Mapping review

The mapping review of studies of interventions to reduce preventable admissions published since 2010 identified 569 publications, predominantly related to heart failure or COPD. The interventions identified by Purdy et al.³ as having the best evidence of effectiveness (or no effect) were well represented in the map. The largest group of studies originated from the USA. The included studies from the UK showed a similar distribution of studies by intervention and population to that of the map as a whole, but there was evidence of some country-specific features, such as the prominence of studies of telehealth. The findings of the mapping review were helpful in informing the sampling frame for the subsequent realist synthesis.

The mapping review and subsequent analysis of UK studies for each intervention (see Chapter 5) revealed that, in many cases, a strong international evidence base for effectiveness of an intervention existed alongside limited evidence specific to the UK context. Cardiac rehabilitation (see Chapter 4, Cardiac rehabilitation) perhaps illustrated this phenomenon most clearly. We often found limited information to help understand how particular interventions have been implemented in the UK NHS and which approaches to implementation work best in NHS contexts. Information on the role of leadership and approaches to facilitating the implementation of interventions was often lacking.

Implementation framework

Within the PARiHS framework, successful implementation is represented as a function of the nature and type of evidence (to be examined from the mapping review), the qualities of the context in which the evidence is being introduced and the way the process is facilitated¹⁷ (to be extracted from included UK studies, both quantitative and qualitative). We found that interventions with strong evidence of effectiveness overall had not necessarily demonstrated effectiveness in UK settings; that the largest proportion of the evidence came from the USA, where the context for delivery of health care is very different from that of the UK; and that facilitation of the implementation of interventions was often not reported or inadequately reported in UK studies, which generally focused mainly on effectiveness or qualitative evidence of the patient and HCP experience of service delivery.

Descriptive framework

The TIDieR-Lite framework provided a useful descriptive framework for recording key elements of the interventions and their delivery. Many of the included interventions were highly diverse in the way they were delivered, the main exceptions being cardiac and pulmonary rehabilitation. There was also considerable overlap between interventions in terms of their key components. The role of specialist nurses in providing continuity of care and links between primary and secondary care was highlighted in multiple studies.

Programme theories

We identified and tested five programme theories (using the sequence CMO) to explain how the interventions might work (see Box 1). The programme theories, expressed as scenarios, were refined and endorsed by our PPI group. We found evidence to support PT1, which suggests that hospital admissions could be reduced by optimal self-management. Considering PT2, we did not find substantive evidence to suggest that patients may seek hospital admission primarily on the basis of relative advantage. It seems that concerns associated with anxiety and risk may constitute a more important driver, hospitals being seen as safe places that can offer security and reassurance. However, the presence of perceived, implicit or indirect pressure cannot be ruled out. PT3 relates to clinicians’ confidence in their own diagnoses and ability to refer appropriately to services that might avoid admission. In the context of cardiovascular and respiratory disease, this is relevant to patients with symptoms, such as breathlessness, that could result from various underlying causes.

Direct evidence for PT4 (admissions resulting from patient delay in seeking treatment) was limited in our sample of studies. One study¹²⁸ noted that a lack of timely and accurate diagnosis contributed to exacerbations of heart failure. Finally, PT5 (influence of the broader health system context) addressed the limitations on rational decision-making around hospital admissions. This was reflected in our studies. For example, heart failure care delivered across multiple services, confusion about eligibility for specialist care and relational/managerial discontinuity of care increased the likelihood of suboptimal management and unplanned admissions.¹²⁸

Overall, we believe that the programme theories considered in this realist synthesis are valuable for understanding why unplanned and avoidable admissions occur and the facilitators of and barriers to reducing them.

Mid-range theories

We found numerous examples (both descriptive and empirical studies) of mid-range theories relevant to the interventions under review. The largest group focused on the patient, for example factors influencing adherence to recommended interventions, but theories related to HCPs’ behaviour and the overall health system were also located. Some theories were cited in relation to several interventions (e.g. Bandura’s self-efficacy theory). However, it is unclear whether the prevalence of efficacy/coping theories reflects their greater utility in this context or simply their higher profile. Nevertheless, self-efficacy was revealed as a key component to PT1, not only determining the level of comfort that a patient had with coping with their own condition but also, more importantly, how well equipped the patient feels they are to be able to manage their own exacerbations without resorting inappropriately to primary care health provision or to admission to an acute hospital.

Overarching theories

The overarching theories discussed in Chapter 5 may be considered as more exploratory than the programme and mid-range theories. In general, these theories may help to understand the underlying mechanisms at the level of the patient and HCP in the presence and absence of interventions designed to avoid admissions.

The double evidence mismatch

In theory, seeking UK-focused implementation evidence to accompany interventions that had previously been demonstrated as effective in preventing inappropriate hospital admissions for cardiorespiratory conditions appears both a coherent and an easily manageable literature review task. In practice, however, we encountered a logic problem that we have labelled the ‘double evidence mismatch’. First, the 2010 review by Purdy¹ demonstrated the average effectiveness of the candidate interventions based on a comprehensive sample of international studies, within which UK studies constituted a small and largely insignificant part. The quest for implementation evidence might therefore largely focus on UK studies that had a small effect, negative effect or a statistically insignificant effect. Second, the fact that our review team focused on studies published since 2010 means that interventions being implemented during this period may not resemble the interventions being trialled during evaluation of the original intervention. Indeed, this situation was compounded by the profound shortage of implementation evidence for UK initiatives between 2010 and 2018. Qualitative evidence constituted a large part of our recent evidence base and may not necessarily be intervention focused, nor linked to relevant trials. This type of evidence ‘mismatch’ is considered rare; we have previously encountered such a mismatch only when trying to map international RCTs from a Cochrane review against a qualitative research evidence base restricted to the UK and similar health systems. We therefore make the methodological recommendation that the scope of intervention and implementation evidence, or indeed quantitative and qualitative evidence, seeks to be as coterminous as possible so as not to artificially constrain the evidence base. However, we also acknowledge that, by its nature, implementation evidence may need to be more current than the original effectiveness studies.

Value of the PARiHS framework

We were unable to exploit the full value of the PARiHS framework, not through any limitations of the framework itself but because of constraints in the reporting of the individual interventions. Limitations encountered tended to fall into three categories:

1. Published reports did not articulate implementation issues in terms of evidence, context and facilitation or did not cover the range of aspects of these dimensions included in PARiHS (e.g. ‘evidence’ is not limited to research evidence).
2. Published reports covered one or more of these issues but reports were not structured in a way in which such data were straightforward to extract.
3. Published reports focused on implementation at a level that excluded some of the important PARiHS concepts. For example, only a study specifically on organisational aspects included data on organisational culture, whereas leadership issues were almost entirely absent.

This confirms observations previously made in relation to frameworks (i.e. that frameworks derived for use in primary research studies may not represent a good match for the level of granularity sustained by their use in synthesis activities, with syntheses typically covering issues at a broader level of detail). The version of PARiHS that we chose was not the most recent but we believe that limitations in the reporting of implementation in the included studies mean that little would have been gained by using a more refined version of the framework.

We had reviewed a range of candidate frameworks before selecting the PARiHS framework as our eventual choice. Analysis of included literature during the subsequent realist process revealed use of alternative frameworks,⁹⁸ specifically when exploring implementation issues, notably May et al.s’²⁹⁴ normalisation process model²⁹⁵ and a framework of barriers to and facilitators of quality improvement projects from a systematic review led by Kaplan et al.²⁹⁶ It is presently unclear whether or not these frameworks would share similar limitations with regard to the granularity of synthesised data or, indeed, if selection of these alternative evaluation ‘lenses’ would have yielded more insightful observations.

Value of the TIDieR-Lite template

In contrast, the TIDieR-Lite framework proved particularly useful both as a means of exploring variation within intervention groups and when seeking to delineate between interventions. It became apparent that several interventions labelled in their own right were also included within other interventions; for example, patient education is a key component in self-management. Similarly, self-management is a key function of much telehealthcare. Specialist clinics may hold a primary function in providing patient education and supporting self-management. Use of the framework therefore revealed the difficulties in attributing an effect to particular interventions and in isolating which components are most essential in delivering an effect. The limitations of the classification of interventions further confirmed the value of moving from an intervention-based approach towards the realist synthesis based on mechanisms. Phenomena such as clinical inertia and the default position, the effects, intended and unintended, of candidacy and the pivotal role of continuity and personalisation of care transcend the seven intervention types and therefore offer more transferable messages.