Measuring and optimising the efficiency of community hospital inpatient care for older people: the MoCHA mixed-methods study

Article history

The research reported in this issue of the journal was funded by the HS&DR programme or one of its preceding programmes as project number 12/177/04. The contractual start date was in July 2014. The final report began editorial review in November 2018 and was accepted for publication in July 2019. The authors have been wholly responsible for all data collection, analysis and interpretation, and for writing up their work. The HS&DR editors and production house have tried to ensure the accuracy of the authors’ report and would like to thank the reviewers for their constructive comments on the final report document. However, they do not accept liability for damages or losses arising from material published in this report.

Declared competing interests of authors

Claire Hulme was a member of the National Institute for Health Research (NIHR) Commissioning Board over the life of the project (from 2013 to 2017). Mary Godfrey reports grants from NIHR for Health Services and Delivery Research (HSDR) 12/209/66, HSDR 12/5003/01, Health Technology Assessment 11/143/01 and Programme Grants for Applied Research RP-PG-1210-12017 during the conduct of the study.

Copyright statement

© Queen’s Printer and Controller of HMSO 2020. This work was produced by Young et al. under the terms of a commissioning contract issued by the Secretary of State for Health and Social Care. This issue may be freely reproduced for the purposes of private research and study and extracts (or indeed, the full report) may be included in professional journals provided that suitable acknowledgement is made and the reproduction is not associated with any form of advertising. Applications for commercial reproduction should be addressed to: NIHR Journals Library, National Institute for Health Research, Evaluation, Trials and Studies Coordinating Centre, Alpha House, University of Southampton Science Park, Southampton SO16 7NS, UK.

2019 Queen’s Printer and Controller of HMSO

Background and rationale to study

Nearly 12 million people in the UK are aged ≥ 65 years, which is projected to increase by 5.5 million in the next 20 years, and the number of people aged ≥ 85 years is projected to double to 4 million. 1 An increasingly older population, along with improvements in the management of chronic conditions, will result in more people living to very old age with one or more chronic conditions. 2 As multimorbidity increases with age, over half of those aged ≥ 85 years will have four or more chronic conditions (complex multimorbidity). 3 Furthermore, among more than one-third of those with complex multimorbidity, the burden is exacerbated by mental ill health (e.g. depression and dementia). 3 Multimorbidity is associated with secondary care utilisation, complications during a hospital stay, reduced functional capacity and increased need for co-ordinated care and support to facilitate safe hospital discharge. 4 Multimorbidity therefore poses considerable challenges for appropriate models of treatment and rehabilitation during and following an acute illness and for ongoing support from health and social care services. This is a major issue and decision-makers across the NHS are attempting to reconfigure services to deal with the year-on-year increase in hospital admissions, often with inadequate evidence to inform their decisions.

Community hospitals are a long-established health-care facility in the UK, as well as internationally. They can be loosely defined as small hospitals that provide a wide range of locality-based outpatient and inpatient services such as health promotion, rehabilitation, day hospital and diagnostic services. Specific service configurations differ between community hospitals as a result of local history, proximity to a larger general acute hospital and perceived need. The policy direction for community hospitals was contained in the white paper Our Health, Our Care, Our Say: A New Direction for Community Services. 5 It was expected that primary care trusts would engage in an ‘ambitious’ shift of resources from secondary care to community services over a 10-year period. A new generation of community hospitals was proposed as an important component of the vision for community services within the umbrella term, ‘intermediate care’. The new format for the community hospital was that they should be locally led, of high quality and focus on older people. It was expected that they would provide new care pathways, adopt new technologies, minimise acute hospital admissions, promote integrated solutions and be affordable for the whole health economy. 5

There is high-quality evidence that supports the community hospital as an effective bed-based rehabilitation service for older people. A multicentre study in England randomised patients (n = 490) either to remain in a general hospital ward or to transfer to one of seven community hospitals. 6 The main finding was a significantly greater functional independence at 6 months for the patients allocated to the community hospital group,6 with similar cost-effectiveness. 7 A similar single-centre randomised controlled trial in Norway (n = 142) reported lower mortality and indirect evidence for increased independence for the community hospital group than for the group receiving continued care in a general hospital. 8 There is also evidence that older people are in favour of and, if given the choice, would prefer community hospital ward care. 9–11

This evidence suggesting superior outcomes for older people receiving rehabilitation in a community hospital, at an affordable cost, is encouraging, particularly as most of ≈400 community hospitals in the UK provide inpatient rehabilitation. 12 Thus, community hospitals remain an important provision for health care in the NHS. Considerable additional information describing the contemporary community hospital is available from two other national surveys from the NHS Benchmarking Network (NHSBN): the NHSBN Community Hospitals Project (begun in 2012);13 and the National Audit of Intermediate Care (NAIC). 14 These complementary surveys provide information on 180 community hospitals (approximately two-thirds of all community hospitals in England). Two findings are prominent. First, it is clear that a core function of the contemporary community hospital is rehabilitation for older people: ‘general rehabilitation’ is provided in 97% of the community hospitals. This is in accord with strong policy directives from the Department of Health and Social Care relating to intermediate-care services for older people. 15,16 Second, the contemporary community hospital is characterised by extreme variation in provision [e.g. bed provision per 100,000 weighted population (range < 10–70); clinical leadership (50% nurse led; 50% consultant led); average length of stay (LOS) (11–58 days); cost per admission (£3700–17,500); and cost per day (£140–450)].

This variation in the community hospital was highlighted by the NHS Institute for Innovation and Improvement but, despite provider engagement with their Productive Community Hospital initiative,17 the wide variations have continued. There is evidence that some features might be associated with performance, for example the lack of ward leadership,18 excessive waiting times for admission19 and underprovision of the Comprehensive Geriatric Assessment. 20 However, our knowledge of the features that characterise optimum community hospital ward performance is incomplete as no large-scale study has been designed to systematically investigate the issue. In addition, there is a paucity of information available to service planners about the comparative outcomes and efficiencies of community hospitals in relation to alternative forms of community services, such as home-based rehabilitation and social service enablement services. Thus, when this study began, nearly 10 years on from the white paper Our Health, Our Care, Our Say: A New Direction for Community Services,5 the ‘ambitious’ shift of care focus for older people from secondary care to the community had not been realised. 21

Community hospitals and intermediate care

A prominent and sustained health-care policy for older people during the last decade has been that of intermediate care. Consistent with the established policy direction of increasing the provision of community-based services for older people, intermediate-care policy has been partly driven by the acknowledgement that large general hospitals comprise an environment in which unintended harms (e.g. increased risk of falls, delirium and increased dependency) are common for this vulnerable patient group, and the proposition that intermediate care might be less expensive.

Intermediate care comprises a range of service models designed to provide community-based, short-term rehabilitation with the aim of preventing admission to hospital and/or facilitating early discharge. Intermediate care was first proposed in the National Beds Inquiry,22 became a formal health-care policy in the NHS Plan of 2000,23 and was implemented as part of the National Service Framework for Older People in 2001. 15 Subsequently, intermediate care formed part of a wider health and social care policy loosely summarised as care closer to home. 16 The new role for community hospitals was specifically highlighted in these policy documents and, later, in implementation guidance. However, despite the successful achievement of intermediate-care expansion targets, and release of substantial capital monies for community hospital new-builds, the strategic objective of intermediate care to reduce pressures on acute hospital beds was not realised. ‘Refreshed’ guidance was produced that, once again, encompassed community hospital provision. 24 Acute services continue to be hard pressed, largely owing to a continuing year-on-year increase in acute admissions of older people. 25,26 Despite a decade of consistently applied policy, investment in intermediate-care services to provide short-term rehabilitation for older people appears to have stalled. The NAIC obtained information in 2014 from 62 commissioning and 112 provider organisations (327 intermediate-care services). 27 An important finding was that intermediate-care capacity nationally was around half that required to address the anticipated need for inappropriate admissions and post-acute care for older people. This mismatch of demand and supply has been a constant finding in the subsequent rounds of the national audit. 28,29 In addition, it was apparent that there is considerable scope to free up capacity in many services by simply addressing excessive lengths of intermediate-care stay. The average LOS was considerably longer than those given in the Department of Health and Social Care guidelines: ‘They (intermediate care) are time-limited, normally no longer than six weeks and frequently as little as 1 to 2 weeks or less’ (© Crown copyright 2009. Contains public sector information licensed under the Open Government Licence v3.0). 24 In the national audit,27 75% of patients in bed-based intermediate care had a LOS of > 2 weeks, and 9% of patients had a LOS of > 90 days. Specifically, for a community hospital with 20 beds and 90% occupancy, if the hospital is performing at the 75% quartile (LOS = 31 days) and improved to the 25% quartile (LOS = 21 days), it would be able to treat approximately 100 more patients per year (48% increase). Alternatively, if a 20-bedded community hospital is performing at the 75% quartile for cost per occupied bed-day (£200 per day) and improves to the 25% quartile (£110 per day), the annual cost saving is approximately £650,000.

From a whole-systems perspective, subsequent rounds of the NAIC reported no relationship between the scale of local intermediate-care provision and its impact on secondary care utilisation. 28,29 It was concluded that the case for further investment required new local evaluations to provide better evidence that future increases in intermediate-care capacity would indeed favourably affect secondary care utilisation. Until, and unless, this evidence is available, it will be difficult for commissioners to justify pro-active reductions in the secondary care bed base. In relation to community hospitals, this means that capital investment in new community hospitals will continue to be sporadic and opportunistic rather than strategically planned, and that some community hospitals will close or have bed reductions because of uncertainty over their contribution to the whole-system approach.

Community hospitals are part of this general uncertainty about intermediate care. Indeed, even producing a unifying definition for community hospitals has proved problematic, mostly because there are different models or subtypes of community hospital. This lack of agreement for community hospitals is important because it limits the potential of the commissioning process to shape the service to be delivered and leaves unanswered questions such as ‘what are the critical parameters of a community hospital?’ and ‘what are the tolerances to these parameters?’. For example, some community hospitals are nurse led and some medically led, but nurse-led rehabilitation units (where the admission, assessment, monitoring and discharge of patients are organised by senior nurses, and doctors have an advisory role) have been associated with weak evidence of increased mortality, and may operate at considerably higher unit costs. 18 Not surprisingly, this leads to uncertainty among commissioners and planners about the possible contribution of community hospitals. The number of community hospitals declined slightly between 1998 and 2008 (334 to 296),12 and again up to 2017, but this headline finding disguises a mixed picture of closures offset by newly opened facilities, which reflects an inconsistent approach to the provision of community services for older people.

A popular alternative bed-based form of intermediate care is the model of care home-based rehabilitation. Although widely used (as evidenced in the NAIC), little is known about its effectiveness, cost-effectiveness or safety. Only one randomised controlled trial has been conducted. 30 The NAIC reported that care home-based intermediate care was provided at considerably lower unit costs than community hospital care (£136 compared with £252 per day). 28

Robust comparison studies of different models of community hospital care are needed to provide confidence to patients, practitioners and commissioners in an evidenced-based selection approach that favours community hospital with specific criteria known to be associated with cost-effectiveness. Improved understanding of the role, external relationships, staffing structures and care processes in community hospitals providing intermediate care would clarify their contribution.

Aims and objectives of the research programme

Overview of method

The underpinning objectives of the project were addressed through a mixed-methods project design, consisting of five interdependent studies. This approach used existing data alongside complementary prospective data to produce insights that could not be achieved by the study of the components alone, and represents an efficient and effective research design.

• Study 1: an analysis of cost efficiency of community hospitals (see Chapter 2).

• Study 2: a national survey of community hospitals (see Chapter 3).

• Study 3: comparative case study of community hospitals (see Chapter 4).

• Study 4: development of toolkits for commissioners and community hospital providers to optimise performance (see Chapter 5).

• Study 5: an analysis of the impact of intermediate-care services on secondary care utilisation (see Chapter 6).

Study team

The co-applicants comprised a team of health economists and efficiency analysis experts led by Claire Hulme; the Patients Association (led by Heather Eardley); the Community Hospitals Association (CHA) (led by Helen Tucker); a social scientist with experience in older people’s services (Mary Godfrey); and academic clinicians with experience in community hospital care delivery (John Young, John Gladman, Elizabeth Teale and Pam Enderby).

Patient and public involvement

Through the participation of the Patients Association and CHA, patient and public involvement was an integral feature of the study from conception and development of the research design to conduct of the research and interpretation and development of the interactive toolkit. Representatives of both (Heather Eardley and Helen Tucker) were co-applicants in the research and their contributions were critical in ensuring that the study engaged with the needs of patients and current quality and performance issues in this service sector. We worked with the Patients Association to produce a ‘primer’ that described the concepts that underpinned the economic efficiency work, concepts that would be unfamiliar to most people (see Appendix 1). This was used as an introduction during two workshops we held at consecutive CHA annual conferences during 2015 and 2016 attended by practitioners and NHS managers.

For all aspects of the study, we engaged with members of the CHA through sequential participation in their annual conferences from 2015 to 2017, where we delivered presentations and workshops, inviting feedback to inform the content and format of the interactive toolkit. Study progress was also conveyed on the CHA website.

In addition, for study 3, we involved our well-informed local older person’s group (approximately 15 people) with which we have a long-standing relationship. This is a subgroup of a forum established by the local authority that comprised approximately 300 retired local people who have been consulted on health and social policy issues. The subgroup has been acting as an advisory body in all aspects of our research with older people since July 2007. Meetings of the group took place to inform data collection methods and contribute to the ethics application, and following conclusion of data collection to reflect on the emerging findings and their implications. We also presented our initial findings to staff groups involved in each of our three case study sites towards the end of the study to further inform our interpretations and conclusions.

Ethics

Ethics approval was required only for study 3 (ethnographic case studies); all of the other studies were based on anonymised, aggregate data. Ethics approval was obtained via the Health Research Authority that governs research ethics in the UK (reference number: 15/YH/0062). The main ethics issues were securing the consent of patients to take part in specific ward observations, conversations and interviews, especially those who might lack capacity on account of dementia or delirium; not seeking formal consent for the general observations of ward routines; and maintaining anonymity of hospitals, wards, trusts/health boards, staff and patients.

Summary

Community hospitals continue to provide an important function for the NHS. There is high-quality evidence that supports the community hospital as an effective bed-based rehabilitation service for older people. Nevertheless, the contemporary community hospital is characterised by extreme variation in provision and there is evidence that some features may be associated with ward-level performance. Our knowledge of the features that characterise optimum community hospital ward performance is incomplete. In a number of sequential and overlapping studies, and employing mixed methods, this research aimed to understand the features of community hospital ward care that optimise or detract from optimum performance, through the lens of rehabilitation for older people, the core unifying function of the contemporary community hospital, and to provide a resource for community hospitals to inform practice. It also sought to investigate the impact on secondary care of community hospital inpatient care for older people. Patient and public involvement was an integral feature of all aspects of the study through the participation of the Patients Association and CHA and a local older person’s group.

Introduction

As discussed in Chapter 1, there is a growing body of evidence to support community hospitals as an effective bed-based alternative to rehabilitation in the acute setting in terms of patient outcomes. However, in terms of economic outcomes, the evidence base is much weaker. One study reported community hospital rehabilitation to be cost-effective relative to secondary care in England. 7 Elsewhere, a national study of 180 community hospitals found considerable variation in the provision of services. 26 However, there is only speculation as to the causes of the variation. Finally, on the basis of cost comparisons, it was estimated that there were achievable savings of £1.2–1.8B in community services in general. 33 With a total spend of £8.4B across these services, this represents a potential for cost reduction of around 15%.

In the study presented here, we sought to extend the understanding of community hospitals in respect of the drivers of costs, economies of scale and bed occupancy. Furthermore, we assessed the relative efficiency of ward providers. In collaboration with the NHSBN, we made use of the NHSBN Community Hospital Programme data set encompassing hospitals in England, Scotland and Wales. 13 We adopted a stochastic frontier framework approach, which allowed both the cost structure and cost efficiency to be analysed.

Specifically, study 1 considered the following objectives:

• study objective 1 – to measure the current relative performance (cost efficiency) of community hospital inpatient care for older people

• study objective 2 – to identify the characteristics of community hospital inpatient care for older people that optimise performance.

We start with a general description of cost efficiency analysis based on frontier techniques that includes a discussion of the econometric approach to efficiency analysis. We then describe the data and categorisation of the variables to construct the final data set used for modelling. The specific methods and empirical specification of models are described in detail. Finally, we report the results of the analyses and diagnostic statistics with a discussion of interpretation and limitations.

Stochastic frontier analysis for community hospital costs and efficiency

Frontier techniques have a long history of use in efficiency analysis, following pioneering work by Farrell. 34 The principle behind frontier techniques is very simple: an efficiency or ‘best-practice’ frontier is defined, representing full efficiency, and then distances from the frontier are measured, representing inefficiency. That is, the further from the frontier, the higher the inefficiency. This is shown for the case of community hospitals in Figure 1.

FIGURE 1.

Conceptual features of frontier-based efficiency analysis as applied to community hospitals.

Frontiers are regarded as the foremost hospital efficiency analysis tool and are seen to have great potential in health-care markets more widely. 35,36 Frontier techniques have been applied widely in health for measuring efficiency. 36–41 There are, broadly, two approaches to frontier analysis, both of which have been deployed repeatedly to analyse efficiency in health-care markets. The first is an approach based on mathematical programming, data envelopment analysis (DEA), and the second is based on econometric estimation, stochastic frontier analysis (SFA). We make use of SFA in this study because, unlike DEA, it enables us to analyse both costs and efficiency, which are of interest to managers and policy-makers. Furthermore, it enables removal of statistical noise from the predictions of efficiency, which is a key advantage over DEA. 41

The stochastic frontier (SF) model for costs defines the efficiency frontier as a cost function (based on outputs, input prices, environmental variables and quality, thus allowing for variables not in control of the decision-making unit. The frontier assumes the shape of the cost function. The frontier represents the minimum attainable cost for a firm, given its levels of outputs and input prices (and other features defined by the cost function noted above). Deviations of firms from the frontier are considered to be, in part, due to inefficiency, where the distance to the frontier represents the magnitude of the inefficiency. In addition, the SF model allows for the removal of random statistical noise in the data from inefficiency estimates.

The SF model can be computed mathematically as:

where care costs are a function represented as f (. . .) in Equation 1 of output (y), input prices (w), environmental influences (z) and quality (q). The ε term is the observed deviation from the frontier; it comprises both inefficiency (u) and random noise (v) to a firm’s (in this context, a hospital ward’s) production that impinges on its costs, as in Equation 2. Figure 2 shows the conceptual features of the SF model.

FIGURE 2.

Diagrammatic representation of the stochastic cost frontier model.

The gradient of the frontier in Figure 2 is defined according to the cost function and is the deterministic element of the model. In this case, costs (c) are defined as a function of outputs (), namely c=α0+y’β.

We consider observation A, representing a specific community hospital A. Here, as shown, community hospital A has output y_A, at which the minimum attainable costs for community hospital A (shown via the cost frontier) are cA=α0+ y’Aβ. The observed value is in fact point A, which is higher than the deterministic frontier for community hospital A, given its level of output. Then, the community hospital-specific observed deviation ϵ_A comprises both the community hospital’s inefficiency u_A and random statistical noise v_A. In this case, both noise and inefficiency have a positive influence on costs.

In some cases, community hospitals are observed to be below the deterministic frontier, as in observation B for which the observed deviation from the frontier ε_B is negative. This is the result of a noise component v_B, which is negative and greater than the community hospital’s (positive) inefficiency u_B. Here, community hospital B has output y_B at which the cost frontier is cB=α0+y’Bβ.

Random noise can encompass a number of features. In non-health-care settings, these typically comprise random ‘shocks’ to production that include untoward events, such as strikes, unusual weather, force majeure, etc. In addition, this component can account for measurement error and approximation error (from the choice of functional form). 41 The defining feature of the SF model is its ability to remove these factors so that they do not distort the underlying metric of interest: inefficiency.

Although we had originally planned to use DEA as a cross-check for ranking efficiency scores, we did not apply DEA because the method does not deal well with random noise in the data and also because we were interested in establishing the relationship between costs and cost drivers, which DEA does not support. We therefore decided to focus on econometric methods as the most appropriate method for the research objectives.

NHS hospital efficiency studies

Although there is no previous evidence on cost efficiency for community hospital wards, there is a wealth of econometric evidence for the general hospital setting. We did not undertake a full systematic review. Instead, a scoping search was carried out in MEDLINE in 2015. The search included all relevant publications published up to 30 September 2015. In addition, research papers already known to the authors of this study were included and citation tracking of identified papers used. A list and summary review of these studies42–53 is provided in Table 1. This will facilitate later comparisons between the two types of inpatient ward provision.

We note the link between the cost function methods used and the SFA literature. In several studies, a hospital-specific effect from a cost function is used as the measure of hospital efficiency (some authors use the term performance). 47,49,50,52 In this sense, they are akin to the panel data SF models that measure time-invariant inefficiency. 41,47

Community hospital data set

The NHSBN is the in-house benchmarking service of the NHS. The Network has 270 members including Clinical Commissioning Groups (CCGs), acute and mental health trusts and community providers, and runs a range of benchmarking projects across all sectors of NHS provision. Members submit locally collected data to participate in the projects. Participation is not compulsory and incomplete data submissions are accepted.

The NHSBN Network Community Hospital Programme was instigated at the request of the network members to provide a structured comparison using key metrics that include workforce, activity, investment levels and quality measures. However, the granularity of the NHSBN Network Community Hospital data set at the ward level for older people was insufficient for the analysis. Further questions were therefore added to the data collection process for the 2013/14 survey13 to enable capture of data at a level to facilitate the analyses, namely at ward level. However, this meant that only this version of the Community Hospitals Project was used for this study, which collected data on 158 community hospitals from the financial year 2013/1413 (not the 3 years originally anticipated).

We started by categorising the variables into general features of the community hospital, information on patient access, services provided, workforce characteristics, activity of community hospitals, financial data and service quality. The nature of the variables includes continuous, categorical, binary and qualitative data.

Our unit of analysis was the community hospital ward for older people. Provision of services on older people’s wards is the core activity of community hospitals. We adopted this approach for a variety of reasons. First, it allowed us to maximise the sample size, as there was a high number of missing data for variables at the whole-hospital level (Table 2 shows a variable summary of missing data). Second, this allowed us to better control for the issue of heterogeneity between providers than would be the case at the whole-hospital level 54 (there is less heterogeneity in the provision of services on the ward). This is important as we have a limited number of variables on patient characteristics, quality, hospital characteristics and input prices. Third, following an initial modelling exercise, results indicated that models at the whole-hospital level yielded questionable results in terms of parameter estimates and, consequentially, efficiency predictions, probably because of the high number of missing data. Fourth, by focusing on wards, managers are better placed to understand the source of inefficiency than with whole-hospital analysis, in which the source of inefficiency may be unclear.

Our data set comprises 158 community hospital wards from 158 NHS providers and 18 variables per ward. Of these, 11 were continuous variables, five were binary variables and two were categorical variables. The continuous variables comprise what can be seen as core variables (i.e. the dependent variable and variables on outputs and input prices) and some environmental variables (these include, for example, patient selection policy and provision of diagnostic services). There are also continuous quality variables. See Table 2 for the list of variables, their descriptions, classification and missing data.

Some of the variables we used were constructed from other variables in the data set. Materials input prices were defined as the non-labour operating costs divided by the number of beds. Staff costs were calculated by the total labour costs divided by the number of staff, in whole-time equivalent (WTE) terms. The proportion of physiotherapy/occupational therapy staff and senior staff was defined as the ratio of each of these to the total number of staff.

We removed 14 outliers from the data set. For materials input prices, there were two clear outliers (around £100,000; the mean for this variable is around £9000). For staff mix, observations in excess of £50,000 were removed (i.e. where average wages were > £50,000). In this case, 12 observations were removed. All other variables appeared to have plausible distributions. Furthermore, we tested models that included the outliers and estimates did not appear to be sensitive. After cleaning the data set for missing values and outliers (some remained unverified), 101 community hospital wards were available for analysis. Thus, we undertook a complete-case analysis (having also removed outliers from the data set).

Descriptive statistics

The descriptive statistics for the 101 community hospital wards are presented in Table 3. For each variable, the category (see Equation 1), mean, standard deviation, minimum, maximum and number are reported.

We can see that, on average, the annual expenditure for the provision of a community hospital ward is £2.3M.

For the output variables, on average, 417 patients were treated each year, accounting for around 11,000 occupied bed-days.

The London weighting variable suggests that around 10% of our sample are based in London (we do not have any further geographical information).

For the environmental variables, average bed occupancy reveals that community hospital wards are close to capacity, with an occupancy rate close to 90%. The National Audit Office (NAO) suggest that bed shortages may occur when bed occupancy is in excess of 85%. 55

Community hospital age is a categorical variable for which categories reflect incrementally older community hospitals.

The remaining environmental variables show various environmental aspects of the sample of community hospitals. Around 70% have a patient selection policy, and around half provide diagnostic services.

For the measures of quality, on average, senior staff occupy 16% of the workforce. There is a range from around 2% up to around 46%. Although the mean is approximately the same for the ratio of physiotherapy and occupational therapy (18%), the range is much greater: from 0% up to 91%. It appears that most community hospitals hold regular governance meetings (88%) and that around one-quarter have been involved in research programmes (24%).

Methods for model specification and estimations

Estimation results, community hospital ward costs and efficiency

Number of log-likelihood versus ordinary least squares is a statistical test of the stochastic frontier model versus the basic cost function

Community hospital ward efficiency

Descriptive statistics for efficiency predictions from the range of models are presented in Table 5. Corresponding kernel densities are presented in Figure 3.

TABLE 5 - Descriptive statistics for efficiency predictions

Mean does not include COLS estimates.

Efficiency specification	Meana	Standard deviation	Minimum	Maximum
COLS	0.560	0.138	0.263	1.000
Half normal	0.808	0.087	0.514	0.940
Truncated normal	0.823	0.083	0.524	0.943
Exponential	0.878	0.063	0.575	0.952
Gamma	0.921	0.049	0.645	0.975
Rayleigh	0.724	0.087	0.458	0.888
Mean	0.831		0.543	0.939

FIGURE 3.

Kernel densities for efficiency predictions. HN, half normal; TN, truncated normal.

In Table 4 we can see that, in three out of five specifications, the SF model was preferred to OLS. As expected, the COLS average efficiency is much lower than the remaining models, owing to the failure to separate statistical noise from efficiency. Indeed, the minimum value is much lower than those of comparator models. Specifically, a value of 0.26 suggests that this community hospital ward could reduce its costs by 74% if it brought its performance in line with best practice. This result seems rather implausible, particularly as the analysis is focused on the ward rather than the whole community hospital. For these reasons, the efficiency predictions from the COLS model have been discarded.

For the remaining models, the mean efficiency predictions are in the range of 0.72–0.92, with an average of 0.83. The interpretation of this number is that, on average, community hospital wards could reduce operating costs by 17% if the performance of all community hospitals was brought in line with best practice. The minimum ranged from 0.46 to 0.58 with an average of 0.54, and the maximum ranged from 0.89 to 0.98 with an average of 0.94. There was some variation in the standard deviations between model predictions, which can be seen in the kernel densities in Figure 3. Some of the models yield predictions with similar average efficiency, whereas others do not. The range in mean predictions of efficiency of 20% (0.72–0.92) is rather large. Therefore, there is a question as to how the efficiency predictions correlate between models. We examine this issue in detail in Tables 5 and 6, by looking at models’ efficiency prediction correlation in terms of absolute values and rank. Figure 4 shows the correlation matrix graphically.

TABLE 6 - Pairwise correlation between efficiency predictions

Correlation	COLS	Half normal	Truncated normal	Exponential	Gamma	Rayleigh
COLS
Half normal	0.91
Truncated normal	0.91	0.99
Exponential	0.84	0.96	0.98
Gamma	0.77	0.91	0.93	0.97
Rayleigh	0.96	0.98	0.99	0.95	0.89

FIGURE 4.

Correlation matrix absolute efficiency predictions. Note that all of the cells in the figure are actually an individual graph themselves. All are showing efficiency: both x- and y-axes. HN, half normal; TN, truncated normal.

It is unsurprising that the Rayleigh model predicts the lowest efficiency given that this distribution allows for a non-zero modal value. Although this is also true for the truncated-normal model, this does not appear to be the case. We note, however, that the mu parameter is not significant in this model, which may explain this finding. Indeed, using the LR test, the half-normal model is preferred to its truncated-normal alternative.

Tables 6 and 7 and Figure 4 show pairwise correlation between the predictions of efficiency, which is high. Farsi et al. 54 suggest that values in excess of 0.90 denote concordance between models. That almost all pairs of efficiency predictions are over 0.90 in Table 6 reflects similarity between models’ predictions. Moreover, in terms of rank correlation, as shown in Table 7, not only are correlation coefficients high, they are statistically significant across all pairs of predictions. This high correlation is visible in Figure 4. These results hold for the COLS model, which may be expected to give different results because of its composed error. We do not find any evidence of this in this instance.

TABLE 7 - Pairwise rank correlation using Kendall’s tau between efficiency predictions

*, **, *** denote significance at the 10%, 5% and 1% level, respectively.

COLS estimates.

Kendall’s tau	COLS	Half normal	Truncated normal	Exponential	Gamma	Rayleigh
COLS
Half normal	0.93***
Truncated normal	0.95***	0.97***
Exponential	0.96***	0.96***	0.99***
Gamma	0.80***	0.77***	0.80***	0.80***
Rayleigh	0.95***	0.97***	0.99***	0.98***	0.80***

Overall, although the average predicted efficiency spanned a range of 20%, the correlation, and in particular the ranks, between model predictions were very similar. The ranks of the predictions are particularly important because the goal of this exercise is to identify high and low performers. We turn to this issue in the discussion.

Discussion

Results from the models are, overall, encouraging from several perspectives. First, the parameter estimates are sensible and plausible. Importantly, the results appear to be consistent over a range of specifications and with the OLS base case, which does not impose the distributional assumptions that the SF models do. The base model, a limited number of variables notwithstanding, had considerable explanatory power, which is a good sign for the reliability of the estimates.

Our testing procedure indicated that our functional form was appropriate. All but one of the models detected the presence of inefficiency. Predictions of inefficiency are within a plausible range, and the models’ predictions of inefficiency are highly correlated. This finding supports previous studies in that ranks are not affected by distributional assumptions. This is key here: our cross-sectional results appear robust to the specification of efficiency distribution. This further implies that models agreed as to high and low performers and it proved possible to rank-order the cost efficiency of the included community hospital wards, but the rank-order was not declared as we received deidentified data for this research study. Nonetheless, in a situation of formal benchmarking, providing a rank order would be a simple procedure based on the models and data inputs described here. In terms of other efficiency studies, the results from our SFA analysis were comparable with those from previous studies43,44,46,51,53 of NHS general hospital efficiency (0.61–0.92; see Table 1). This suggests that the cost efficiency of community hospital wards is similar to that of general hospital wards. The COLS estimates were below those presented in studies elsewhere. Our COLS result was 0.56; results from other studies42–44,46 ranged from 0.69 to 0.74. However, the COLS models are more susceptible to outliers. Moreover, our testing procedure favoured the use of SFA, the results of which accord with prior studies. Importantly, the results here seem to accord with those of community hospitals more generally. 33

There are some methodological points for discussion. First, being a cross-sectional analysis, there are concerns over uncertainty; as we have one observation only per ward, we do not see the precision of the efficiency estimates increase per ward as the sample size increases. 44,60 We did not calculate prediction intervals around our estimates, although this may be an important avenue for future development so as to assess the extent of the uncertainty. However, that the ranks of providers are consistent across models helps to allay this issue insofar as we are able to assess the general level of efficiency and identify high and low performers.

Second, there is an issue around unobserved heterogeneity (when the variance, or standard deviation, of a variable varies across the range of the variable), particularly as the model did not support the inclusion of variables as proxies for case mix, patient severity and quality. We note that we are analysing costs at ward level, where there is likely to be much less heterogeneity than if we were modelling at the whole-hospital level. Indeed, the OLS results indicate that much of the variation has been explained. Moreover, by allowing for noise and removing it from our measure of inefficiency, it may well be that some unobserved influences are captured here. 41

We were able to utilise a unique data set on community hospitals that facilitated an econometric efficiency analysis on this part of the NHS for the first time. The data set suffered, to an extent, from missing values leading to a reduced sample with which to estimate models. We do not know the reason for the missing data; we have assumed it to be missing at random for the purposes of this analysis. We attempted an imputation exercise in preliminary work, which did not yield plausible results.

There may also be other more practical considerations. For example, although the NHSBN Community Hospital Programme included WTEs for staffing levels, there is the potential for over- or underestimation for staff who have responsibilities across sites or hospitals.

We considered that there might be selection effects of hospitals that chose deliberately not to report on the basis of their own poor performance. But to know this, they would need to have an intimate knowledge of their own and other wards’ operations in order to make a reasonable determination of this. Thus, it seems unlikely that they would be able to do this (although they might have considered themselves to be a poor performer). Moreover, if this was the case, we surely would not observe the range of efficiency estimates that we did: those that genuinely performed poorly would drop out of the sample.

The restricted sample size may be one reason why it was not possible to use a translog functional form and, thus, allow some flexibility in the cost function. In particular, no evidence was found to support variation in the cost elasticity over the output range. Of course, it may be that the cost elasticity is constant over the output range, but this view was not supported by community hospital staff. The other complication to this issue is the lack of a capital component in our cost variable. We note that, in this regard, we are not truly measuring the economies of scale, more a measure of size. This may an explanation for the finding on the cost elasticity with respect to output.

Our general view is that, given the data available, the models are well specified and are a good approximation of the underlying cost function. We thus have confidence in our predictions of inefficiency. The level of efficiency is high and the results are in keeping with those found elsewhere in the general hospital sector. Furthermore, we are able to identify high and low performers from the models, meaning that managerial effort can be targeted effectively to maximise gains across the sector.

The average predicted efficiency among community hospital wards was 0.83, implying that, on average, community hospitals could reduce operating costs in the provision of service on older people’s wards by 17% if performance was in line with best practice. If community hospitals were able to extinguish this inefficiency, there would be around £47M of available gains; total spend is around £227M. There was a range of performance, with the lowest-performing community hospital operating at around 0.55, whereas the highest-performing community hospitals were operating at around 0.94. That is, the models indicate that all community hospitals can improve efficiency on their wards.

Using a price year of 2013/14, in terms of input prices and staff mix, the results were unsurprising. Those community hospitals that faced higher input prices appeared to have higher costs; those community hospitals that employed a more experienced staff force also appeared to have higher costs. In terms of output, the results appeared to suggest that there were reasonable economies of scale at the sample mean, with a beta coefficient on output of around 0.85. This suggests that, for wards with bed size below the mean, marginal costs can be reduced by increasing the scale of the operation; for those above the mean, the scale at which the all-important cut-off point of 1 is reached is unclear. We note a slight limitation here, which is that, as suggested by staff working in community hospitals at a workshop held at the 2016 CHA conference (13–15 May, Bristol, UK)), we did not find variation in economies of scale between smaller and larger community hospitals (we did not find a statistically significant squared term on the output variable). However, the main finding was in keeping with what had been proposed: that there were economies of scale present. For bed occupancy, it appears that increasing bed occupancy rates are associated with lower costs. This may reflect that spare capacity is costly. However, this needs to be balanced against the effect on the service of high occupancy, which may be disruptive to service provision; we note that our sample mean is around 90%. Indeed, the NAO have suggested that services in which bed occupancy levels are in excess of 85% may experience disruption (in the general hospital setting). 55 This would be an interesting area for future work.

Summary

In the analyses described in this chapter we have applied econometric techniques to NHS community hospitals for the first time using a unique data set collected by NHSBN. The analyses suggest the potential for economies of scale in the provision of community hospital services for older people. As expected, higher input prices and a more experienced staff mix were associated with higher costs, and costs appeared to be lower when bed occupancy was higher.

We used a range of models to predict efficiency in our sample of community hospitals. Under differing assumptions, models’ predictions varied somewhat. Their average was 0.83, implying that, on average, community hospitals could reduce operating costs in the provision of service on older people’s wards by 17% if performance was in line with best practice. There was a range of performance, with the lowest-performing community hospital operating at around 0.55, whereas the highest performing community hospitals were operating at around 0.94. The predictions of efficiency were found to be highly correlated between models, in terms of both absolute values and ranks. Policy implications of our findings are discussed in Chapter 7.

Introduction

According to the CHA, there are > 500 community hospitals in the UK: 346 in England, 101 in Scotland, 44 in Wales and 13 in Northern Ireland. However, not all community hospitals have inpatient beds and the numbers fluctuate because of closures and newly commissioned facilities. 12 From this national pool of community hospitals, 158 elected to participate in the Community Hospitals Project run by the NHSBN for the financial year of 2013/14. 13 Of these 158 hospitals, 101 returned questionnaires with the required information for the health economics analyses to determine key performance features (see Chapter 2, Community hospital data set). This sample comprises approximately 25% of community hospitals across England. Community hospitals across England will have a range of (unknown) reasons to participate in the NHSBN Community Hospitals Project and this may lead to a sample bias, for example over-representation of more adventurous community hospitals with high cost efficiencies. To address this, and to obtain more generalisable estimates of cost efficiency, we designed a national survey of community hospitals.

The intention of the survey was to augment the NHSBN Community Hospitals Project data set through a greater representation of community hospitals not only in England but across the UK. This study therefore contributed to the following project objectives:

• to measure the current relative performance (cost efficiency) of community hospital inpatient care for older people

• to identify the characteristics of community hospital inpatient care for older people that optimise performance.

Method

The focus of the national community hospital survey was to collect data from a larger and more representative sample of community hospitals that provide rehabilitation to older people in the UK. The aim was to develop a brief postal survey instrument that focused only on the four key cost efficiency parameters identified in study 1: bed occupancy, input prices, LOS and staff mix (see Chapter 2). Questions relating to these four parameters were extracted from the existing items in the NHSBN Community Hospitals Project, to ensure comparability. The survey instrument therefore comprised a covering letter and four questions presented in four A4 pages (see Appendix 3).

Stage 1

The first stage entailed the receipt of a list of community hospitals in the UK from the University of Birmingham community hospitals project team. The list had been sourced by the University of Birmingham from NHS Digital (formerly known as the Health and Social Care Information Centre) and was an accurate listing of community hospitals in 2012/13. This initial list contained 404 hospitals (559 rehabilitation/older people’s wards) of which 258 were in England, 11 were in Northern Ireland, 40 were in Wales and 95 were in Scotland.

Stage 2

The second stage involved a reconciliation process to ensure that the list excluded hospitals that had been closed or lost their beds or changed ownership. This was done in conjunction with the CHA, which provided business intelligence on recent hospital closures, and supplemented through internet searches and telephone calls to 291 individual community hospitals. The final list to be used for the postal survey contained 423 community hospitals (284 in England, 10 in Northern Ireland, 91 in Scotland and 38 in Wales), and an estimated 582 rehabilitation/older people’s wards in the UK.

Stage 3

To develop a contact list for the community hospitals, we sought the name of the matron/charge nurse/ward manager at each of the 423 sites. This information was obtained through internet searches and telephone calls.

The intention of this was to personalise the covering letter that accompanied the survey to facilitate improved completion and return rates.

Results

The pilot of the questionnaire had three returns (response rate 33.3%) from a sample of 10 community hospitals.

The initial mailing of the questionnaires to 587 rehabilitation/older people’s wards (423 community hospitals) resulted in only 17 responses (2.9%). We therefore contacted, by telephone, the ward managers/matrons for a selection of the non-responding sites to establish reasons for not completing the survey. The reasons provided included:

• ‘too busy on the ward to keep up with “non-essential” admin[istration]’

• ‘survey isn’t a priority’

• ‘no incentive to complete the survey’

• ‘don’t have information readily available to enable easy completion’

• ‘financial information not immediately available’

• ‘terminology (e.g. what do we mean by indirect costs?)’

• ‘needed manager‘s approval to complete survey’

• ‘information requested on finances is confidential’.

We subsequently modified the administration method by adopting a more focused approach that included only the non-participants of the NHSBN Community Hospitals Project. In this way, we hoped to learn something about these ‘non-responders’ and to determine if they were systematically different from the responding hospitals across our four key variables.

As part of this method we contacted approximately 10% of the non-responding hospitals (30 in England, 11 in Wales and 27 in Scotland). This resulted in an additional seven questionnaires from wards being returned.

Thus, we received a total of 24 survey questionnaires from the first and second methods of administration: 17 from the first method and seven from the second method. This represented a return rate of 4.1% from the overall 587 community hospital wards surveyed (Figure 5).

FIGURE 5.

Number and response rates for the national community hospitals survey.

Discussion

Our ambition had been to augment the NHSBN Community Hospitals Project data set using updated information obtained from a new national survey. It was hypothesised that the ‘more engaged’ hospitals might be over-represented in the existing NHSBN Community Hospitals Project data set. It was envisaged that the administration of the specifically developed MoCHA survey questionnaire would provide a more representative sample of community hospitals wards across the UK.

However, despite the active and helpful assistance of the CHA, and using the questions that had been previously tested in the NHSBN Community Hospitals Project, and despite a reasonably successful pilot survey (33% return rate) that allowed us to improve our methods, the main survey response rate was insufficient to allow us to re-calculate the estimates of cost efficiency – the main purpose of the national survey. This result was both disappointing and unexpected. National surveys of community hospitals have been conducted regularly by the CHA and there have been excellent response rates. It might have been that the purpose of our survey was either misunderstood, or understood but perceived as being of low value to the respondents. Feedback from some respondents supported the latter, which included comments such as: ‘too busy on the ward to keep up with “non-essential” admin[istration]’, ‘survey isn’t a priority’ and ‘no incentive to complete the survey’. It is also possible that the data we were seeking were not easily or consistently accessible at ward level.

Introduction

We anticipated that some components of community hospital ward care would not be captured or adequately summarised through the quantitative indicators of performance available for the cost-efficiency analysis described in the preceding chapter. Study 3 therefore sought to provide insight into the dimensions of rehabilitation practice as accomplished on community hospital wards with older people.

Community hospitals provide an important intermediate-care function within a repertoire of services aimed at providing transitional or time-limited support to people at a critical point on either side of the policy fulcrum that is the acute hospital (i.e. on the cusp of a shift from independence to dependence, following an acute event or exacerbation of chronic illness or illness in context of frailty). Rehabilitation to facilitate recovery is a major part of the intermediate-care function. Nevertheless, research on patients’ experiences of community hospitals primarily focuses on factors such as the ‘homely’ nature of the environment (physical and social: size and atmosphere); the localism or closeness of the community hospital to their own home and family; and the interpersonal, social and psychological aspects of care delivery (i.e. care as human connection). 9,10,61,62 Although central to the patient experience of care, less clear is how the work to facilitate recovery is actually accomplished. 63–66

There is broad agreement in research and policy on what is good rehabilitation practice. 67,68 It should be multidisciplinary, co-ordinated and patient centred: encompassing comprehensive need assessment; collaborative active engagement of the person in establishing goals for recovery in their terms; and provide enabling support, skills development, and emotional and practical assistance. Understanding what this ‘good practice’ actually looks like in the community hospital setting in respect of older people and how it is organised, would provide new insight into qualitative aspects of community hospital performance and what shapes it (objective 2). This understanding could also be incorporated into the interactive toolkits for use by local commissioners and community hospital teams to support operational changes to optimise performance (objective 4). To achieve these objectives, we undertook in-depth case studies of four wards in three community hospitals.

Methodology

We performed a multimethod, comparative case study design. Case study research69–72 is characterised by the depth and detail of information collected within naturally occurring, bounded settings, although the boundary between the setting and focus of analytic interest is permeable. 71 In this study, analytic focus was care delivery to older people with rehabilitation needs on community hospital wards. This is one aspect of the work of community hospitals that is also affected by external factors (e.g. referral practices and resources within the health and social care system). The distinctiveness of the case study method is its systemic and holistic perspective. 69 It facilitates in-depth and context-specific understanding of practice and how it is experienced in real time.

Methods

Findings

In the first section, we present the descriptive context of the three community hospitals. We then examine staff conceptions and delivery of rehabilitation and care, including with respect to people with complex needs, and identify quality dimensions of practice that emerged from the study. Finally, we draw together how therapy and care supported recovery and the processes for effecting ‘safe’, timely discharge, including how transitional and follow-on support was negotiated and organised.

Summary

In this study, we sought to provide insight into the less tangible components of community hospital working that might be incorporated into the interactive toolkits for use by local commissioners and community hospital teams to support operational changes to optimise performance. Our conclusions are based on the analysis of 226 patient observation episodes obtained from three study sites, and a total of 66 interviews with staff, senior managers and patients/carers. In many aspects, the three community hospital wards were very different: in the model of operation of the community hospital, in their physical environments and in the resources and support available to facilitate transitional and ongoing support, particularly for older people with frailty. This is not surprising, as a multimethod review of community hospitals in the UK and in other high-income countries concluded: ‘such hospitals exist in a variety of forms and encompass a varied range of services, diverse models of ownership and management and different degrees of co-ordination with general hospitals on the one hand, and with primary and community services, on the other’. 85

Even so, despite their differences, each community hospital was undergoing a process of change as a consequence of several linked strategic drivers: more effective and efficient use of acute hospitals; delivering health care ‘closer to home’, particularly for older people with complex needs; and an integrated response between sectors and services for those needing time, space and transitional support to recover from illness. Responding to the specific needs of patients with multiple and complex comorbidities made particular demands on community hospital ward staff’s knowledge, skills and resources in terms of how the clinical, care and therapy work with patients on the ward was organised and the engagement of family members and patients was pursued. Each of them adopted systems and mechanisms to facilitate interdisciplinary working, care co-ordination and patient/family engagement, which varied in content reflecting local circumstance, although their underpinning purpose and functions were similar. Nevertheless, forging an enabling and patient-centred rehabilitation ethos and practice across the whole of their patient population was a work in progress. In each case study, features of the physical environment, knowledge and skill deficits vis a vis the emerging patient group affected rehabilitation practice in ways that did not clearly differentiate the ‘cost-efficient performers’ and the less ‘cost-efficient performers’.

In each community hospital, there existed formal structures reflecting and reinforcing distributive forms of leadership stretching across the medical, nursing, therapy and care functions of the setting. These extended beyond the traditional MDT meeting that typically involved medical expertise and oversight to encompass arenas for in-depth progress review, concerted, interdependent action and co-ordination of work to effect movement in the patient ‘recovery journey’, including organising support to manage transition beyond the hospital.

Introduction

Continuing pressures on limited NHS finances and increased demand for services have led the NHS to consider ways by which to improve operational efficiency. The NHS Five Year Forward View86 sets out an ambition to achieve 2% net efficiency gains across the NHS each year for the rest of the decade.

A report by Monitor in 2013 estimated achievable savings of £1.2–1.8B in community services in general. 33 Given a total spend of £8.4B across these services, this represents around a 15% potential for cost reduction.

In this chapter we set out how we developed a toolkit for commissioners and community hospital providers to support operational changes to optimise the cost efficiency of community hospital ward care for older people. This is the final stage of our programme of work and related to objective 3 (study 4): to develop toolkits for commissioners and community hospital providers to optimise performance.

Methods

A web-based interactive toolkit for use by local commissioners and community hospital teams was formulated by integrating the outputs from the econometrics analysis (study 1, see Chapter 2), the data collected by the NHSBN and the outputs from the qualitative case studies (study 3, see Chapter 4).

The content of the toolkit was co-produced and iteratively modified by the Project Management Group, the CHA (lead Helen Tucker) and the Patients Association (lead Heather Eardley). The toolkit was based on previous similar work conducted by NHSBN for other clinical service areas. The toolkit comprises a database in structured query language server, an application interface built in a Node.js framework and a front end (user interface) built using Angular 2 technology. The toolkit is hosted on a Windows^® (Microsoft Corporation) server within Internet Information Services (Microsoft Corporation).

Initial testing was conducted by the NHSBN analytics team. The web pages were then published to the testing area of the NHSBN server. A further round of testing took place with three community hospital sites. Amendments were made at this stage to ensure that the toolkits functioned correctly as specified and met user requirements.

The toolkit was launched for discussion and feedback at the 2017 CHA Annual Conference (2–4 June, Manchester, UK) in the format of an interactive workshop. Delegates’ suggestions for further modification were recorded and incorporated into the final version.

Once testing was complete, the toolkit was published on the NHSBN live web and database environment and is now being promoted nationally through the CHA.

Toolkit content

The toolkit comprises three key components:

1. an efficiency calculator

2. scenario calculator

3. case studies.

Efficiency calculator

The efficiency calculator was developed to facilitate the assessment by local teams of their relative community hospital ward cost compared with other community hospital wards in the UK.

The calculator generates an ‘efficiency score’ for any community hospital for which data are entered into the tool by the user. The resulting efficiency score is then benchmarked against efficiency scores of other community hospitals derived from the analysis of community hospital data from the NHSBN’s 2014 Community Hospitals Project.

The formula used to calculate the efficiency score is:

c = f ( y , w , z , q , ) + ε ,

ε = u + v .

Formula reference examples:

• c – costs of staff, materials, and so forth

• y – a string of cost drivers such as the number of patients treated

• w – input prices, wages, capital costs

• z – environmental variables such as bed occupancy, mix of services provided

• q – quality of clinical audit results, research activities, meticillin-resistant Staphylococcus aureus (MRSA) occurrence, pressure sores, patient-reported experience measure

• ε – variables of inefficiency.

Scenario-based calculator

The key characteristics (indicators) of community hospital wards that optimise performance, as identified in study 1, are analysed in detail by the scenario calculator.

The characteristics considered in the calculator are pay cost per occupied bed-day, bed occupancy, nursing ratio, staffing levels and LOS. The tool allows the user to enter values for a community hospital ward for each of the indicators to assess the resource implications of moving to national average or target values.

The calculator shows the gap between local and national average or target levels for community hospital ward performance and the investment required to meet these levels or the potential savings available.

Case studies

A workshop was arranged at the 2016 CHA Annual Conference (13–15 May, Bristol, UK) attended by community hospital staff of various grades. Participants were clear that aspects of quality of care were important and needed to be incorporated into the proposed toolkit. Different methods of presenting the findings from the case studies were discussed including self-assessment questionnaires that explored ward culture and processes, checklists and training programmes. The favoured solution, however, was a series of case vignettes that demonstrated the findings from the case studies. The participants felt that this would be a familiar format for them and would allow the staff to reflect on their own practices in relation to the key rehabilitation care characteristics that had emerged. Subsequent meetings and feedback on the content and presentation of illustrative vignettes involved reciprocal exchange between researchers and leading members of the CHA. These were further tested out in workshops at the 2017 CHA Annual Conference.

Notwithstanding the diversity of their service models and the dynamic of health and social care systems within which the community hospitals operated, the critical features of quality practice, which framed the vignettes, resonated with participants. As elaborated on in Chapter 4, these embraced mechanisms for interdisciplinary working and communication, meaningful engagement with patients and families and goal-directed, purposive action to move patients forward to secure a ‘safe’ discharge.

The vignettes have been provided in Figures 6–8. The vignette in Figure 6 is intended to set the scene by describing the organisational context, model of service provision, patient profile and patient and service outcomes for the three case studies. It highlights the ways in which community hospitals appear to be undergoing a process of change in some respects, although there is continuity in others. It also sets out the quality criteria that we employed (i.e. criteria that broadly reflect a consensus in policy and research), although here specifically pertaining to the work with older people with rehabilitation needs in a community hospital environment. The scenarios conveyed in Figures 7 and 8 set out practices, patient experiences and discussion points relating to features of the work of community hospitals that staff indicated were challenging: the increased complexity of their patient profile, the dynamic and non-linear nature of recovery of patients with complex needs, and the demands on them in terms of skill and creativity in responding to the multiple and complex needs of these patient. They are not intended to convey one ideal solution; on the contrary, they are intended to prompt discussion of locally relevant solutions to similar issues encountered.

FIGURE 6.

Qualitative case studies: organisation, model of delivery, patient profile and outcomes, and quality criteria.

FIGURE 7.

Rehabilitation: a patient with frailty. A&E, accident and emergency; OT, occupational therapist; PT, physiotherapist.

FIGURE 8.

Complexity of patient need: a patient with coincidental dementia. OT, occupational therapist; PT, physiotherapist.

Publication

A full version of the toolkit can be found at http://mocha.nhsbenchmarking.nhs.uk/. 87

Discussion

The toolkit contains an important practical output available from the MoCHA study. It enables a summative, user-friendly resource, which allows the project outputs to be more accessible to a wider audience. Feedback on the toolkit was regularly sought from commissioners, community hospital teams and other stakeholders throughout the toolkit development process. Hence, the finalised toolkit will be in line with the users’ requirements.

The efficiency page of the toolkit enables community hospital staff to produce an efficiency score, using the equation developed during study 1 of the project. It was considered that a simplified approach to presenting this information was preferable. Thus, the efficiency calculation is hidden and users are required to enter only the relevant data for their community hospital. The score is then generated by the user selecting a ‘calculate’ tab. The efficiency score provides the percentage of savings that could be achieved if the community hospital was working in line with the most efficient hospital. The limitation of this approach, however, is that a proportion of the toolkit users may be interested in viewing the efficiency score calculation and information on how this calculation was produced. To mitigate against this limitation, a link with further information on the calculation is provided on the toolkit’s efficiency page.

The economic analysis used data sets collected from the NHSBN 2013/14 Community Hospitals survey13 and the 2014 NAIC survey27 to produce the efficiency score. However, the toolkit requires the community hospital staff to input the most recent data available to calculate a local efficiency score. This discrepancy is outlined in the toolkit. Annual updates are possible but would require a source of funds.

The toolkit’s calculations page reviews the metrics that were identified in the economic analysis as having the greatest influence on a community hospital ward efficiency score. These were input prices, nursing staff skill mix, bed occupancy and LOS. The resources required to bring the index community hospital ward current values in line with a national average or target rate is assessed. The nursing ratio in a community hospital ward is also included in the calculations page. To easily assess the impact of changing certain metrics, they are reviewed in isolation. The impact on a wider range of metrics needs to be fully considered before any changes in service delivery are made.

The case studies developed to fulfil the requirements for study 3 of the project identified working practices and service delivery that may have contributed to differences in cost efficiency that were not accounted for in the quantitative health economic analyses. The practices identified were distinctive for each of the three case studies but have not been formally linked to cost efficiency. They were, however, greatly valued by the community hospital staff involved in developing the toolkit.

The qualitative nature of the case study material was not easily captured for use within the otherwise quantitative toolkit. The case study information is supplied as downloadable Portable Document Format (PDF) files. 87 Alternative formats were considered but they were not achievable because of the limited resources available. A more interactive presentation of the qualitative information might be considered for future projects.

Introduction

Increasing emergency admissions to hospital is an established issue facing health-care systems internationally. A key factor underpinning the increase is the complex needs of older people with frailty and multimorbidity. 88 The continuing global demographic transition89 implies that urgent solutions are required. In England, the NAO has reported that over half of the growth in emergency admissions is related to older people. 55 More detailed analysis demonstrated that an important subgroup of emergency admissions comprised patients with very short LOSs: 79% of the increase in emergency admissions between 2013/14 and 2016/17 was caused by people who did not stay overnight, such that nearly one-third of emergency admissions in 2016/17 did not stay overnight. 90 However, there is evidence from national data in England that this subgroup of emergency admissions involving older people varies substantially across local areas with a 9.3-fold variation. 91 This suggests that some areas have developed services, or have care systems in place, that are capable of addressing the needs of patients who would otherwise be admitted to hospital for < 24 hours.

The term intermediate care refers to a range of community services that have formed part of national health-care policy in England since 2001. 24 The defining characteristics of intermediate care are that they are community-based services, they are multiagency and multidisciplinary, and they provide a support and rehabilitation model of care that is time limited (generally up to 2 weeks and no more than 6 weeks). 24

A practical typography of intermediate care has been developed27 that identifies four common types of intermediate care:

1. bed-based intermediate care comprising community hospitals and short-term care home admissions

2. home-based intermediate care comprising ‘hospital at home’ services

3. crisis response teams (CRTs) that aim to respond within 48 hours of referral

4. reablement services, which are similar to ‘hospital at home’ services but primarily funded by local councils rather than health-care commissioners.

Department of Health and Social Care policy stipulated that one function of intermediate care is a focus on admission avoidance. 24 It is therefore plausible to consider that one explanation of the national variation in emergency admissions with short LOSs might relate to the deployment of these services in terms of scale and type. However, the effectiveness of intermediate-care services for hospital admission avoidance is unknown. 92 To investigate this aspect of intermediate care, we used national data sets to explore the relationship between the provision of intermediate care in local areas and emergency admissions to hospital.

This study related to:

• research objective 4 – to investigate the impact of community hospital care for older people on use of secondary care

• research objective 5 – to determine if there is an association between short-term, community-based services and reduction in secondary care utilisation by older people.

Method

Our plan was to use multiple regression analyses to investigate the relationship between the dependent variables (indicators of secondary care utilisation) and independent variables characterising the quantum of intermediate-care service types in local areas. We anticipated that the independent variables would relate to measures of intermediate-care service type activity and that the data used in this work were aggregated to the level of a CCG.

Results

The Atlas reported the emergency admission indicator (‘Map 61’) for all CCGs (n = 211) with a mean of 5261 admissions per year (standard deviation 2046 admissions per year; range 1186–11,011 admissions per year) per 100,000 population per CCG. However, data on intermediate-care service provision from the NAIC were restricted to the CCGs that had enrolled in the audit and not all of the participating CCGs provided the full range of intermediate-care service types (Table 11).

Discussion

This study investigated the impact of different types of intermediate-care services (including community hospitals) on secondary care utilisation. Intermediate-care services have been established in many health-care systems internationally. They operate in the ‘virtual space’ between primary and secondary care and are directed especially, but not exclusively, on the needs of older people with, or recovering from, acute illness or injury. Prevention of emergency admissions to hospital is a stated ambition for intermediate-care services in the NHS in England and Wales. 24 It is unclear to what extent they fulfil this function. We believe this to be the first large-scale study to investigate this issue. It was considered an important question for service commissioners who wish to understand not simply the effect of services on clinical outcomes, but how services might be inter-related and the extent to which services might operate as a substitutes for one another.

Our starting point was a national data set: the Atlas of Variation. 91 This reports population-level outcome indicators for each CCG in England. We selected Map 61 as the one most likely to be influenced by the provision of local intermediate-care services. This indicator, ‘emergency admission with a stay of < 24 hours for people aged ≥ 75 years per 100,000 population’, reports on a group of patients who will be familiar to health-care professionals working in acute care. These patients present to hospital acutely but are well enough to return home the following day. We hypothesised that this patient group would represent an ideal population for intermediate-care services and that increased intermediate-care activity might decrease these brief admissions. However, we recognised that intermediate care is an umbrella term that includes several service models. For this reason we investigated the four most common forms of intermediate care, as described and defined in the NAIC. 28 These were bed-based intermediate care (community hospitals and care home beds); CRTs; home-based intermediate care; and reablement services.

Our main finding was that the CRT type of intermediate care, and only the CRT type, had a statistically significant negative association with the emergency admission indicator. Thus, as CRT activity increased (quantified by referrals to the CRTs), there was an associated decrease in emergency admissions with a LOS of < 24 hours for people aged ≥ 75 years. This result is based on CRT activity data from 42 CCGs. We anticipated a non-linear relationship and indeed this was so. But the final regression model we developed dealt effectively with the critical statistical issues of heteroscedasticity and goodness of fit and the results can therefore be considered robust. It is reasonable to assume that the CRT form of intermediate care can function as a substitution service for brief emergency admissions to hospital for older people.

It should also be noted that we did not find an effect of community hospitals (bed-based intermediate care) on admission avoidance. In National Audit of Intermediate Care 2014: Key Findings,27 it was reported that the largest source of referrals for bed-based intermediate-care services was acute wards (60%), followed by GPs (18%) and social care (6%). Thus, admission avoidance represents a minority function for bed-based intermediate-care services.

Introduction

Community hospitals are small hospitals with a range of locally provided inpatient and outpatient services95 and are established health-care facilities in the UK and internationally. 85 Although they can provide a range of local services (e.g. minor injury units, day hospitals, satellite secondary care clinics), 98% of community hospitals in the UK provide inpatient rehabilitation for older people. 26–29 In the context of an ageing population, this is a potentially valuable health service function. Indeed, high-quality evidence supports community hospitals as a clinically effective bed-based rehabilitation service for older people when compared with larger general hospitals,6,8 and there is evidence that inpatient community hospital rehabilitation for older people is cost-effective relative to secondary care in England. 7 However, survey information from 180 community hospitals in England indicates considerable variation in operating costs and outputs of these hospitals. 13 This is an important consideration for health-care planners who need to make decisions on how to deploy resources optimally for local populations.

We developed several inter-related studies to explore and understand more fully the nature of these variations in community hospital performance. The previous chapters have presented and discussed the constituent studies that constituted our programme of work. They have also conveyed the reasons why we were unable to pursue aspects of the intended study that are set out in the published protocol. Here, we return to our research objectives to discuss the key findings and implications for community hospital delivery and organisation.

Research objective 1

To measure the current relative performance (cost efficiency) of community hospital inpatient care for older people (studies 1 and 2).

Cost efficiency

Efficiency represents how well inputs such as labour, capital and materials are used to produce outputs. In the case of inpatient community hospital rehabilitation, outputs are the treatment of patients. Cost efficiency is producing a given output at lowest cost. Frontier techniques have been applied widely in health care to measure cost efficiency,36–40 including estimates of the cost efficiency for inpatient care in large hospitals. 43,44,46 To our knowledge, this is the first study to investigate the cost efficiency of inpatient community hospital rehabilitation using frontier analysis techniques.

The modelling process can be thought of conceptually as having two stages (although estimation proceeds in a single stage). In the ‘first stage’, an equation is estimated that describes the relationships between the costs of a case and its outputs, inputs, environmental influences and quality variables – these relationships are referred to as elasticities (how much costs change as the variables change). The ‘second stage’ calculates the efficiency of each case, using the elasticities (the relationships between costs, outputs and other variables) from the first stage. The SFA techniques attempt to mathematically distinguish the efficiency of a case from random factors that might also have affected its outputs (e.g. bad weather, industrial action). Figures 1 and 2 provide a visual representation of the concept with an efficiency ‘frontier’ demonstrating the minimum possible cost for different output levels (bed-days). Each of the points represents a hospital. The vertical distance of each hospital from that frontier is made up of deviations resulting from random factors (noise) and also lower efficiency that may affect costs in a given year, as noted above.

Because the ‘second stage’ of the analysis (i.e. the estimation of efficiency) is dependent on the assumed distributions of efficiency and random factors, it is usual in this second stage to use several different sets of assumptions about these distributions, thereby producing several different estimates of efficiency, testing whether the estimates are stable over a range of assumptions, or whether they are dependent on certain assumptions. We therefore developed six different models (COLS and five SFA models), each with a different assumption about the distribution of efficiency of community hospitals (the distributions used for the SFA models were half normal, truncated normal, exponential, gamma and Rayleigh) to examine whether or not the estimates were robust to changes in the distributional assumptions.

The results of this SFA analysis suggest that cost efficiency in community hospital wards in the UK is comparable with that of the NHS acute hospital sector (albeit previous studies have considered the whole hospital rather than analysis at the ward level). 43,44,46 Overall, the results from the models were robust. The parameter estimates were within plausible ranges and the results appeared to be consistent over a range of specifications and with the COLS base case, which does not impose the distributional assumptions that the SF models do. The models’ predictions of efficiency were highly correlated. Furthermore, our finding of economies of scale was supported when results were presented to community hospital staff workshops. The estimated frontier model reported enables community hospitals in the sample to observe their relative efficiency against the estimated cost frontier (representing the most cost efficient hospitals). It should also be noted that those hospitals operating at the highest cost efficiency can be readily distinguished (but were not identified in our study because of our use of anonymised records), which is useful to facilitate learning across the sector. Moreover, the cost frontier was translated into a practical web-based tool (study 4), meaning that community hospitals can now benchmark their performance after inputting their individual characteristics.

The findings have several important policy implications. The efficiency estimates imply that, on average, community hospitals could reduce operating costs in the provision of inpatient wards by 17% if performance was in line with the most cost-efficient hospitals in the data set. This would result in around £47M of available annual gains in relation to our sample of 101 community hospital wards. There was a range of performance, with the lowest-performing community hospital operating at around 0.55 efficiency, although the highest-performing community hospitals were operating at around 0.94 (Figure 9). That is, the model indicates that all community hospitals can improve efficiency on their wards.

FIGURE 9.

Distribution of the community hospital ward (n = 101) efficiency estimates. Dotted lines are minimum, mean and maximum values. HN, half normal.

The economies of scale findings suggest that, for wards with size below the mean, average costs can be reduced by increasing the scale of the operation. We found that a 1% increase in bed occupancy is associated with a 0.94–1.24% decrease in operating costs. However, this is not simply a question of increasing bed occupancy. The NAO has suggested that services in which bed occupancy levels are in excess of 85% may experience disruption (in the general hospital setting)55 and we note that our sample mean for bed occupancy is around 90%. More importantly, we found that a 1% rise in output is associated with an average 0.85% increase in costs, that is, output can rise more quickly than costs. This again points to opportunities for efficiency gains.

The interpretation of the efficiency estimate is as follows. An estimate of 1 means that the community hospital is fully efficient, otherwise it is not. So, taking 0.8 as an example, a community hospital could reduce its costs by 20% (i.e. 1 – 0.8 = 0.2) if it were to improve its performance in line with best practice.

Research objective 2

To identify the characteristics of community hospital inpatient care for older people that optimise performance (studies 1 and 3).

Research objective 3

To develop web-based interactive toolkits for use by local commissioners and community hospital teams that support operational changes to optimise performance (study 4).

From a community hospital care delivery perspective, this was regarded as an important output of our programme of work. Essentially, this is a practical tool for use by front-line community hospital staff and service managers that operationalised the evidence base developed in our research studies. This work was led by NHSBN which has considerable experience in building, testing and launching this type of interactive toolkit. Involvement of the membership of the CHA and subsequent endorsement and promotion of the toolkit was similarly vital to our shared work. The toolkit is now available at http://mocha.nhsbenchmarking.nhs.uk/. 87

Research objectives 4 and 5

To investigate the current impact of community hospital inpatient care for older people on secondary care and the potential impact if the community hospital care was optimised to best practice nationally (study 5).

To determine if there is an association between the configuration (capacity and proportions) of short-term, community-based services (i.e. community hospital wards, home-based rehabilitation, care home rehabilitation and enabling services) and reduction in secondary care bed utilisation by older people (study 5).

These were demanding and ambitious objectives for which we made important but incomplete progress. 96 We successfully utilised data from the Atlas of Variation91 and linked this to the NAIC data set available to us from NHSBN. Thus, we took a broader perspective than solely the contribution of community hospitals. Rather, we wished to consider the range of community services under the umbrella term intermediate care. Many definitions of intermediate care have been proposed. For the English NHS, the Department of Health and Social Care published a definition of intermediate care:15

. . . a range of integrated services to promote faster recovery from illness, prevent unnecessary acute hospital admission and premature admission to long-term residential care, support timely discharge from hospital and maximise independent living.

Department of Health and Social Care. 15

© Crown Copyright. Contains public sector information licensed under the Open Government Licence v3.0

These services have undergone rapid expansion nationally since they were promoted in the National Service Framework for Older People in 2001. 15

The NAIC data set is a unique community service data set that contains descriptive, activity, process and outcome variables on four well-categorised and nationally common types of intermediate care: bed-based intermediate care (community hospitals and care home beds), CRTs, home-based intermediate care and reablement services (delivered by local councils). Thus, the audit encompasses the core community services that are essential to improve and/or maintain the independence of older people with dementia, multimorbidity and frailty after an acute illness or injury, or potentially to provide an urgent response to prevent hospital admission.

Although our eventual research focus on Map 61 was pragmatic, it also made good sense from a health services policy perspective. One of the founding ambitions of intermediate care was that it would lead to reduced demand for emergency hospital admissions for older people. 15 As far as we are aware, this ambition for intermediate care has not been subjected to robust scrutiny. And if intermediate-care services are indeed able to reduce demand for acute hospital care, it seemed it would be most likely to affect transient emergency admissions (LOSs of < 24 hours), as this population of older people might be most readily supported in the community (i.e. admission avoidance).

Our main finding was that the CRT type of intermediate care had a statistically significant negative association with the emergency admission indicator. Thus, as CRT activity increased, there was an associated decrease in emergency admissions with a LOS of < 24 hours for people aged ≥ 75 years. It should be noted that none of the other types of intermediate-care activity was associated with a reduction in emergency admissions. Thus, for bed-based (that includes community hospitals) and home-based intermediate care, and reablement services, no statistical link to admission avoidance was found in our analyses.

The modest sample size of 42 CCGs precluded a reliable calculation of effect size. This would be a useful future extension of our work. A further extension would be to examine a broader definition of secondary care utilisation. We requested a bespoke data set from NHS Digital to obtain CCG-level activity data for numbers of emergency admissions for people aged ≥ 75 years with a LOS of > 24 hours, and duration of admission. We believed this might lead to a better understanding of the contribution of the four types of intermediate care separately, and in combination, on secondary bed use. For example, it might then be possible to understand the role of community hospitals in supporting the general hospital in terms of releasing beds through discharge support (‘step-down’ care). However, negotiating the data specifications proved a laborious process that ran over the time window we had available to us.

There is much uncertainty about how best to address the pressing issue of admission avoidance. 97 Our finding that the CRT form of intermediate care can function as a substitution service for brief emergency admissions to hospital for older people is therefore an important one for commissioners. It provides greater confidence that investment in these types of services can indeed reduce secondary care utilisation for a growing and vulnerable population group. Our findings are in accord with findings from the NAIC, which enrolled 60 CRTs with 60,384 associated patients and reported that only 10% subsequently required admission to hospital. 28 This was interpreted as evidence that CRTs can support the majority of patients referred to them within their own homes.

Conclusions

The main results suggest that efficiency of community hospital wards in the NHS is comparable with that of the NHS acute sector. Furthermore, they suggest economies of scale in the provision of ward services and potential efficiency savings that might be augmented by economies of scale. As a top-down analysis, it therefore serves as a challenge for further research as to how less efficient community hospitals might reduce costs. We also present the first evidence to confirm the admission avoidance function of the intermediate-care service model of CRTs. This finding did not apply to the current configuration of community hospitals, which are currently more frequently used as step-down facilities. Given the still increasing demands on acute care internationally, the strategic role of community hospitals for admission avoidance remains an important area for future research. The work described here indicates the practical and methodological challenges involved in combining data sets that employ different service models and different geographical footprints. The NAIC currently combines data for all types of bed-based services. Future rounds of the audit might usefully separate out the different bed-based service models to facilitate learning.

The case studies reinforced the significance of a whole-system approach to understanding community hospital performance. How people were routed into community hospitals, the characteristics of such patients, the performance standards they were expected to achieve and the resources available to carry out their intended purpose were affected by dynamic national and local health and social care factors. The increasing complexity of patients, in turn, shaped a conception of rehabilitation as contributory to a process of recovery that would continue in the person’s own familiar environment or provide opportunity and space for patients and families to come to terms with a move to long-term care.

The quality dimensions of practice identified were integral to securing ‘timely’, safe discharge: conjoint, interdependent and interprofessional working and knowledge-sharing drove goal-planning and action forward, informed by engagement of patient and family members. These mechanisms and practices established a framework for ongoing progress review and identifying support necessary to facilitate ‘safe’ discharge, both an end point in the hospital episode and a critical juncture in the patient journey.

Research implications

1. Future research should be based, when possible, on panel data sets that follow key variables over time. This would allow increased analytic power and should be possible given the NAIC is now in its seventh iteration, and that the Atlas of Variation is regularly updated.

2. Future national surveys of community hospitals should include our four key variables so that cost efficiency can be calculated as a new summary outcome for community hospital ward performance.

3. A programme of development work could be initiated to examine the extent to which community hospital wards are able to optimise their quality and efficiency using the published toolkit. This might include new training programmes for staff working in this sector that take account of the increased complexity of patients.

4. Community hospitals tend to be considered in isolation. We recommend more whole-system analyses, possibly using other national ‘Maps’ available in the Atlas of Variation, to further investigate the impact of community hospital activity on secondary care utilisation. Our preliminary econometric analyses suggest that this is methodologically feasible.

5. At care-delivery level, there is need for in-depth research into the relationship between patient profile, mechanisms and practices for delivering rehabilitation, patient-recovery trajectories and destination outcomes.

6. There is need for the development of a more granular understanding of intermediate-care service use in respect of different service models and their combinations to better understand how to optimise outcomes for patients with particular characteristics.

Contributions of authors

John Young (https://orcid.org/0000-0003-4085-9306) prepared the Abstract, Plain English summary, Scientific summary, and Chapters 1, 3 and 7, which were then improved and edited by the other authors. He also edited Chapters 2, 3 and 6.

Claire Hulme (https://orcid.org/0000-0003-2077-0419) co-wrote a larger report from which Chapter 2 was drawn, and also wrote a larger report (alongside the health economics team) from which Chapter 6 was drawn.

Andrew Smith (https://orcid.org/0000-0003-3668-5593) co-wrote a larger report from which Chapter 2 was drawn.

John Buckell (https://orcid.org/0000-0002-4157-4217) co-wrote a larger report from which Chapter 2 was drawn.

Mary Godfrey (https://orcid.org/0000-0002-2408-534X) wrote Chapter 3, with Helen Tucker (https://orcid.org/0000-0003-2829-9115) contributing to the writing of the case study scenarios.

Claire Holditch (https://orcid.org/0000-0002-9250-952X) and Jessica Grantham (https://orcid.org/0000-0003-0759-676X) co-wrote Chapter 5.

Pam Enderby (https://orcid.org/0000-0002-4371-9053) and John Gladman (https://orcid.org/0000-0002-8506-7786) provided specific input into Chapters 1 and 7.

Elizabeth Teale (https://orcid.org/0000-0002-5923-3170) provided specific input into Chapter 1 and the Scientific summary.

Jean-Christophe Thiebaud (https://orcid.org/0000-0003-0331-4753) wrote the econometric analysis section of Chapter 6.

All chapters were reviewed by all authors.

Data-sharing statement

No new quantitative data were generated by this programme of work. The data sets we used for the econometric analyses are the property of the NHSBN. The qualitative (case study) material is held by Mary Godfrey (Academic Unit of Elderly Care & Rehabilitation, University of Leeds). The toolkit is published on the NHSBN website and available to use after prior registration. All data requests should be submitted to the corresponding author for consideration.

Patient data

This work uses data provided by patients and collected by the NHS as part of their care and support. Using patient data are vital to improve health and care for everyone. There is huge potential to make better use of information from people’s patient records, to understand more about disease, develop new treatments, monitor safety, and plan NHS services. Patient data should be kept safe and secure, to protect everyone’s privacy, and it’s important that there are safeguards to make sure that it is stored and used responsibly. Everyone should be able to find out about how patient data are used. #datasaveslives You can find out more about the background to this citation here: https://understandingpatientdata.org.uk/data-citation.

Disclaimers

This report presents independent research funded by the National Institute for Health Research (NIHR). The views and opinions expressed by authors in this publication are those of the authors and do not necessarily reflect those of the NHS, the NIHR, NETSCC, the HS&DR programme or the Department of Health and Social Care. If there are verbatim quotations included in this publication the views and opinions expressed by the interviewees are those of the interviewees and do not necessarily reflect those of the authors, those of the NHS, the NIHR, NETSCC, the HS&DR programme or the Department of Health and Social Care.

Introduction

This paper provides an introduction to the economic analysis tool SFA that is being used in the Community Hospitals Research Programme on Efficiency by a team of researchers at Leeds University. This briefing paper provides supporting information for a presentation and workshop prepared for the CHA conference.

The researchers are interested in feedback from practitioners working in community hospitals on their view of which factors can influence efficiency in practice. We are also interested in ‘variables’ that may affect the service. We are available for any questions and further discussion on this tool and its application to community hospitals in practice.

Please contact: XXXX.

Fundamentals of economics

In the language of economics, firms are said to be operating a production process that uses inputs (labour, capital, materials) to produce outputs. In the case of community hospitals, inputs might include the costs of employing staff, buying medicines and running the premises, whereas outputs will typically include the health service provided (i.e. treating patients) and will incorporate aspects of service quality. We can examine this ‘production process’ through cost and other key data. Some community hospitals are adept at converting their inputs into outputs, that is, they are efficient. Other community hospitals are less so, that is, they are inefficient. We are interested in quantifying this efficiency/inefficiency spectrum. If a community hospital’s costs are our measure of interest, then we are measuring cost efficiency.

Measuring cost efficiency

In order to measure cost efficiency, we can construct an efficiency line or a ‘frontier’ that represents the minimum possible cost for a community hospital given its output. The line advances upwards as the number of cases treated increases and, therefore, some of the costs also increase (e.g. medicines used). This line would represent the most cost-efficient community hospital and can be viewed as a benchmark against which other community hospitals can be compared. We can therefore plot other community hospitals in relation to this most cost-efficient line. The further the particular community hospital lies above this line or frontier, the more that particular community hospital can be said to be cost inefficient. If we plot data from lots of community hospitals, we can measure relative cost efficiency. This is shown in Figure 10.

FIGURE 10.

The efficiency frontier line. For simplicity, the frontier is shown as a straight line but it is more commonly a curved line.

The cost function

The shape of the efficiency frontier is defined using a cost function that shows how costs (c) are related to a string of cost drivers including outputs (y) (e.g. the number of patients treated), input prices (w) (e.g. wages, capital costs), environmental variables (z) (e.g. bed occupancy, mix of services provided) and quality (q) (e.g. clinical audit results, research activities, MRSA occurrence, pressure sores, patient-reported experience measure)

Distributional assumptions

One issue with the SF model is the need to make distributional assumptions and the effect of the assumption on the prediction of inefficiency. Because of this, there has been some concern around the prediction of inefficiency. 42 It is possible to use a number of different assumptions for the inefficiency term and, therefore, to make a number of assumptions about the form of inefficiency and then to test between models. Moreover, it is possible to test the correlation between models’ predictions, both in terms of absolute results and rank, to assess robustness. Therefore, the following models are tested (Table 15).

For further discussion of distributional assumptions, see Coelli et al. 41 For the Rayleigh distribution, see Hajargasht. 98

To test for inefficiency, we test each model against the OLS as a base model. We adopt a LR for this purpose.

In total, we estimate 12 models (six specifications of the distribution of inefficiency; two functional forms). We report the results for six models, having applied the test for functional form.

We use LIMDEP version 10 (Econometric Software, Inc., Plainview, NY, USA) for estimations. 99

The full report of this analysis, written by Claire Hulme and economics colleagues, may be obtained if requested from the authors.