From programme theory to logic models for multispecialty community providers: a realist evidence synthesis

Multispecialty community provider outcomes

Despite the different approaches to care ‘integration’, the policy outcomes that policy-makers intended MCPs to produce most resemble those of the PACSs and were:

• 7-day access to services5,7

• the House of Commons Health Committee mentions the Improved Access to Psychological Therapies programme waiting time standards5 in ways that hint that they should apply to mental health services generally

• ‘measurable reduction in age standardised emergency admission rates and emergency inpatient bed-day rates; more significant reductions through the New Care Model programme covering at least 50% of population’8

• significant measurable progress in health and social care integration, urgent and emergency care (including ensuring a single point of contact), and electronic health record (EHR) sharing, in areas covered by the New Care Model programme7

• better access to care nearer to home5 (e.g. more convenient care). 8

Multispecialty community provider mechanisms

The 5YFV6 itself describes certain mechanisms that MCPs ‘will’ or ‘would’ use. ‘Expert generalists’ (i.e. GPs) will work more intensively with patients with complex needs (e.g. frail older people; chronic conditions). Nurses, therapists and other CHS professionals will be included in MCP ‘leadership’ (management). There will be a wider range of primary care services. MCPs will draw on the ‘renewable energy’ of carers, volunteers and patients.

Multispecialty community providers ‘may include a number of variants’. 9 A longer list describes mechanisms that MCPs ‘could’ use, hinting that different variants may involve different combinations of the following:

1. fuller use of digital technology

2. fuller use of ‘new skills and roles’ (i.e. new divisions of labour)

3. extended group general practices, ‘either as federations, networks, or single organisations’6

4. general practices employing consultants or making them partners

5. such consultants (and by implication GPs) ‘work[ing] alongside’ CHS staff, pharmacists, psychologists, social workers and others

6. running local community hospitals and perhaps expanding their diagnostic services

7. GP-admitting rights to acute hospitals

8. ‘in time’, GPs managing the NHS budget for their patients

9. care hubs, which perhaps may also provide OOH services.

Within MCPs, small independent general practices will continue while GPs wish it, which implies some form of networked rather than line-management relationship between these practices and the rest of the MCP.

A wide range of MCP sizes (the first wave served populations ranging from 63,000 to 330,000) and of possible governance structures is envisaged. Perhaps the most obvious are networks of independent general practices, possibly perhaps with a strong central co-ordinating body (a ‘federation’). MCPs are described as ‘extended group [GP] practices’, which might be ‘federations, networks or single organisations’ (p. 20). 4 The House of Commons Health Committee5 argued that federations allow specialised development of services and care teams while retaining the existing scale of general practices. However, MCPs might also commission specialist providers, implying a potential role for governance and co-ordination through quasi-markets. New hierarchical organisations (e.g. on the lines of NHS foundation trusts) are also foreseen. Potentially, they might also organisationally integrate general practice and CHSs, which the so-called ‘integrated’ care pilots never did. Another expected kind of single organisation is the enlarged professional partnership. The 5YFV comes close to implying that a MCP might also have the structure of a social enterprise or co-operative.

Multispecialty community provider context

Multispecialty community providers’ external relationships with the rest of the NHS will, above all, be through monitoring and a contract. The 5YFV6 expected that standardised data will enable real-time monitoring and evaluation of MCPs’ quality, outcomes, costs and benefits. NHS England is establishing a new operational research and evaluation capability to support this activity.

A ‘new voluntary contract for GPs (Multispecialty Community Provider contract)’ will be MCPs’ main financial link to NHS commissioners. Its three varieties10 are:

1. a ‘partially integrated’ contract [i.e. an additional contract supplementing the General Medical Services (GMS) contract]

2. an integrated single contract based on the GMS contract but excluding the QOF element – a whole population budget for all primary care home (PCH) and CHS services for perhaps 10–15 years

3. ‘a virtual, alliance contract’.

The contractor and, by implication, overall co-ordinating body of a MCP might be a community interest company or limited liability company (both wide categories), partnership (including GP federation) or a statutory NHS provider. 4 MCPs will receive capitated payments, but not fees for service (which general practices now do, although it is not usually the main element of their income). 9 The new, longer-term contracts could follow the outcomes-based commissioning approach already being tried elsewhere in the NHS. 11

As usual for NHS organisational innovations, MCPs will be introduced in waves. For the first wave, ‘The purpose of becoming an initial site is not simply to address local needs, but to become a successful prototype that can be adapted elsewhere, designed from the outset to be replicated’. 9

Definition by example

Policy documents and recent plans for the first wave most often characterise MCPs in structural terms (which organisations will participate and collaborate) and, to a lesser extent, in terms of certain cross-organisational care processes. However, these documents expressly leave many possible varieties and options open. 6 Therefore, another way of defining a MCP is ostensibly by considering what common characteristics the first wave of MCPs have12 (see Appendix 1).

Across the 14 first-wave MCPs, the participating organisations (mostly health-care providers) will, in 11 cases, be networks (e.g. federations) of, and individual, general practices (including a super-practice and a proprietary one). Eleven MCPs will also include a NHS hospital trust, nine will include a mental health trust and eight will include a CHS trust. Eight also include one or more CCG. Local authorities, or departments thereof, are included in nine MCPs, in particular, social services (in four). Five MCPs include umbrella organisations for local voluntary organisations and another two MCPs include ‘groups’ of the same. Three MCPs involve urgent-care services (OOH, ambulance). Other, more disparate participants include Healthwatch, one local medical committee, one hospice, commercial pharmacy, NHS England and the Local Government Association (LGA).

As mechanisms, the first-wave MCPs most frequently mention establishing, or strengthening, existing multidisciplinary teams (MDTs) (eight projects). The specific composition varies, but across the projects the team participants include GPs, advanced nurse practitioners, social workers, mental health services, voluntary sector link workers and pharmacists. Next most frequently, five MCPs mention various forms of shared care planning (one of them a GP-led complex-care management service). Another partly overlapping set of five projects plans to create a physical location (‘hub’) in which to combine services and provide a single point of access to them. Three projects mention preventative care, three information technology (IT)-based mechanisms (shared health records, digital access to health care) and three preventative care (including for children and self-care). Two mention care co-ordinators or navigators and two propose to enhance local referral networks and pathways. Various other mechanisms are mentioned by only one prospective MCP (new forms of contract, extended access to GP services, mobile clinics, recruitment of hospital consultants and, contingent on projects outside the NHS, a ‘health and care garden city’).

Working definition of ‘multispecialty community provider’

The foregoing suggests that the essence and function of a MCP is horizontal ‘integration’ among the various primary care providers (general practices, CHSs, mental health, OOH, ambulance, urgent care, etc.) and related non-health services (primarily, social services and residential care). Functional (as opposed to organisational) ‘integration’ will typically mean closer care co-ordination across still-separate provider organisations rather than organisational integration, although even the latter may occur in the future. 6 However, in the meantime, MCPs will be interorganisational networks.

We put the term ‘integration’ within quotation marks because research and policy documents often conflate three distinct concepts:

1. co-ordination – the deliberate combination, connecting and sequencing of separate but interdependent resources,13 above all, individuals’ care activities, into a single care process14

2. continuity – a term covering the cross-sectional, longitudinal, flexible, informational and relational continuities of care;15–17 the common element is the non-interruption of care co-ordination

3. integration – use of a single organisational structure to co-ordinate care. 18

Research and policy documents are especially prone to saying ‘integration’ when referring to (closer) co-ordination.

Gesundes Kinzigtal (Germany)

Gesundes Kinzigtal GmbH, two-thirds owned by a network of local doctors and one-third by a health-care management company, has a shared savings contract between with one large social health insurer [Allgemeine Ortskrankenkasse (AOK)] and one small one [Landwirtschaftliche Krankenkasse (LKK); for farmers]. This contract gives both sides incentives to make and share savings. Some 33,000 people (about half of the area’s population) subscribe to the scheme. Its models of care are based on the collaboration (still unusual in Germany) of doctors, hospitals, social care, nursing staff, therapists and pharmacies. The project offers ‘a set of community initiatives’,19 preventative, patient self-management and health promotion activities. 20 It has been described19 as an accountable care organisation (ACO). It provides individual treatment plans, post-discharge follow-up care and case management. It focuses on removing care pathway bottlenecks (e.g. waits for physiotherapy) and uses a single system-wide EHR.

Buurtzorg (the Netherlands)

Buurtzorg originated as a proprietary CHS nursing and allied health professional (AHP) service provider, but a very mission-led one. It now has 630 work teams whose work includes house cleaning for disabled people (Buurtdiensten), services for young people (Buurtzorg Jong), home-based rehabilitation (Buurtzorgpension) and hospice care (Buurtzorghuis). Buurtzorg charges a flat hourly fee for its work, with self-managed local teams deciding the skill mix ad hoc. The managerial infrastructure is very small. Work co-ordination relies heavily on an IT system based on spreadsheets devised by the teams themselves and a shared EHR. 21 Reflecting practice in the Netherlands generally, the teams do not include doctors (separately organised in small general practices, much as in the NHS).

Swedish vårdcentral (Sweden)

Swedish primary health care clinics (PHCCs) (vårdcentral: ‘polyclinic’) traditionally provided both primary medical care and home nursing care services (i.e. a similar function to NHS community nurses). Some offer OOH emergency services, but not OOH home visits by doctors. For-profit providers have about a 15% market share, as does Praktikertjänst, a medical co-operative. As in the UK, local authorities provide social services, with client copayment. 22 In mental health services, nurses are often the care co-ordinators but, in acute primary care, it is often the GP. Some PHCCs host outreach specialist services (e.g. neurology, geriatrics), therapy services and diagnostics. Multiprofessional teams often operate within each PHCC but generally rely on informal co-ordination. There is no universal EHR, and usually only partial data sharing among health-care providers (among which nursing homes or social services are not included). 23

In Norrtälje, Sweden, the vårdcentral model has been extended. An integrated financial administration (TioHundra Forvaltningen) administers combined (pooled) budgets for all health and social care. They commission a single publicly owned not-for-profit company, TioHundra AB, to provide integrated primary care, hospital and social care services for the whole population. Its PHCC provides medical, nursing and speech therapy services, including post-hospital nursing services for up to 2 weeks after discharge. A separate division within TioHundra AB provides all other community nursing and social care. 24,25

Accountable care organisations (the USA)

The US government’s Centers for Medicare and Medicaid Services (CMMS) defines ACOs as voluntary associations of hospitals, doctors, other health-care providers and professionals who together co-ordinate the care they provide for Medicare patients, in particular chronically ill patients. These ACOs aim to prevent medical errors and service duplication and to make cost savings that the ACO will share with the participating providers. 26

Varieties of ACO programmes have included a Medicare Shared Savings Program (as an alternative to fee-for-service payment), an Advance Payment ACO Model (supplementary incentive programme for selected participants in the Shared Savings Program) and a now discontinued Pioneer ACO Model for early adopters of co-ordinated care.

The NHS now uses the phrase ‘ACO’ to mean the commissioning of a single contract and lead contractor for most of the primary and secondary care health services in a CCG or other wide area. However, in the USA most providers that join an ACO also have non-Medicare (and, in that respect, non-ACO) patients. Provider membership of an ACO is voluntary and, therefore, providers require an incentive to join, usually the financial incentive of sharing the savings.

Awareness of these differences is necessary when interpreting findings about American ACOs for NHS use.

Patient Medical Home (the USA)

In US settings, the term ‘patient-centred medical home’ or ‘primary care medical home’ (PCMH) means something very close to group general medical practices with a stable list of registered patients (as opposed to episodically caring for patients) and providing holistic, co-ordinated, accessible, comprehensive care and also some non-medical services, that is, something similar to the UK model of general practice, with its system of patient lists, since the 1940s. 27 However, recent NHS guidance sees the PCH (a namesake of the US models) as serving a patient list of 30,000–50,000 people, having an integrated workforce, focusing on both population and personalised care, and with ‘alignment of clinical and financial drivers’. 4

Identifying core multispecialty community provider policy sources

The original call for proposals for this research, and the research protocol itself, focused on the 5YFV6 as the main policy source about what policy outcomes MCP are intended to produce, and the means by which policy-makers assume MCPs will do so. For this reason, we used the 5YFV as one of our focal documents in step 1.

We conducted a literature search to identify additional English policy statements on care models for people with chronic conditions. The aim of this search was to find a core set of policy documents in order to identify policy-makers’ assumptions about MCPs. This search used the Health Management Information Consortium (HMIC) database (via Ovid), which indexes policy content from the Department of Health and Social Care (DHSC) database (DHSC Data) and The King’s Fund database. HMIC was the only database we searched, as it was found to index all the authoritative policy papers and web pages. Search terms were selected by inspecting the titles and abstracts of the known relevant policy documents mentioned. The search used generic terms describing generic and specific interventions that appeared functionally equivalent to MCPs (e.g. ‘integrated primary and community care’) and particular international examples, such as Buurtzorg and the Wiesbaden Network for Geriatric Rehabilitation. These terms were combined using the Boolean operator ‘AND’ with terms for older people and people with chronic and complex conditions. Both sets of search terms were represented by free-text terms and indexing terms. The search was conducted on 25 August 2016 and date limited to after January 1991 (the foundation of the NHS quasi-market).

The 5YFV used partly different terminology from the other key policy documents identified by web searching. The 5YFV focused on developing ‘sustainable’ ways of organising care to tackle health inequalities, rather than on models of care to tackle chronic conditions. Therefore, we made a supplementary search of HMIC using search terms for ‘inequalities’, ‘health care’ and ‘sustainability’ – a more focused search than the first. Search terms were limited to the notes field of HMIC records, which is used to summarise the contents of a report as a supplement to the abstract that is often not included with policy literature. The search was conducted on 25 August 2016 and no date limit was used.

The results from both searches were exported to an EndNote X7 (Clarivate Analytics, Philadelphia, PA, USA) database. The search strategy and the number of hits for each search are presented in Appendix 2.

Only a handful of policy documents were identified that explained MCPs in much detail (see Chapter 1). The most informative were the ‘logic model descriptions’, which each of the first-wave MCP Vanguard sites prepared and which, as first-hand accounts endorsed by NHS England of what the MCP Vanguards were attempting to do, were especially relevant and important. The focal policy documents used from which to extract policy-makers’ assumptions about MCPs are cited in Chapter 4 and listed in Appendix 3.

Connecting and mapping mechanisms and outcomes

We first elicited as many of the policy-makers’ assumptions about MCPs as we could from the identified sources. In order to elicit the IPT, we framed or reformulated policy-makers’ assumptions in realist terms as context–mechanism–outcome configurations (CMOCs), or parts thereof, with the terms ‘context’, ‘mechanism’ and ‘outcome’ (defined as in Research design).

We articulated these CMOCs in ‘if–then’ statements, that is, statements of the assumed context and mechanism (‘if’) and outcome (‘then’); for example, if MDTs are established in the context that patients want to maintain their own health, then preventative health care will develop. This was a practical necessity as it was rare for the policy statements to specify a context (in the realist sense, i.e. under what conditions the proposed mechanism would or would not work) in addition to mechanism and outcome. Some if–then statements were describing essentially the same CMOC in different words (e.g. about electronically sharing patient information between organisations). We treated these statements as multiple formulations of the same assumption and, in effect, merged them. Many other statements referred (again, sometimes using different words) to essentially the same mechanism (e.g. ‘multi-disciplinary team’, ‘inter-professional team’, ‘cross-professional group’). In some cases, the mechanism of one statement was a subset, component or special case of the mechanism of another (e.g. ‘primary care’ and ‘primary care close to home’). Therefore, we grouped these under the same overarching mechanism. Similarly, many statements referred (again sometimes with different words) to essentially the same outcomes (e.g. ‘patient self-care and activation’ and ‘patient engagement in care and self-care’). We also grouped these accordingly.

This grouping of the if–then statements by mechanism and outcome identified from the policy-makers’ assumptions what the core MCP ‘components’ [mechanism, outcome or context (as the case might be)] were, and the ‘causal links’ between them. We identified 13 components of MCPs and 28 causal links between them, and numbered each linked mechanism and outcome so that the links between components could be traced back to the initial if–then statements.

The MCP components are inter-related in complex ways. Many MCP components were mechanisms for producing several outcomes. Many components were also assumed to be the product of several other components. Often, mechanisms were linked together in chains (‘concatenated’): the outcome of one mechanism was to set up another mechanism producing a further outcome. Producing the second mechanism was, thus, an intermediate outcome.

Next, we mapped what the policy documents assumed the causal links between the MCP components to be, which revealed the assumed chains of MCP components and their complex inter-relationships; in particular, the ways in which some mechanisms were assumed to produce or trigger others. Throughout, and, in the following chapters, we have maintained the same system of numbering for these causal links. For example, ‘(3:10)’ means that component 3 is assumed to be a mechanism that produces component 10. One way of showing the relationships between the mechanisms is by using graphics. In these graphics, we have represented each MCP component by a box containing its constituents and numbered to indicate its source(s). Arrows between components showed the causal links that the policy-makers assumed. Figure A (see Report Supplementary Material 1) shows the first such graphic, based on only the policy documents mentioned previously.

Patient and public involvement

In this study, PPI was through participation in stakeholder ‘think tanks’ (see Stakeholder think tanks). This method of participation was co-designed with PPI during the submission of our research proposal to the National Institute for Health Research (NIHR).

The stakeholder group included four members from the wider Peninsula Collaborations for Leadership in Applied Health Research and Care (CLAHRC) Patient Involvement Group who expressed interest during their involvement in the preparation of the research proposal (see Appendix 4).

Stakeholder think tanks

To check our understanding of the programme theory of MCPs for any missing or misinterpreted elements, we consulted a think tank of patient and NHS ‘stakeholders’. The latter included people who would be implementing MCPs. We used these meetings to:

• check our interpretation of the initial MCP programme theory

• resolve ambiguities

• add any missing components

• advise as to which MCP components were most important and should, therefore, be prioritised in the evidence review (step 2).

Senior researchers identified stakeholders at the research group meetings. Mark Pearson provided names of service users, Helen Lloyd gave a list of policy-makers and academics and Richard Byng supplied a pool of GPs and managers working in GP surgeries. The final list encompassed stakeholders across England, including senior staff from NHS England.

We held three think tank meetings in October 2016. Participants were general practice members (GP, practice manager), service users, policy-makers (including NHS England) and a minority of academics. The researchers made field notes and (with the participants’ consent) audio-recorded the meetings in order to return to key points, if necessary. After each meeting the if–then statements and map were successively modified.

We held a further meeting with our stakeholders in March 2017 in order to check our understanding of the linkages between MCP components and we will meet the stakeholders again to discuss further how to disseminate our findings.

From the included policy sources and the think tank interpretations we arrived at 242 if–then statements (see Appendix 5).

Deduplicating and consolidating the conceptual map

Given the number of if–then statements, data reduction was necessary. When we had one link A–B–C and one A–C; the first was more informative (about mediating steps) and so we removed the second as a duplicate. We also removed non-redundant, but trivial, links (e.g. ‘if there is scope for local innovation in creating MCPs, then MCPs will be created’).

Even after consulting the stakeholder think tank, many of the if–then statements still explicitly stated only one or two of the context–mechanism–outcome (CMO) trinity, which previous studies43–47 have already shown is often the case with policy sources. In developing the conceptual map, the researchers imputed the missing assumptions from our background knowledge of the English health system and clinical practice within it. In doing so, we:

• clearly differentiated the imputed assumptions from those explicitly stated in the policy sources

• selected, when alternative assumptions might be imputed, those that have the strongest evidence base and were most consistent with those explicitly stated in the policy sources, avoiding the construction of a ‘straw man’ or unfairly weak interpretation.

Adding these connections produced a second graphic, Figure B (see Report Supplementary Material 1). The graphic includes the numbered ‘ifs’ and ‘thens’ behind each component in order to illustrate some (but certainly not all) of the complexity of their inter-relationships, the direction of ‘flow’ from input to output (showing which were intermediate and which were final outcomes), but also removing redundant links as explained above. This method ensured that the fully articulated initial MCP programme theory remained as comprehensible as possible while remaining as close as possible to the original policy statements. Chapter 4, The initial programme theory: multispecialty community provider components and the causal links between them formulates the IPT taken forward into step 2. Chapter 4 describes (both verbally and with a graphic), in detail, the mechanisms, intermediate outcomes, final policy outcomes and contexts that, together, comprise the fully articulated IPT for MCPs early in 2017.

Exploration and search strategy development

The aim of the realist evidence review was to discover an evidence base against which to ‘test’ IPT (see Step 3: building a revised logic model) and reveal whether or not the IPT omitted any important MCP components or causal links between them. Owing to the size and complexity of the corpus of relevant studies, we were also aware of the necessity for a well-defined and focused search strategy. We focused the search by:

• Searching for concepts and terminology from the main components of programme theory, starting with the formation of MCPs and its subcomponents (see Chapter 4). This search covered all 13 components of the IPT. The search concentrated on the connections between the 13 components rather than on each component in isolation from its effects and contexts.

• ‘ANDing’ these with names of MCP-equivalent organisations, networks and projects. Chapter 1 defined ‘MCP equivalent’ as performing a similar function of horizontal co-ordination between primary medical care, domiciliary health care, other primary care health services, and social care. Sarah L Brand and Simon Briscoe assembled a list of MCP equivalents [including Chronic Care Models (CCMs)], drawing on the whole research team’s knowledge.

We developed a search in MEDLINE (via Ovid) using the above sets of terms. Search terms were represented by free-text terms and indexing terms. The final search was translated for use in a selection of topically appropriate databases, including MEDLINE, MEDLINE In-Process & Other Non-Indexed Citations, PsycINFO (all via Ovid), Cumulative Index to Nursing and Allied Health Literature (CINAHL; via EBSCOhost) and Applied Social Sciences Index and Abstracts (ASSIA; via ProQuest). The search was conducted on 5 December 2016 and no date limit was used. We exported the search results to EndNote X7 and deduplicated them using automatic and manual checking. The search strategies and number of hits are presented in Appendix 6.

Studies were also identified through opportunistic finds from e-mail updates from relevant journals.

Selection

Five reviewers (RS, MP, SLB, MF and AW) between them screened 1319 titles and abstracts in the EndNote database. There were two rounds of screening. For each round, we developed a screening tool (Appendices 7 and 8), each of which went through two rounds of piloting on 10 studies (20 in total) by all reviewers. Discrepancies in tool use and include/exclude decisions were discussed and resolved after each pilot to achieve consistency in its use.

Data extraction and quality appraisal

The aim of data extraction was to extract evidence about the 28 causal links in the IPT (see Chapter 4). Four reviewers (RS, MP, SLB and AW) extracted data from the included studies. Each reviewer was allocated 1–4 of the 13 components. The data extraction tool (see Appendix 9) was piloted on two studies by two reviewers (SLB and RS), followed by discussion to resolve any discrepancies or other problems. For each of the 28 causal links, we sought to:

• extract data tending to corroborate the causal link

• extract data that were evidence against the causal link

• extract evidence of new causal links or components not in the IPT

• specify context(s), that is, evidence specifying the circumstances under which one component would produce another, or fail to

• note the quantity and strength of evidence about the causal link

• note any qualifications or limitations to the findings reported in the study from which data were being extracted

• note which kind(s) of MCP equivalent(s) the study described, in which country and serving which care group(s).

For studies allocated to more than one reviewer (i.e. relevant to more than one component, which was most of the studies), the first reviewer extracted data and saved the data extraction form, the next reviewer extracted data from that study and then checked the first reviewer’s data extraction and added their own data extraction (if any) to the first reviewer’s form, and so on. In this way, 26 out of the 116 included studies were data extracted by two reviewers (22.4%).

Each included study was assessed for methodological quality using the Mixed Methods Appraisal Tool (MMAT). 48 We used the MMAT because, reflecting their complex objects of study, we expected most of the studies to use mixed or qualitative methods, with some quantitative studies. Two reviewers (SLB and RS) piloted MMAT scoring on two studies, then discussed the discrepancies with the wider team to ensure consistency. Any issues arising in quality appraisal assessment were raised and discussed in team meetings during the quality appraisal stage. The data extractor(s) for each study also assessed its MMAT quality score. The MMAT provides a standardised appraisal checklist of four items (hence, scores of 0, 1, 2, 3 or 4) for qualitative studies, and the same for RCTs, non-randomised trials and descriptive quantitative research. For mixed methods, it provides a three-item checklist. Criteria for all the checklists are detailed and well specified. To assess the quality of the included SRs, we used the Assessment of Multiple Systematic Reviews (AMSTAR) quality appraisal tool for SRs. 49 AMSTAR also provides well specified, consistently structured checklists for 11 characteristics indicative of, in this case, the quality of a SR, giving a total score of between 0 and 11 for each SR. MMAT and AMSTAR ratings were conducted by one reviewer. Ten per cent of these were then rated by a second reviewer (SLB, n = 9; RS, n = 3) and one discrepancy was resolved by a third reviewer (HL). The MMAT or AMSTAR rating and a narrative summary of any methodological quality issues with a study were also recorded on its data extraction form.

Collating and coding data

The data extraction tool (see Appendix 9) was structured according to the 28 causal links between MCP components in the IPT, in 11 groups according to which component was the mechanism, as opposed to the outcome, in that causal link. This structure was also the overall coding framework for the data. To automate data sorting and retrieval, we created an NVivo 11 (QSR International, Warrington, UK) database, with a node for each causal link and, therefore, the corresponding section of the data extraction tool. Within each node, subnodes corresponded to the lower-level links between the subcomponents of each MCP component. Data from all the data extraction forms were imported into the corresponding NVivo node(s). When no suitable node existed, we created new nodes, as necessary, during data extraction. These were where additions to and elaborations of the IPT began to emerge.

Comparing the initial programme theory with the evidence review findings

The 28 causal links between the 13 MCP components were the analytic framework for this comparison. We collated the relevant contents of the completed data thematically into that framework. For the 28 causal links in the IPT, we:

• assessed the overall evidence for the causal link

• inducted patterns and subthemes

• noted strengths of evidence and gaps in the evidence, including any absence of contextual information about each causal link

• noted new causal links not in the IPT

• noted any contradictions or ambiguities in the evidence about particular causal links.

Synthesis

For each causal link, we summarised the number and quality of studies supporting, refuting or qualifying it (see Chapter 7). We categorised the strength of each causal link’s evidential support as one of (in descending order):

• ‘substantial’ [i.e. a combination of primary studies and SR(s)]

• ‘supporting’ (i.e. multiple primary studies)

• ‘minimal’ (i.e. a single primary study)

• ‘partial support’ (i.e. some supporting evidence for parts of the underlying programme theory about that causal link – that it only operates in certain conditions, or with certain populations)

• ‘equivocal’ [i.e. evidence both for and against (but we also noted whether or not, in such cases, the evidence was predominantly on one side)]

• ‘none’ (whether evidence to the contrary or just the simple absence of any supporting evidence in the studies available to us).

A single working instance of a causal link between two components (‘minimal’ evidential support) does at least give evidence of the feasibility (‘proof-of-concept’) for that component operating as a mechanism to produce that outcome in another setting provided that the destination context has similar moderating characteristics to the original ‘proof-of-concept’ context. Equivocal evidence is, to the realist mind, a clue to the possible presence of contextual factors, which determine whether or not that mechanism will produce that outcome in different contexts for different populations and what kind or size its impact will be (e.g. the mechanism ‘works’ for one care group or in one kind of health system but not another).

Revising the initial programme theory

To convert the IPT into a revised, more strongly evidence-based logic model, we removed the causal links with no supporting evidence or where evidence existed but was against them. For causal links that had only partial support, we removed the unevidenced elements. These subtractions produced a truncated, but more strongly evidence-based, programme theory.

To that truncated version, we next added:

• relevant causal links found in the body of evidence but omitted from the IPT

• contextual statements of the circumstances that qualify the causal link between two MCP components, because certain specific conditions strengthen or weaken the outcome produced.

In places, the IPT was formulated ambiguously (see Chapter 4). To test it as it stood, we left these formulations untouched when comparing it with the secondary evidence. To produce a more coherent, less ambiguous, more evidence-based MCP programme theory, we separated out those concepts (e.g. ‘co-ordination’ and ‘integration’; see Chapters 4 and 6) that the policy sources had conflated.

Adding further contexts and mechanisms made an already complex programme theory more complex. It would be an exaggeration, but one with a grain of truth, to say that the initial MCP programme theory had come close to assuming that, in MCPs, every component helped to produce every other component (see Chapter 4). To differentiate the critical from the non-critical causal links, we used two methods. First, using the categories described above, we also categorised the strength of evidence for each subsequently added causal link, from ‘minimal’ to ‘substantial’ (see Chapter 7). As a graphical representation, we redrew the map of the revised logic model so that the width of each link reflected the ‘strength’ of evidence for it (see Figure 5). Second, to simplify the graphical representations, we eliminated redundant links in the revised logic model, applying the same principle as previously. However, in reviewing the evidence, we included all the links, both direct and indirect.

The product of these subtractions, additions, qualifications and definitions was a revised, more strongly evidence-based programme theory for MCPs, articulated in correspondingly revised verbal, tabular and graphical presentations (see Chapter 6).

Table A (see Report Supplementary Material 2) itemises, in detail, how our methods complied with the RAMESES quality standards. 41

Underspecified policy-makers’ assumptions

The 5YFV6 and related policy documents stated in general terms that MCPs will promote the ‘integration’ of care for older people with multiple morbidities by partly replacing hospital care with non-hospital care. However, for the most part, MCP policy was unclear about which components might act as mechanisms to produce which specific outcomes. At times, policy statements asserted what should be done without expressly stating how and/or what effects doing this would have. For example, at one Vanguard site there would be ‘more ways for people to digitally access health care (including online directories of local services, and a library of helpful health apps on its website)’,8 but this idea was not explicitly connected to any policy outcomes it would produce or to contextual requirements for it to work. Other statements were so broad as to be difficult to interpret concretely (e.g. that ‘artificial boundaries between hospitals and primary care, health and social care, between generalists and specialists are broken out of’ and ‘long term conditions are better cared for’). 6

Policy-makers may have left these points underspecified so as not to foreclose MCP design options or for other reasons (as with other policies10). Policy documents said that different types of MCP might emerge but not what these variants were or what might differentiate them. They suggested possible MCP contractors,5 but proposed a wide range, including Community Interest Companies, Limited Liability Companies, partnerships (including GP federations) and statutory NHS providers. MCPs were also described as ‘extended group [GP] practices’ that might be ‘federations, networks or single organisations’ (p. 20). 6 Concomitantly, ‘general practice at scale’ might, according to the 5YFV,6 be networks of independent general practices, perhaps with a strong central co-ordinating body (a ‘federation’). The House of Commons Health Committee argued that federations allow specialisation of service and care team development but retain the existing scale of general practices. 5 Although policy documents stated that MCPs will also have an element of vertical integration, or rather co-ordination, of care, short of structurally integrating primary and secondary care, they also usually discussed MCPs separately from PACSs.

Relationships between mechanism and outcome in the policy documents were often underspecified, and there was often ambiguity over whether the terms referred to a mechanism and its context or (without differentiating) both a mechanism and its outcome:

1. ‘MCP setup’ – ambiguity between a mechanism (i.e. actions by NHS managers) and a context (favourable or unfavourable background conditions).

2. ‘Demand management’ – ambiguity between a mechanism for managing demand (e.g. referral screening, risk stratification) and the outcome of doing so managing demand (fewer referrals and admissions to hospitals).

3. ‘Patient diversion’ – ambiguity between mechanisms for diverting patients away from hospital (e.g. providing alternative care outside hospital) and the outcome of doing so (e.g. fewer hospital admissions).

Multiplexity

If–then relationships between MCP components in the policy documents were successively linked (‘concatenated’) and multiplex. They rarely assumed that one mechanism produced just one final outcome, but more often that different mechanisms were concatenated so that the output of one was to create, or to trigger, the next. For example, policy statements expected the creation of MCPs to strengthen the management of provider networks, which would then strengthen referral networks, and then the referral network would lead to patients being diverted from hospital, and so on. The if–then relationships were multiplex in that a single mechanism was sometimes assumed to trigger several others. IT-based care co-ordination would, the policy statements jointly assumed, divert patients away from hospital, strengthen care planning at patient level and make urgent care more responsive. In reverse, the policy statements also assumed that one policy outcome would result from many mechanisms. For example, improved care planning at patient level would be the joint effect of IT-based care co-ordination plus referral networks plus demand management systems (themselves also resulting from care planning at organisational level) plus preventative care.

Translations and nomenclature

The policy documents made little explicit reference to evidence bases beyond some local evaluations. However, they often referred to two main international prototypes for MCPs, the US ACOs and PCMHs.

In NHS policy documents, the term ‘accountable care organisation’ means an ‘overarching organisation that sits above a joined up health and social care system made up of a number of different providers, from health services to the local council’. 50 Such an ACO would be certainly the predominant, perhaps sole, contractor for NHS-funded services with its local commissioner. The US Government’s CMMS defines ACOs as ‘groups of doctors, hospitals, and other health care providers, who come together voluntarily to give co-ordinated high quality care to their Medicare patients’26 (i.e. not for all patients) and provider membership is optional – differences to bear in mind when interpreting the ACO model and research for NHS use. NHS policy statements also borrow the term ‘Patient Centred Medical Home’ which, in the USA, formulates an ambition to construct something similar to what current NHS general practice (usually) is, that is, primary medical care based on the:

underlying principle of a single physician who coordinates the patient’s care and engages a team of health care providers and their patient in an individualized treatment and management plan.

Kash et al. 51

As the Royal College of General Practitioners (RCGP) points out, general practice is (already) ‘the natural medical home for patients’. 52 In the USA, a ‘medical neighbourhood’ is understood as a group of ‘medical homes’. Finally, ‘integration’ in NHS policy documents almost never means ‘organisationally integrated’ (as it would in some countries) but, rather, the closer co-ordination of services provided by separate organisations. Many health systems pursue that aim by creating referral networks53 of providers, with each network having a ‘network administrative organisation’54 doing much of the actual co-ordinating work. Policy documents also implicitly used the term ‘prevention’ in a hitherto non-standard way to mean long-term self-care, ‘activation’ and ‘empowerment’, and patient education, rather than clinical prevention or intersectoral activity, for health promotion.

Apparent omissions

Many policy statements were implicitly in mechanism–outcome configuration, rather than a CMO configuration. From a realist perspective, it was noticeable that policy sources seldom made assumptions (even implicitly) about what contextual factors would moderate the many assumed causal links between MCP components and outcomes. Nevertheless, a few contextual assumptions were stated and are outlined below.

Compared with the policy issues covered in Chapter 1, the policy statements said little about:

• organisational integration, in the sense of GPs, CHSs and other staff being members of the same organisation

• lack of residential and social care

• risk stratification and case management

• MCPs’ relationships to the other six new models of care.

Imputing the missing causal links and contexts in the policy-makers’ assumptions

To make the policy-makers’ assumptions empirically testable, one has, at times, to impute the necessary missing definitions and terms and operationalise them. As Chapter 3 explained, we did so by:

• asking our NHS think tank to interpret what, in practical terms, the policy statements appeared to mean to them as NHS clinicians and managers

• cross-referring between policy statements (at the cost of assuming that the same word means the same thing in different statements)

• exploiting the textual setting. For example, a statement about information-sharing in the context of hospital referrals and discharges was taken to refer to information sharing between hospitals and GPs, and between hospitals and CHSs

• referring to the international prototypes that policy documents cite (see Namesakes and equivalents in other health systems), although with due attention given to differences between the original and the NHS settings

• referring to particular named examples of plans for, or evaluations of, existing proto-MCPs. From these descriptions, the researchers abstracted the more general assumptions about how this MCP would work from its local particularities

• calling on the researchers’ (who included clinicians) background knowledge of primary care in the NHS and of relevant research to infer what such statements were most likely to allude to, and interpret, such euphemisms as ‘leadership’ for ‘managers’ or ‘expert generalist’ for ‘GP’.

By these means, we, so far as possible, formulated and elaborated the policy statements as ‘if–then’ statements (‘if’ = M–; ‘then’ = –O, provided that C–), which realist synthesis (indeed any empirical test) requires as raw material.

The initial programme theory: multispecialty community provider components and the causal links between them

We grouped by mechanism and by outcome the 242 if–then statements obtained from the policy sources (described in Chapter 3). These 13 groups were named as MCP ‘components’ and were linked by 28 ‘causal links’. Together, these make up the top level of the initial MCP programme theory. Underlying or, rather, composing each causal link are single or multiple if–then statements, making the more detailed content of the IPT. Appendix 11 summarises the 13 MCP components in our interpretation of the policy-makers’ assumptions about MCPs.

Table 1 and Figure 1 illustrate the IPT that we took forward to the evidence review. The IPT is our interpretation of the policy-makers’ assumptions (developed as described in Chapter 3, and glossed at points to explain our reasoning). This IPT is made up of 13 components and 28 causal links between them. As this review is focused on exploring the evidence for how MCPs work and not what MCPs are, it is the causal links between the 13 components and not the 13 components themselves that guide our evidence review (step 2; see Chapter 5 for results of the evidence review). Figure 1 illustrates the 28 causal links. Note that the 13 components (see Appendix 11) include the two main outcomes of MCPs [(1) component 12: patient experience and care will improve; and (2) component 13: NHS costs will reduce). Because these two ‘components’ are the intended end result of MCPs in NHS policy, in the IPT they are not the mechanism for producing any of the other 11 components, and, hence, appear only in the right-hand (‘then’) column in Table 1.

FIGURE 1.

Causal relationships between the 13 MCP components in the policy-makers’ IPT.

Figure 1 shows the relationships between these overall groups of causal links. Each arrow represents a generalisation from the if–then relationships stated or assumed in the policy documents. In realist terms, each arrow with its left-hand box represents a mechanism and the box at the right-hand (destination) end of the arrow represents its outcome. Table B (see Report Supplementary Material 2) shows the same relationships in tabular form.

Figure 1 shows the ‘flow’ or sequenced linkage (‘concatenation’) of the assumed causal links between components 1 and 11, through to outcomes 12 and 13 (improved care and reduced cost). Each component is assumed to be the mechanism to bring about change in later components, and those components are then assumed to be the mechanism to bring about change in yet later components; these eventually jointly lead to the MCP outcomes (far right).

Contexts

As noted, policy sources contained fuller accounts of assumed mechanisms and outcomes, and some mediating linkages, than of the contexts that might moderate the achievement of those outcomes. However, they did include some detailed assumptions, outlined below, about what initial conditions favour the establishment of MCPs, in particular the first wave (i.e. that concerned only links 1:2 and 1:7). In addition to some managerial mechanisms (a vision of a model of care;9 effective managerial and clinical leadership;9 standardised data to enable real-time monitoring and evaluation of quality outcomes, costs and benefits;9 and planning how to provide care for people with long-term conditions in primary care settings and in their own homes, with a focus on prevention9), the contextual conditions likely to be critical to enable the first wave of MCPs to bring about their intended outcomes were assumed to be:

• existing progress towards new ways of working9

• a financial situation that allows start-up money to be found for MCPs – local commissioners support already-agreed funding for the MCP9 existing ‘partners’, such as voluntary and community sector organisations, and ‘communities’9 are supportively engaged with the MCP. Organisations relate to each other in a collaborative, mutually helpful way. Local relationships are good

• local NHS leadership focus on MCPs and care integration generally

• the populations served are of a size and type likely to benefit, which we interpret as being large enough to allow economies of scale and scope in collaborative working, and with a health profile and socioeconomic mix that the MCP services can accommodate

• a population who desire autonomy and control over their health and health care, and are likely to participate (‘engage’) in activities to maintain their own health and to help care informally for those experiencing chronic ill health

• health professionals and organisations view those whom they care for as people, not patients

• sufficient staff inputs (time, skill mix)

• the responsible CCGs show engagement and flexibility, and are not excessively risk averse towards the risks of procuring new organisations and/or networks to operate a MCP

• a well-functioning GP network (group or federation)

• the corresponding social services are capable of providing the services needed to sustain and MCP.

Policy statements and informants did foresee certain general contextual problems in establishing interorganisational level care co-ordination, but did not clearly link them to any specific relationships between any of the 13 main components of the policy-makers’ IPT:

• tension between clinical and financial imperatives

• the necessity of moving the pressure of demand to new points in the local health system and of removing some roles

• difficulty in moving beyond information distribution to ensure that organisations within the MCP use that information effectively

• increased pressure on carers and voluntary sector. If not a context, this might be understood as a side-effect, or perhaps a feedback effect, of links 5:8 and 5:9

• an initial dip in patients’ experience because some patients would be resistant to the changes in care provision.

However, the fundamental contextual assumption was that a substantial proportion of unplanned admissions of older people with multiple morbidity are clinically unnecessary, even iatrogenic; hence, preventable. 32 This implicitly applies to all the links involving patient diversion (6:11, 7:11, 8:11, 9:11, 10:11, 11:12 and 11:13).

Causal link 1:2 – if NHS managers establish multispecialty community providers, then multispecialty community provider-wide management structures will develop

Unless they fail even to get MCPs started, it is a near-tautology to say that if NHS managers establish MCPs, then MCP-wide structures for planning, developing and operating the included services will develop. Nevertheless, the studies we found included some indicating how this mechanism worked and options for the structures to set up. Three main mechanisms contributed to the creation of a MCP-like structure, two of them being motives of the providers joining it.

First, provider organisations wished, but were unable, to provide all the needed health services. The prospect of providing such services on site (e.g. dental and vision care, and specialty medical care) motivated clinics in Boston to form ‘strategic partnerships’ to meet the needs of a complex patient population. 55 Providers with a large share of patients with mental health needs were more motivated than other providers to use an ACO to improve mental health services, both to meet patients’ needs and to reduce the burden on providers themselves. ACOs in regions with a low supply of mental health specialists were also looking for ways to integrate mental health care into other settings, typically primary care, to meet patient demand. 56 That is, the MCP-like structure (for instance, an ACO) appeared to those provider organisations joining it to be relevant to their care group(s) and clinical work. 57 In the USA, the criteria of ‘relevance’ included whether or not the patients would have insurance coverage. 57

Motivation to pursue the member organisations’ interest as organisations was a second mechanism. In another study, the majority of respondents (15/25) gave joining like-minded organisations and minimising future risk (18/25) as ‘important’ or ‘very important’ reasons for forming their ACO. 58 For some, an important consideration was that joining an ACO offered participating organisations the prospect of clinical ‘integration’ without corporate ‘consolidation’ (takeover). Hence, physician practices that participate in ACOs are likely to be large and/or be members of an independent practice association or physician hospital organisation and unlikely to be hospital-owned. 11 Health centres and other ACO members retained their independence but worked together under an ACO contract in new partnerships. 59 Endorsing or improving the member organisations’ internal organisation was also a motive for joining. Some US doctors perceived organising as a patient-centred medical home as key to providing high-quality care and as ‘the right thing to do’. Others described recognition as acknowledgement for how they organised their practice. Physicians described participation in PCMH demonstration projects, quality improvement (QI) initiatives and external support for seeking recognition as key motivating factors. The extent to which substance abuse treatment services’ staff were clinicians with professional degrees predicted these organisations’ likelihood of participating in an ACO. Some also said that participation gave them access to external data sources (such as insurance companies) and to health information exchange, which enhanced their QI strategies and ability to function as PCMHs. 59 Physician groups played a more prominent role than other provider types (including solo-practice physicians) in forming and managing rural ACOs. 58 Organisations’ financial motives for joining ACOs were to increase activity (hence income), in contrast to a NHS context. Preparing for value-based purchasing (14/25 respondents) and getting paid for quality (10/25 respondents) were the most frequently cited ‘very important’ reasons for the ACO formation. 58 Among substance abuse treatment centres specifically, those who reported a greater local competition were more likely to have signed a contract with an ACO. 57

Both of these mechanisms imply that member organisations in a voluntary network are self-selecting, which, a Canadian study suggests, will of itself stimulate evolution towards a more integrated network. Patients may also be self-selecting. Québec clinics’ improvement in ideal type integrated care (ICIT) scores was partly due to a ‘natural selection’ effect of clinics that closed, and the effect was mitigated by clinics that opened after the 2005 survey. Change in ICIT score was associated with both this evolutionary trend and central reform policies. 60

Decisions from higher authority were the third mechanism for the formation of MCP-like entities. Decreed ‘top-down’ reform was instrumental and an obvious prerequisite for initiating change in the Québec health-care system. Coercive and mimetic factors influenced primary care provider organisations’ ICIT score to shift towards greater ‘integration’. These primary care organisations did not regard health and social care centres’ support in creating PCMH-like organisations as very substantial. 60 The Ontario Health Ministry’s use of ‘simple rules’ encouraged change in the desired direction without stifling creativity and innovation. 61 In Australia (a similar primary care system to the UK in many respects), HealthOne Mount Druitt needed sustained support at higher governmental levels (New South Wales Health and the regional Local Health District management), but in a form enabling policy change without attempting to micromanage local developments, which would have ended all chance of GP participation. 62 In general, efforts to improve outcomes by exerting top-down control were often intrusive and futile, slowing down the inherent capacity of the system to adapt and evolve. 62

The studies that we reviewed described different structures that emerged, in different contexts, as the outcome of managers’ attempts to set up MCP-like organisations and networks. In ascending order of common managerial control, the simplest was information managing across the member organisations. Thus, ACOs must report on 33 quality metrics across patient/caregiver experience, care co-ordination/patient safety, preventative care and at-risk population. 11 EHRs are an important means of managing and co-ordinating patient care for effective ACO performance; substance abuse treatment services’ use of EHRs predicted how likely they were to participate in an ACO. 57 Other studies corroborate this pattern63 (see also Chapter 6).

Contracts have also been used to establish MCP-like networks on a quasi-market basis. One example, despite linking only mental health services, are the Integrierte Versorgung networks of primary, secondary and social care services in cities such as Hamburg. In the Netherlands, bundled payments result in a principal contracting entity (provider) being lead contractor for numerous other subcontractors. 64 Billings and de Weger11 describe and distinguish three other contract-based structures: (1) outcomes-based commissioning (an existing NHS approach), (2) the ACO model and (3) an alliance model of a network of providers making a joint contract with a payer (in that respect, closer to the NHS idea of an ACO). However, some ACOs themselves use joint payment contracts. Such ACOs are more likely than others to include community health centres (CHCs; ‘safety-net’ primary care providers), hospitals, medical groups, nursing facilities and specialty groups, but not to have more primary care and specialty clinicians. 59 In passing, Billings and de Weger11 mention four more models: (1) a ‘Partnering Model’, (2) ‘Value-based Health Care’ (which is not specific to MCP-like organisations65), (3) ‘Incomplete Contracting’ and (4) the Alzira Model (which is less relevant here because it focuses on vertical integration).

Managers have often established a central co-ordinating body (‘network administrative organisation’54) to co-ordinate MCP-like networks and quasi-markets. Thus, Intermountain Healthcare in Utah had a central medical and administrative team in which research groups gave economic and organisational support for running clinical programmes. The support elaborated good practice recommendations and the corresponding indicators, which providers followed and measured. 66 In establishing PCMH, one problem for managers, as for researchers and other interventions too, was that different organisations varied considerably in their definitions of a PCMH, so in practice they were not all trying to implement the same intervention. 67

Some health systems have pursued the functions for which MCPs were designed through organisational integration, that is, amalgamating the separate components of primary care (i.e. primary medical care, community nursing, therapies and perhaps mental health services and/or social care) into one organisation, as do the Swedish23,24 and Finnish polyclinic model and Italian Unità Sanitarie Localis (USLs).

We found no evidence either for or against the policy assumptions that population attitudes and beliefs about actively maintaining one’s own health and helping to care informally for people in chronic ill health or about autonomy and control over one’s health and health care were mechanisms that contributed to establishing MCPs. The same applies to whether or not health professionals and organisations view those whom they care for as people not patients. We also found no direct evidence about the requisite state of social services.

Causal link 1:2 – contexts favouring the creation of multispecialty community provider-like networks or organisations

Evidence from several countries suggests that good pre-existing interorganisational relationships facilitate the establishment of interorganisational co-ordination mechanisms, which then reinforce the good relationships in a virtuous circle68 (as has also been reported in studies of hospitals69). This pattern recurred in several studies of ACO formation. Although the joint payment system was new to the ACOs that used it, the organisations that participated had existing informal partnerships. 59 A study of the formation of four rural ACOs also found that prior experience with risk sharing and provider integration facilitated ACO formation. 58 An Australian study corroborated these patterns. Planning for HealthOne Mount Druitt was led by a steering committee with links into the local community through strong representation from local GPs (71% of them were single-handed), community nursing services and the Western Sydney Medicare Local. 62 In Ontario, Canada, organisations with a history of collaboration, pre-existing relationships among partners and a pre-existing focus on integrated care saw the Ontario Health Links model as a natural step forward and found the transition into it easier than for organisations without existing collaborative relationships, knowledge and resources to draw on. 61

Implementing ‘top-down’ decisions to create PCMH-like organisations in Québec also required their internal ‘receptivity’ to joining a network, including a ‘mimetic’ context of other exemplar PHC organisations also participating [but the admired prototypes were family medicine groups (FMGs) and/or network clinics rather than the new health and social services centres], and the presence of ‘local champions’ advocating the new models and demonstrating their feasibility and desirability. 60

A realistic time scale was also required. PCMH programmes typically took a few years to reach maturity and produce measurable effects. 70 In the USA, time was required to obtain ‘bureaucratic’ approvals and check conformity with antitrust regulations. 11 Similarly, in Australia, the planning stage of HealthOne Mount Druitt took 2 years; the greatest challenge was building relationships between the key partners, especially overcoming strong established barriers to trust between GPs and community health. 62

Several studies (e.g. Billings and de Weger11) report substantial ACO start-up costs. Just one ACO in New Jersey (covering 2 million people) required US$2.8M start-up funding, after which it was to become self-sustaining from the savings generated. 71 Patient Care Medical Homes also required start-up funding. For example, participation in a state-funded interpractice collaboration to improve quality helped 20 medical practices to implement a PCMH approach internally. 72 ‘Technical support’ (character unstated) was also required. 72 Conversely, another study70 noted as unusual that the ‘CareFirst’ PCMH project did not require ‘up-front investment’. When that investment has to come from the participating provider organisations themselves, smaller organisations are at a disadvantage. 70 To address that obstacle, the ACO Investment Model programme provided initial investment capital and variable monthly payments to ACO participants in rural and underserved areas that might not otherwise have access to the capital needed for successful ACO formation and operation. CMMS also contracted 32 organisations under a special ‘Pioneer ACOs’ demonstration project. 58

When statute guarantees patients a choice of provider (as in the USA and, in theory, the UK), it will be difficult to steer patients to particular providers,11 weakening the evolutionary pressures towards MCP-like ‘integration’.

We found no direct evidence about maximum or minimum viable size of a MCP, economies of scale or scope, or the demographic or social character of places where it might be easier or harder to establish MCPs. However MCP-like networks appear harder to establish in rural areas, where general practices are small and isolated68 and where providers cannot contribute to MCP start-up costs.

Causal link 1:7 – if NHS managers establish multispecialty community providers, then planned referral networks develop

As regards referral networks specifically, our evidence suggested that some MCP-like networks do indeed develop referral network planning at organisational and/or interorganisation level. Physician practices that participate in ACOs are more likely than non-participating practices to use more care management processes. 11 One study describes a PCMH that negotiated 50 ‘compacts’ (agreed procedures for referring patients between providers) with specialist providers while other nearby PCMHs negotiated few or none, but does not report any contexts explaining why these differences arose. 68 Furthermore, the PCMH is designed to co-ordinate patient care mainly within a primary care team (within a general practice, in NHS terms) rather than across care teams. 68 A limitation to establishing referral networks is that many ACOs do not formally cover post-acute care (the function of CHS in England). A total of 87% of the ACOs that did cover post-acute care included a hospital (compared with 41% of ACOs without post-acute care). CHCs were also more likely to be integrated into ACOs that included post-acute care (58% vs. 49%). 73 Small, isolated rural practices were less likely to establish care compacts. 68

However, all of this does not resolve the issue of whether or not prior collaboration favours the initial formation of a MCP-like organisation or network, or whether or not stronger referral networks result from forming such an organisation or network, or, in a virtuous circle, both. Although it gives proof of concept that it is feasible, the evidence of ACOs also suggests that ‘horizontal’ PHC networks do not automatically develop interorganisational referral networks, in particular between GPs and CHSs (or the local equivalents). This suggests that further specific contexts are required, as yet unidentified in the published research.

Causal link 2:3 – if network management develops, then multidisciplinary teams will develop

The studies also reported many instances of MCP-like networks setting up MDTs. One aim of the Utah Mental Health Integration (MHI) programme was to orient patients towards support by a MDT (general physician, care manager, psychiatrist and psychologist) in ambulatory care, or hospital for the most severe cases. 66 The Versailles geriatrics network brought multidisiciplinary expertise together at the local information and co-ordination centres (Centres Locaux d’Information et de Coordination), Homes for the Autonomy and Integration of Alzheimers’ Patients (Maisons pour l’Autonomie et l’Intégration des malades d’Alzheimer), and the mobile geriatrics teams (Équipes Mobiles Gériatriques), which worked with local hospitals and doctors to avoid hospitalisation. 74 In general, MDTs require clear boundaries, a need to be collectively accountable for patient care, a need to be highly interdependent and a stable membership. 67

Multidisciplinary teams varied in their occupational membership and, therefore, the services that they could provide without external referral. The studies reported MDTs that included:

• CHSs but not doctors. Buurtzorg nurses worked with community volunteers, social workers, physiotherapists, occupational therapists and community psychiatric nurses21 but, although Buurtzorg also encourage collaboration between their own staff and the patients’ doctors75 (whom they do not employ), their MDTs do not usually include doctors.

• Doctors but not CHSs, as in a large number of American ACOs (see Causal link 1:7 – if NHS managers establish multispecialty community providers, then planned referral networks develop).

• Both doctors and CHSs. Swedish and Finnish polyclinics, some primary care providers in Spain and Italian USLs employ doctors, nurses and therapists together within a single organisation. Some NHS Integrated Care pilots relied heavily on MDTs, although the ‘virtual ward’ involved hospital doctors rather than GPs. 76 A survey of ACO-employed social workers found that 65% worked with primary care physicians, 55% with specialty physicians, 74% with nurses or nurse practitioners and 31% with psychologists. A total of 48% had a nurse or nurse practitioner as their immediate supervisor, 25% had a social worker and 4% had a manager. 77 One US variant was a ‘physician-led’ team such as the ‘perioperative surgical home’ (PSH) in which activities included patient ‘rehabilitation’ before surgery and transitions to home or post-acute care designed to reduce complications and readmissions. 51

• Mental health services. Lewis et al. 56 describe the addition of mental health-workers (e.g. social worker, psychiatrist) to existing primary care teams, so that care management remains with just one provider.

• The patient and/or informal carer(s). 67

Some MDTs were ‘virtual’ (i.e. co-ordinated by teleconference, video conferences or other HIT systems). 62,78 Between them, the studies that we found reported that MDTs were based (moving from ‘virtual’ to ‘real’) on:

1. The consulting model, in which one clinician consulted another without actually referring (i.e. temporarily transferring) the patient. In some ACOs, the role of consulting mental health specialists included coaching primary care providers in the use of psychiatric medications, assisting with diagnoses, and making appropriate referrals to specialised mental health care services. 56 In the HealthOne Mount Druitt project (Australia), consultations also included ‘more informal exchanges of information.’3

2. A dispersed, partly remote, team linked by IT. Thus, a paediatrics MDT in five English CCGs involved members (including GPs) by teleconference. 76

3. Colocation (e.g. of mental with physical care clinicians). 46,56

4. ‘Embedding’ of, for example, mental health-care clinicians within primary care teams;56 in effect, seconding staff from one organisation to another.

5. ‘Huddles’, that is, informal, ad hoc, but frequent (e.g. daily) staff meetings, reported in 73% of practices in a survey of 40 small primary care practices (PCPs) in Texas. 79

6. Formally structured cross-organisational teams.

7. Staff all employed by the same organisation.

Care co-ordination and communication sometimes required MDTs to adapt health professionals’ traditional roles. 66

Causal link 2:3 – contexts

Barriers to including pharmacists in MDTs in PHC medical practices in Vermont were the pharmacists being employed by a separate organisation, with pharmacists and physicians being unfamiliar with each other’s scope of practice and roles. 80 Different patients might also have different but often overlapping provider networks, and these overlaps offered the greatest scope for strengthening care co-ordination. 67

Most papers described what care co-ordination activities MCP-like interorganisational networks undertook rather than how these co-ordination arrangements were created or (in the realist sense) their contexts. Many of these arrangements were reported in just one study, but, when there were several reports, they were mutually consistent.

Causal link 2:6 – if network management develops, then care co-ordination supported by health information technology will develop

We found substantial evidence of interorganisational care networks establishing structures and work processes to co-ordinate care across multiple provider organisations, so that clinicians and organisations adapted their work routines and practices to network standards, shared information, created ‘boundary objects’ such as care plans, and standardised organisations’ and clinicians’ roles, and care pathways across organisations. 67

Six out of the 13 sites in the Alidina et al. 68 study reported that implementing co-ordination mechanisms increased communication and trust. Such routines included primary care doctors ‘feeding back’ to specialists. 68 ‘Care compacts’ assisted communication, decision and negotiation between organisations and improved care access and quality. 68 A survey of rural pioneer ACOs found that managing care across the continuum and meeting quality standards were what the respondents most frequently reported as ‘very important’ to the ACO’s success. 68 Initially, maintaining good relationships between the member organisations was important for ACO success, pending the development of more standardised and contractual relationships. 68 Some Ontario Local Health Integration Networks (LHINs) pooled resources across partners and standardised structures and processes related to governance, accountability and administrative functions in an attempt to avoid duplication and waste. 61 Several respondents in those networks suggested that the type of lead organisation mattered less than that organisation’s reputation, existing relationships and partnerships, and leadership style (e.g. having a positive image in the community and among providers and a track record of innovating and following through on commitments, and for tolerating change, risk and ambiguity). 61 The Kinzigtal network (in Germany) jointly developed care pathways across providers and synchronised hospitals’ and ambulatory care providers’ formularies across all care sectors. 81 In the Netherlands, care standards with a modular structure (general and disease-specific elements) were jointly negotiated among providers, an arrangement that routinised collaboration among doctors. 64 The HealthOne Mount Druitt project (in Australia) used case conferences to co-ordinate services at patient level, in particular with non-health-care services, such as social care. 62

Assuming that standardised care pathways and quality standards do, indeed, define the character of patient needs more clearly, these studies tend to support the policy assumption that MCP-like networks will lead to clearer definitions of patient needs and promote evidence-based targets for managing long-term conditions.

Causal link 2:6 – care co-ordinators

The studies that we found neither used the term ‘care navigator’ nor described similar advocates or helpers for individual patients. Instead, many of them reported how MCP-like networks had used care co-ordinators, for example by creating dedicated care co-ordinator positions. 68 Nurses working as care co-ordinators were reported in Texas and Colorado. 82 Community health workers (CHWs) recruited from the local population were used (in Texas) to help bridge gaps between patients and organisations, and between organisations, to enable PHC teams to connect patients with resources that patients need. 83 In New York, social workers were ‘embedded’ in PCPs and included in all practice-based meetings and other aspects of patient care. ACO quality metrics meetings were critical to developing working relationships with PCPs and other members of the care team, and with care co-ordination staff in other programmes. 78 The HealthOne Mount Druitt project (in Australia) also recruited GP liaison nurses for co-ordinating services in ways that the GPs could not because of lack of time or knowledge of the services available (e.g. home care, counselling and other allied health services). 62 These nurses managed communications, case conferencing, case management and overall care co-ordination, and allocated the case management of individual patients to the most appropriate person in the MDT. 62

The foregoing evidence corroborates the policy assumption that primary care provider networks are capable of co-ordinating inputs across multiple services. Contrary to UK policy assumptions, Alidina et al. 68 concluded that the above changes did not require culture changes or payment reform, but Wholey et al. 67 argued (corroborating UK policy assumptions) that they do require large numbers of clients so as to allow economies of scale.

Causal link 2:6 – contracts

Nevertheless, several health systems have attempted to use contractual mechanisms to strength care co-ordination between separate providers. The Kinzigtal network (Germany) co-ordinating body made contracts with the two main social health insurers involved (AOK and LKK) and over 100 local providers to implement various programmes for individual treatment plans, patient self-management, follow-up care and case management. 81 Two complications are the ‘hangover’ of existing contracts and technical difficulties of contract monitoring. In the US ACOs, providers’ decisions whether or not to pursue integrated models depended powerfully on the design of the ACO payment model (implying, at one remove, patients’ insurance status), details of contracts and the quality measures used in contracts. Contract design appeared to influence the extent to which ACOs integrated mental care. 56 In practice, the English NHS has so far had little success in commissioning through outcomes-based incentivised contracts for these purposes because of the difficulty in specifying and measuring the relevant outcomes, and then in knowing whether or not to attribute any changes to the providers, care co-ordination or extraneous confounding factors. 11

Interorganisational co-ordination mechanisms are especially required when patients have highly complex health problems and providers have a low level of knowledge about the patient’s condition,68 as often applies to patients with long-term conditions. In combination with other (unspecified) enabling ‘changes’ within the local health district and the wider New South Wales health sector, the HealthOne Mount Druitt network (Australia) began delivering services through two main streams: (1) chronic aged and complex care and (2) child and family care. 62

Causal link 2:6 – network membership

A limiting factor is what organisations (hence, services) a network contains. In the studies, we found that MCP-like primary care networks varied in whether or not they included:

• Mental health services – in 2014, 42% of ACOs included mental health-care providers. ACOs with ‘a comprehensive, chronic care management program’ were more likely than others also to have integrated mental and physical care. 56 In Utah, MHI’s co-ordinating approach allowed it to replace the traditional model of partitioned-off, sectorised psychiatry with a co-ordinated combination of ambulatory care, specialist secondary, and first-recourse care, which, in turn, allowed territory-wide, whole-population planning of its services (reducing ‘medical deserts’), organising support networks to promote preventative care, and developing ambulatory services that linked hospital, medicosocial work and social care. 66 Lewis et al. 56 describes two main approaches to overcoming the traditional separation of primary and mental health care –

  1. expanding primary care to cover mental health conditions (9 out of 16 PCMHs in that study)

  2. integrating primary care providers into existing mental health programmes (2 out of 16 PCMHs).

• Children’s services – for the NHS, Woodman et al. 76 report four ways of bringing paediatric expertise into primary care and/or improving joint working –

  1. telephone-based MDTs

  2. hospital at home

  3. outreach clinics

  4. paediatrician advice and guidance to GPs.

     These initiatives work by promoting shared responsibility, upskilling GPs, establishing relationships between paediatricians and primary health-care professionals, and by taking specialist care to the patient.

• Community health services (or the equivalent) – 48% of ACOs in the Colla et al. 73 study did not include post-acute care. Those were more likely to be physician-led. ACOs that did include post-acute care were more likely to have programmes to reduce preventable hospital admissions and for end-of-life care. 73

• ‘Safety-net’ services – a substantial number of ACOs included CHCs. A greater proportion of those ACOs with a centre reported experience of public reporting, of having patient-centred medical homes and holding other risk-bearing contracts. ACOs that included at least one federally qualified health centre among their participating provider groups were more likely to report complete integration of services and to offer less common services, such as health coaches and case managers. 56

• Primary medical care – the studies mentioning general practice engagement in MCP-like networks reported that GPs (or the equivalent) valued the access to additional resources that such networks gave. Versailles doctors (including GPs) participating in a geriatrics network reported being satisfied with the way it provided expert advice and access to hospital-like support for patients at home. 74 Similarly, physicians within integrated health systems in Texas and Colorado frequently discussed the value for care co-ordination purposes of shared resources across sites, for example nurse care co-ordinators, nurses providing advice during and after office hours, enhanced access through expanded office hours, electronic communication, ‘virtual visits’ (to patients), access to hospital records, referral tracking, physical workspaces organised to facilitate team-based care and access to non-physician providers (e.g. dieticians, psychologists). 56 However, it was not always easy for GPs to participate in network activities. In the Versailles study,74 one-third of doctors did not wish to participate in network meetings at patients’ homes, judging them too time-consuming. These studies did not directly report whether or not MCP-like networks reduced general practice overload. Indeed, the Versailles study74 implies the opposite. These studies tend to corroborate the assumption that smaller general practices find it difficult to contribute to networked care co-ordination activities. We found no studies reporting whether or not the creation of MCP-like networks leads to improved infrastructure management in primary/community care.

A network’s membership constrains that of the MDTs within it.

Causal link 2:6 – health information technology adoption

A shared patient record promotes informational connectivity68 and, by implication, informational continuity of care. 15–17,84 We found recurrent accounts of primary care networks attempting to increase staff access information needed for making referral decisions. The Kinzigtal network (Germany) introduced common EHRs across all care sectors. 81 However, initially, many US ACOs did not uniformly have developed, interoperable IT systems. 68 Assuming that shared information will help networks and providers define the character and scale of patient needs more clearly, these studies tend to support the policy assumption that MCP-like networks will lead to clearer definitions of patient needs. Although we found no counter examples, these studies also indicated that such information sharing is not easily achieved.

Causal link 2:6 – contexts

Just as prior collaboration assisted the formation of MCP-like networks, so it facilitated network management. In the 13 ACOs that Alidina et al. 68 studied, more complex co-ordination mechanisms (i.e. communication, decision and negotiation) complemented, rather than replaced, existing ones. 68 Conversely, lack of trust was an initial challenge in setting up the HealthOne Mount Druitt (Australia) system. 62 Irrespective of their profession, uniform training for care co-ordination staff in New York (covering ‘Basics’, Practice, Psychosocial Domains, Disease Conditions, and Medical Services) helped to ensure a consistent approach to care co-ordination. 78 A study of PCPs in Colorado and Texas found that the use of practice facilitators to visit primary care physicians was significantly correlated with the use of sustained chronic care management strategies. Despite external facilitation, it remained difficult for the smaller PCPs practices to implement the CCM. 82 In using contracts to co-ordinate care, pre-existing carve-outs, in which a commercial payer had already contracted mental health care to a separate provider, practically excluded those services from an ACO in the short term. 56

Case-mix was another important context. High patient complexity and low knowledge about the patient’s condition is the situation that, the Alidina et al. 68 study suggests, most requires ‘boundary spanners’ for enabling reciprocal co-ordination between providers.

In New York, preparation for sharing medical records across providers involved extensive training, work-group activity and software development (for reconciling the different PCPs’ discrepant EHR systems). 78 In Virginia, the obstacles appeared to include lack of internet access and computer literacy among the target populations. Even in the Netherlands, where internet usage is extremely high, patients showed a lack of awareness and motivation to hold their own health records, and there were usability problems in the systems for accessing them. 85 There have been similar experiments in Sweden, with mixed success. 23

Causal link 3:7 – multidisciplinary team working produces planned referral networks

Multidisciplinary team working produces care network planning at whole-organisational and at interorganisational levels by facilitating co-support and decision-making across disciplinary boundaries. These activities are enabled by:

• the development of new or expanded boundary-spanning roles that enable fuller formal and informal communication across the MDT, and joint support for decision-making across disciplinary boundaries

• inclusion of colleagues from a range of disciplines and interprofessional relationship building

• addressing barriers (e.g. traditional hierarchies, lack of role clarity, divergent expectations) to awareness and understanding of the knowledge, training and benefit of working in an interprofessional way when dealing with complex, multimorbid patients.

Contexts that facilitate this mechanism are reported to be managerial recognition and support of MDT working, and cultivating trust in place of resistance towards other professions.

Causal link 3:7 – co-support and decision-making

Qualitative (five) and mixed-methods (one) studies in the USA (five) and the UK (one) show that, in addition to promoting care planning for individual patients (see Care planning for individual patients), exposure to multidisciplinary working [e.g. through ‘embedding’ (seconding) or colocating staff] creates more opportunities for different professions to improve understanding of each other’s treatment approach. 86–88 It also shifts providers’ expectations for communication and increases their awareness of the importance and benefits of involving other primary care providers in complex cases,89 upskills primary care providers and promotes shared responsibility. 76 A narrative case study of mental health integration in a CCM showed that this co-support across disciplinary boundaries helps members of each profession not to feel alone in the face of complex multimorbidity issues about which they are not specialists and to make shared decisions on complex problems. 66

Causal link 3:7 – boundary-spanning roles

Many studies about how MDTs surmount organisational barriers described new or expanded boundary-spanning roles as a key mechanism. These boundary-spanning roles improved co-ordination and integration of services through improving communication (formal and informal) between the various other care providers, through co-ordinating multiple services, addressing psychosocial as well as physical health issues, providing the conduit for GPs, community health, and other health and social care providers to work together more closely. 90 They also provided support for clinical and administrative staff. 87 MDTs are also part of the health-care delivery system redesign and connection to the community care resources involved in the CCM. 83 A study of focus groups with 387 people from 10 US communities suggested that having non-medical staff in boundary-spanning roles helped to co-ordinate patients’ care and to address barriers to care co-ordination. Patients appreciated having individual care plans with a holistic orientation, including a personal physician providing access to continuous comprehensive care. Two other studies86,91 reported similar findings.

Although MDT members might see these roles as the ‘glue’ holding care co-ordination and care teams together,90 interviews with 25 clinical pharmacists and 17 primary care clinicians found that traditional status hierarchies could be a barrier to effective collaboration and communication in PCMHs, where there were new roles for some or all professionals. 92 An online focus group of people self-identifying as care co-ordinators in PCMHs described primary care doctors as the biggest such barrier. The MDT had to win them over by strong self-promotion if these resistant doctors were to become a resource to the rest of the team. 87 This focus group also indicated the importance of boundary-spanners being embedded within a primary medical care practice. 87 Interviews and a survey of people in different MDT roles in the PCMH (the USA) found that the benefit of these new roles was maximised when there were loosely specified implementation protocols and a vision of the roles’ full potential. 93 Similarly, policy- and decision-makers in a chronic aged and complex care network suggested that boundary-spanners need to have the seniority and expertise to be leaders who earn and maintain the respect of the MDT by initiating culture change, and to have sufficient flexibility in their role to work with GPs to support and add value to the care they provide. 90

Causal link 3:7 – role clarity and expectations

Multidisciplinary teams often involve team members taking on new roles. This creates the potential for lack of clarity about roles and expectations between MDT members and, thus, strained relationships across disciplinary boundaries. In a case study, Matiz et al. 94 observed that responding to PCMH team members’ concerns and clarifying roles by educating teams about each profession’s strengths and limitations proved essential to integrating the MDT. 94 Interviews with primary care providers and clinical pharmacists in PCMHs showed that, despite frustrations between professionals with different opinions about new roles within PCMHs, being exposed to the other professionals’ reasoning improved understanding, respect and communication. 92

Causal link 3:7 – relationship building

Most of the barriers to, and facilitators of, care co-ordination at the organisational level identified in an online focus group of self-identified care co-ordinators in PCMHs by Friedman et al. 87 related directly to relationship building in MDTs, which was facilitated by boundary-spanning roles, enhanced communication (e.g. on-site mental health services), EHRs that interfaced well with outside organisations and training in motivational interviewing. Interviews and a survey with mental and primary care staff in PCMHs showed that mutual familiarity across disciplines through the use of a staff directory (with picture and contact information for each clinician), cross-disciplinary training and a mail server for ongoing, informal, patient non-specific consultation all improved interprofessional relationships. 89 A focus group of 17 primary care clinicians from different ‘integrated’ care models in the USA showed the importance of staff perceptions and knowledge about the training of other disciplines. At first, most doctors seemed reluctant to consider pharmacists as providing patient care but reviewing their training and knowledge led some physicians to value pharmacists’ contribution to patient care. 95

Causal link 3:7 – inclusion of new roles in multidisciplinary teams

Structured team communication in the PCMH in the USA facilitated the inclusion of the new members as part of the MDT and improved recognition of other MDT members’ value. 93 Two studies of the US PCMH model found that facilitators of improved communication are clearly defined expectations with agreed time frames for written updates, judicious use of HIT, electronic information exchange that met confidentiality requirements, jointly determining key information to be shared and frequency of updates89 and using faxes for routine updates so as to reserve the use of telephone calls for urgent matters and pre-planned consultations. 93 Faxable forms worked better if they were concise, easy-to-use, included the desired data and clinical impressions, used tick-boxes to document information and contained a pre-agreed expected minimum level of information to be shared by each discipline. 89 Clinical pharmacists and primary care providers in the PCMH described how delays caused by communicating back and forth electronically, the absence of real-time (or face-to-face) explanations and diverging inferences about each other’s intentions could impede communication within the team. 92

Causal link 3:7 – contexts

Contexts facilitating the operation of the above mechanisms included management, skills development and professional attitudes. One such context was for managers to encourage mutual support between staff of different professions. Interviews with 12 participants in US PCMHs found that giving clinic administrators protected time for interdisciplinary meetings or consultation and allowing for warm handoff in clinicians’ schedules facilitated interdisciplinary working. 88 A qualitative study of US integrated care models found that those integrated care models that were successful on at least one of clinical outcomes, satisfaction and spending, managers had found successful boundary-spanners and facilitated their relationship with other staff (clinical and non-clinical). 96 Conversely, lack of understanding of the integrated care model and the roles of other professionals within it prevented MDTs from surmounting organisational barriers. 88 Training programmes can increase such understanding and address the scope of practice for each profession, liability and confidentiality issues. Interviews with people in ‘successful’ integrated care models in the USA suggested that people in boundary-spanning roles need to be able to be assertive when necessary, to understand practice culture in its setting and to maintain good relationships with everyone caring for the patient. 96 Clinicians needed to be aware of their own limits of expertise and of the skills and limits of each professional, and to consult and refer when a clinical problem was beyond their scope. 88 A mixed-methods study86 of 18 complex care management organisations in the USA found that educating providers about the roles and responsibilities of care managers and providing complementary services that fill patient care gaps helped generate trust and support within MDTs. 86

There was a fragile balance between resistance to including new disciplines (e.g. pharmacists) as MDT members and acknowledging the need for them. In one study,95 some doctors expressed concerns about having pharmacists challenge their prescribing decisions directly or overstepping their professional boundaries, while others valued having pharmacists work with them as team members and saw them as a critical piece of a patient-centred medical team. Single-handed doctors, doctors not affiliated to physician networks or doctors who had never worked with clinically trained pharmacists in primary care had more difficulty envisaging collaborations with pharmacists than doctors in group practices or a hospital–physician network, who had previous working experience with clinical pharmacists. 95 In general, the studies we found suggested that it was necessary to work around or weaken defensive professional perceptions of other professionals;89 and around doctors’ and patients’ resistance to boundary-spanners cultivating cross-professional and cross-organisational relationships. 87

Causal link 3:9 – multidisciplinary teams produce health planning and better preventative care

A narrative case study of an integrated mental health service in the USA66 found that MDTs allow better territorial planning of health as a whole regarding:

• the health needs of the whole population

• reducing ‘medical deserts’

• organising support networks which promote preventative and ambulatory care offering medico-social work, social care and at need hospital care. 66

Coleman and Phillips97 created a ‘teamness’ index based on whether or not non-physicians shared responsibility for managing and co-ordinating care. Practices that scored high in ‘teamness’ were more likely than low-scoring practices to report well-functioning processes to support communication and access to care, and to connect chronically ill patients to self-management programmes. 98 Hong’s mixed-methods study86 of 18 American complex care management organisations found that care co-ordinators negotiating a care plan that reflected the individual patient’s, and their family’s, priorities and preferences facilitated various actions including identifying patients’ behavioural health and social service needs, and using motivational interviewing to encourage patient activation and self-management.

A Canadian survey99 of adult patients and administrators found that MDT working produced better preventative care through better first-contact accessibility (FCA) and accessibility accommodation (AA), which increased equity of access to such services. AA was the way PHC resources were organised to accommodate a wide range of patients’ abilities to contact health-care clinicians and reach health-care services. FCA was the ease with which a person could obtain required care (including advice and support) from the practitioner of choice within a time frame appropriate to the urgency of the problem. Carroll et al. 99 found that FCA was better in clinics with ≤ 10 doctors, a nurse, telephone access 24 hours a day and 7 days a week, and evening walk-in services.

Further US studies corroborated the finding that integrating community and/or mental health professionals into MDTs improved preventative care. Matiz et al. 94 found that doing so made care delivery more comprehensive and identified high-risk populations for care co-ordination. Such organisations had decreased emergency department (ED) utilisation and hospitalisations for asthma, resulting in overall improved outcomes. 94 Briot et al. 66 found that mental health professional integration offered good-quality ambulatory care to more patients at a lower cost and better managed complex family health problems66 than traditional forms of organisation did. Similarly, in a descriptive quantitative study100 including pharmacists in PCMHs allowed screening of diabetes mellitus and hypertension patients, care reviews, inclusion/exclusion decisions and provision of preventative pharmaceuticals.

Causal link 3:9 – patient engagement, patient self-care, activation and empowerment

Evidence from two qualitative studies of PCMHs87,88 suggested that mechanisms by which MDTs improved patient engagement were care co-ordinator roles and capitalising on the primary care relationship. An online focus group of care co-ordinators in PCMHs reported improving engagement of patients by using motivational interviewing, being patient but persistent, keeping promises, listening carefully, using humour, sharing personal anecdotes and earning trust with small gestures so that larger problems could be tackled later. 87 In interviews with five medical and seven mental health clinicians in PCMHs, Rajala88 found that capitalising on a patient’s relationship with their primary care office to connect them with mental health services was one of the largest factors in increasing patient engagement and access to mental health care. 88 By making their health care more coproductive, a US learning network (ImproveCareNow) of patients and health workers increased remission rates from 60% to 79% for children and adolescents with irritable bowel disease. 101

Six qualitative studies (four of PCMHs,87,91,93,94 one of mental health services in the USA66 and one of a geriatric network in France74) evidenced how MDT working produced patient self-care, activation and empowerment through social prescribing, integrating community and mental health into primary care teams and better informed physicians.

Social prescribing appeared to be more acceptable to patients than other prescriptions in the retrospective observational study of the EPSILON geriatric network in Versailles,74 with compliance rates of 72% for medical prescriptions, 74% for paramedical prescriptions and 100% for social prescribing. Focus groups with 387 people from 10 US communities found that patients appreciated PCMH models that included access to education, social and support resources (e.g. nutritionists, smoking cessation classes, exercise and fitness programmes, weight loss classes, meditation, counselling services, religious groups, peer-support groups) to help patients manage their care better. 91

Four qualitative studies in the USA also provided some evidence that integrating community and mental health workers into primary care teams produces better patient self-care, activation and empowerment. For mental health services, Briot et al. 66 found that integrating mental health professionals into the MDT promoted families’ capacity to mobilise themselves if a family member was in distress. A case study found that integrating CHWs into PCMHs is a means of providing support and education to hundreds of patients. 94 Furthermore, in the USA, Collinsworth et al. 83 found that these workers improved patient knowledge and activation levels, primary care providers’ ability to identify and address specific patient needs, and patient outcomes. Preventative care improved if a MCP-like model of primary care enabled CHWs to undertake disease/illness education, nutritional counselling, patient follow-up; to identify patient barriers to care or self-care, patient activation, social and self-management support (e.g. for diabetes mellitus control); to link patients to community resources; and to co-ordinate care. 83 These boundary-spanning roles directly facilitated patient activation through trust, cultural understanding, common language, manageable goals and a team approach and availability. They did so indirectly by making primary care clinicians more informed about patient goals and barriers and preparing patients more for meeting primary care clinicians. 83 Another study corroborated that MDTs improved patient confidence through making doctors better informed. In interviews with people working in PCMHs, Grace et al. 93 found that routine structured communication facilitated continuity of care and improved co-ordination among team members, which made physicians better informed on the status of shared patients. Well-informed physicians communicated more effectively with patients and increased patient confidence, trust and satisfaction. 93

Causal link 3:9 – contexts for patient engagement through multidisciplinary teams

Studies showed that patients’ own responses were a context determining whether or not MDTs succeed in promoting preventative care. At times, the expectation of greater involvement in their care could be a barrier to patient engagement and could create discomfort for them. The online focus group of care co-ordinators reported by Friedman et al. 87 highlighted patients’ lack of trust, insufficient understanding of the care co-ordinator’s role and inability to take responsibility for self-management of chronic conditions as barriers to improving patient self-care, activation and empowerment. Some patients who agreed to work with care co-ordinators continued to call multiple people in the clinic and attend the ED for needs best treated in the clinic. They ‘technically have a [care co-ordinator] but they continue to have fragmented care’. 87 Patients could feel scared to express their views in front of a MDT and, when asked, might interpret this as an admission from the MDT that they do not know what they are doing: ‘it’s tricky, you know – [clinician] was trying to be patient-centred, but [patient] didn’t have a context for it’. 88 When Rajala88 interviewed five medical and seven mental health clinicians in PCMHs, she found that patients were often surprised when a mental health provider was invited into their appointment, but typically came to appreciate it, for instance when mental health clinicians were introduced in terms of how they could help treat the patient’s particular symptoms. 88 However, some patients experienced integrated care as a loss of control over their information. 88

Our review also discovered evidence for additional causal links (Table 5) to those in the IPT in which MDTs are the mechanism to create change in other MCP components.

Causal link 3:4 – multidisciplinary team working produces culture change in the health system

One Australian study62 concluded that MDT working has the potential to change the culture of the health-care system. This qualitative study62 of a chronic aged and complex care service model found that the creation or expansion of roles to work across traditional boundaries between other members of the primary care team instigated or enabled system-wide culture change through improving communication (formal and informal) between the various care providers.

Causal link 3:5 – multidisciplinary team supports voluntary involvement

Just one study76 suggested that MDTs encourage family and carer support for patient care. In five instances of MDT initiatives in the UK, enhanced access strategies of telephone MDT, Hospital at Home, and Advice and Guidance services produced better patient experience and less inconvenience and disruption for patients and families, and gave them extra skills and confidence to look after their unwell child without professional support. 76

Causal link 3:6 – multidisciplinary team working produces informational continuity of care

One quantitative descriptive research study100 of pharmacist recommendations and physician responses related to 954 complex patients in a PCMH found that MDT working produced better use of EHRs and electronic communication. However, for remote electronic communication to be successful, face-to-face contact was also needed to build the relationships required. 100

Causal link 3:8 – multidisciplinary teams produce better demand management systems

Three case studies in the UK76 and the USA66,102 provided evidence that MDTs could strengthen demand management systems and redistribute workload pressures across the care system. A multiple case study of five NHS vertical integration projects for paediatric/young persons’ services76 found that MDT working produced better gate-keeping and need- and/or risk-stratification. GPs with access to advice and guidance from a consultant developed specialist expertise and could manage more complex cases without referring to secondary care, thus easing the workload there. 76 A study of the American Veterans Health Administration (VHA)102 found that nurse visits in primary care were associated with a decreased risk of all-cause hospitalisation for veterans > 65 years of age. The Briot et al. 66 case study of MHI in Utah66 found evidence that MDT working redistributed workloads. When consultations were multidisciplinary, health professionals jointly put into effect care strategies individualised and co-ordinated (through a case manager) for the user and their family, using the family’s own health and social networks. That gave users good overall care by a better team at lower cost, reduced GP workload and freed specialists to support more severe cases. 66

Causal link 3:10 – multidisciplinary teams produce care planning at the patient level

Ten studies21,66,68,76,78,83,90,91,93,94 suggested that MDT working facilitated care planning at the patient level through the operation of boundary-spanning roles and by giving greater access to enhanced primary care.

Causal link 3:10 – care planning and boundary spanning

Nine studies21,66,68,78,83,90,91,93,94 provided further evidence that boundary-spanning roles facilitated many forms of care ‘integration’. These roles may be filled by care co-ordinators, nurse practitioners, CHWs and many other occupations. These roles increased awareness and use of care plans in the MDT, organising access to types of care that patients need and desire. 66,78,83,90,91 Alidina et al. 68 found that high performing PCMHs typically had at least one dedicated care co-ordinator position. Lower-performing PCMHs typically had none (care co-ordination responsibilities were shared between staff).

Several studies21,68,78,83,90,93,94 reported how MDT members in boundary-spanning roles helped co-ordinate the MDT to effect individualised care strategies co-ordinated around the patient and her family, make use of a health and social network to provide education for patients and their families, and put their counsellors at their disposal, providing the patient with good overall care, at the right moment, by a better team, at lower cost. Briot et al. 66 describe this in mental health services in Utah. Similarly, for physical health, boundary-spanning MDT members (e.g. embedded CHWs) with close contact with patients found what barriers to treatment patients faced, and communicated these barriers to other MDT members who could then work with patients to overcome them68,78,83,94 and increase the use of care plans (from < 5% to 39%94). Although the GPs in a study in Australia did not have the time or resources to deal with psychosocial aspects of patients’ health, the general practice liaison nurses were able to arrange case conferences between all necessary professionals and develop care plans for patients. 90 A case study of Buurtzorg21 found that, by working in this way, a MDT was able to deliver more person-centred care by allowing staff to organise care that made sense to them and the patient, which made them feel able to deliver good-quality, holistic care and allowed the MDT to organise itself so as to achieve the best possible outcomes for patients.

In the HealthOne Mount Druitt project (Australia), McNab et al. 90 found that primary care providers appreciated the familiar face and voice of the boundary-spanner, with whom they felt they could, over time, build an ongoing relationship of mutual trust. The boundary-spanners’ local knowledge of services and time to liaise with them benefited the GPs because it allowed more efficient and effective liaison than they could themselves provide and made a huge difference to service provision and support for chronically ill patients. 90 Many primary care professionals in a US study acknowledged spending more time co-ordinating care for patients before these roles were implemented, and saw the time savings as allowing them to communicate more effectively with patients. 93

Causal link 3:10 – multidisciplinary team working gives patients access to a wider range of primary care services

In their case studies of five integrated care initiatives, Woodman et al. 76 described how enhanced access strategies used by MDTs improved patient care. If MDTs discussed complex cases at high risk of needing secondary care by telephone each month, GPs became more motivated and confident to manage these patients, gaining skills and access to specialist support to do so, so that patients received higher-quality care from their GP. MDT members better understood their colleagues and service thresholds, established professional relationships and shared norms. Families perceived and patients experienced a more ‘joined-up’ health-care service, trusted the care they received from the GP, felt motivated to seek help from primary care, became confident in managing their own chronic conditions, experienced fewer exacerbations of chronic illness and experienced less inconvenience and disruption. 76 In a qualitative study in Australia, people working in new boundary-spanning roles were found to make a broader range of services available to patients through case conferencing, care planning, liaison and information provision, and being a single point of contact for GPs to access all the other services and professionals in the community. 90

Causal link 3:10 – contexts for multidisciplinary teams facilitating patient-level care planning

In the above studies, the most important context is case-mix. MDTs are particularly necessary for stimulating the use of individual care plans when patients have complex conditions about which clinicians have a lower level of knowledge than for more common conditions and for which reciprocal co-ordination of treatments is necessary. 68

Other contexts were similar to those facilitating MDTs in undertaking network-level care planning. One interview study of embedding CHWs in a CCM in the USA highlighted the importance of other team members’ trust in care co-ordinators as a context supporting better care co-ordination, but also noted that it could take a year of working together to establish this trust. 83 Primary care doctors said that they gained trust in the CHWs as they recognised their many competencies and saw the positive impacts that were had on patients. After recognising their value, these doctors sought to provide the CHWs embedded in a CCM with ‘plenty of support’ to address patients’ clinical needs and helping them to deal with challenging situations. 83 A qualitative study of five PCMH pilot sites93 found that primary care doctors did not always value the boundary-spanning roles. Some reported only ad hoc meetings with the boundary spanners to discuss specific complex cases, ambiguity about the appropriate tasks to delegate to them and indicated that more structured communication was needed. 93 Another study103 provided evidence that trust in the sense of willingness to delegate work within the MDT was another aspect of this context:

When we first started putting care co-ordinators in the offices, we got pushback from the doctors that we were taking away some of the things they do. But after they got familiar with it and realised that these aren’t things that you really need a medical degree for and it actually means that the minutes I’m in the room with the patient I can talk to the patient about their health, they were OK with it.

ACO interview, Shortell et al. 103

Used with permission of SAGE publications, Inc.

Limited evidence from two studies in the USA78,86 suggested that training of staff working in MDTs and care co-ordination roles supported patient-centred care. Hong et al. 86 described how ‘successful’ CCMs offered customised training, including didactic experiences, mentoring and shadowing. A uniform training and education platform for all new and existing care co-ordinators, irrespective of profession, was found in another large ACO to ensure a consistent approach to providing care co-ordination services to patients. Training, together with recruitment difficulties and the retention and cost of care co-ordinators were other barriers to MDTs’ care co-ordination work. 68

Causal link 3:11 – multidisciplinary teams divert patients from hospital to primary care

Two SRs104,105 and one quantitative study106 found that MDT working reduces hospital readmission rates. Two104,106 of these studies described the importance of specialist involvement in the MDT, and two105,106 described that of care co-ordinators. An umbrella review104 (a SR of SRs) of case management found that the CCM, discharge management, complex interventions, patient self-management and MDTs, particularly when they focused on one specific health condition [in particular heart failure and chronic obstructive pulmonary disease (COPD)] and included condition-specific specialists (medical, nursing, pharmacist), together decreased emergency admissions. Half of the SRs quantified the reductions, giving figures ranging from 25% to 43%. 104

In a quantitative study of PACT PCMH implementation by the US VHA, Nelson et al. 106 found that greater continuity of care (i.e. all other providers all working with and communicating with patient’s primary care provider) was associated with lower likelihood of hospitalisation and mortality. Nurse visits in primary care were associated with a decreased risk of all-cause hospitalisation for veterans > 65 years of age. 106 As less direct evidence, a SR105 of RCTs of transitional care interventions that aimed to improve care transitions from hospital to home and to reduce hospital readmissions for chronically ill patients found that a home visit within 3 days, care co-ordination by a nurse and communication between the hospital and the primary care provider were significantly associated with reduced short-term readmission rates. 105 Kinjo et al. 107 describes zaitaku primary care MDTs, as yet on a small scale, replacing hospital end-of-life care in Japan.

Causal link 3:11 – multidisciplinary team working diverts patients from inpatient to primary care services

Three studies21,66,103 found that, where MDTs enabled the flexible mobilisation of a range of professional expertise, training and knowledge, including from community providers, care was more centred around the patient’s goals and needs. In a historical narrative case study of a programme for clinical integration of mental health specialists with community primary care medicine, Briot et al. 66 found that the use of the MDT members adapted according to the severity and complexity of the pathology in order to cosupport in a scalable way. MDT care adapted flexibly to the service users’ mental and physical health, family circumstances, medical and social comorbidities and fed into the provision of specialised care. 66 A similar example was Buurtzorg (the Netherlands), which used self-managed teams to produce and plan patient care, with teams of 8–12 nurses and nurse assistants covering a geographical patch that they themselves choose. 21 A mixed-methods study of 11 purposively sampled ACOs in the USA103 provided another example of how flexible mobilisation of community resources by a MDT supported patient-centred care and, thus, reduced demands on hospitals: a physician-led ACO network in the north east of the USA used an interdisciplinary care team to work with patients with complex needs. One was a patient:

. . . who went 132 times in 12 months to the emergency department. She is . . . in a wheelchair . . . lives in a house with no ramp. She doesn’t have much social support, doesn’t have any food. A diabetic, out of control. She doesn’t have a refrigerator for insulin. From one visit, we engaged our team of care management [who] . . . built her a ramp, donated a refrigerator, and hooked her up to an equivalent of Meals on Wheels so she has food, and arranged for transportation to get her to regular visits to her primary care physician. And in the past ten months . . . she’s not been back [to the ER] one time.

Shortell et al. 103

Used with permission of SAGE publications, Inc.

Causal link 3:12 – multidisciplinary team improves patient experience

Our secondary evidence suggested that MDTs that included pharmacists, nurses and CHWs can improve patient experience, outcomes and continuity of care. We found some evidence that MDT working improves patient experience through boundary-spanning roles (see Causal link 3:7 – boundary-spanning roles), enhanced access to primary care (see Causal link 3:10 – multidisciplinary team working gives patients access to a wider range of primary care services), better communication between providers and, thus, more patient confidence, trust in, and satisfaction with, care. Twenty-eight patients and informal caregivers and 20 health-care providers in community-based PHC in Canada described MDTs as providing a holistic care experience to their patients. 108

A virtual MDT (team members linked remotely by telephone or HIT) and hospital-at-home schemes produced better patient experience with less inconvenience and disruption for the patient and family receiving paediatric health care. 76 In ‘successful’ US primary care–integrated complex care management programmes, the MDTs’ key role was to build trusting relationships between patients and families, and primary care providers and their staff. 86 Routine structured communication in MDTs facilitated continuity of care and co-ordination so that the doctors were better informed on the patients’ status and thus communicated more effectively with patients, which increased patients’ confidence, trust, and satisfaction. 93 In another study, patients indicated that the boundary-spanners were able to bridge the gap between them and the doctors by talking to them on a level they understood, understanding cultural barriers and patiently answering questions. 83 Conversely, a sample of US patients said that a lack of boundary-spanning MDT members tended to leave patients lacking understanding about what was going on with their care, feeling left out of the dialogue and decision-making and feeling vulnerable as a result of their uncertainty. 91 In the Australian HealthOne Mount Druitt project, boundary-spanner care co-ordinators made patients feel more supported and less anxious and thus reduced hospital visits. 90

A reanalysis of administrative data in PCMHs and ACOs that involved pharmacists80 found that pharmacists identified 708 drug therapy problems through direct patient care (336/708; 47.5%), population-based strategies (276/708; 38.9%) and education (96/708; 13.6%). Pharmacists combining academic detailing with direct patient care and population-based medication management probably helped to optimise patient outcomes. 80 Woodman et al. 76 found that UK nurses making home visits in a Hospital at Home team improved child safeguarding and heightened awareness and paediatric referral to all community nursing services. In two cases, informants reported that commissioners and providers had warned of potential harm to children. 76 Similarly, in a study of PCMH implementation by the American VHA, nurse visits in primary care were associated with greater continuity of care and lower mortality rates among a patient cohort. 106 Interviews with CHWs, patients and primary care providers in CCMs found that CHWs facilitate trust, communication, understanding of roles and PCP support, leading to such patient outcomes as improved glycated haemoglobin (HbA_1c) control. 83

Causal link 4:3 – if culture changes occur in the participating organisations, then multidisciplinary teams will develop

The programme theory first assumed that a shift in the culture of care delivery organisations and professions would include shifts in their assumptions about desirable models of care, interorganisational and interprofessional working practices, all of which would produce workforce development and engagement in ways that promoted the development of MDTs.

Causal link 4:3 – workforce development and engagement

Two studies provided evidence that culture change supports different professionals to work together across disciplinary boundaries. Greene et al. 89 conducted qualitative interviews with, and a survey of, providers and staff in mental health and paediatric PCPs in the USA and found that culture change enabled new ways of working and communicating that dismantled a key barrier to collaboration, including improving shared expectations, increasing awareness of what other professionals within the wider care team have to offer and building better understanding of the culture of other professions. 89 Conversely, Bergman et al. 92 interviewed key informants in PCMH and team-based care models and found that those working in more traditional roles can feel defensive around their boundaries and roles and that their expertise or specialism is under threat. 92 Interviews with five medical and seven mental health workers (PCMH, USA) showed that the latter could also be culturally resistant to practising in an integrated model. 88

People working in new boundary-spanning roles may attempt to prevent other team members from feeling threatened by their recommendations or opinions by using indirect, non-threatening forms of communication such as gentle hints, suggestions and questions like ‘are you sure that’s really what you wanted?’. These indirect communications risked important information not being effectively communicated in safety-critical situations. 92 Bergman et al. 92 concluded that reducing these risks and helping the new roles become a driver for culture change could occur by:

• creating a culture of openness (feeling comfortable speaking up to reduce error when problems are suspected) through training to improve communication across hierarchies, for example the Crew Resource Management training adopted in some US medical and pharmacy training programmes109–111

• agreeing, at the outset of their collaboration, clear (e.g. written) scopes of practice between different professions, to cultivate awareness and shared expectations of each other’s duties and responsibilities. 92

Two studies found that respect could overcome or bypass the perceived threat of new boundary-spanning roles. McNab et al. 90 found that other members of certain primary care teams came to respect people in boundary-spanning roles when they saw the latter changing culture. That respect enabled further culture change by supporting formal and informal communication between the various clinicians. 90 Bergman et al. 92 provided an example: when primary care doctors working with pharmacists in new expanded roles were exposed to situations in which their opinions conflict, they came to recognise that the pharmacists were ‘usually right’ (about pharmacy-related matters), learned to respect them and see value in their expanded role, which facilitated multidisciplinary working. 92 Producing trusting working relationships between primary health-care doctors and people in boundary-spanning roles has been found to take around 1 year. 83

Causal link 4:3 – ‘joined-up’ working

A web-based survey of ACOs, in the USA, found that shared culture was necessary for their success. 58

Two studies offered evidence about how to create culture change so as to improve primary care teams’ integration. In their qualitative interviews with, and a survey of, staff in mental health and paediatric PCPs in the USA, Greene et al. 89 found that shifting shared expectations, improving awareness of other professionals’ roles in the primary care team and understanding the culture of other professions enabled ways of working and communicating to change, dismantling a key barrier to collaboration. 89 The Weldon et al. 112 study found that sequential sequencing workshops, in which workshop participants’ experienced and then discussed in groups ‘real world’ examples of their role within the health-care system and how what they did impacted on collaborative (person-centred co-ordinated) care, improved staff knowledge and understanding of the impact on collaborative care. In one workshop with GP receptionists in the UK, a new professional structure for GP receptionists appeared to be emerging, with receptionists empowered to see the importance of their role within the wider context of health-care system, as well as how crucial they were for integrated care to work. 112

McNab et al. 90 found that in systems in which there is no system-wide culture change in support of integrated multidisciplinary working across teams embedded in the partner organisations and established throughout the primary health-care sector to support integrated MDT working, there remained a heavy dependence on leadership from the GPs and CHWs on the network steering committee. 90

Causal link 4:3 – contexts for culture change producing multidisciplinary teams

The same study found that one way to support the above culture changes across professions was by creating boundary-spanning roles that themselves instigated or enabled system-wide culture change by improving both formal and informal communication between the various care providers. 90 Wholey et al. 67 argued that tasks are the functions that a team has to perform to achieve its goals (e.g. care co-ordination) and so they, and not culture, are the logical starting point for MDT design. Chapter 7 considers this apparent contradiction more closely.

Interviews and a survey in 13 PCMH practices in the USA found that existing cultures of individual excellence, individual accountability and established practice norms were an obstacle to collaboration. 68

Causal link 4:8 – if culture changes, then demand management systems develop

We found no published research about whether or not or how culture change in an integrated model of care makes demand management systems develop.

Causal link 4:9 – if culture changes, then preventative care develops

We found a little evidence from one study that culture change increased preventative care. In a SR, the creation of a non-intimidating environment/culture was reported to be an enabler of improvements in patient knowledge, self-care behaviour and self-efficacy. Busetto et al. 113 reported that a CHC collaborative could not have led to increased patient self-management without changing the health centre philosophy towards more patient-centredness and empowerment. However, another study suggested that other, less resource-intensive mechanisms for improving prevention may be more acceptable and feasible. 79 None of these studies described what contexts (in the realist sense of the term) were required.

Beyond the IPT, the secondary literature reported further ways in which culture change might be a mechanism for creating change in other MCP components (Table 7), although none of the studies found stated what contexts (in the realist sense) were required.

Causal link 4:7 – if culture changes occur in the participating organisations, then planned referral networks develop

As noted, the Australian study by McNab et al. 90 found that culture change itself resulted in part from introducing boundary-spanning work roles, but also that a wider culture change was needed to ensure that practices and processes are embedded in the member organisations of a MCP-like network, including, by implication, any interorganisational referral networks. Two further studies88,89 reported the particular need, in setting up ACOs, to have a cross-cultural dialogue between medical and mental health providers.

Causal link 4:12 – if culture changes occur in the participating organisations, then there will be better patient experience and outcomes

Demiris and Kneale’s85 narrative literature review found that implementation of patient-centred care depended on culture change in health-care organisations and among health-care consumers. In their SR, Busetto et al. 114 reported a study by Borgermans et al. in which the presence of interdisciplinary diabetes mellitus care teams was associated with significant improvements in HbA_1C and low-density lipoproprotein (LDL)-cholesterol levels, and increased statin and anti-platelet therapy use, which were attributed to the quality and task orientation of the teams, shared leadership and shared group norms. Busetto et al. 114 also reported a study in which Yu and Beresford found three critical success factors for their chronic illness model that led to improvements in HbA_1c, blood pressure, LDL and urine albumin-to-creatinine ratio, namely (1) leadership commitment to change, (2) increased clinical staff involvement and (3) residents acting as change agents. They also found that the same shift in the culture produced a non-intimidating environment that facilitated better-co-ordinated patient-centred care and improved health workers’ mental health and well-being. 114

Causal link 5:8 – if there is voluntary sector involvement in multispecialty community providers, then demand management systems will develop

Noël et al. 79 found that community linkages are utilised less often than the other components of the CCM. Bodenheimer et al. (2002; reported in Lafortune et al. 108) described linkages between clinical settings and community health resources as highly important, particularly for health-care professionals who are not operating as part of a large team-based organisation and for those treating patients with chronic illness (and, we add, may, in the USA, have difficulty obtaining health insurance). These apart, we found no studies reporting how, or even if, voluntary sector involvement in MCP-like networks helps them manage the demand either for hospital services or for formal primary care services, carers or voluntary organisations.

Causal link 5:9 – if the voluntary sector becomes involved in multispecialty community providers, then preventative care will develop

Although the MCP programme theory emphasises access to a wide range of resources around a person’s goals, studies reporting whether or not and how voluntary sector involvement strengthens preventative care were sparse. From focus groups with 387 participants in 10 US communities, Mead et al. 91 described the barriers to involving the voluntary sector in the PCMH model. Despite patients reporting a need for community resources, such as education classes, diet and exercise groups, and peer support groups to provide additional support to help them deal with the burdens of managing chronic illness, the PCMH model was limited to formal services within the health-care system and lacked having processes to support, pay for or even refer patients to resources outside the health-care system that could be useful for their health. Hence, this paper suggests that providers who treat disadvantaged populations need training to develop relationships with service providers that will take on low-income, uninsured or underinsured patients, and to be innovative. Understanding each patient’s personal constraints and not just the typical medical history is a critical aspect of patient-centred care but is not highlighted as a key component of the PCMH model. Participants highlighted the importance of religious organisations and community-based organisations and cited several examples of how these resources provided important support in the managing their overall health and well-being. 91 A qualitative study of community-based PHC in Canada found that self-management support groups and resources allowed patients to be more engaged in maintaining their own health and helped to prepare them for discharge or care transitions. 108 Neither study reported what contexts (in realist terms) favour voluntary sector involvement.

Causal link 5:12 – if voluntary sector becomes involved in multispecialty community providers, then patient outcomes improve

We found little evidence consistent with the IPT that involving the voluntary sector in MCP-like networks might improve patient outcomes. In comparative case studies of a German scheme (Kinzigtal), a Netherlands-wide programme (one care group) and 16 English pilot schemes, Busse and Stahl64 report that, in the Kinzigtal care model, multisectoral collaboration had, after 2.5 years, reduced mortality rates by half (from 3.74% to 1.76%) for those enrolled in the programme compared with those who were not. Although the network had voluntary sector input, these results are attributed to the network as a whole, leaving it uncertain whether or not the voluntary sector input contributed to these mortality improvements and, if so, to what extent and through what mechanisms and contexts. 64

Causal link 6.7 – if health information technology is used to strengthen informational continuity of care, then planned referral networks will develop

Many MCP-like organisations used HIT effectively. A national survey of US physicians found that EHR use in ACO or PCMH settings was associated with increased activity in health management at population level, quality measurement, patient communication and care co-ordination. Two other studies (reported in King et al. 115) found that PCMH doctors who used EHRs had ‘greater quality improvements and changes in utilisation over time on some measures’. One SR (Fontaine et al. , reported by Lafortune108) found evidence that electronic health systems could function as a way to improve patient safety, reduce medical errors, improve access to data and decrease staff time spent on administrative tasks. Through semistructured interviews with physicians in PCMHs, Petersen et al. 82 found that well-designed EHRs allowed them to better co-ordinate care and share information. Links with hospitals were also important. A study of six ACOs found that timely, consistent information about patients’ admissions and discharge enabled the planning of follow-up services that patients might need within 30 days of discharge. 63

Two additional studies described individual projects that effectively used HIT to co-ordinate care in MCP-like contexts. The Gesundes Kinzigtal project reported improved patient and health-worker experience, and reduced costs and mortality. The project relied on sharing an EHR system across providers to co-ordinate care. 64 Buutzorg used a simple, web-based solution designed by nursing assistants, nurses and back-office employees to communicate and share information in real time between locations such as the patient’s home, in the office or on the road. 21

However, in many cases, HIT systems that had not been carefully designed and implemented hindered health professionals in communicating and sharing information. Recurrently reported barriers to effective HIT implementation included lack of interoperability between HIT systems [see Causal link 6.7 – lack of interoperability between health information technology systems (both within and between organisations)], lack of necessary data analysis tools (see Causal link 6.7 – lack of necessary data analysis tools), lack of workflow tools (see Causal link 6.7 – lack of workflow tools) and the limitations of current technology (see Causal link 6.7 – limitations of current technology).

Causal link 6.7 – lack of interoperability between health information technology systems (both within and between organisations)

Almost every study discussed the importance of HIT connectivity both within and between provider organisations. Participants across different studies emphasised the importance of using a common health information system between services or redesigning systems so that they communicated with one another. 88,98 In many cases, HIT systems within an organisation were flawed. Two studies described how care managers needed to use a completely separate system from physicians, resulting in clunky ad hoc systems to collect and manage data. 87,116

Causal link 6.7 – lack of necessary data analysis tools

Research participants across studies lamented the inability of their IT systems to do basic data analyses such as risk-stratifying patients, tracking subpopulations of patients, determining which patients need follow-up, generating relevant reports and tracking hospitalisations. 116

Causal link 6.7 – lack of workflow tools

Many studies78,108,116,117 reported that health workers wanted an IT system that would track patients more effectively. A number of studies mentioned such tools as task management systems, care planning systems, standardised care pathway templates for physicians, notification systems for changes in patient status and individual patient tracking through the health-care system. A recurring frustration was an inability to get the right information at the right time, which resulted in participants assembling ad hoc systems to piece together different software systems to generate needed reports. Richardson et al. 117 describes the ‘shadow system’ of data captured through ‘homegrown’ methods, which was often used when EHRs failed to adequately meet an organisation’s needs. 87

Many organisations reported difficulty in extracting and piecing together data even from EHRs that complied with the continuity of care document standard for interoperability, suggesting that these standards may be insufficient for MCPs’ needs. 117

Causal link 6.7 – limitations of current technology

Two studies discussed the limitations of current HIT for MCP purposes. Bauer et al. 118 noted that traditional HIT tools were not built to monitor populations and subpopulations of patients, actively flag patients for follow-up or to respond to real-time data on patient progress. Rudin and Bates (reported in Richardson et al. 117) concluded that the current HIT marketplace ‘has failed to provide adequate solutions’ for care co-ordination. Another study68 noted that PCMHs tended to use IT systems for more straightforward uses, but more complex patients were dealt with offline owing to underdeveloped technologies. Finally, one study119 found that having an EMR did not automatically improve care co-ordination. These two studies suggested that previous generations of EHRs may not be suitable for new models of care, as the difficulties that many organisations have faced in implementing them symptomatised.

Many studies in our review corroborated that integrated IT systems alone would not lead to co-ordinated care systems. Other mechanisms, such as reworking staff roles and a shared physical space, were also likely to be required.

Several studies specified staff attitudes and skills that were important for a successful EHR. One SR114 found that personal barriers to integrated care interventions included staff reluctance to use HIT, unawareness of system features, unwillingness to share data and lack of IT skills. Another paper92 implied that the structure of the EHR inadvertently made it a battleground between physicians and pharmacists. The same study92 found that 68% of clinical pharmacists who were surveyed in the PCMH context referred to examples of problems with electronic communication in their relationships with primary care physicians.

Many studies emphasised the importance of task delegation, workflows and routines. One study116 found that primary care teams that used EHRs consistently for data entry and agreed on communication methods between staff members were more likely to score high on the National Committee for Quality Assurance (NCQA) 2011 PCMH recognition tool. Best practice in the use of EHRs to facilitate communication between staff members included access to patient information for all staff members, instant messaging, within-chart notes, telephone templates that could be routed to team members’ inboxes, task assignments and ‘huddle sheets’ for the day embedded in the EHR. 116

Finally, care managers emerged as important facilitators of effective HIT use. Morton et al. 120 found that practices with a non-clinician member of staff who was responsible for co-ordinating care were much better at care co-ordinating activities and conducting these activities electronically. Other studies118 also noted that the care manager provided much-needed support to ensure smooth operations.

Causal link 6:7 – contexts

Overall, the evidence suggested that HIT systems can support communication and data sharing between health professionals at MCPs, but only provided that these HIT systems be designed and implemented with care. Otherwise, they risked being a barrier to effective MCP working.

Causal link 6:10 – informational continuity of care produces care planning at the patient level

The evidence from our review supported the assumption that EHR systems, when set up to support co-ordinated care processes, can improve patient outcomes. Two different studies85,114 reported in SRs found that effective EHR use enabled teams to increase quality of care for diabetes mellitus patients. Several studies85,118,121 have also found that electronic patient registries can improve patient quality outcomes.

Across studies, participants agreed on features that increased the effectiveness of EHRs for patient care. A recurring theme was the importance of using the EHR to guide physician practice and workflow, and provide reminders for actions. One practice used 200 different symptom-specific templates. The template system increased productivity and allowed physicians to focus better on patient needs during their appointment. 116 Xenakis’s study78 corroborated this, and participants in other studies108 lamented the lack of a template system in their EHR.

Other EHR features that study participants repeatedly requested included the ability to create care plans (recording goals, barriers and specific steps to that goal) and notification systems to help staff engage patients when the patient’s status changed (e.g. following hospital admission, no-show at follow-up appointments). 116,117 Yet, despite widespread agreement about the characteristics of an ideal EHR, we found only limited evidence of its ability to improve patient care, probably because many provider organisations did not yet have the requisite features for its optimal use.

Many studies described patient-facing electronic tools. One SR85 found very little evidence that an electronic personal health record (accessible by patients) increased care outcomes, care co-ordination or patient engagement. By contrast, another found that patients had very positive responses to a patient portal. 103 These contradictory findings may be explained by the slow uptake of personal health records, and a lack of studies connecting personal health records to patient outcomes. 85 Another study (reported in Bauer et al. 118) described technology-enabled delivery of mental health interventions, such as mobile devices assisting self-management. However, the same study cautioned that patient-facing tools are most effective when combined with a relationship with a health worker, such as a counsellor. Using technology to build a relationship can provide more accountability and support patient engagement, whereas stand-alone interventions require patients to be much more self-motivated. 118

Overall, there is evidence that EHRs can improve patient outcomes, but only when they include robust functionality such as care planning and tracking population- and individual-level data over time. However, HIT use alone does not guarantee improved care. Instead, HIT, whether an EHR or patient-facing tools, must be carefully designed to complement interpersonal relationships.

Causal link 6:11 – informational continuity of care helps divert patients from hospital to primary care

Our review found mixed evidence for the assumption that effective use of data in MCPs can lead to reduced unnecessary A&E admissions. Kaushal et al. 122 found no difference in ED visits and hospital admissions, or hospital readmissions between PCMH and non-PCMH settings over a 3-year study period. However, other studies found mixed or inconclusive evidence. Two studies reported by Demiris and Kneale85 found opposing results for emergency admission rates for home telehealth programmes but not in MCP-like settings. Two additional studies found weak evidence. One survey56 found that ACOs were slightly more likely to track inappropriate ED use than their non-ACO counterparts, whereas another123 concluded that PCMHs with patient registries have the potential to use data to reduce unnecessary ED admissions. The same study recommended network analysis for tracking patients’ movement between providers so that care and resources can be better co-ordinated, possibly leading to reduced admissions.

This review also found evidence for additional outcomes of HIT beyond those in the IPT (Table 10).

Causal link 6:3 – information continuity of care produces multidisciplinary team working

Although most research on MDT working focused on face-to-face meetings rather than virtual communication,85 several studies noted that well-designed HIT can support effective communication both within organisations and across service providers. 73

We found examples of HIT supporting relationships between physicians and pharmacists, for instance of pharmacists having shared access to the EHR to approve drug requests or, in one case, select patients for further physician screening. 80,100

Many studies noted the importance of creating a shared understanding between staff about routines, roles and processes. Some studies reported confusion as to the proper use of the EHR:

. . . like tasks you put in the EMR [electronic medical record], where do you put it, how do you write it, what do you say, what language do you use, what format, all that stuff.

Rajala88

Other studies85,88,116 reported best practices that worked in particular organisations, such as the ability to send instant messages for informal communication (e.g. for a ‘warm’ hand-off), creating task lists and delegating roles in the EHR, ability for notes to be embedded in a patient chart, telephone templates that could be routed to team members’ inboxes, virtual ‘huddle sheets’ with patients scheduled for the day in the EHR. Systems for accomplishing this shared understanding varied between practices, but all studies emphasised the importance of being able to communicate informally through the EHR and for each staff member to use the EHR consistently.

There were many examples of positive working relationships facilitated through EHRs in the evidence. However, these relationships may be strengthened through opportunities for in-person communication and free-text notes built into the EHR.

Causal link 6:3 – contexts

Using HIT to mediate interprofessional relationships must be done carefully. Bergman et al. 92 describes a particularly complex EHR causing poor relationships between physicians and pharmacists because they negotiated drug approval requests without the support of informal communication (such as free-text explanations for approvals or rejections). Conversely, personal relationships, for example team huddles or informal chats, could make virtual communication more effective. Two studies83,116 emphasised the importance of primary care staff being able to communicate both online and offline.

Causal link 6:8 – informational continuity of care produces demand management systems

One study96 in our review found that one attribute of a successful ACO programme is that it can stratify patients by risk. Many MCP-like networks and organisations did so (although the methods and risk groups differed) but most did not report whether or not this helped providers to manage resources better. 64 However, the Mount Sinai (New York) ACO did report successfully using risk stratification data to guide staff workflow in different ways depending on identified care gaps and whether risk was categorised as high, rising/moderate or low. 78

Causal link 6:9 – informational continuity of care produces preventative care

Care processes in ACOs or PCMHs were more likely than those in their standard counterparts to:

• create lists of patients due for tests or preventative care, and

• provide patient reminders for preventative follow-up care.

Accountable care organisations and PCMHs that used EHRs were more likely than providers to carry out these tasks without such records. 115 Xenakis78 described an ACO with workflows in its EMR to support disease prevention, and Johnson et al. 124 described a case in which patients received automated text message reminders about recommended preventative services. Overall, there was a little evidence to support the assumption that HIT and EHRs can assist preventative care, but more research is needed to test these claims.

Causal link 6:13 – informational continuity of care produces cost savings

We found some evidence that HIT can increase organisational efficiencies. Colla et al. 73 concluded that ACOs with HIT investment was likely to save post-acute care costs. Another study108 found that telemedicine-based collaborative care was more cost-effective than a practice-based model in medically underserved areas. Several studies found that HITs could increase administrative productivity, thereby saving costs. A qualitative survey116 of PCMHs found that electronic systems reduced administrative burden and increased data accuracy for physicians when teams had specific role definitions stating who recorded what onto the system and how they recorded it. One prospective cohort study122 described how IT in the PCMH context led to a reduction in specialist visits.

Although these studies describe MCP-like organisations or networks using HIT to reduce costs, few of them clearly explained the links between the two. Overall, they suggested that organisations can reduce costs through using EHRs but only in certain contexts.

Causal link 6:13 – contexts

Colla et al. 73 found that HIT investment probably saved post-acute care costs, but this finding reflects a context of US incentive structures that reward or penalise ACOs according to their costs, and in which private hospitals can make large investments in data analytics. A multisite ethnographic study found that organisations that used a combination of electronic and paper chart systems increased the time demands on staff, suggesting that IT systems need to be fully electronic to be cost-effective (McMurray et al. , reported in Lafortune108) and not duplicated through a shadow paper system of files.

A common context for the above mechanism (HIT) to bring about the other MCP component outcomes described was that HIT must be well designed and mirror the care processes that health workers use in practice. It was consistently reported that technology that was bespoke to the organisation(s) and designed with the users in mind had better outcomes on a variety of measures.

Causal link 7:8 – if planned referral networks develop, then demand management systems will develop

We found no evidence to support (or refute) the IPT assumption that referral networks produce better demand management systems, nor did we find any evidence as to whether or not referral networks produce preventative care.

Causal link 7:9 – if planned referral networks develop, then preventative health care will develop

Shortell et al. 103 reported a respondent from an American ACO saying that installing a patient portal had made patients more willing to ‘engage’ with planning their own care. That finding would be relevant to this link only if ‘engaging with care’ included ‘engaging with preventive care’, which the article does not report. Thus, we found no evidence unequivocally corroborating this link.

Causal link 7:10 – if planned referral networks develop, then care planning for individual patients will become more prevalent

We found some evidence that establishing a referral network produces greater use of care plans and more patient-centred care generally.

Colla et al. 73 evaluated the impact that ACOs had on care co-ordination and care management for older populations by exploring the extent to which ACOs incorporated post-acute care into their referral networks. Although the associations were not all statistically significant, Colla et al. 73 concluded that doing so resulted in more comprehensive CCM programmes, and the creation of systems to assure smooth transitions of care across different organisations and settings (ACO, USA). ACO referral networks that included post-acute care services were more likely than those without to have established processes for identifying, counselling and planning for end-of-life care across settings of care. 73

Alidina et al. 68 carried out a mixed-methods study of 13 PCMH ‘medical neighbourhoods’ (local referral and care co-ordination networks of PCMHs) to understand what role co-ordination mechanisms play in them. These networks used communication, negotiation and decision mechanisms through which neighbouring PCMHs agreed how to co-ordinate care and explicitly allocated mutual responsibilities for communication and care co-ordination for shared patients. Such mechanisms included care compacts and agreements negotiated through local Independent Physician Associations. High-performing PCMHs typically had written care compacts with specialists, low-performing PCMHs did not. 68 For care co-ordination at patient level, the most important activities were interorganisationally agreed common working routines, information connectivity and (again) the creation of boundary-spanning roles. A combination of these mechanisms, adjusted to the contextual conditions noted in the following subsection, could improve interorganisational care co-ordination. 68 There was a little evidence from a qualitative study of ACOs and PCMs that care was more patient-centred when referral networks existed. 73

Causal link 7:10 – contexts

As previously noted, Alidina et al. 68 provided important information about what contextual factors call for communication, negotiation and decision mechanisms. These referral network mechanisms are more necessary for patients about whose condition staff have low levels of knowledge, for more complex patients and when reciprocal co-ordination is required (i.e. patients transfer from one organisation to another and back again). Barriers to establishing care compacts were geographical (small or isolated communities), small general practices (small referral base), misaligned payments and time costs (e.g. search costs to find ‘good neighbours’, bargaining and decision-making costs, time to build relationships and costs of internal reorganisation). 68

In a qualitative study of integrating mental health into a primary care setting under the PCMH model, Rajala88 described the operational barriers to care co-ordination through a referral network in the PCMH in the USA: providers having different workflows and expectations, separate medical records or limited access to records and a separate referral process for mental health services. The last barrier resulted in long waiting lists, poor follow-up and less patient centeredness. If mental health services functioned as their own separate subsystem within primary care, there was increased difficulty co-ordinating services. 88

Patients’ own behaviour may be another relevant context. A cross-sectional national survey73 of ACOs defined self-referral as an indicator of ineffective care co-ordination. It found that the trend in the weighted absolute number of self-referred visits among Medicare and private-insurance beneficiaries remained generally stable from 2000 to 2009. Aliu et al. 125 concluded that, whatever attempts ACOs had made at care co-ordination, patients had bypassed them by making self-referrals as well.

Causal link 7:11 – if planned referral networks develop, then more patients will be diverted from inpatient to primary care

Five studies62,73,105,119,126 provided evidence supporting the IPT that referral networks can divert patients from inpatient care. Reassigning care to the PCMH enabled primary care teams to take on additional tasks, reducing specialty visits for low- and, to a limited extent, medium-morbidity patients. 127 In a cross-sectional analysis of the national survey of ACOs, Colla et al. 73 found that ACOs that included post-acute care providers were more likely than those that did not to report a fully developed programme to reduce preventable hospital readmissions. The six components of Wagner’s CCM are (1) community resources and policies, (2) health care organisation, (3) self-management support, (4) delivery system design, (5) decision support and (6) clinical information systems. Five studies28,30,31,49,51 reported in a SR of integrated care models for patients with chronic diseases found that projects that incorporated at least two of the six components had significantly fewer admissions and fewer inpatients days than other integrated care projects. 126

Huber et al. 128 attributed reductions in the likelihood of hospital admission for cardiovascular and COPD, but not respiratory disease, patients to the introduction of care co-ordination and care guidelines in Swiss primary care networks.

A case study of a complex care referral network in Australia62 found that it increased referrals across organisational boundaries and reduced ED use. Referrals to physiotherapy, podiatry, occupational therapy, dietetics and psychosocial services rose and there were fewer referrals to less specialised community home nursing. People enrolled into the programme with chronic and complex conditions had, in the following 12 months, fewer ED presentations and significantly reduced lengths of stay in the ED compared with the 12 months before. Almost 30% of participants had no hospital presentations. 62 In Canada, a cross-sectional patient experience survey99 for ambulatory sensitive conditions found that FCA and AA created a referral network in which care planning process dealt with preventative and acute emergency care in ways that diverted patients from acute secondary services. 99 A meta-analysis of RCTs105 found that transitional care interventions were associated with reduced intermediate-term (31–180 days) and long-term (181–365 days) all-cause hospital readmissions of chronically ill patients.

A set of UK case studies76 highlight reduced tariff income for NHS hospitals when patients were diverted from hospital to primary care as a barrier to patient diversion but, against this, patient diversion was also an opportunity for the hospitals (in any event not lacking work) to free up staff time and beds and meet performance targets. One enhanced access initiative (advice and guidance) created additional capacity in outpatient hospital clinics, which could help hospitals reach performance targets as well as generate additional income by hosting tertiary care clinics. In another example, clinicians spent less time transferring patients to other hospitals after the hospital at home initiative reduced admission and length of stay for other patients. 76

We found only one study104 about whether or not in-reach into hospitals ensures timely discharge of patients, but it was a SR. It found that transition from hospital to home was most effective when interventions to expedite it were initiated during the inpatient phase and continued post discharge.

Causal link 7:11 – contexts

Two studies71,127 stated a specific contextual requirement for referral networks to work as a mechanism for diverting patients from hospital to primary care. The mechanism worked best for high users of acute care and for low- and medium-morbidity patients who could be effectively managed in the community, but high-morbidity patients still required more intensive comanagement by primary care teams and specialists. An analysis71 of routine administrative data from 380,000 records for high users of A&E living at the poverty level found that referral network schemes for diverting patients from hospital had the greatest effect when targeted on the 1% of heavy users (five or more hospitalisations per year). Another study,127 a 48-month interrupted time series from a baseline through PCMH implementation and post-implementation periods for 36,805 hypertension patients, also found reductions in specialist use, but only for low- and medium-morbidity patients. Indeed, high-morbidity patients made significantly increased use of specialist use after PCMH implementation. The study authors127 concluded that referral networks between primary care teams and specialists in the ‘medical neighbourhood’ should cater above all for high morbidity, clinically complex patients. The increased referrals of high-morbidity patients highlighted that primary care teams and specialists also need to sustain effective comanagement of these patients.

Causal link 8:9 – demand management systems enable preventative care and vice versa

Mead et al. 91 reported evidence from 10 focus groups (n = 387 participants) in purposively sampled US communities that supported the assumption that people with complex health conditions seek ED care when problems with access to preventative services make it difficult for them to manage their health. One apparent solution was to ‘empower’ patients to manage their own condition. However, the few studies that we found were equivocal about what effects that mechanism had.

A German study81 using routine data described how, after 2004, osteoporosis prevalence increased (+ 18%) faster in Kinzigtal than in Baden-Württemberg (+ 6%) as a whole, but the available data were insufficient to determine whether the Kinzigtal increase was an epidemiological trend or resulted from a screening and prevention programme.

Before primary care physician reimbursement was linked to patient quality outcomes, a survey by Hibbard et al. 129 (no sampling strategy reported) in the USA found that only 10% of physicians intended to develop patient self-management as a way of improving incomes. The authors’129 follow-up survey after reimbursement had been linked to patient quality outcomes that used different variables, so an exact comparison was not possible, but it found that 60% of primary care physicians had made little or no increase in their support for patient self-management. Hence, reimbursement changes alone were insufficient to incentivise physicians to develop and support patient self-management.

A non-SR85 of informatics support for patient-centred care identified how communication and information-access portals could facilitate patient-centred care, but were not themselves sufficient to enable it.

Causal link 8:9 – contexts

In the above studies, doctors’ and patients’ characteristics were an obvious context for mechanisms ‘empowering’ patients to manage their own health. The survey by Hibbard et al. 129 in the USA found that some physicians expressed frustration at their inability to change patients’ unhealthy behaviours regarding diet, inactivity, smoking and so on. A total of 70% of the primary care physicians surveyed identified ‘patients’ unwillingness to change behaviours’ as an obstacle to achieving care quality metrics (compared with 65.1% identifying ‘lack of time to spend with patients’, 47.7% identifying ‘lack of high-quality support resources’ and 24.8% identifying ‘not knowing how to support patients in behaviour change’). Among the 15.3% of primary care physicians who, nevertheless, reported that they had increased their support for patient self-management, there was double the number aged < 35 years than there were of an older age. 129

Causal link 8:10 – if demand management systems are established, then care planning for individual patients develops

We did not locate any evidence about whether or not, or how, demand management systems such as risk stratification affected the use of care planning for individual patients.

Causal link 8:11 – if demand management systems are established, then more patients are diverted from inpatient to primary care

Taken together, the two relevant studies130,131 in our review were both equivocal about whether or not demand management systems (as opposed to individual care plans; see Care planning for individual patients) diverted patients away from hospital and into primary care.

One cross-sectional study in the USA,130 which used a convenience sample (n = 150), compared ‘comprehensive care’ (i.e. one physician managing both the primary and tertiary health-care needs of a child: a form of gatekeeping) with usual services. Under comprehensive care there were fewer ED contacts (incidence rate ratio 0.51, 95% confidence interval 0.33 to 0.78) and a lower hospitalisation rate. Without directly comparing it with comprehensive care, this study130 also reported that ‘co-ordinated care’ (defined as a provider sharing information and communicating effectively with child, family and consultants, as well as linking to community resources) did not have either effect.

In a retrospective analysis of longitudinal routine data (2,607,902 patients from 796 clinics), Yoon et al. 131 reported how a PCMH model in the USA increased the use of primary care services. The increase arose from practice reorganisation rather than from patient-facing efforts to increase access to care (e.g. by offering flexible and same-day appointments and non-face-to-face services, such as telecare). However, more granular analysis showed that certain elements of practice organisation, such as team huddles and tracking laboratory tests, were associated with fewer primary care visits per patient, which Yoon et al. 131 characterised as greater ‘efficiency’.

Causal link 8:11 – contexts

Being about the USA, these studies presupposed a particular context. As Chapter 1 explained, the concept of PCMH corresponds to the principles under which NHS general practice has already been organised in principle, and often in practice, since 1948. Therefore, even if the above changes to (the equivalents of) general practice co-ordination did divert patients from hospital to primary care, they may already have been adopted in much of the NHS. So, the scope for marginal gains in patient diversion in the NHS may be less than the findings of Yoon et al. 131 suggest. The practice of the same doctor providing (and co-ordinating) a patient’s primary, secondary and tertiary care occurs only under the ‘admitting rights’ model of hospital medicine that exists in much (although not all of) the US health system but hardly at all in the NHS.

Besser’s132 before-and-after study in the USA found that increased service provision led to increased service use. The introduction of a mental health provider in a PCMH led to an increase in the percentage of visits for depression (2010, 0.86%; 2011, 0.54%; 2012, 1.02%; and 2013, 1.26%) and a significant increase (from 3% to 33%) in the percentage of depression visits seen by mental health specialists. Besser132 concluded that these increases occurred owing to services addressing hitherto unmet needs (and, therefore, also preventing use of other services in the future), but the paper reported no data substantiating that. ‘Roemer’s law’ that ‘a built bed is a filled bed’133 is well established with regard to the USA and, with qualifications, to many European health systems. 134–136 It implies that, even if demand management methods do reduce admissions from existing care groups, other hospital admissions are likely to take their place. In part, this is a consequence of per-patient payment systems used by sick funds, corporate insurers and those public bodies that have copied from them a diagnosis-related group-like payment system.

Causal link 9:11 – preventative care enables referrals to be diverted from inpatient to primary care services

We did not locate any evidence directly reporting whether or not preventative health care enables referrals to be diverted from hospital, or any about whether or not patient self-care activation produces better demand management systems for general practice.

Causal link 10:9 – care planning at the patient level produces preventative care

The IPT assumed that having a patient care plan builds patient confidence and their capability for making good decisions about their self-care, and so improves preventative care. One US PCMH study91 confirmed that having an embedded case manager using joint care planning and motivational interviewing resulted in more trusting relationships between the patient and doctor, care that is more customised to their patient’s individual needs and in patients making the most of their appointments and taking a more active role in their care. 91 Similarly, a case study in a PCMH found that case managers using motivational interviewing, assessment skills and joint care planning enhanced the value of primary care visits for patients and engaged patients more in their own care. 137

In contrast, a survey of 10,990 adults with asthma, diabetes mellitus or chronic heart disease in a US PCMH138 found that having an individual treatment plan was not associated with patient empowerment. Nevertheless, these adults were more likely to report using preventative and ambulatory care when their care involved at least two of care co-ordination, care continuity and a care plan. In this study,138 unlike the others, the unit of analysis was the patient rather than the provider organisation.

Causal link 10:9 – contexts

Taken together, the findings from the three relevant studies91,137,138 on preventative care imply that patient empowerment is sufficient, but not necessary, to stimulate increased use of preventative and ambulatory care.

Causal link 10:11 – care planning diverts patients from inpatient to primary care

Various studies62,87,131,137,139 describe mechanisms by which individual care planning diverts some patients from inpatient to primary care. MDTs’ boundary-spanning roles assist in identifying patients in need of care co-ordination. A UK study76 of paediatric care identified daily specialist community nurses visits to acutely unwell complex patients at home and telephone consultations between the community nurse and hospital duty consultant as means whereby a MDT could support joint care planning so as to avoid emergency hospital admissions and, if they should occur, enable the child to be discharged earlier.

As regards the outcomes so produced, four studies62,131,137,139 found that care planning for individual patients increased patient diversion from hospital to primary care. One study138 did not (described below). Two studies105,140 found that care planning during and after the transition from secondary to primary care reduces readmission rates, and another two86,105 that having operational facilitators to support care planning reduced readmission rates.

In a matched case–control study of American PCMHs, Clarke et al. 139 implemented and evaluated a programme that embedded non-licensed comprehensive care co-ordinators (CCCs) in 14 PCMHs to help primary care doctors execute care plans in order (inter alia) to extend each practice’s ability to support patients before, after and between primary care visits. This intervention reduced ED admissions by 20% annually compared with the control practices. Treadwell and Giardino137 also found that ‘embedding’ a case manager in a PCMH reduced admissions per 1000 patients over the following 18 months. Similarly, the Yoon et al. 131 retrospective longitudinal study of 2,607,902 patients from 796 VHA primary care clinics in the PCMH model found that creating individualised treatment plans, assessing treatment barriers, and better co-ordinating visits, to other physicians decreased the mean number of ED visits by 0.04 visits per patient (p = 0.018). A panel study141 of patients in Philadelphia found that care management by PCMHs (rather than improved access to primary care) reduced ED attendances by between 5.24% and 7.78%, but only for chronically ill patients (especially those with coronary artery disease, hypertension, congestive heart failure, COPD or asthma). A case study62 of a chronic aged and complex care service model in Australia found that the number of ED presentations and length of stay in the ED fell significantly in the 12 months following enrolment compared with the previous 12 months. Almost 30% of participants had no hospital presentations after enrolment. Referrals to non-hospital physiotherapy, podiatry, occupational therapy, dietetics and psychosocial services (but not to community nursing) increased. 62

Against this, Pourat et al. 138 found that the combination of care co-ordination, continuity of care and care plans did not decrease the likelihood of ED use, although it did increase use of preventative and ambulatory care and improved clinicians’ communication with patients. This is an absence of evidence for the outcome that NHS policy-makers had assumed, not evidence of an opposite effect. This study138 also implied that care planning at the patient level is most effective when combined with co-ordination and activities to increase the continuities of care plan.

Other studies argued that individual care planning and co-ordination during and after the transition from hospital to home reduced readmission rates to EDs. One specific mechanism was illness-specific specialised education to support patients in self-management post discharge (i.e. telephone advice on how to monitor one’s weight and look out for warning signs that would prevent an ED visit or hospitalisation). This intervention did indeed make patients more likely to monitor their weight and change their health behaviours, but no less likely to be admitted or re-admitted to hospital. 140 A SR105 of transitional care interventions found that three components of care were associated with reduced short-term admission rates: (1) care co-ordination by a nurse (most frequently a registered nurse or advanced-practice nurse), (2) a home visit within 3 days and (3) communication between the hospital and the primary care provider. Most of the interventions in the review that reduced intermediate- and long-term readmissions involved care co-ordination. 105 A mixed-methods study86 of 18 complex care management organisations suggested that CCM teams that receive timely notifications of their patients’ ED visits could intervene to avoid hospitalisations. Methods for ensuring safe transitions included medication reconciliation and developing contingency plans in case certain trigger events occurred. 86

Causal link 10:11 – contexts

Only two studies about diverting patients from hospital to primary care identified contexts in the realist sense. To minimise re-admissions, CCM teams must help patients find the resources they need in local health systems and communities. 86 Verhaegh et al. 105 observed that developing a valid and reliable method to measure the preventability of a readmission was important to enable clinicians to implement targeted readmission policies and penalties for preventable readmissions. Case management is one form of care planning for individual patients. An umbrella review by Damery et al. 104 found that only one out of eight SRs showed that case management reduced hospital admissions, that is, diverted patients from inpatient to primary care. The exception was a review showing a 49% relative risk reduction of hospital admission for patients with heart failure.

Causal link 10:12 – care planning improves patient experience

The evidence that we found supported the assumption that care plans for individual patients improve patients’ experience of care (10:12), but the studies were few. Just one study137 reported US focus groups’ opinions that joint care planning with patients gained for health-care providers a comprehensive understanding of the client’s or family’s health-care needs, barriers to and potential action regarding a patient’s social support, health literacy, understanding of the care plan, plan adherence, and care preferences. 137

We found limited evidence from three studies91,103,108 that patient involvement in decision-making about (i.e. planning) their own care improved patients’ experience of care. When patients felt they had been left out of decision-making about their own care, they felt vulnerable. 91 A mixed-methods study103 of American ACOs suggested the value of involving in care pathway redesign:

We gave them (patients) an initial care pathway as we saw it and had them fill in what we missed. Every single interview raised using catheters as a point of anxiety for the patient and the urologists didn’t realize that was a point of anxiety.

Shortell et al. 103

In three Canadian focus groups108 with 28 patients and informal caregivers in community-based PHC, patients expressed support for the role of patient advocates who helped them navigate the care system and participate more fully in decision-making. This was particularly important for patients without a family member to bring to appointments. 108

Causal link 10:12 – contexts for improved patient experience

As the context for care planning to improve patient experience, a study of focus groups across 10 communities found that trusting and open relationships between patients and providers created the conditions for care customised to the patients’ specific needs,91 suggesting that there may be a virtuous circle between trusting and open patient–provider relationships and patient-centred care.

Causal link 11:12 – diverting patients from inpatient care improves patients’ experience of care

Three studies66,74,81 published since 2014 reported evidence about what effect diverting patients from secondary care had on their experience of care. A ‘qualitative analysis’ of documentation66 described how a programme that integrated mental health care into ambulatory care enabled practitioners to support those with acute mental health needs and promote a more holistic and empowering approach to self-care that encompassed welfare and healing. The context for these perceived effects is noted in Causal link 3:9 – patient engagement, patient self-care, activation and empowerment. An analysis of patient experience indicators compared the Integrierte Versorgung Gesundes Kinzigtal (IVGK; ‘Healthy Kinzigtal Integrated Care’) programme with ‘usual care’. 81 The adjusted comparison showed that approximately one-third of the indicators were significantly better for people in the IVGK programme. Another third of the indicators changed in the desired direction, but not statistically significantly. The remaining third did not change. 81 In some areas of care, such as osteoporosis treatment, important outcomes such as the number of fractures (closely related to quality of life and of patient experience) were significantly lower for people in the IVGK programme than for those in the control programme. 81

In a descriptively analysed study74 of telephone interviews with patients (convenience sample, n = 15) who received care from the EPSILON geriatrics team network (France), patients reported being satisfied with the way the network enabled access to expert advice and support that would otherwise require hospital admission. ‘Compliance’ with medical and paramedical prescriptions in this small sample was reported as 72% and 74%, respectively, but Canali et al. 74 reported no comparisons with patients outside the network.

Causal link 11:13 – diverting patients from inpatient care reduces costs

We found few post-2014 studies reporting how diverting patients from inpatient care to outpatients departments or nursing homes reduces costs. Regarding the use of ED services, one US case study66 ‘qualitatively analysed’ Institute for Health Improvement documentation and publications to examine the cost impact of integrating mental health care provision into ambulatory care for mental health service users with light, moderate and severe levels of complexity. This integrated approach cost less than a non-integrated approach. 66 Users of the integrated service used urgent care services 54% less than non-integrated service users, although the study66 did not quantify how much money was saved. In primary mental health services the per-patient costs of care increased for both integrated and non-integrated services, but less for integrated care. 66

A matched case–control study in the USA139 evaluated a programme that embedded non-licensed CCCs in 14 PCMHs to help primary care doctors execute care plans so as to extend each practice’s ability to support patients before, after and between primary care visits. This intervention reduced ED admissions by 20% annually compared with the control practices, saving payers approximately US$2000 per ED visit, which implied an estimated total annual cost reduction of US$1.4 million for the whole programme. Salary and benefits costs of the personnel dedicated to the programme (but ignoring the hidden costs of medical directors’ and other support staff time) were approximately US$950,000 annually. 139 Treadwell and Giardino137 also report the equivocal finding that, in two out of five case study sites, PCMH-based care co-ordination by an ‘embedded’ case manager reduced the number of admissions per thousand patients and, therefore, reduced claims costs by US$7 per member per month in one site and US$14 in the other. The range of costs narrowed across all five sites, suggesting stronger managerial control of costs. 137

A German study81 used a difference-in-differences approach to compare the care costs (based on routine data) for people who were in the IVGK programme with those who were not. 81 Care costs for people in the IVGK programme were €322 less per annum than the ‘usual care’ group, with the greatest savings occurring in relation to hospital care (€179 per person), other services (€93 per person) and medicinal products (€37 per person). 81 Huber et al. 128 attributed annual cost savings of CHF440 (cardiovascular patients), CHF780 (diabetes mellitus) or CHF200 (respiratory illnesses) to the introduction of care guidelines in consequence of ‘integrated care’ in Switzerland.

Damery et al. 104 found that, in general, the evidence that ‘integrated care’ (mechanisms not specified) reduced health-care costs was ‘poor and heterogeneous’ and equivocal, with some SRs reporting cost savings, especially for the CCM, and others not.

Causal link 11:13 – contexts

Treadwell and Giardino137 did not report what contexts (in the realist sense) differed between the PCMHs where patients had been diverted away from hospital care (reducing costs) and those where they had not. Briot et al. 66 emphasised the necessity of integrating mental health care provision into (general) ambulatory care. Clarke et al. 139 gave no contextual information (in the realist sense). Huber et al. 128 described capitated (as opposed to activity-based) payments to providers as a favourable context. Kinjo et al. 107 found that replacing hospital with PHC end-of-life care reduced costs only if community care began > 30 days before the patient’s death, although that finding reflected the Japanese structure of GP payments.

How do policy-makers and top NHS managers predict multispecialty community providers will generate the policy outcomes stated in the Five Year Forward View?

Chapter 4 answers this question more fully, but in brief there was no simple answer to this question. The policy-makers’ programme theories proposed a large number of links, all originating from three starting points (see Figure 1):

1. NHS managers’ action in setting up MCPs as network co-ordination structures for (at least) general practice and CHS, and for a complex of other services that varied between MCPs but usually included social services and urgent care, and (less often) mental health services

2. changes in the culture of these organisations and across the whole MCP

3. voluntary sector willingness to contribute to the activities noted in the following bullet points.

Between them, these three would be the mechanisms producing (as first-wave intermediate outcomes):

• network management of the above constellation of organisations

• formation of multidisciplinary care teams

• referral networks, now planned and managed rather than emergent

• more active development of preventative care

• ‘demand management’ systems.

Combined with other linkages, these outcomes would then launch a further set of mechanisms, producing as (second-wave) outcomes:

• care co-ordination by means of HITs

• further development of ‘demand management’ systems

• care planning for patients

• diversion of patients from hospital into primary care.

Finally, the last outcomes would become the mechanisms for reducing the cost of NHS care and improving patients’ experience of care. These effects would result from various parallel, mutually supporting and parallel mechanisms in which main linkages only are summarised above. NHS cost reduction was assumed to be the outcome of (depend on) 39 prior mechanisms in all, improved patient experience on 40.

What variants of multispecialty community provider are policy-makers creating?

So far (spring 2017) neither the policy materials we analysed, including the first-wave MCP logic models, nor professional press rapportage suggest that there are any groups of MCPs in which shared characteristics can be contrasted with those of other groups of MCPs with different shared characteristics. Rather, MCPs at this stage all serve essentially the same function (according to 5YFV6) of horizontally co-ordinating managed referral networks across general practices (and/or general practice ‘at scale’), CHS, social services, mental health, urgent care and (varying by site) miscellaneous other services. Accordingly, MCPs have a similar architecture, with a central body (perhaps one their member organisations) co-ordinating the aforementioned activities and mechanisms across the network as a whole. When MCPs do vary, it is in how each is adapted to its particular local setting and assemblage of member organisations. At most, one might say that the 14 first-wave MCPs represented 14 variants. However, MCPs are still at an early stage of development, so the question of whether or not distinct types (groups) of MCPs will develop remains open. In particular, the relationship of PCH models to MCPs is at present uncertain. In their size, structure, function and governance they are quite different from MCPs. Their policy relationship to MCPs (whether PCHs are an alternative or a part of MCPs) remains, at present, an important undecided aspect of NHS policy. The relationship of MCPs to the English version of ACOs is also unclear at present.

Clientele

What population is served defines the scale, range and geographical distribution of the services a MCP has to co-ordinate, and what constraints (e.g. extent of patient choice) that apply. Our review showed that MCP-equivalent organisations or networks were designed to cater for either of two kinds of clientele:

1. Individual subscribers to particular social health insurers (‘sick fund’), public (e.g. Medicaid), mutual or corporate insurers. Then, MCP-equivalent entities can plan for only part of the population of a locality, must plan for a clientele that may be widely geographically dispersed and may be unable to decide which providers their clients use (e.g. in Germany). Then, for example, they must recruit patients voluntarily to any programme that selects service providers with a view to reducing referrals and costs. Unless they can informally negotiate other arrangements,46 they are obliged to pay hospitals for each referral the hospital can attract. This applies to the Kinzigtal project (Germany), French and German primary care providers generally, and most American ACOs and PCMHs. In the USA, providers can and do select their clientele by insurance status.

2. Similar to the NHS as a whole, MCPs cater for whole populations defined by place of residence. The same applies to MCP-equivalent bodies in Sweden (e.g. the Norrtälje project24) some Catalan primary care providers, USLs in Italy, LHINs in Canada and the (in that respect atypical) population-based ACO in New Jersey. 71 Planning and referral network management mechanisms developed in this context are more likely to be directly relevant to MCPs.

Within either type of clientele, some MCP equivalents serve everyone, whereas others serve specific care groups defined by morbidity, such as people with multiple long-term conditions, frail older people or people with mental health problems (again, Chapter 6 gives some details). The literature we reviewed focuses less on groups of people with a single major condition (but when they do, coronary heart disease is often studied), but more often on morbidities with functional impact morbidities (e.g. diabetes mellitus) and some forms of high functional impact multimorbidity (e.g. dementia).

Services

Which MCP-equivalent organisation or network is the most relevant proof of concept, even prototype, for a particular MCP will depend on which services it is especially attempting to co-ordinate (i.e. which interorganisational boundaries it is attempting to surmount). Interface by interface, that approach suggests the following international partial equivalents to MCPs:

• Primary medical care (GP or equivalent) with CHSs – in the existing NHS, these organisations are separate. International equivalents include Kinzigtal (Germany), the Versailles geriatrics network, LHINs in Canada, certain ACOs (e.g. in Texas and Colorado82) and some instances of PCMH (e.g. in Manhattan94). As proof of concept examples of what can be done to integrate these services organisationally, the most relevant international equivalents are Swedish PHCCs (vårdcentral, ‘polyclinic’). Between the two sets of equivalents are the Italian USLs, in which domiciliary nursing care is organisationally integrated with primary medical care, but in a structure of separate subhierarchies (‘silos’) with a common manager only at the most senior level. Catalan primary care centres typically organisationally integrate GPs with CHSs, with specialised services for women’s health and paediatrics and, on occasion, other specialties.

• Primary medical care (excluding CHS) with mental health services – international equivalents include the Norrtälje project (Sweden24), Inter-Mountain Health, Utah66 and (more narrowly because essentially contract driven) the Integriete Versorgung mental health schemes in Germany. 150–152

• Primary medical care with social care – for this interface, international equivalents include Local Community Services Centers (Québec), public clinics providing health with social services60 and the Mount Sinai organisation in New York. 78 Italian USLs organisationally integrate primary medical care and social care, but (as noted) in a structure of separate subhierarchies (‘silos’) with a common manager only at the most senior level. Primary care centres in Catalonia offer strong co-ordination by including the provision of social care services, although health and social services remain managed through different hierarchies.

• Primary medical care with community pharmacy – accounts of this interface were rare. We found just two studies80,100 that described how these two services were co-ordinated, although older studies153,154 about the UK also exist.

• CHS with social care – for organisational integration between CHS and several forms of social care, some of them (including nursing homes, hospices) residential, an obvious international equivalent to a MCP (including separation of CHS from general medical practice) is Buurtzorg in the Netherlands. 21 In the Italian USLs, social care is organisationally integrated with primary medical care, but in a structure of separate subhierarchies (‘silos’) with a common manager only at the most senior level, much as in Northern Ireland in the past. 155,156

• CHS with mental health services – again, Inter-Mountain Health instantiates such co-ordination, but Canadian LHINs (for instance in Toronto) developed in a context more similar to that of the NHS and the Norrtälje project in a context more similar still, with a substantial local government role and mental health services previously organisationally separate from other health services.

• General practice ‘at scale’ – if this means many small, even single-handed, practices co-ordinated through a hub, the HealthOne Mount Druitt project in New South Wales provides the equivalent for those parts of the NHS with many single-handed, general practices dealing with deprived populations. If ‘at scale’ means employing large numbers of PHC doctors in one organisation, more relevant MCP equivalents include some large US providers (e.g. Group Health, Kaiser Permanente). Swedish primary care clinics also tend to be managed in large groups, either by a municipality, a corporation or, in the case of Praktikerjänst, a health worker co-operative that supplies > 15% of primary medical care in Sweden. 23 The PCMH concept derived originally from US perceptions of NHS general practice has now been reimported back to England as a ‘new’ model of care. It is probably intermediate between general practice and a MCP with a larger population and ambitions to incorporate community services and reach across the interface to hospital, but could also be a constituent of a large MCP operating across the catchment of one district general provider hospital.

The above are possible prototypes for interorganisational co-ordination (through referral networks) or organisational integration of primary care. For MDT prototypes and for successful instances of using HIT for care co-ordination, one has to look to the particular studies cited (see Chapter 5).

Governance structures

The types of governance structures within which MCP equivalents are embedded are important contexts defining their relevance (equivalence) to MCPs. Governance structures within them are another point of equivalence (or not) for MCP mechanisms. Modifying Thomson’s157 categories, the three types of governance structures relevant to MCPs are quasi-markets, hierarchy and networks.

1. Networks: the more numerous and varied the member organisations a network contains, the more numerous and varied are the interorganisational boundaries it has to surmount. Similarity in the number and mix of member organisations are, therefore, important criteria of equivalence (relevance) to a given MCP. Hence, two important equivalents (or approximate equivalents) to MCPs are:

   1. Networks of doctor-owned practices, including substantial proportions of both single-handed practices and partnerships (see the discussion of ‘general practice at scale’). In these respects, the nearest equivalents to MCPs are Medicare Locals in Australia; in particular, the more ‘integration’ minded ones (e.g. HealthOne Mount Druitt) and local networks (a category partly overlapping with ACOs and the PCMH, but also with Health Maintenance Organizations) in the USA

   2. Mixed networks of doctor-owned practices, corporations and voluntary organisations. The relevant examples here are networks such as the Kinzigtal and the Versailles examples, many ACOs in the USA and LHINs in Canada.

2. Quasi-markets: for attempts to govern MCP-equivalent groups of organisations by means of contracts (quasi-market governance), the relevant examples are the American ACOs and the Integrierte Versorgung mental health networks in Germany. Although these instances suggest that contracts could finance, even incentivise, organisations to join a network, pursue common goals and MCP-equivalent care co-ordination, US legal studies158,159 also suggest that contracts alone are both too incomplete and too inflexible to establish, by themselves, the mechanisms of action described below. Contracts are between payer and provider organisations, not between clinicians or (even in France and the USA) between clinicians and patients. Other governance and co-ordinating mechanisms are required to supplement them.

3. Hierarchy: as the external context of MCPs, hierarchy would imply either organisational integration into, say, a municipality or, as in the current NHS, a highly centralised and centrally controlled network of formally independent organisations: a ‘quasi-hierarchy’. 160,161 As already noted, the nearest equivalents to a hierarchically structured MCP are the Swedish and Finnish primary care clinics, USLs, some US organisations (Group Health, Kaiser Permanente), and some primary care providers in Catalonia.

The above lists are not exhaustive, even for the MCP-equivalent entities that our secondary data covered. Many studies say only vaguely that a given project ‘brings together’ different providers, omitting the all-important (for present purposes) details of the mechanisms used and whether or not the providers remained organisationally separate. However, the above lists do suggest starting points for developing sampling frames for more detailed research into the mechanisms used.

Does evidence about how multispecialty community provider equivalents ‘work’ support our revised logic model?

The exact mechanisms vary equivalent by MCP equivalent. Chapter 7 itemises the causal links between the components of MCPs in our revised logic model. Nevertheless, certain triggers of care co-ordination mechanisms recurred across many MCP equivalents and contexts.

Tables 17 and 18 in Chapter 7 illustrate the number of studies providing supporting evidence for each mechanism. The six most frequently mentioned mechanisms (each across their different causal links) are what we next report. Five of the six most frequently mentioned mechanisms were also the ones with ‘substantial’ evidential support (i.e. both SRs and additional primary research) according to our realist review (see Tables 17 and 18). The exception was network management, based on ‘supporting’ rather than ‘substantial’ evidence. Most of the studies we reviewed were non-realist and, therefore, present their findings in terms of what we have called the ‘components’ that are antecedents to or triggers of the mechanisms that make up our revised logic model. For short, we use the term ‘MDT based’ to indicate the set of mechanisms for which MDT is the antecedent or trigger (and analogously for the other groups of mechanisms).

Health information technology-based mechanisms

A recurrent theme was that HIT, in particular the EHR, had an effect through the impact that it had on work processes such as task reminders, delegation, workflows and informal communication among staff, instant messaging, within-chart notes and telephone templates that could be routed to team members’ inboxes, task assignments and ‘huddle sheets’ among others, in addition to data retrieval and communication (Box 1). One study116 found that primary care teams that used EHRs consistently for data entry and agreed on communication methods between staff members were more likely to score highly on the NCQA 2011 PCMH recognition tool. This compound mechanism relies heavily on the quality of HIT design in terms of functionality (whether or not a HIT can perform risk stratification, manage workflows, etc.) and interoperability between IT systems, with the prior requirement that such systems are actually available at all.

Therefore, we conclude that HIT-based mechanisms will underpin achievement of multiple MCP-like functions and that they can operate whatever the structure of teams and organisations.

Multidisciplinary team-based mechanisms

A similar theme recurred in studies of MDTs (Box 2). MDTs improve patient experience through the impact that it has on many everyday clinical working practices: enhanced patient access to services (e.g. to primary care as an alternative to unnecessary hospital admission); better communication between providers and, thus, more patient confidence in, trust in and satisfaction with care; and a more holistic approach to care. For MDTs to work, it is necessary that they focus on tasks of practical value to their members, include the relevant services, and actually work in a collaborative, interprofessional way within the team itself. An essential component of this mechanism is boundary-spanning roles, such as that of the care co-ordinator whose professional origin appears less important than his/her capacity to support care planning for individual patients (see Care plan-based mechanisms), improve the continuity of care, make care more person centred and promote shared decision-making. Although rare in practice, patient participation in the MDT facilitates all this. When different professions work for different organisations, the boundaries to be spanned are simultaneously interprofessional and interorganisational. Several studies reported the value of face-to-face communication within teams, which implies a practical value for care co-ordination in colocating MDT members.

Although the formalising of multidisciplinary working into teams, with clarity about roles and boundary-spanning activity, is likely to contribute to MCP objectives, it is unclear in what contexts new teams should form or existing ones be enhanced, and which functions (admission avoidance, proactive care planning or enhancing social connectivity) are particularly supported by MDTs. There is no clear guidance on how much to focus on protocolised role clarity or on flexibility and reducing differentiation between occupational groups.

Care plan-based mechanisms

As a mechanism for diverting patients from hospital to primary care, care plans work by being implemented, above all, by a boundary-spanner (e.g. care co-ordinator). This implies a single care plan (not multiple duplicating plans, as often happens in practice23) for each individual covering all their health-care needs. The care plan can be disease oriented or address an individual’s more social and emotional goals as well as aim to reduce burden of care. As a mechanism for diverting patients from hospital to primary care, care plans work by being implemented above all by a boundary spanner (e.g. care co-ordinator). One component of this mechanism is to develop patients’ self-care and self-management of their condition, which may itself require patient education and indeed patients’ and/or informal carers’ participation in the care planning, shared decision-making and even patient advocacy. Another component is real-time information about what is happening to the patient (see Health information technology-based mechanisms) so that the care co-ordinator can plan and manage the transitions between hospital and home, and other changes in the patient’s condition or circumstances.

The studies available to us contained little evidence about how clinicians or a MDT might use the making of a care plan as a means of deciding with the patient, or at least among themselves, whether or not the patient needs certain kinds of more intense care (e.g. medications, hospital admission). Although enhancing care planning activity (both the interactive decision-making itself and then making shared, comprehensible documentation of the decisions available) appears to be key to generating better outcomes, there is little evidence to guide the level of complexity and multimorbidity that necessitates a shift towards more complex, multidisciplinary plans.

Culture change-based mechanisms

Of all the mechanisms in the IPT, these were the most obscure (Box 3). Many studies examine organisational cultures and cultures of multiprofessionality or collaboration in other health-care settings. Among the studies that we found, many invoked culture change as a mechanism that organisations or networks exploited but few explained how that culture change was produced. Those that did mentioned interprofessional and/or interorganisational training. Some appeared to assume that ‘leadership’ was responsible, perhaps for culture change but certainly for setting up the boundary-spanning mechanisms described above. Two studies62,68 implied that culture change was not the original change-driving force, but perhaps a part of a virtuous circle driven by other causes.

Despite this lack of evidence, we did not interpret culture change as being unimportant; rather that we need more research to define what aspects of culture (e.g. interprofessional equality, person-centredness, positive risk taking) are most important and whether they should be the direct subject of training or seen as indicators of success.

Planned referral networks-based mechanisms

This component was one of designing referral pathways for the main care groups, establishing agreed divisions of labour and working practices across different provider organisations (‘care compacts’), criteria of appropriate referral and, for patients who do not need hospital admission, alternative destinations than hospital, including what in the UK is called ‘social prescribing’ to voluntary sector resources.

Network management-based mechanisms

A network managing (or co-ordinating) body is the mechanism for managing the care ‘continuum’ (i.e. the patient’s experience of care as a whole and over time) (Box 4). Critical components of this mechanism are shared goals and boundary-objects (i.e. objects used in common by all the member organisations at their interfaces), such objects as care compacts, EHR, patient care plans, formularies, agreed care standards and interorganisational care pathways (in addition to any that are used just within a single organisation). Such a network co-ordinating body deliberately supports the production of these goals and objects for the network as a whole, whether by creating them itself from scratch or by adopting and developing any such goals and objects that have already spontaneously emerged ‘bottom up’ from within, and between, the network’s member organisations. Boundary-spanning staff roles are one essential component of this mechanism too. Another is referral network planning (see Diverting patients from secondary to primary care, thereby reducing costs).

How do these mechanisms depend on specific contexts

Chapters 6 and 7 itemised which specific contexts each MCP component requires when operating as a mechanism to produce other components. Nevertheless, certain contexts recurred across more than one causal link between the 13 components. Briefly, they were:

• prior collaboration and mutual trust between provider organisations

• funding for the start-up costs (network formation, HIT, training) and to establish primary care alternatives to hospital, include payment to enable patients to access voluntary sector support

• clinician time for setting up and then participating in MDTs

• status differences between professions and professionals are weak, or deliberately weakened, to facilitate the culture changes mentioned above (see Culture change-based mechanisms)

• lack of health-worker resistance; GP (or equivalent) participation in particular is indispensable

• patient’s active participation in the co-ordination of their care and in self-managing their condition, when feasible

• suitable HIT systems exist (or can be constructed) and are obtainable

• alternative PHC services to hospitals exist, and are of the necessary types and scale

• a suitable case-mix of patients, that is, patients who –

  • are heavy users of hospital services (five or more admissions annually)

  • have complex, not well-understood health problems, whose management often requires informal discussions among health workers

  • have chronic single conditions with well-defined treatment plans; hence, are more suitable for HIT-based methods of care co-ordination

• colocated staff, whether out-posted, ‘embedded’ (i.e. seconded) or all employed by the same provider organisation. Colocation requires a suitable clinic (or similar) as the place of colocation or, failing that, organising base for virtual MDTs. 76 It could also be the place for the co-ordinating body of the MCP as a network of provider organisations, and as the central ‘hub’ for a network of general practices. Not least, colocation provides the opportunity for interprofessional working not only in formal meetings but in everyday, informal working practices such as ‘huddles’.

Some evidence about contexts was conflicting; in the realist view, this was a possible marker for as yet undiscovered contextual moderators of the mechanisms mentioned above (see Network management-based mechanisms).

Some of the component mechanisms were mutually reinforcing and had common elements (e.g. boundary-spanning staff); such were MDTs and HIT, and cultural change and MDTs. Notwithstanding the ‘mechanism’ metaphor, all of the above components of MCPs when acting as mechanisms consist (we reiterate) of the individual actions, understanding and resource use (in short, working practices) of the clinicians, managers, other staff and the other agents involved, not least, patients. It should also be noted that the above components act as mechanisms for cross-organisational provision and co-ordination of care, typically for people with chronic, and often multiple, health problems. They are not necessarily needed to provide more casual, non-complex episodic care.

What policy outcomes are these equivalents reported to produce?

Our evidence review provided evidence about MCP equivalents and whether or not, and how, they bring about the two central outcomes of the initial MCP programme theory (i.e. cost reduction and good quality of patient experience of care).

Diverting patients from secondary to primary care, thereby reducing costs

A number of studies reported MCP-equivalent organisations and networks diverting patients from secondary back to enhanced primary care. A few of them suggested what mechanisms and contexts had produced these outcomes. Again, Chapter 6 provides further detail. These studies offer proofs of concept that the mechanisms can produce these intended outcomes, provided the mechanisms are correctly implemented and provided the relevant contextual conditions are present.

Across several countries, the balance of evidence tended to suggest that more active care co-ordination across organisations (and, for emergency admissions, home telehealth programmes)85 can reduce ED use, hospital admissions or readmissions. Studies from Australia,62,90 Canada,99,119 England76 and the USA66,130,131,137,139,162 reported various combinations of such reductions and greater use of (enhanced) primary care services. A SR126 and a meta-analysis of RCTs105 both suggested that transitional care interventions tend to reduce hospital readmissions of chronically ill patients. The three exceptions to this pattern were only partial exceptions. One US study showed no decrease in ED use but did show greater use of preventative and ambulatory care. 138 Two US studies71,127 showed reductions in specialist use for low-and medium-morbidity patients but the opposite for high-morbidity patients. Therefore, the overall pattern suggests that MCP-like interventions can, in favourable contexts, produce the desired outcomes but with two important caveats. First, we have to be aware of publication bias; failed attempts may be less likely to be published. Second, the devil in these studies is in the detail of what specific mechanisms and contexts were necessary.

Supposing that in favourable contexts these mechanisms do reduce unnecessary referrals, we found less evidence for whether or not overall costs of care consequently fall. Several studies (see Chapters 6 and 7) attributed cost reductions through HIT to the partial automation of work, provided the conditions mentioned in Chapter 6 were satisfied. HIT was also an element of the Kinzigtal project, which achieved cost reductions for the social health insurers. 81 However, there was also a little evidence that HIT in the PCMH context reduced specialist visits. 122 However, one study139 did estimate cost savings arising from stronger care co-ordination reducing ED visits (in that study, US$1.4M annually across 14 medical practices serving 25,356 patients). So, although the evidence base is smaller and weaker, this overall pattern also suggests that such MCP-like interventions can, in favourable contexts, reduce the use of hospital services in a suitable context, but the requisite context is, Chapter 6 suggested, narrowly defined unless the savings per episode accrue directly to the primary care provider or payer in the form of reduced tariff bills. Furthermore, this evidence comes from health systems facing less severe budgetary constraints than the current NHS.

Patient experience

For conditions in which the very occurrence or exacerbation is itself an outcome, and for which evidence-based treatments exist, some studies of MCP-like schemes did report improved outcomes, for instance fewer ED and hospital admissions for asthma94 or increased screening of diabetes mellitus and hypertension patients leading to the prescription of preventative pharmaceuticals. 100 Diabetes mellitus is one such condition in which improved outcomes were, according to several studies (including one SR), associated with MDTs, ‘leadership’ (managerial) commitment to changed working practices, shared goals and staff involvement (in designing and implementing the care pathway). The use of EHRs has also been reported to accelerate ‘quality improvements and changes in utilisation over time on some measures’,115 again in several countries: Germany, the Netherlands and the USA. Two other characteristics that are shown in several studies to improve patients’ experience of care in MCP-like organisations and networks are (1) the use of patient panels to strengthen trust in patient–provider relationships91,163 and (2) personalised care and support from people working in boundary-spanning roles. 66,164

However, the studies we reviewed generally lacked evidence about how to evaluate, monitor and adjust the overall flow of patients within a MCP equivalent in order to ensure that it can achieve its aims of improving care within tight resources.

Perverse or unforeseen outcomes

The studies that we reviewed also reported certain perverse outcomes from MCP-like networks and organisations, unforeseen in the UK policy documents:

• More efficient demand management systems increase case finding, leading (at least initially) to more rather than fewer hospital referrals.

• Increasing hospital and PHC efficiency increases the total costs of care for the reasons noted in Chapter 6.

• Roemer’s law133 increased provision (in this case, enhanced primary care and reduced pressure on hospital beds) leads to increased service use,132 whether by lowering referral or treatment thresholds, meeting hitherto unmet needs (see point 1), adding preventative to existing curative services or making it easier for patients to access enhanced primary care. 131

To these outcomes must be added the perennial uncertainties of implementation, especially when changes (such as revising occupational roles) are contentious and may be resisted or renegotiated. 95 As Pineault et al. 60 observed in Québec, modifications of structures and resources come first, with new working practices always lagging behind.

Models of care

Further primary research would be required to test elements of the revised programme theory. In the research that we reviewed, a number of gaps were apparent. They indicate further research needs. We judge them to be in the following descending order of importance. They concerned:

1. How, and what circumstances, MDT-based locality teams and enhanced general practice (the PCMH and general practice ‘at scale’) compare and interact, or can be combined, in managing referral networks so as to reduce workload for other health-care providers.

2. Whether or not and, if so, how and in what circumstances, diverting patients from hospital into enhanced primary care does indeed:

   1. reduce the overall cost of health care

   2. improve patients’ experience of care.

3. How general practices are affected and have to adapt if larger numbers of patients are diverted from hospital to enhanced primary care.

4. How the other new models of care (above all, PACS) being developed concurrently with MCPs interact with MCPs. The work would compare and synthesise the findings from this studies with those from the concurrent studies of the other new models of care.

5. How urgent care services will be affected and have to adapt if more patients are diverted from hospital to enhanced primary care.

6. How care co-ordination through HIT supports (or not) the:

   1. – management of interorganisational referral networks

   2. – diversion of suitable patients from hospital into enhanced primary care services

   3. production and use of care plans for individual patients.

7. How the resources and mechanisms deployed in MCPs will contribute to changing care for different groups of people [defined by morbidity, e.g. single major condition (such as cancer), multiple low functional impact morbidities (e.g. diabetes mellitus, HT), high functional impact multimorbidity (e.g. stroke, arthritis, dementia)].

8. How referral networks are established and managed in such a way as to establish referral management systems.

9. How and under what circumstances the management of referral networks promotes (or not) the use of care plans for individual patients.

10. How and under what circumstances the voluntary sector and MCP-like networks and organisations collaborate in pursuit of the ends for which MCPs were set up.

11. How organisational culture is produced and changes in MCP-like contexts (an area lacking research despite the abundance of studies in hospital and non-health-care settings).

As previously noted, equivocal research findings suggest (to realists) areas in which as yet unknown contextual factors might be strongly influencing the effects that component mechanisms of MCPs have. The main ambiguities, requiring further research to resolve them, concerned the contexts in which:

1. ‘horizontal’ MCP-equivalent networks develop interorganisational referral networks, in particular between GPs and CHSs (or the local equivalents)

2. care co-ordination through HIT supports (or not) the:

   1. management of interorganisational referral networks

   2. diversion of suitable patients from hospital into enhanced primary care services

   3. production and use of care plans for individual patients

3. the management of referral networks promotes (or not) the use of care plans for individual patients.

Methodological research recommendations

• Comparative research to establish an optimal (i.e. accurate, usable within a reasonable time frame) critical appraisal tool for the study components necessary for refining programme theory.

• Exploratory research into ways in which consistent definitions of key realist concepts (in particular, ‘mechanism’) can be applied by those whose experience of applying realist methods ranges from ‘novice’ to ‘expert’.

• Exploratory research into how researchers and stakeholders apply mutable realist concepts in a way that is consistent with complex systems concepts.

• Evaluation of complex review knowledge translation strategies (e.g. tailored prompts, infographics, workshops, coaching, and so on) for different groups of knowledge-users.

Integrated care and chronic conditions

Database: HMIC.

Host: Ovid.

Data parameters: 1979 to July 2016.

Date searched: 25 August 2016.

Searcher: Sarah L Brand.

Hits: 3667.

Health-care divisions and strategies

Database: HMIC.

Host: Ovid.

Data parameters: 1979 to July 2016.

Date searched: 25 August 2016.

Searcher: Sarah L Brand.

Hits: 357.

Database: MEDLINE

Host: Ovid.

Data parameters: 1946 to November week 4 2016.

Date searched: 5 December 2016.

Searcher: Sarah L Brand.

Hits: 676.

Database: MEDLINE In-Process & Other Non-Indexed Citations

Host: Ovid.

Data parameters: 2 December 2016.

Date searched: 5 December 2016.

Searcher: Sarah L Brand.

Hits: 162.

Database: PsycINFO

Host: Ovid.

Data parameters: 1806 to November week 4 2016.

Date searched: 5 December 2016.

Searcher: Sarah L Brand.

Hits: 265.

Strategy: see MEDLINE in-process search strategy

Database: Cumulative Index to Nursing and Allied Health Literature

Host: EBSCOhost.

Data parameters: not applicable.

Date searched: 5 December 2016.

Searcher: Sarah L Brand.

Hits: 756.

Database: Applied Social Sciences Index and Abstracts

Host: ProQuest.

Data parameters: not applicable.

Date searched: 5 December 2016.

Searcher: Sarah L Brand.

Hits: 44.