas, case management interventions were area-based – in response to funding and local health service management organisations – and did not map onto the entire footprint of the ambulance service. There was a common theme of change over time, with short-term funding allocations and changes in staffing.
All three groups of respondents spoke of the complexities of needs among frequent callers, with awareness of medical presentations often manifesting as a result of long-term needs. There was acknowledgement from all three groups that emergency care was a default for those who – for whatever reason – had not received appropriate healthcare support through other routes.
Service users
Across the two service user groups of respondents, there were common issues in their perspectives. All said they called for emergency care as a last resort, because they had no other routes to access help. Their perception of their situation was that it was precarious and their health posed a danger to their life, or to the life of the person they cared for. Many of the service users felt demoralised that repeated contacts with health professionals had not reduced or cured their symptoms. They seemed to use the emergency services as a gateway to accessing better care and resolving their problems.
Except for the carer and family group, all respondents were dissatisfied with the way they were treated and that the emergency issue remained unresolved. Some of the emotions expressed included feeling ignored, judged, brutalised, rejected, dehumanised and threatened by the emergency services and clinicians who dealt with them. Instead of feeling flagged for tailored and co-ordinated support by health professionals, they felt labelled as troublemakers and nuisance callers. Even though some individuals received support – from clinical specialists, mental health teams, a third-sector organisation or a GP – respondents felt isolated and powerless. This contrasts with the perspective of service providers, who generally seemed to regard only a minority of frequent callers as ‘nuisance’ and were sympathetic to the unresolved needs of the majority.
Both groups of service users expressed anger and dissatisfaction. They appeared to be victims of a broken system and had run out of options for help. Communication between them and the healthcare system appeared to have broken down and they felt unsupported and unheard. Those who, according to service records, had been offered case management seemed not to be aware of this, and none spoke of any improvement in their lives as a result. If they had any experience to describe in the way of interventions for frequent callers, it was to do with caller management initiatives limiting contact with providers – as for many of the service provider respondents, these two approaches seemed to blur.
Our respondents in the two service user groups prioritised receiving judgement-free reassurance when experiencing acute health needs. Person-centred care is one component of multidisciplinary working aiming to improve quality of care and effective use of resources to support patients.72,73 Those receiving case management seemed unaware that different organisations and care professionals were considering their situations, potentially making decisions about what happened to them. Nor did recipients report any changes in their experience of making frequent emergency calls. If healthcare professionals do not engage closely enough with selected individuals to understand motivators and barriers accessing effective health care, people are unlikely to feel or act differently.
Equality, diversity and inclusion
The methods of participant recruitment for this study did not proactively target specific groups other than on the grounds of clinical or self-reported healthcare need. Due to the nature of the study, participants were often marginalised and resident in relatively deprived socioeconomic areas.
Study limitations
In this natural experiment, we took advantage of the situation across many ambulance services where practice varied between Local Health Board or CCG areas. We asked participating ambulance services to identify ‘control’ CCG areas, using the traditional model of within-service response, that were as similar as possible to the identified intervention areas. Options were limited by the rapid and disparate spread of case management/high-intensity user models of care. While this allowed comparison within services of processes, costs and outcomes of care between areas with contrasting models of care, it also meant that there were differences both in care model provided and population, between and within services.
As we were not recruiting prospectively, we estimated sample sizes based on data provided by each service before study set-up. In practice, there was much more variation than expected between sites and services. Although we almost reached our overall total predicted sample size, recruitment was dominated by one service, with more than half of intervention arm patients recruited from AS4. We included site-level secondary analyses, with results broadly consistent with our primary analysis, although STRETCHED was not powered to detect effects at site level, especially with such variation in recruitment between sites and between intervention and control arms (particularly AS2 and AS3).
In this study design, we relied on routine data outcomes for patients – which is efficient but does not provide findings related to patient well-being or experience. We proposed this study design as we were not confident of achieving a reasonable response rate to questionnaires in this study population. There may be benefits or harms that did not show in our outcomes – we based our outcomes, and expected effect size, on those used to justify the provision of case management, for instance in funding requests. These outcomes may be too blunt, or the actual impact may be much smaller than predicted, and we could have missed a smaller impact on routine outcomes, or effects on patients that would need to be self-reported. A longer follow-up period might have been captured the effects of ‘behavioural change’, but with further delays in routine data provision.
We chose to include the patient voice through qualitative interviews but were unable to achieve these at three of four sites due to internal processes with some reluctance to identify patients for inclusion (gatekeeping). Issues with recruitment of people who call ambulance services frequently is not unique to this study.74 Interviews were undertaken where a parallel project allowed these interviews to be carried out by a member of the research team, rather than relying on research paramedics at sites to undertake the interviews. We conclude that our decision to include routinely collected outcomes only in the quantitative study design was justified, but it does mean that we do not have evidence about effects of cross-service case management approaches on patient reported outcomes or experience.
Some data quality issues affected key outcomes – notably for outcomes that were not part of routinely collected data sets but were defined locally and were provided by ambulance services for the study. Improved harmonisation and completeness of data sets relevant to this patient group would facilitate further cross-service evaluation and comparisons. In particular, we were only able to retrieve partial data on referral for, and provision and costs of, case management, and other non-NHS outcome data, such as police arrests and convictions.
STRETCHED was carried out across the initial part of the COVID-19 pandemic; site activity was paused during the early months, and there were significant gaps in intervention provision and delays in data collection which resulted in some mismatch between routine data and qualitative data.
Patient and public involvement and engagement
Throughout the design, conduct and dissemination of STRETCHED, we involved public contributors with experience of accessing emergency ambulance services and the complex health conditions experienced by people making frequent emergency calls. We created different forums for involvement – strategic, managerial, oversight and implementation – to enable individuals with diverse backgrounds and perspectives to collaborate in this study. We ensured opportunities were accessible and named an experienced academic to lead public involvement in the research team. We established trust and worked in a mutually supportive way in our collaboration.
Public contributors ensured service users’ issues remained at the forefront of team members’ awareness. They highlighted the importance of language, advised on recruitment and data collection and provided rich insight into interpreting and presenting the findings. They have an active role in our dissemination strategy. We have reported our findings to contribute knowledge about processes and experiences to inform the growing evidence base about public involvement in research.69
Implications for policy practice and research
We do not believe that our findings mean that case management never works, or is not worth providing, but we found that expectations of impact based on advice and previous small-scale evaluations were unrealistic. This matches existing evidence, which is somewhat mixed on whether case management for people who frequently use healthcare services is able to reduce service use or lead to other outcomes.75
When considering the logic model, stakeholders agreed that clear criteria for referrals to case management intervention were needed. This was in addition to appropriately trained case managers who had ring-fenced time to dedicate to patients and case management. On an organisational level, the stakeholders emphasised the importance of information resources on service availability and referral pathways, better information sharing protocols and infrastructure within and across organisation under effective leadership for change management. These inputs could in turn aid in flagging and identifying high-intensity user patients and better identification of patient needs. This could also help in the development of a personalised assessment and care plan for the patient which in turn would lead to fewer adverse events, contacts with services, increased patient satisfaction, increased sense of control within the patient, and in ultimately fewer emergency ambulance service calls.
At the synthesis stage, with the results of quantitative and qualitative analysis available, the research team revisited the logic model developed with stakeholders at the outset of the study. We questioned some of the framing of the predicted mechanisms of change and some of the expected impacts. In particular, we noted that there had been an expectation that the cross-service case management approach would help to achieve a change in patient help-seeking behaviour that would be the mechanism that would lead to reduced calls to the emergency ambulance service as well as a range of other outcomes both for patients, the NHS and ambulance staff. This reflects existing theory of how case management works in primary care, based on the relationship between the case manager and the patient, with the patient being empowered to self-manage their condition as key mechanisms.76 On reflection, the research team considered that this logic was too linear, oversimplified and unrealistic for this patient group, and also that too much emphasis was put on the responsibility lying with the patient to change. Instead, theory of how case management works, for whom and in which circumstances needs to include greater consideration of the complexity of patients’ lives and their particular circumstances. Moreover, our data showed that the patient group is varied and had high needs for care – around one in eight patients had died within 6 months, approximately two-thirds of patients were admitted to hospital at least once during that 6-month period, and the most commonly recorded reason for calling was that the patient had fallen.
There is a common characterisation – within services and in the media – of people who call frequently needing to increase their resilience or ability to self-manage their problems which are often described as social, emotional or ‘not coping’. Our quantitative and qualitative data show that many patients are calling with persistent clinical needs, sometimes compounded by challenging life circumstances and hardship or lack of family/community support. Although case management offers an attempt to provide holistic care to people who make high use of emergency health services,20 in practice we have found, in common with other researchers in this area, that the patient group is diverse; given its heterogeneous nature, it is likely that improved strategies for the identification of individuals most likely to benefit are needed for case management to be successful.77 Furthermore, this group includes a high proportion of patients who are very unwell, with a high level of clinical needs which may not be well matched to the service provided. In other cases, problems may be intractable and not amenable to simple, short-term solutions.14,43
Supporting and treating people with complex needs involves building trust and working flexibly with individuals – as service providers acknowledged – yet these service users felt they had no say in processes and decisions affecting them. There are intrinsic challenges when an emergency ambulance service – designed to provide an immediate and short-term curative response – becomes engaged with needs which are long term, multifaceted and often requiring care, or basic human services such as housing, and responses are often on an ‘as needed’ basis, rather than addressing underlying issues.14 This tension is present in many aspects of emergency ambulance service work, but particularly apparent among the group of callers who meet the frequent caller threshold.
Services need to be provided in an accessible and relevant way to create genuine opportunities to resolve demand.78 For example, mental health issues can present as inability to follow routine or attend structured support sessions, yet be judged as failing lifestyle behaviours. Many of the service providers talked about challenges of engagement with service users; from the perspective of service users, the issue may not be so much about engagement as to do with access and availability.
Conclusions
We did not find any reductions in deaths, emergency ambulance calls or other emergency healthcare contacts associated with case management. The patients classified as frequent callers had high and varied needs. We do not believe that our findings mean that case management never works, or is not worth providing, but we found that expectations of its impact, based on advice and previous small-scale evaluations, were unrealistic.
We have not been able to answer the question of whether case management should be commissioned and supported for the care of a subgroup of people who make high use of 999 emergency ambulance services. While we have found that outcomes were not improved across the whole cohort of patients in intervention sites, we found that the group is heterogeneous in nature. Lack of data prevented an assessment of opportunity costs to services beyond the ambulance service. We cannot rule out there being benefit for subgroups – for example, patients with mental health or social care needs – or that any benefit would be worth the investment (costs). We are therefore unable to make recommendations about whether case management should be introduced or continue to be supported, for defined groups. This question requires further study.
We recommend the following further research activity, in order of priority, be undertaken or commissioned on this very important topic:
- Prospective evaluation of care options for this patient group – that is, ‘what works for whom’, with outcomes measured over a longer follow-up period – probably 1 year. A natural experiment design could be used to undertake this evaluation, with criteria for inclusion matched to those for referral to case management teams rather than for all patients meeting the ‘frequent caller’ definition.
- Exploration of different forms of case management, tailored for different patient groups.
- Data sets important for the care of people who call 999 frequently are defined and agreed across ambulance services in order to harmonise these data sets and allow cross-service evaluation and comparisons. This might be facilitated by FreCaNN.
- To ensure in future research through early work with ambulance services that processes of care can be retrieved: in particular, detail of referrals for case management and adverse events, for example arrests; and for including patients in the research through qualitative study.
- Investigation of the feasibility of obtaining self-reported outcomes from this patient group and appropriate methods for doing this.
- Exploration of methods for including in qualitative research those patients judged by services to be vulnerable or high risk – that is, how best to navigate ‘gatekeeping’.