National trends and local delivery in old age mental health services: Towards an evidence base. A mixed methodology study of the balance of care approach, community mental health terms and specialist mental health outreach to care homes

Search strategy

A systematic literature review was undertaken following established guidance. 84,85 An initial search for existing reviews was executed in the Cochrane Database of Systematic Reviews, Database of Abstracts of Reviews of Effectiveness, Health Technology Assessment database, National Health Service Economic Evaluation Database and Social Care Online in June 2008. The following databases were then searched for individual studies on 22 and 23 October 2008, from the earliest to the most recent dates available: MEDLINE, PsycINFO, Applied Social Sciences Index and Abstracts (ASSIA), HEMBASE, Health Management Information Consortium (HMIC) and Web of Science. No attempt was made to limit the searches by language (although the search terms were in English), nor to a particular geographical location or time period.

The search strategy aimed to capture not only those studies that explicitly employed the BoC approach, but applications based on the same principles, and an iterative approach was taken to the identification of potential search terms to identify that combination yielding the greatest number of relevant publications. The final strategy sought references containing any of the following phrases:

• ‘balance of care’

• ‘margin(s) of care’

• ‘marginal analysis’ or ‘marginal analyses’,

in their title or abstract, as well as work citing Mooney, an early key exponent of the approach. 71,82,86 The Web of Science search also contained terms to limit the topic area to health/social care. An example is given in Appendix 1.

Additional searches for the term ‘balance of care’ were subsequently undertaken in the System for Information on Grey Literature in Europe; the websites of a number of specialist research centres and Google; the reference lists of relevant publications were scrutinised for further studies; and experts were asked to identify missing studies.

Study selection and data extraction process

The process of selecting studies had two stages. First, one researcher screened the titles and abstracts of all citations against the initial inclusion criteria (Box 1), while a second researcher confirmed the exclusion of each rejected reference. Where decisions were not clear, the full text was reviewed, and, if uncertainty persisted, this was resolved through discussion. One researcher then read the full text of the retained references and extracted data about the key characteristics of studies meeting the full inclusion criteria (in summary, empirical studies providing data about client characteristics, service use and costs). A second researcher confirmed the inclusion of, and independently extracted data from, approximately one-third of these references, and double-checked the inclusion of, and data extracted from, all other retained publications. They also confirmed the screening-out of each excluded reference, with any inconsistencies/disagreements again resolved by discussion.

A standardised electronic form based on guidance from the Centre for Reviews and Dissemination85 was used to extract all data. This was tested and refined on a sample of five studies before full data extraction began. It contained 15 domains covering studies’ aims, settings, populations, data collection processes, analyses and conclusions (see Appendix 2).

Quality assessment

Two researchers working independently and then together determined the extent that the reported studies exhibited a number of features of good practice. These were drawn from established criteria for systematic reviews and economic evaluations, reporting standards for economic submissions to major health and social science journals and expert opinion85,87,88 (Box 2) and included questions about studies’ design, conduct and analysis. Clear coding guidance was given (see Appendix 3). Where two or more publications related to a single study, these were considered both separately and together, resulting in two sets of codes, by reference and by study.

Lastly, four summary measures were constructed to depict the proportion of good practice indicators

• exhibited by

• not applicable to

• not clearly described in, or

• not exhibited by,

for each study. In essence, these counted the number of responses in each category and expressed them as a percentage of the total.

Included/excluded literature

The search for systematic reviews identified no relevant publications. However, a small selective review of past BoC studies undertaken by one of the authors helped conceptualise the current review. 72

The electronic database search identified 328 references, of which 16 met the inclusion criteria. A further 26 citations were identified by reference tracking and experts, giving 42 citations in the final review (Figure 3). Of these, 22 were published in 22 different journals, whereas the remainder constituted a disparate mix of monographs, book chapters, discussion papers and reports. Slightly more references were identified from the 1980s and 1990s than from previous or subsequent decades. However, given the potential delay in recent reports reaching electronic databases, the general picture was suggestive of a steady flow of publications (Table 1).

FIGURE 3.

Flow diagram of the study selection process.

Thirty-three discrete studies were described in this data set, which included multiple reports of the same study and single publications describing multiple studies. Given that the review sought to elucidate the BoC methodology, it was the studies per se that were of interest, and this is the unit of analysis reported in the rest of this chapter. As there was too much information to present everything, the following material and references have been chosen to illustrate important points. Additional information is given in Appendices 4 and 5, and in Tucker et al. 89

Coverage of included studies

Table 2 confirms the longevity of the BoC approach and highlights its limited geographical employment. The origins of the BoC model have been attributed to the British government,81,90 and the vast majority of subsequent studies have been undertaken in the British Isles. 91–96 Nothing in the approach, however, limits its use to this policy context, as demonstrated by studies from Italy97 and Canada,98 nor to the national level, with local studies predominating. 71,97–101

The original model’s applicability across multiple client groups was not found in later studies, with all but two focusing on just one population, older people. 10,102–104 Again, however, there seems nothing to restrict the model’s use to this group (as illustrated by the diversity of groups represented in Table 2), nor to a particular setting. Thus, while more than half of studies echoed the then DHSS’s interest in shifts between hospital, residential and community services,105–108 almost one-third focused simply on the residential/domiciliary margin. 96,99,109,110 The remainder included studies of the potential to locate renal dialysis services in three alternative hospital settings111 and divert acute psychiatric inpatients to supported hostels. 112 Furthermore, although most studies focused only on downward shifts from supposedly more costly, institutional settings to cheaper, community provision,91,97,98 a handful considered moves in both directions. 71,113,114

Approaches to profiling clients

One key feature of BoC studies is their depiction of the needs of people supported in different settings, and in the majority of studies (n = 20) this information was collected via some form of local survey, typically completed by practitioners110,115–118 and/or researchers. 119–121 Those studies that employed secondary data generally obtained this from national data sets or surveys in other areas. 96,122,123

Although all bar two studies provided information about their data sources, only just over a half provided enough detail to judge whether or not the cases forming the basis of their analyses were (a) sufficient in number to instil confidence in the results and (b) broadly typical of the population of interest (n = 18 in each instance). In the vast majority of studies where this information was provided, cases seemed valid and representative. Nevertheless, one study’s sample was judged too small (13 people in each setting),95 and another failed to address an important subsection of the target group. 109

Detailing the original BoC philosophy, Arthur Andersen and Company81 state that when considering alternative care options it is preferable to look at groups of clients, not individuals. In practice, this means dividing the population into categories of clients (case types) with similar needs for support on the basis of those characteristics deemed most significant in determining the locus of and/or costs of their care. Of the 33 studies in this review, 23 took this approach. Table 3 lists the variables most frequently used, with the person’s ability to undertake daily activities of living,103–105,120,121 the extent of their informal care93,96,101 and the impact of their mental state/behaviour108,117,118,124 commonly thought important. Those studies concerned with the location for a specific treatment (as opposed to where different groups of people might reside) also considered such factors as the distance a person would have to travel97,111 and the severity of their illness. 97,125,126 Less frequently mentioned variables included age, gender and level of risk.

Most studies used between three and five attributes, each with two or three levels (e.g. the presence or absence of cognitive impairment), resulting in between 16 and 48 possible case types. The subgroups used in four studies were, however, considered too broad to identify clinically recognisable groups. For example one broke the population into just three ‘standard’ groups of children,127 which would not be clinically recognisable groups.

Approaches to profiling services

Although a comparison of the services people currently receive and alternative ways of meeting their needs is central to the BoC approach, the very first DHSS studies assumed that the total amount of resources available would be curbed by limits on the overall supply of services. 90,119 By contrast, later adaptations of the DHSS model ran both with and without resource constraints,128 whereas other studies tended not to restrict resources to pre-specified levels,98,122,123 though sometimes suggested that account be taken of likely financial constraints. 101,117 In order to estimate the total resource requirements of the various options, the aggregate resources proposed for the care of different individuals/case types were then compared with the resources actually used by the same population.

The range of resources considered varied according to studies’ aims, populations and margins of interest, but generally included those public services likely to account for a significant proportion of total client group spending. Hospital and care home beds, as well as commonly utilised community services, therefore featured frequently. 10,80,113,114,124 The sources of service receipt data were generally poorly detailed. However, some studies employed aggregate measures of available resources taken from routinely collected statistics,80,93,94 whereas others undertook individual-level data collections similar to those above. 71,92,99

Approaches to identifying alternative care arrangements

In the original DHSS studies, alternative care services were identified by modellers after consultation with a team of medical, nursing and social work advisors and ‘mathematical programming’ was used to estimate how practitioners might allocate these based on existing patterns of resource allocation. 80,105 Subsequent BoC studies have, however, generally taken a simpler approach. Many asked practitioners (individuals, monodisciplinary or multidisciplinary groups) to identify the most appropriate form of care for particular case types/individuals (facilitating the incorporation of new services),101,117,118,120,121 whereas others asked practitioners to identify those people in location A who could be cared for in location B. 71,91,113,114 An alternative was to draw on policy documents,103 research recommendations10,121,122 or comparative provision elsewhere. 93,94

Approaches to costing

A comparison of the costs of current and alternative care forms a key element of BoC studies. Few studies provided detailed descriptions, and it was not always clear what costs were included. Furthermore, there is not necessarily one single ‘right’ concept of costs. 129 Thus, in studies exploring how public expenditure might be reduced, it could be argued that only public costs are relevant, whereas this is inadequate when valuing wider social opportunity costs. 130 The consideration of costs undertaken in this review therefore addressed a number of different questions (Table 4).

Less than one-sixth of studies undertook a comprehensive costings approach encompassing not only those costs incurred by public agencies, but also the costs of housing, personal consumption/living expenses and informal care. 92,95,112,122,123 A further fifth incorporated some of these elements. 71,91,110 The remainder considered only public expenditure which, depending on their foci, covered the costs incurred by health and/or social services. 103,106,118,120,121 Interestingly, although many aspired to a comprehensive costing approach, there was little evidence that one framework had come to dominate the field. However, there appeared to be an order in which non-public costs were considered with more studies including housing than living expenses, and informal care costs least likely to be examined.

In all except one study the data used appeared valid, that is related to costs drawn from empirical sources in keeping with the study’s coverage. The costs of local public services were typically supplied by the relevant agencies’ finance departments, whereas national costs were calculated from statistics provided by the Chartered Institute of Public Finance and Accountancy131 or the average local costs in the studied sites. Living costs were generally taken from the Family Expenditure Survey,132 whereas housing cost sources included the estimates of an experienced valuer, national survey data and rateable values. Just two studies described the valuation of informal care costs, with one basing these on the replacement costs of formal care services,10 and the other considering the costs of foregone paid work, non-market work and leisure time. 113,114

Where costs were valid and in keeping with studies’ aims they were deemed ‘fit for purpose’. Thus, studies undertaken from a provider/commissioner perspective (interested in public expenditure) that included the most important health and/or social care costs and used valid data were scored positively, as were those that sought to calculate comprehensive costs and included the four elements detailed above. In almost two-thirds of cases, however, there was insufficient information to make this judgement. Furthermore, a substantial number of studies failed to report the year to which costs referred, whether appropriate adjustments were made for inflation or the extent to which any reallocation of resources would change the distribution of the cost burden between health and social care/the public and private sector. There were no reports of the transaction costs that might be incurred in reallocating resources between care locations or creating new services.

Approaches to the inclusion of outcomes

Consideration of the relative benefits of alternative care options was widely advocated, although only four studies reported collecting any outcome information. 97,99,101,127 A further four demonstrated awareness of existing outcome data. 10,91,102,122,123 Not all of those studies that collected outcome information explained how it was used. However, one used data on fostering breakdown to explore concerns that greater use of fostering might increase placement failure,127 whereas another used information on (among other things) the extent to which people’s needs were met and how satisfied they were to indicate care quality. 99

Most of those studies that drew on existing evidence were more explicit. One cited research on the relative benefits of community and institutional care to support its plans to increase domiciliary support,91 whereas another discussed possible changes to the support of older people with cognitive impairment based on evidence of the quality and effectiveness of each option. 122,123 More recently still, Wanless et al. 10 created an outcomes-led estimate of the costs of addressing the social care needs of vulnerable older people. This work aside, several studies used practitioners to make explicit judgements of best placements, on the assumption that ‘best interests’ evaluations of outcome were included,96,117,118 whereas others simply presented decision makers with the relevant cost data alongside a description of the individuals likely to be affected by any reallocation of resources, leaving them to judge the relative benefits of any proposed transition in terms of, say, equity, continuity, normalisation and/or effectiveness.

Good practice indicators

The preceding sections have summarised the extent to which studies exhibited the criteria of good practice identified in Box 2 and demonstrated how key elements of the BoC framework have been operationalised, but have not addressed the overall design or reporting of individual studies.

To assist such a comparison, Figure 4 ranks the studies by the number of good practice indicators exhibited (most to least). As can be seen, not one study exhibited all 16 components, whereas approaching half (n = 15) exhibited < 50%. As one moves down the figure, however, the proportion of items not reported increases more than the proportion definitely not exhibited, suggesting that study designs may be less variable than reporting standards. This distinction is reinforced by Figure 5, which ranks the studies from least to most good practice indicators not exhibited, and although the upper sections of Figures 4 and 5 are similar, the ordering of those studies in the middle and lower sections changes considerably. Interestingly, there is no indication that studies with multiple publications or described at more length performed systematically better than studies described in single publications or at less length. Moreover, there is no straightforward association with year of publication, since most of the more recent studies are in the upper half, as are some of the earliest.

FIGURE 4.

Ranking of studies by good practice criteria exhibited (most to least).

FIGURE 5.

Ranking of studies by good practice criteria not exhibited (least to most).

Methodological considerations

This review faced a number of methodological challenges, not least of which was the desire to examine not only that work that explicitly used the BoC model, but that sharing the same approach. Although the selected search terms may not have captured all relevant studies, experts highlighted only a handful of additional publications suggesting the final list was relatively complete, and just two unsourced references related to studies not captured elsewhere. The lack of methodological detail given in many publications was also problematic, and no suitable validated quality assessment tool was identified. The components of the constructed checklist were, however, selected with due consideration for the scope and purpose of the exercise (if not necessarily all of equal importance), and highlighted a number of areas in need of methodological refinement.

Lessons for future applications

Key lessons arising from this review focus on the need to minimise bespoke data collections, the formation of case types, the choice of margins, the delineation of care alternatives, the measurement of costs and the inclusion of outcome data.

Activity 1.1

Multiple, secondary data sources were used to develop an overview of the existing distribution of resources for older people with mental health problems, providing a baseline against which future changes in service provision could be considered. These data were compared with published national findings in order to benchmark service provision in the study sites against other areas, providing a wider context for the research.

Activity 1.2

Front-line practitioners completed bespoke surveys providing information about the sociodemographic, functional, clinical and service receipt characteristics of service users in each setting (the baseline data collection) drawing on routinely collected data (Box 3). In the ECH, care home and inpatient samples, they also indicated which of a list of factors identified from the literature contributed to the admission. The questionnaires (see Appendices 6 and 7) were designed to collect the minimum amount of information necessary to profile the populations of interest, including standardised measures such as the modified Barthel Index,138–140 the Cognitive Performance Scale (CPS)141 and, where available, the Mini Mental State Examination (MMSE),142 Geriatric Depression Scale (GDS)143 and Hospital Anxiety and Depression Scale (HADS). 144

All information supplied to the research team was provided in a fully anonymised form. However, in order to facilitate the identification of people for whom outcome data would be sought in Activity 4.4, each proforma included a unique study or non-sensitive agency number. As the participating social services’ documentation did not reliably distinguish older people with mental health problems from other older people, the relevant forms also contained a brief screen to identify the study population for whom the full data set was then collected.

The sampling strategy aimed to identify individuals at a point on the care pathway where real life choices are made between settings, while simultaneously generating sufficient numbers for the formation of case types (see Activity 2.1) and the power required by the planned outcome analysis (see Activity 4.4). The focus on people with an organic mental illness in the CMHT sample was also influenced by the need to match CMHT clients with care home admissions (see Activity 4.4), whereas the small size of the ECH sampling frame necessitated a whole population approach.

Data were initially entered onto SPSS for Windows (version 19; SPSS Inc., Chicago, IL, USA) and checked for errors, while subsequent analyses were conducted with Stata (versions 10 to 12; StataCorp LP, College Station, TX, USA). Where data permitted, critical gaps were filled by model-based imputation routines and scores for summary measures were calculated from their constituent elements (see Appendices 8 and 9). Differences between groups were explored using appropriate statistical tests.

Activity 2.1

The study samples were divided into relatively homogeneous subgroups (case types) on the basis of those client characteristics deemed likely to be important in determining the locus of/and/or costs of their care. The attributes employed (Box 4), were informed by the literature review (see Chapter 2), the wider literature on care home and inpatient admission predictors and exploratory analyses of the empirical data. Further details are given in Appendix 10. As each of the three variables used in the domiciliary, ECH and care home settings had three levels, this gave 27 possible different combinations or case types, whereas the four variables used in the inpatient classification generated 72 possible subgroups.

Activity 2.2

Vignettes were formulated to exemplify the most prevalent case types for care home admissions in site X (see Appendix 11). Based on real individuals, they took the form of brief case histories which systematically incorporated information about the three key variables as well as individuals’ falls risk, physical health, living situation, service receipt, location immediately prior to admission and preferences. A few control vignettes were also constructed to depict two of the same case types in the CMHT sample as well as the most commonly populated CMHT case type, and the homeogeneity of the needs of the people represented by each vignette was checked.

A similar process was used to formulate vignettes representing the inpatient sample, but depicting the most prevalent case types across the three sites to facilitate wider exploration of the appropriateness of inpatient admission. They also included information about individuals’ mental health history (see Appendix 12). In one instance, where the needs of the individuals represented by a particular case type appeared too heterogeneous, the relevant case type was divided according to the individuals’ usual place of residence (i.e. home or care home).

To adhere to the study timeline, this exercise was started approximately half-way through the baseline data collection exercise.

Activity 3.1

First, an empirical analysis was undertaken of the extent to which the same case types were found in the two domiciliary, ECH and care home samples.

Activity 3.2

Second, local staff attending a series of practitioner workshops identified those commonly populated care home and inpatient case types whose needs could be met by other services, scoping the potential for downwards substitution from the two most intensive settings. Box 5 sets out the workshops’ key characteristics.

Participants in the care home workshop worked in five small groups, with each group pseudo-randomly allocated a set of pre-selected care home and domiciliary control vignettes.

Working alone, practitioners first indicated where they believed the depicted ‘client’ would be most appropriately cared for and, where care home placement was felt preferable, recorded the main reason for their decision. They also indicated if their decision would be different if (a) the service user lived with or without a coresident carer and (b) was at home or in hospital at the point of assessment. Working collectively, each group then identified between two and four cases that could appropriately be diverted from care home admission (‘marginal care home cases’) and compiled care plans specifying the alternative community resources required.

In this exercise participants were asked to put aside short-term constraints in current services and be creative, while remembering that all services inevitably have funding implications. To help practitioners to think creatively, they were given a care planning sheet containing a list of services already available in other areas.

Employing a similar process, participants at the inpatient workshops first individually indicated whether it was ‘completely’, ‘possibly’ or ‘not’ appropriate to admit the ‘client’ to a mental health bed. To construct a hierarchy of appropriateness, case types were then scored: two points were given for each respondent stating ‘completely’; one point for each respondent stating ‘possibly’; the points were totalled; and totals expressed as a percentage of the maximum possible. Working in small groups, practitioners then formulated care plans to enable the four to six lowest scoring admissions in each workshop (‘marginal inpatient cases’) to be diverted from hospital care.

Activity 3.3

Statistical modelling was undertaken to illuminate the decision-making process. The individual decisions made during workshops (clustered by rater) formed the analysis data set, and a series of logistic regression analyses were undertaken to explore the factors associated with those care home and inpatient case types that were definitely not divertible, as well as rater characteristics that may have influenced these decisions.

Activity 4.1

A comprehensive approach was employed in the costing exercise for the marginal care home case types. This covered the most important (expensive or commonly incurred) public sector and service user costs that might be expected to vary between settings.

The main health and social care costs stemmed from the receipt of services, information on which was drawn from the baseline data collection exercise (see Activity 1.2) and practitioner workshops (see Activity 3.2), with any gaps filled by estimates from published evidence. Wherever possible, service-related costs were calculated from data provided by the participating agencies in the main research site. Where local costs were unavailable, however, figures were based on national sources. Information on the receipt of benefits (including attendance allowance, housing benefit and council tax benefit) were similarly obtained from external sources, whereas data on expenditure retrieved through charging for formal social care services was obtained from the LA.

The estimation of private costs (i.e. costs to service users and their carers) was similarly derived from information collected in the baseline data collection exercise, including where service users lived, with whom they lived, and the extent of their informal support. The living expenses of people residing at home were informed by data from the Family Expenditure Survey, and figures for the opportunity costs of owner-occupied housing were calculated by imputing a rent from comparable market values. Informal carers’ time was also costed using an equivalent market value approach.

Costs were adjusted to take account of the real distribution of service receipt within case types; confidence intervals were calculated to reflect uncertainty about the quantity and cost of resources used in the alternative planning scenarios; and a range of sensitivity analyses were undertaken to explore the effects of different costing assumptions. The potential for cost-shifting between health and social services, the public sector and private individuals was also explored. Further details of the specific costs and general approach used in this exercise are given in Appendices 13 and 14.

A similar approach was taken to the costing of the marginal inpatient case types. However, in the absence of good information on the general health care, private and other government costs for these service users, this focused mainly on the potential trade-offs between mental health inpatient, mental health community and social services costs.

Activity 4.2

Three approaches were employed in the identification of outcomes of people with similar needs for care supported in different settings. The first was a selective literature review that focused on the outcomes of older people with mental health problems supported in ECH as compared with domiciliary or care home settings. This involved interrogating several electronic databases [including PubMed, PsycINFO and Cumulative Index to Nursing and Allied Health Literature (CINAHL)] using search terms such as ‘care home’, ‘extra care housing’, ‘dementia’, ‘older persons’, ‘outcomes’ and ‘effects’, with similar search terms used to identify relevant grey literature. Studies that examined the effects of ECH on residents’ outcomes or investigated the margins between ECH and domiciliary or care home placements were included, and data was extracted on all reported outcomes, including social engagement, activities of daily living (ADL) and cognitive status.

Activity 4.3

The second approach involved a search for secondary national data sets that contained the study target populations, such as the English Longitudinal Study of Ageing. Relevant publications and sources such as the UK Data Archive were also considered.

Activity 4.4

The third approach explored the feasibility of collecting primary outcome data. It was designed as a matched cohort study, and aimed to compare the 4-month outcomes of 120 marginal care home cases, represented by a small number of case types, with those of a matched sample of 120 service users who received CMHT support at home. Examples in Box 6 illustrate how the sample size was determined.

To be eligible for inclusion in the cohort study, service users were required to:

• have remained in their respective care setting for the 4-month period since baseline data collection

• be physically well enough to participate in an interview

• be mentally well enough to participate in an interview (e.g. not unduly anxious or agitated) and

• be able to meaningfully answer questions about their care experience as judged by their keyworker/care co-ordinator.

These staff also judged service users’ capacity to participate in the study, identifying a family member/informal carer possibly willing to act as personal consultee where capacity was lacking.

Potentially eligible service users (or consultees) were first approached about the study by their key worker/care co-ordinator and, if interested, were sent a study information sheet and invitation to participate by the research team. A Personal Social Services Research Unit (PSSRU) interviewer subsequently telephoned to provide further details and, subject to verbal agreement, arranged to visit. Written consent or a written record of consultation was gained prior to the formal interview. A summary of the measures this encompassed is given in Box 7 and the full interview schedule forms are in Appendices 15 and 16.

Activity 5.1

Details of the marginal care home case types were presented at a specially convened meeting of the site X Resource Allocation Management (RAM) Panel which, in effect, acts as gatekeeper for access to care home placement. These utilised local proformas and included information on service users’:

• situation, presentation and needs

• original care home placements (including their cost to social services) and

• alternative care plans (including their cost to social services) as devised by local practitioners in Activity 3.2.

The RAM Panel then decided whether or not the alternative support was appropriate; sufficient to enable the people represented by each case type to stay in the community; and fundable, mirroring local decision-making processes.

Activity 5.2

The RAM Panel was next given further information about two case types and asked whether or not this would affect their placement decisions. In essence, the additional information explored the weight the RAM Panel gave to service users’ preferences, carers’ preferences and carers’ stress levels and evidence about service user outcomes.

These exercises were recorded and the discussion transcripts were subjected to a thematic analysis using a framework approach to identify both the manifest and latent content of the conversation. Three researchers from different professional backgrounds independently coded the entire script to give greater reliability to the development of themes and increase the trustworthiness of the analysis. 157

Activity 6.1

The acceptability of the RAM Panel’s decisions was reviewed at two workshops attended by a mix of older people who either had, or supported someone with, dementia. Participants first read a small random selection of (appropriately modified) vignettes about recent care home entrants for whom the Panel had approved an alternative community care package. Working in small groups they decided whether or not the original care placement or the suggested alternative was preferable, noted the main reason for their decision and listed any other support they felt was needed.

Activity 6.2

In a similar exercise, a number of acknowledged experts in the care of older people with mental health problems reviewed a sample of the marginal care home cases by examining details of actual care placements and proposed alternatives. Participants were asked to identify their preferred care option, the main reason for their decision, and to specify any further support they deemed necessary.

An equivalent exercise was undertaken for marginal inpatient case types, with respondents also asked to indicate how appropriate they felt inpatient admission was.

Activity 7.1

The final stage of the study consisted of an analysis of the potential for local agencies to shift the BoC provided for older people with mental health problems. In essence this explored the resource implications of caring for different combinations of those care home entrant and inpatient case types thought to have the most potential for diversion from institutional to community care and the effects of any change on the wider care system. Although the care home analysis related to site X alone, the inpatient exploration encompassed all three sites.

In order to understand the full cost implications of different BoC scenarios, deterministic sensitivity analyses were used to explore the impact on the results of changing a number of structural aspects of the model. For the care home analysis, these included the proportion of service users in each case type that might realistically be cared for in alternative settings and the length of time it might be possible to divert them from care home entry. This latter factor had important implications for the attribution of so-called ‘one-off’ costs (e.g. an expensive inpatient stay for assessment purposes). In order to standardise cost components all costs were annualised. First, the likely annual prevalence of potentially divertible care home entrants was estimated from the baseline data collection, assuming current conditions prevailed. Second, total annual costs were calculated for all cost components by combining the alternative costs incurred during the chosen diversion period with the ‘original’ costs incurred during the remainder of the year. For example, where it was thought a case type could be diverted for 12 months, only alternative costs were incurred, whereas where it was thought a case type could be diverted for only 6 months, 6 months’ worth of alternative and 6 months’ worth of original expenditure were incurred. Third, the linear cost trade-offs associated with increasing diversion ‘success rates’ were illustrated by means of cost tables and charts.

The inpatient analysis followed a similar approach, albeit this considered only those costs directly associated with the inpatient stay and/or the alternative packages provided to prevent admission. An adjustment was also made for the possibility that service users in any marginal inpatient case type may have more than one admission per year based on information from the baseline data collection, whereas estimates of the likely length of inpatient stay for each case type were informed by the discharge data collection.

The aggregate costs associated with the inappropriate admission of service users in each case type were then calculated: N × weekly costs × time on ward × repeat admission multiplier. In order to calcuate the costs of the alternative care needed to prevent such admissions, however, it was necessary to make a number of assumptions about the likely duration of the proposed care packages. The aim was to include only those inputs that acted as direct substitutes for inpatient admission. Thus if, for example, a social care package was recommended to prevent an inpatient spell of 50 days and it seemed likely that the patient would have received the same or a very similar package on discharge from hospital anyway (based on their obvious lack of critically needed formal support prior to admission), then only the ‘extra’ cost of the 50 days diverted from hospital was considered as an alternative social service cost. Following a similar principle, the alternative community mental health-care package was costed exactly as specified at the beginning, but assumed to reduce over time to a point at which it would not differ from the ‘normal’ level of support that might be expected after discharge from hospital anyway.

Activity 7.2

A range of key stakeholders from the participating sites, including service commissioners, managers, providers, front-line staff, older people and carers were invited to a presentation and discussion of the study’s main findings. This acted as a form of respondent validation and provided further understanding of the main issues. Attendees were also invited to comment on the impediments to health and social care staff participating in research and ways of enhancing service user recruitment.

Older people living at home with social services support but no specialist mental health input (site X)

Local authority information indicated that, in June 2011, there were 2375 older people (aged ≥ 65 years) recorded on the locality teams’ caseloads, of whom over three-quarters lived at home. Of the latter, almost 4 in 10 were open to a named practitioner. The original target sample size was 100, and baseline information was sought for 199, which although considerably less than originally requested, represented the maximum number of cases local managers deemed acceptable at a time of great organisational change given the lack of financial incentives for local authorities to participate in research.

Completed forms were returned for 128 service users, of whom 38 were excluded because they had no indication of mental health problems; a further 37 received specialist mental health input; and three no longer lived at home. This gave a study sample of 50 (i.e. approximately 64% of the maximum achievable number of eligible cases).

The first column of Table 7 details the sociodemographic characteristics, dependency and general health of this service user group. Mirroring the sampling frame, females outnumbered males by roughly two to one, while more than four-tenths were in the oldest age category, over one-quarter were very/totally dependent and a similar proportion had poor health. Over half lived on their own compared with some 35% of older people in site X generally.

As shown in Table 8, just under half had a known diagnosis of dementia. However, given their lack of contact with specialist mental health services (who often make such diagnoses), this may well be an underestimate. Indeed, Table 9 indicates that over three-quarters had at least mild cognitive impairment. Furthermore, although we lack data on severity, approximately half (53%) displayed indicators of low mood or anxiety. Similarly, over half had moderately or highly challenging behaviour.

Given this picture of general frailty plus cognitive impairment, the high proportion of people considered at risk of falls or self-neglect in Table 10 is perhaps not surprising. Almost one-quarter were reported to be of medium or high risk of accidental self-harm, with a similar proportion considered at risk of abuse by others, illustrating the vulnerability of this group.

Although the vast majority of service users had at least some informal care, and approximately two-thirds received ≥ 8 hours per week (Table 11), the nature of this help varied. Around half of carers (typically service users’ children) provided assistance with general tasks (e.g. shopping, housework) and emotional and social support, whereas the other half also undertook personal care. As expected, the overwhelming majority received formal social care, though only just over one-quarter received an intensive care package.

Older people living at home with community mental health team support (sites X, Y and Z)

The original target sample size was 900 and, based on information supplied by practitioners, at the point of sampling there were 1277 people on the CMHT caseloads. However, of these, only 926 (73%) were aged ≥ 65 years and lived at home [significantly less than the number indicated by the participating trusts’ information technology (IT) systems at the study design stage].

Baseline information was sought for 542 cases and completed forms returned for 501. Of these, 14 were excluded because they no longer lived at home/had moved out of area, 10 were duplicates, eight had not been assessed/seen in the past 3 months, seven were in hospital and five had died. This gave a study sample of 457 (i.e. approximately 93% of the maximum achievable number of eligible cases).

Of the 141 who were from site X, the proportion of women on the CMHT caseload (see column 2, Table 7) was similar to that in the social services sample. However, the CMHT clients were significantly more likely to be physically independent (design-based chi-squared: 95.34, F = 1,187, p < 0.001) and were in better physical health (design-based chi-squared: 8.41, F = 1,190, p = 0.004). They also appeared more likely to live with others, although this was not statistically significant (design-based chi-squared: 0.87, F = 1,190, p = 0.351).

Such differences are, at least in part, likely to be age related, for the CMHT sample were significantly younger (mean age of 77.3 years) than the social services caseload [mean age of 83.6 years; t-test/weighted regression: t = –5.09, degrees of freedom (df) = 186, p < 0.001]. However, they may also be influenced by the different referral criteria the two agencies operate. Thus, while over a third of the CMHT caseload had a known organic disorder (including people with co-existent functional mental health problems), more than four-tenths had a primary diagnosis of anxiety and/or depression (see Table 8).

Table 9 confirms that the CMHT caseload were significantly less likely than the social services sample to be cognitively impaired (design-based chi-squared: 16.78, F = 1,190, p < 0.001) and displayed significantly lower levels of challenging behaviour (design-based chi-squared: 4.89, F = 1,189, p = 0.028). However, they did not have more indicators of low mood or anxiety (design-based chi-squared: 0.60, F = 1,183, p = 0.439). This variation is again likely to be related to their respective diagnostic profiles. Indeed, further analysis of just cases with dementia showed that, although the CMHT clients had a younger, more independent profile than their social services counterparts, their affect and behaviour were very similar.

Given these differences, the significantly smaller proportion of the CMHT than of the social services caseload considered at high risk from falls or self-neglect is understandable, for both are likely to be frailty/dependency related, whereas the CMHT caseload contained a greater proportion of people considered at moderate risk of deliberate self-harm (see Table 10). The caseloads’ differing sociodemographic and clinicial profiles are also likely to explain the different patterns of support shown in Table 11, with a significantly higher proportion of the CMHT caseload receiving no informal (design-based chi-squared: 22.0, F = 1,188, p < 0.001) or formal social care (design-based chi-squared: 48.63, F = 1,187, p < 0.001). Not surprisingly, however, the vast majority of the CMHT caseload had received CMHT input in the previous month, and approaching one-sixth had a mental health inpatient admission in the previous 6 months.

In comparison with the site X CMHT sample, the sites Y and Z caseloads had a somewhat older, more dependent profile. The site Z sample also contained a higher proportion of people with dementia.

Extra care housing residents (sites X, Y and Z)

Information was collected about 32 older people with mental health problems living in ECH (original target sample size n = 30). Of these, nine were generic ECH tenants known to site X’s council, whereas the remainder were tenants of specialist mental health facilities identified by sites Y and Z’s CMHTs. Although not a random sample, these arguably represent the range of older people with mental health problems for whom ECH was provided in the study area.

Given their small number and mixed provenance, no statistical analysis was undertaken on these data. Nevertheless, the fifth column of Table 7 suggests that although this group had a similar age profile to the site X CMHT caseload, they were in poorer health and more dependent, if not as dependent as the social services sample. As might be expected, the vast majority of tenants (87%), about whom CMHTs provided information, but fewer site X cases (67%), about whom generic social care teams provided data, had a formal psychiatric disorder. Moreover, whereas most of the site X tenants had dementia, the majority in site Y had a functional mental illness and the site Z sample had a more mixed profile (see Table 8). Understandably, then, the site X tenants were rather less likely to display indicators of anxiety or depression and rather more likely to be cognitively impaired than tenants in other areas (see Tables 9 and 10). However, levels of challenging behaviour (which were somewhat higher than in the previous samples) and risk were simliar in all sites.

In terms of their level of support, Table 11 shows that, although the majority of ECH tenants (all in site X) received informal care, the level of this was generally lower than that in the social services or CMHT samples, while in no instance were family or friends providing personal care. On the other hand, all bar one tenant received formal social support, and roughly one-third received an intensive care package (similar to the social services sample), with the risk of self-neglect said to have contributed to approximately four-tenths of moves to ECH (Table 12).

Perhaps not surprisingly given the nature of the sampling frame, the most frequently reported reason for moving into ECH was the person’s mental health. Indeed, this was the main reason in approximately one-quarter of cases (Tables 13 and 14). Physical health problems and an inability to perform daily activities of living were also commonly given reasons for moving into ECH, although these were rarely the main reason, with other, often related, factors such as the risk of self-neglect or breakdown of care, carrying more weight. Service users’ preferences were said to have contributed to the move in four-tenths of cases.

Care home entrants (sites X, Y and Z)

The original target sample size was 350, but recruitment to this strand of the study was severely affected by the unexpected decision of the Council in site Z not to participate in the research, despite prior commitment. Nevertheless, information was collected about 197 care home entrants (including 104 from site X out of a possible 150 in the data collection period). Of these, 28 (14%) were excluded because they had no identified mental health problems. This gave a study sample of 169 (i.e. approximately 69% of the maximum achievable number of eligible cases).

In contrast to the domiciliary and ECH samples, all of whom were resident in their settings, these individuals were described as they presented just before care home entry. Most did not move into a care home directly from home. Indeed, of the 164 people for whom this information was available, 81 (49%) were transferred from a general hospital ward and 15 (9%) from a mental health ward.

The sociodemographic, functional and general health characteristics of the site X care home entrants are detailed in Table 7 (sixth column). Women again outnumbered men by roughly two to one and the mix of people living alone/with others was also similar to the social services sample. The mean age (84.8 years) of the care home entrants was, however, significantly higher than in previous settings (mean of 78.2 years; t-test/weighted regression: t = 6.97, df = 275, p < 0.001). Accordingly, they were more likely to need help with daily activities (design-based chi-squared: 140.46, F = 1,276, p < 0.001) and experienced worse physical health (only 30% reported to be in a least good health; design-based chi-squared: 7.84, F = 1,280, p = 0.006).

As might be expected, approaching nine-tenths had at least mild cognitive impairment, and a similar proportion of those with formal psychiatric diagnoses had an organic mental health problem (predominantly dementia) (see Tables 8 and 9). Indeed, over one-third had severe/very severe impairment, and nearly four-fifths presented moderate or worse levels of challenging behaviour, which typically included agitation, aggression and/or wandering, setting them apart from previous samples. Anxiety was more evident in this group, while the fact that more than half were considered at risk of falls (see Table 12) indicates their physical as well as mental frailty.

This group received similar levels of informal support before admission to that of the social services domiciliary sample, the vast majority having had some help, typically from their children (see Table 11). However, although just under half had an intensive formal social care package, approximately one-fifth received no/minor social care input. Furthermore, less than half had received recent mental health support.

Consistent with this profile, the most commonly cited reason for care home admission was the service user’s inability to undertake daily activities of living (see Table 13), although this was never the only factor implicated, and very rarely the main one (see Table 14). As with the ECH sample, mental health problems were the most important driver for admission, while physical health problems also featured highly as both ‘a’ and ‘the main’ reason for admission. Other factors cited in at least one-third of cases included carer stress and the risk of falls, self-neglect and accidental self-harm.

Compared with site X, site Y entrants were slightly younger and rather less likely to have been living alone, but had similar health and dependency profiles. Reported levels of cognitive impairment and challenging behaviour, and formal and informal care inputs were, however, all higher in this site. Site Z cases, all identified by CMHTs, presented a rather different subgroup, with a higher proportion experiencing functional mental health problems, while remaining relatively independent. Nonetheless, nearly all presented with at least moderately challenging behaviour and over one-third with a recent mental health admission.

Acute mental health inpatient admissions (sites X, Y and Z)

The original target sample size was 300, but recruitment to this strand of the research was affected by a number of issues, including the closure to admissions of one of the participating wards. However information was collected about 216 admissions to acute mental health beds (67 in site X) of which the vast majority (96%) related to people who had a single admission episode in the 6-month data collection period.

Within the site X sample (see ninth column, Table 7) the gender ratio was very similar to that in other settings, though a significantly lower proportion of women had an organic rather than a functional diagnosis (chi-squared: 4.13, df = 1, p = 0.042). Although the whole sample’s age profile was relatively young and independent, the mean age of the organic cases (80.0 years) was somewhat older than that of the functional ones (mean of 77.1 years; one-sided t-test: t = 1.33, df = 64, p = 0.094; note, the effect is highly significant in a two-sided t-test when pooling all three locations). Organic cases appeared to be in better health (chi-squared: 3.39, df = 1, p = 0.066) with 48% of functional cases (as opposed to 24%) reportedly suffering poor health.

Although these findings were not unexpected, the very low proportion of inpatients who lived alone prior to admission was more surprising, possibly arising by chance. An alternative explanation is that the one-sixth of admissions from care homes previously lived alone. Certainly, the admissions from care homes formed a distinct subgroup, with more than one-third being very or totally dependent.

Diagnostic and clinical profiles (see Tables 8 and 9) suggest that people with a functional disorder outnumbered people with organic problems by roughly two to one. A very high percentage of inpatients exhibited indicators of low mood, while the proportion who exhibited at least moderately challenging behaviour was also greater than in any other setting, highlighting the severity of this sample’s presentation. Almost half were described as often agitated/restless; a similar proportion displayed at least occasional delusions/hallucinations/paranoia; still more were sometimes or often disturbed at night; and almost four-fifths were considered at unacceptable risk of deliberate self-harm, harming others or self-neglect (see Table 12).

Approximately two-thirds of the inpatients who normally lived at home received some informal care before their admission, most commonly from their spouse (see Table 11). Indeed, half received more than 20 hours of help per week. However, less than one-quarter were said to have had any formal social support, and only one had an intensive care package. Interestingly, most (over seven-tenths) were receiving community mental health input, and approximately one-sixth had undergone a recent inpatient admission.

Table 15 shows the two most frequent reasons for inpatient admission as behavioural management and medication review, whereas other factors contributing to at least one-third of cases were the risk of self-neglect, the need for general diagnostic assessment and carer stress. As in the care home admission sample, these were not necessarily the main reasons for admission. Although behavioural management was the most frequently cited main driver for admission, this was closely followed by the risk of harm to others for the organic and care home subgroups, and the risk of deliberate self-harm for the functional subgroup (Table 16).

Compared with their site X counterparts, the site Y and site Z samples were younger and more likely to live alone. They also had a higher proportion of men, but a similar proportion of care home residents. The site Y sample included a higher percentage of people with an organic mental health problem (chi-squared: 7.28, df = 2, p = 0.026), whereas the site Z cohort seemed to be in better physical health (chi-squared: 34.52, df = 6, p < 0.001) and more independent. Levels of challenging behaviour were high throughout, and the need for behavioural management and an unacceptable risk of deliberate self-harm featured among the top three reasons for admission in all sites. The proportion of admissions with previous CMHT support or inpatient admissions were also almost identical.

Although the previous four settings may be perceived as people’s permanent residence, an inpatient ward is generally a short-term placement. In resource terms, this raises questions about the appropriateness of admission, the timing of discharge and the extent to which certain inpatients stay longer than necessary. To investigate this, discharge data was obtained for approximately two-thirds of the sample (Table 17). This covered around four-fifths of admissions in sites X and Y, but less than half in site Z where baseline data collection started later.

The mean length of stay (LOS) for completed inpatient episodes was nearly 68 days (Table 18). However, people with organic mental health problems typically stayed longer than people with functional disorders (Mann–Whitney: z = –2.50, p = 0.012), and inpatient stays in site Y were longer than elsewhere (84 days compared with 64 and 57 days) (Mann–Whitney: z = –1.81, p = 0.070).

Of 100 patients admitted from their own home (either directly or via another ward/service), most (81) returned there, while one person was discharged to ECH and 16 others (12 with organic disorders) entered a care home, including four in site X. In the great majority of cases (123/129, 95%), the discharge setting was deemed optimal in which to meet their needs, with four of the remaining six entering a care home when a less intensive care setting was judged preferable.

Only a minority (58 of 132 admissions, 43.9%) were discharged on the date they were deemed medically fit for discharge, with the remainder experiencing some delay. Although in 22 cases (29.7%) this was just 1 or 2 days, while discharge arrangements were confirmed (medications obtained, transport arranged, etc.), 34 discharges (46.0%) were delayed by at least a week, and 12 (16.2%) by more than a fortnight. The most commonly given reason for longer delays was the need to organise a suitable care home placement, including the identification of an appropriate home, funding and assessment by care home staff.

An empirical analysis of the overlap between the domiciliary, extra care housing and care home samples

As seen in Table 19, of the 27 case types used to characterise the domiciliary, ECH and care home samples, 26 were populated. Together they captured 94% of the social services sample, 99% of the CMHT sample, 97% of the ECH sample and 94% of the care home sample. However, some case types were more prevalent than others and their distribution varied across settings.

The most common case types were case type 1 in the two domiciliary settings (low dependency, low cognitive impairment, low challenging behaviour), case type 2 (low dependency, low cognitive impairment, medium challenging behaviour) in the ECH sample and case type 5 (low dependency, medium cognitive impairment, medium challenging behaviour) for the care home entrants. In order to render the text more readable, in the remainder of this report these have been abbreviated to LLL, LLM and LMM, with all other combinations of the three defining characteristics following the same pattern.

Case types commonly found in more than one setting (i.e. containing at least 5% of two or more samples) might theoretically represent people on the margins of care, although given the small numbers in some of these cells, it clearly is not possible to say definitively these are the boundaries offering the biggest scope for service substitution. The emerging distribution (Table 20) nevertheless gives some indication of the likely spread of cases, and the potential for both upward (more intensive) and downward (less intensive) service substitution across settings.

The identification of potentially marginal care home entrants by practitioners

By contrast, Table 21 identifies which of 14 commonly populated care home case types (those representing at least four site X admissions at the point of analysis) local practitioners believed could be more appropriately cared for in other settings. Seventeen staff members participated in this exercise, and each case type was considered by at least six individuals.

Although the degree of consensus between practitioners varied across case types, staff working in small groups concluded that nine case types could appropriately be diverted from care home admission. However, they did not necessarily believe they should remain at home, with ECH often seen as a better option.

As might be expected, the needs of those case types considered suitable for diversion were generally less complex than those for whom care home placement was deemed the best option. For example, the people represented by just two of the nine case types for whom an alternative community setting was preferred were highly dependent, compared with four of the five for whom care home placement was supported. The logistic regression model presented in Tables 22 and 23 shows that case types with high physical dependency were almost five times more likely to be deemed in need of care home placement than those with moderate or low levels of dependency. Case types displaying high levels of two or more of the key variables (physical dependency, cognitive impairment and challenging behaviour) were approximately 10 times more likely to be considered best placed in a care home over and above this. A case type’s age and gender had additional positive impacts on the perceived need for care home placement.

Interestingly, living situation appeared to have a less clear effect. For no case type was it suggested that, had their living situation been different from that depicted, staff would have been more likely to advocate care home placement. Indeed, in one instance where care home placement was felt to be the best option (case type 18, MHH) there was a suggestion that, had the individual lived alone, staff might have been more inclined to try to maintain him at home (the need for care home admission at least partly driven by carer stress). Whether or not the depicted ‘client’ was at home or in hospital also appeared to make litttle difference to staff recommendations.

Although more will be said about the alternative care options suggested for the nine case types discussed in Chapter 5, it is noteworthy that together they constitute more than half (53%) of the site X sample. Of the four group decisions relating to people from the domiciliary samples (i.e. the controls), three confirmed that home was the most appropriate placement, whereas one advocated specialist ECH as a more appropriate setting for people in case type 5 (LMM). Furthermore, although there was some discrepancy between the recommendations made for the vignettes representing the care home entrants in case types 2 and 5 and their counterparts in the community, care home placement was never the preferred option.

The identification of potentially marginal inpatient admissions by practitioners

Of 72 case types used to characterise the inpatient sample, 46 were populated, which between them captured 98% of admissions overall (100% in site X, Table 24). Again, some combinations of characteristics were more prevalent than others and only those 16 case types that contained four or more admissions at the point of analysis were considered for further exploration. However, as one of these contained two slightly different subgroups relating to individuals’ place of residence (home or care home), these were subsequently treated separately and 17 case types were used in the subsequent analysis.

In Table 25 these 17 case types are ranked in order of appropriateness for admission (from most to least) as judged by groups of local staff. Thirty-eight staff members participated in this exercise, and each case type was considered by at least 17 individuals. Thus, of 22 practitioners who rated case type 2 (which was considered most suitable for admission), 20 believed the ‘client’ depicted was completely appropriate for admission, whereas no-one considered the admissions represented by case types 26 or 31 to be definitely suitable for admission.

Logistic regression suggested that a variety of intrinsic and extrinsic factors played a role in predicting those case types considered definitely suitable for admission. Where the main risk of admission was deliberate self-harm (as opposed to any other reason), the likelihood of admission being seen as definitely appropriate increased more than fivefold (Tables 26 and 27). Living alone or lack of any previous contact with mental health services were also positively associated with the probability of being considered an appropriate admission. Other tested intrinsic factors either had no effect (e.g. age and gender) or were already explained by the main predictors (e.g. diagnosis and whether or not the person was admitted in normal working hours). However, certain rater characteristics did affect decisions. Compared with clinical staff, managers and commissioners were only half as likely to perceive inpatient admission as definitely appropriate, while staff who worked all or some of the time in community settings were less likely to fully endorse inpatient care than staff who worked solely in inpatient settings.

As any attempt to change the BoC might logically be expected to focus on those case types considered least appropriate for admission, the potential implications for admission can be estimated. Depending on the level of ambition, diversion of the three case types at the bottom of the distribution (6, 26 and 31) to community care might reduce inpatient admissions in site X by over 10%. If it was also possible to support case types 10 and 28a in the community, a reduction of more than 20% might be feasible. However, whether or not this would be a better option than the status quo would depend on the relative costs and consequences of the alternative options. These issues are the subject of Chapter 5.

Costs of alternative care options (see Chapter 3, Activity 4.1)

The estimated costs of the community care arrangements deemed most appropriate by local practitioners to meet the needs of the nine marginal care home case types identified in Chapter 4 are detailed in Table 28, alongside those of their actual residential care placement. Further information about their constituent elements is given in Box 8. A number of the alternatives involved moves to ECH (variously specialist or generic). Other commonly employed resources included home care (again both specialist and generic) and support from specialist mental health services [predominantly community mental health nurse (CMHN) and psychiatrist input]. Considerable use was also made of telecare, including pendant alarms and smoke, gas and fall detectors. Where more than one alternative was provided these are presented in the table.

Taking a comprehensive costings approach, the total predicted costs of the alternative plans exceeded those of care home placement in every case. However, some were relatively more expensive than others. Additional expenditure per service user ranged from £31 to £1188 per week, with a mean difference of almost £320 per week. Much of this can be attributed to the increased private (personal) costs associated with the alternative packages, which averaged £195 per week more than for residential care placement. In comparison, public expenditure varied less. Seven community packages increased public costs and seven decreased them. There was also a clear shift in the distribution of public expenditure, with NHS and other government costs typically higher for alternative options and social services costs generally higher for care home options.

Although these figures are best or central estimates, the upper and lower bounds in Table 28 represent the possible extremes of any cost differences, allowing for uncertainty in the quantity and cost of resources and the representativeness of vignettes. These, understandably, introduce much more variation into the model. Nevertheless, even after such adjustments, the alternative care packages for case types 6 and 9 remain considerably more expensive than their original care home placements.

Outcomes of alternative care options (see Chapter 3, Activities 4.2–4.4)

A marked lack of relevant literature comparing the relative outcomes of people with similar needs supported at home, in care homes or ECH was identified. Furthermore, although a number of publications addressed the outcomes of older people with mental health problems cared for in particular settings, none provided enough detail about their samples to match them to this study’s case types. Although mostly originating from small, descriptive studies, there is, however, a growing body of evidence to suggest that social well-being and quality of life (QoL) in ECH are generally good. 164–166 In one recent study, people who moved into ECH reported significantly improved levels of QoL and decreased levels of unmet need compared with people living at home. 167

The search for secondary data sets containing comparative information about the outcomes of older people supported in different settings similarly proved unfruitful. Even the English Longitudinal Study of Ageing included relatively few older people who received formal mental health support, and fewer still who had entered a care home. This may reflect the relatively young age of the study’s participants. However, even if the numbers had been larger, the data would have required considerable manipulation to be suitable for our purposes.

The attempt to collect primary data (matched cohort study) about the relative outcomes of care home entrants and older people living at home with CMHT support also encountered difficulties. Not least of these was the Council in site Z’s unexpected late decision not to participate in the study. Both the achieved care home sample and the number of people represented by the care home case types were considerably smaller than anticipated, and the degree of empirical overlap between the two samples was also less than expected.

The plan to investigate the outcomes of people in one or two matched case types was thus abandoned and a broader approach taken in which all potentially marginal care home entrants (i.e. all case types commonly found in both samples) were considered for follow-up. In effect, this amounted to all care home entrants who were less than ‘very dependent’ (Barthel > 40) and less than ‘severely cognitively impaired’ (CPS score < 4) so long as they did not have complex challenging behaviour (behaviour score < 9) and were appropriate for interview (e.g. could express their basic needs). Comparable CMHT clients were then chosen using a manual ‘nearest neighbour’ approach that took into account individuals’ age, gender and diagnostic group as well as case type. As the study progressed, however, additional individuals just outside the margins of care were included in an attempt to boost numbers.

Despite these actions, the number of achieved interviews remained disappointingly small (Figure 6). Only 16% of the care home entrants and 13% of the CMHT clients the study attempted to recruit participated in this exercise. Forty-four per cent of the selected care home entrants and 58% of the CMHT clients were deemed ineligible by their keyworkers, while no response was received for a further 22% and 14% respectively. Of those service users who were considered eligible, however, 73% were happy to be approached by the research team, and 66% were interviewed (11 care home entrants and 12 CMHT clients).

FIGURE 6.

Flow diagram of matched cohort study recruitment. (a) Care home arm; and (b) CMHT arm.

Bearing in mind the small numbers and our intention to capture people with similar needs, the interviewed care home entrants were significantly older than the CMHT clients (mean age 87.8 years compared with 78.8 years, t-test: 3.41, df = 21, p = 0.003). They also needed more help with activities of living, with 9 of 11 care home entrants compared with just 1 of 12 CMHT clients unable to manage stairs (chi-squared: 12.68, df = 2, p = 0.002). Consequently, though matches were achieved for six of the care home entrants, the study lacked community clients to match those admissions with more dependent profiles.

Perhaps unsurprisingly, no statistically significant differences were found in the two samples’ outcomes with regard to QoL or satisfaction, which were generally said to be good. However, when asked if they were basically satisfied with their lives, two CMHT clients but no care home entrants said ‘no’, and the same number indicated that they were not getting enough help. Asked to name the best thing about their current situation, care home entrants typically spoke about security and food, while the worst aspects of care home life were the lack of stimulation and the manner of certain carers. Correspondingly, those interviewees supported at home most frequently praised the help they received from different agencies, whereas any negative comments were particular to individual circumstances.

The incorporation of the cost and outcome data into the balance of care analysis (see Chapter 3, Activities 5.1 and 5.2)

When presented with the details of the marginal care home case types identified by practitioners, their original care home placements and suggested alternative care packages, the site X RAM Panel opted to support all nine cases in the community. As shown in Box 8, however, they did not necessarily agree with the proposed care plans, particularly where these involved moves to ECH. Indeed, the only situation in which this was endorsed concerned a case type whose home was described as physically unsuitable to meet their needs, such that a change of residence was, at least in the short-term, inevitable. They were also inclined to make more use of day care and considerable use of telecare, including the ‘Just Checking’ system.

Although typically lower than the costs of the practitioners’ plans, the estimated costs of the packages of care the Panel recommended exceeded those of care home placement for eight of the nine case types (Table 29). Furthermore, even using the most extreme plausible values, the costs of the alternative care packages for case types 6 and 9 exceeded those of residential care.

As shown in Table 30, much of this difference was again attributable to the greater private costs associated with the enhanced packages, which averaged £153 more per week than those of care home placement. However, even taking a purely public expenditure approach, the alternative package costs for seven case types were similar to, or higher than, those of a care home. Nevertheless, a clear financial incentive remained for the Social Services Department (SSD) to maintain certain groups of service users in the community, as direct savings were predicted for five of the nine case types.

Qualitative analysis of the panels’ decision-making process identified six main themes (Box 9), of which perhaps the most pervasive was their strong desire to maintain people at home. Their explanation to the research team was:

. . . and our key brief and our key principle is that we will always try and support someone in their own home, taking into account the risks that they would actually face in doing that, before we would agree any form of residential care . . .

Indeed, the other five domains could be interpreted as adjuncts to this. The minimisation of risks, for example, seems a necessary step in making home care viable, while the possibility of the service user improving over time may be viewed as strengthening the case for maintaining them at home.

There was surprisingly little discussion of costs, which were only mentioned in relation to one case type. Potential outcomes were also only discussed at a very general level (that staying at home would be in the service user’s ‘best interests’). There was, however, some debate about the best time to enter ECH, acknowledging that for certain individuals it might be better to move while they were still able to adjust to a new environment (although, in only one instance did they opt for such care). Furthermore, the provision of specific evidence about the likely social benefits of ECH compared with care at home, and the advantages of care home placement over ECH for people with advanced dementia, made no difference to the Panel’s decisions. Thus, in the first instance they reiterated that:

Our decision-making will always be, if the person can be supported where they are in their own home and that’s where they would like to be, then that will be our starting point.

They were also inclined to feel that the evidence would not apply to the case type in question, thus relying on maintenance at home as the priority.

Athough the last quote suggests that account was taken of service users’ preferences, the weight given to these appeared dependent on setting. Thus, where service users expressed a positive desire to remain at home, this was seen as justification for maintaining them in the community, whereas if they or their family expressed concerns about staying at home, the Panel tried to find ways of alleviating/overcoming these. Indeed, as one member of the panel put it, although they always tried to:

. . . consider somebody’s wishes . . . in terms of making a decision as to whether or not we can use the public purse to support someone in residential care . . . (we would) differentiate between want and need.

The panel did, however, acknowledge that it could be very difficult for front-line staff to promote the case for home care in the face of strong family opposition, and that for many carers the admission of their relative to hospital served as a ‘tipping point’. The pressure to discharge people from hospital before they had reached their full potential also increased pressure on staff to arrange short-term care home placements, a proportion of which inevitably became long-term. This was particularly true if their stay became protracted, they lost skills or confidence, they or their carers did not engage with plans to return home, and/or the care home wished to keep them.

The validation of local practitioners’ decisions (see Chapter 3, Activities 6.1 and 6.2)

The majority of older people’s groups (34 participants) who reviewed the plans proposed by the RAM Panel favoured the suggested community care packages over care home placement (Table 31). Indeed, the sole exception to this concerned case type 14 (MMM), where opinion was divided. However, complete consensus was only achieved for one case type (9, LHH), while in a number of instances it was felt that the depicted ‘clients’ would need a higher level of community support than had been recommended, including more day care, carer support and telecare services.

Expert opinion was similarly divided among the seven participants. Although for every case type the majority of experts favoured the alternative option, six of the seven experts advocated residential care in at least one instance. As with the older people, many proposed additional resources, echoing the call for more carer support and stressing the value of experienced social work input, the importance of meeting people’s social care needs, and the benefits of short-term placement in a dedicated assessment and rehabilitation unit. When the decisions made by the older people, carers and national experts were included in the logisitic regression analysis presented in Chapter 4 (see Tables 22 and 23), the model suggested older people and carers were considerably more likely than local staff to favour care home placement, whereas experts were also somewhat more likely to support residential care (albeit each individual looked only at a subset of cases).

When asked to state the main reason for preferring community care, both older people and experts stressed their desire to respect service users’ wishes, and a belief that community services could meet people’s needs. The experts also highlighted the need for further assessment at home before care home entry and the desire to avoid potentially debilitating moves, whereas the older people were more likely to point to the potential financial implications of care home entry for service users. Conversely, care home entry was typically predicated on the need for more care than could be provided at home (particularly at night) and the risks service users posed to themselves and others.

Costs of alternative care options (see Chapter 3, Activity 4.1)

Box 10 details the packages of care local practitioners’ believed would meet the needs of the five most marginal inpatient case types. Interestingly, all three plans for care home residents involved input from a specialist care home support team (CHST), whereas those for people admitted from home drew on an intensive mix of primary care and mental health expertise, including frequent mental health support worker input to assist people with their diet, medication, personal and social care needs.

The estimated mental health and social care costs of the alternative arrangements are set out in Table 32, suggesting that in every case the predicted costs of the proposed community care packages were lower than those of hospital admission. Indeed, the mean weekly difference was £1873. Even allowing for considerable uncertainty in the quantity and cost of resources and the representativeness of vignettes, the general picture remained unchanged. The vast majority of this difference was attributable to the relatively low costs of providing community as opposed to inpatient mental health care, even where this involved multiple staff and frequent input. Furthermore, in contrast to the care home entrants modelling, there was no obvious trade-off between health and social care costs.