Comprehensive geriatric assessment for frail older people in acute hospitals: the HoW-CGA 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 12/5003/02. The contractual start date was in November 2014. The final report began editorial review in January 2018 and was accepted for publication in June 2018. 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

Simon Paul Conroy reports working for the Acute Frailty Network, an improvement collaborative focusing on acute hospital care for frail older people. Graham Martin was a member of the Health Services and Delivery Research Priority Research Advisory Methods Group during the course of this study. Stuart Parker reports grants from the Alzheimer’s Society, grants from the British Heart Foundation and grants from the National Institute for Health Research (NIHR) School for Primary Care Research, during the conduct of the study. Helen Roberts is a member of the NIHR Journals Library Editorial Group.

Disclaimer

This report contains transcripts of interviews conducted in the course of the research and contains language that may offend some readers.

Copyright statement

© Queen’s Printer and Controller of HMSO 2019. This work was produced by Conroy 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

About language

The internationally recognised name of a medical specialty (Geriatric | Medicine) and the title of its key technology [Comprehensive Geriatric Assessment (CGA)] are now well established and contain the term ‘geriatric’, the origins of which were described in a treatise by Ignatz Leo Nascher, published in 1913 and entitled ‘Longevity and Rejuvenescence’, as follows:1

Geriatrics, from geras, old age, and iatrikos, relating to the physician, is a term I would suggest as an addition to our vocabulary to cover the same field, in old age, that is covered by the term ‘paediatrics’ in childhood.

Nascher IL. ‘Longevity and Rejuvenescence’ (1913), quoted in Freeman JT. Nascher: Excerpts from His Life, Letters, and Works. Gerontologist 1961;1:17–261

Unfortunately, in the intervening century, and in contrast to the adjective ‘paediatric’, the term ‘geriatric’ has acquired a pejorative meaning when applied to older people (or indeed to objects), implying uselessness and decrepitude, as the following quotations illustrate:

Geriatric judges with 19th century social and political prejudices only bring the rule of law into disrepute. 2

I hear and read such phrases as ‘geriatric old twit’: an expression which would hardly have sprung to the lips of the pious Aeneas. 2

In the preparation of the report for this project, challenges were presented by older people attending a dissemination event, and by the project’s external steering group, to consider alternative language to replace the term ‘geriatric’ throughout the report. This was a challenge, not least because the word appears in the name of the health and care technology that is the subject of the report.

Accordingly, we held a discussion with members of the Voice North group and circulated the content of the discussion for comment by members. All agreed that the term ‘geriatric’ had acquired a pejorative meaning, particularly when applied to older people. Alternative descriptors for older people were suggested by group members, who advised that we should avoid the terms ‘geriatric’ and ‘the elderly’ and should instead refer throughout to ‘older people’, which we have tried to do.

In addition, instead of using ‘CGA’, ‘multidisciplinary assessment and management’ is used, which is consistent with the most commonly used definition of CGA, namely that of a multidimensional, multidisciplinary process that identifies medical, social and functional needs, and the development of an integrated/co-ordinated care plan to meet those needs. 3

For consistency with the existing literature, the abbreviation ‘CGA’ is used to refer to this assessment and management process.

The origins of a multidisciplinary approach being used and applied to vulnerable and older people in the UK can be traced to what has become the foundational literature of the medical specialty of older people’s medicine (OPM) in the 1940s. 4 Subsequent development and, importantly, evaluation in randomised controlled trials (RCTs) was reported in the 1980s. By the 1990s, meta-analyses of multiple RCTs were being reported and are being maintained up to the present day. 3,5,6

The term ‘CGA’ has become established as shorthand for the ‘technology of older people’s medicine’. It is no longer new7 and it has acquired status as a proven, effective and essential component of the assessment and management of older patients in hospital and community settings.

Comprehensive Geriatric Assessment is usually delivered by a multidisciplinary team (MDT), sometimes working in a specific ward environment but more often as part of a mobile or peripatetic consultation service. 3 CGA teams may use specific assessment tools and protocols to aid the assessment process and usually meet regularly to discuss and co-ordinate the assessment and, crucially, the associated treatment goals and management plans of older people. 3

Compared with ‘usual care’ in RCTs in hospital settings, CGA has been shown to have positive effects on key personal and operational outcomes. The CGA process increases the likelihood of being alive and living at home and avoiding institutionalisation, death and deterioration3 in relation to an episode of inpatient hospital care.

The participants in the trials that established the effectiveness of CGA were mostly older people, defined by the norms of the era and location in which the trials were performed. Some may find it surprising that participants in these trials could be as young as 50 years, with the majority of participants being described as in the ‘60+’ and ‘65+’ age ranges. 8

Comprehensive Geriatric Assessment in hospital settings

Structural population change and improved health over the past 100 years has led to older people becoming the predominant users of inpatient hospital services. Older people admitted to hospital often have complex needs, have multiple comorbidities and are at high risk of complications and adverse outcomes. They often present with the clinical syndromes associated with frailty, which include falls, loss of mobility, confusion and incontinence9–13 – the so-called ‘Geriatric Giants’. 14

There is considerable evidence for using CGA in the care of patients with these clinical problems, including in a variety of specialised inpatient settings, and for conditions that are common in old age. 15,16 For example, many hospitals have ‘Evaluation and Management Units’ in which a MDT provides a multidimensional assessment and develops a management plan in collaboration with the patient and carers. These plans include rehabilitation, discharge-planning, co-ordination and follow-up and are personalised for each patient. This model is also seen in single-condition care units, such as orthopaedics (fracture) and stroke units, which were developed using the principles of comprehensive assessment and multidisciplinary care that underpin the CGA process. 17,18

Elsewhere in hospitals, CGA has been less available and understood, perhaps on account of uncertainty about how to direct care to suitable recipients. The clinical trials that showed the effectiveness of CGA did not always stratify participants by characteristics that we consider important today, such as the presence or absence of specific clinical syndromes, or the presence or severity of frailty. Neither were they focused on identifying solutions for the whole hospital,19 for example by providing CGA at the front door,20,21 or the cost-effectiveness of different services and settings.

There is emerging evidence of the development of new forms and settings to deliver CGA. A recent benchmarking survey conducted in 49 acute services in the UK22 showed that 34% of trusts had developed enhanced teams in the emergency department (ED) and 42% of trusts had developed frailty units. About half of short- and intermediate-term hospital-based assessment units were using CGA, with 25–44% having a dedicated team. This survey identified that in 59% of the health and social care economies surveyed, recognised assessment tools and pathways for frailty were used.

Hospital-wide Comprehensive Geriatric Assessment

As population ageing progresses, older people are becoming the majority users of inpatient hospital services. In this context there is a clear need for the hospital of the future to provide services structured around the needs of patients. 23 Any coherent vision of a hospital fit for the future must, of course, include making hospitals ‘good places for old people’. 24

Although a compelling argument can be made for the effectiveness of CGA,3 and we can see that it is beginning to be delivered outside the traditional boundaries of the specialised ‘Evaluation and Management Unit’, the question of its potential beyond specialised inpatient services remains open.

It is possible that optimisation of inpatient care could include the provision of CGA by hospital inpatient services so that all hospital inpatients with the potential to benefit from the process would receive timely and effective CGA to shape the clinical decision-making process to meet their complex needs. For example, for those undergoing elective surgery, CGA might be incorporated into the pre-operative care pathway25 to ensure that decision-making around the procedure itself, as well as post-operative rehabilitation, might be optimised in a way that might improve outcomes, minimise length of stay and facilitate recovery.

To achieve a vision of timely and effective use of CGA on a hospital-wide basis, the development of new and innovative service models and the tools to support its implementation are required.

Overview of the Hospital-Wide Comprehensive Geriatric Assessment study

Consideration of these issues prompted a call for research proposals by the National Institute for Health Research (NIHR), one result of which is the collection of research projects described in this report. In this series of interdependent studies, existing models of care and the development and validation of tools to deliver CGA on a hospital-wide basis are described.

Research questions

The main questions addressed by this programme of research were:

• How is CGA defined and recognised?

• How, and in what forms is CGA currently organised and delivered in the UK?

• Who receives CGA, and can we identify who benefits most?

• How can we develop tools to assist the delivery of CGA on a hospital-wide basis?

Aims and objectives

Presentation of this report

Multiple components of this project are presented in the following chapters. Chapter 2 presents an umbrella review of the literature and an overview of new and emerging service models. Chapter 3 presents the results of a national survey of services providing CGA in acute hospitals. Chapters 4–7 deal with the creation of tools to understand the need for CGA, characterising beneficiaries and testing Hospital Episode Statistics (HES)-based proxies for frailty using a range of approaches, including:

• population segmentation

• diagnoses linked to frailty

• clustering diagnoses linked to frailty

• deriving an ordinal ‘frailty score’ from clusters

• costs of CGA

• combined information tools.

Chapters 8 and 9 present the development of implementation strategies and tools and the evaluation of the toolkits and their implementation. Chapter 10 provides a summary and discussion of the Hospital-Wide Comprehensive Geriatric Assessment (HoW-CGA) project and its findings.

Patient and public involvement has been critical to the project, indeed forming one of the five main workstreams (WSs). In addition to being cited where relevant in the individual chapters, a summary of the PPI activity is included in Appendix 1.

Umbrella review

An umbrella review provides an overview of existing systematic reviews. 28,29 The principal objectives of this review were to:

• define characteristics of the main beneficiaries of CGA

• define key elements of CGA

• define principal outcome measures

• summarise evidence on the cost-effectiveness of models of delivery of CGA

• highlight gaps and weaknesses in the evidence base across relevant inpatient clinical areas.

Methods

The Joanna Briggs Institute (JBI) umbrella review method was used. Systematic reviews and meta-analyses that described the provision of CGA to hospital inpatients aged > 65 years were reviewed. Comparisons of inpatient CGA to usual inpatient care were included.

Five reviewers (PM, SPC, SP, HR and KP) worked in pairs to review titles and abstracts, before the full texts of potentially eligible papers were obtained. Final selection for inclusion in the review was by agreement between both members of the selecting pair. Disagreements were arbitrated by another member of the team.

The following databases were searched: Cochrane Database of Systematic Reviews, Database of Abstracts of Reviews and Effects (DARE), MEDLINE and EMBASE (for search strategy examples, see Appendix 2).

Searches were limited by year (2005 to February 2017) and restricted by the level of evidence to systematic review and meta-analysis, or other evidence syntheses. English-language papers only were selected.

Methodological quality/bias risk was recorded using the JBI critical appraisal checklist for Systematic Reviews and Research Syntheses (see Appendix 3).

The JBI data extraction tool was used to extract data from the included reviews after discussion and piloting its use. Several additions and modifications were demanded by the nature of the review question. A database of evidence tables was created for definitions and key elements of CGA, setting and staff, key participants, outcome measures and costs. The database was then used to create the summary tables that were used to develop a narrative overview of the evidence.

Results

A total of 1010 titles and 419 abstracts were screened for eligibility, 143 full articles were reviewed for relevance and 24 were included in a final quality and relevance check (Figure 1). Thirteen reviews, reported in 15 papers,3,15,16,30–41 were selected for review. The most recently conducted trial included in the reviews was reported in 2014; all other trials were reported between 1983 and 2012.

FIGURE 1.

Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) flow chart detailing the identification and selection of research syntheses for inclusion in the umbrella review and also used for the new and emerging models of care review. Reproduced by permission of Oxford University Press from Parker SG, McCue P, Phelps K, McCleod A, Arora S, Nockels K, Kennedy S, et al. What is Comprehensive Geriatric Assessment (CGA)? An umbrella review. Age Ageing 2018;47:149–55. https://doi.org/10.1093/ageing/afx166. 8

Reproduced by permission of Oxford University Press from Parker SG, McCue P, Phelps K, McCleod A, Arora S, Nockels K, Kennedy S, et al. What is Comprehensive Geriatric Assessment (CGA)? An umbrella review. Age Ageing 2018;47:149–55. doi: https://doi.org/10.1093/ageing/afx166. 8

Ninety-five original articles were cited 166 times. Twenty-six original articles were cited more than once. The most commonly cited articles were Landefeld et al. ,42 Asplund et al. 43 (seven citations each) and Counsell et al. 44 (six citations). Removing all except one of the reviews,3,31 which cited these three most highly cited papers, did not significantly affect our conclusions with regard to the population characteristics, intervention definition, settings and comparisons and clinical outcomes. Some health economics detail was lost in this sensitivity analysis.

The main beneficiaries were older (≥ 55 years) hospital inpatients in acute care settings. In most studies, frailty was not explicitly identified as a characteristic of CGA recipients; however, one review (which included most of the most highly cited trials) attempted to stratify trials by frailty.

The most widely used definition of CGA was that of a multidimensional, multidisciplinary process that identifies medical, social and functional needs, and the development of an integrated/co-ordinated care plan to meet those needs. 3

Dimensions of CGA reported consistently included medical/physical, psychological/psychiatric, socio-economic, function, and nutritional assessment.

Most of the reviews used the same body of literature (from 1983 onwards) to examine some aspect of in-hospital CGA. Reviews citing literature mainly outside this highly cited core included a review of interface care,30 gerontologically informed nursing assessment and referral32 and MDT interventions39 (Table 1).

Key clinical outcomes (Table 2) were mortality (12/13 reviews), activities of daily living (13/13 reviews), cognitive functioning (9/13 reviews) and dependency (6/13 reviews). Principal operational outcomes were length of stay (11/13 reviews) and re-admissions (12/13 reviews). Destinational outcomes included living at home (7/13 reviews) and institutionalisation (11/13 reviews). Four reviews mentioned resource use and costs. Patient-related outcomes (health-related quality of life, well-being and participation) were not usually reported.

Few studies considered costs; none took a broader view and included direct costs (staff and resources), subsequent costs (such as community health and social care costs), costs to the patients and wider societal costs. Furthermore, although the reviews and trials describe multiple intervention configurations, these were mostly not standardised. An exception to this33 removed one outlier study, and the result of meta-analysis showed that the costs of specialised acute unit care were less than those of usual care [weighted mean difference US$245.80, 95% confidence interval (CI) US$446.23 to US$45.38; p = 0.02]. 33 Two studies3,15 concluded that many hospital-based services showed reduced costs associated with CGA. In a review of trials of Acute Care for Elders (ACE) model components, little cost evidence was available to differentiate and compare relative effectiveness between components of the ACE model.

New and emerging models of care review

Methods

Results

Types of services provided

The trusts were asked if they provide a multidisciplinary assessment in acute care for older people who are frail in a number of clinical areas. Responses described 323 services in 10 predefined clinical areas (Table 5). The predefined framework captured the majority of service types. Services recorded in the ‘Other’ category (n = 7) included ‘Enhanced Care’, a dementia and delirium team and a rehabilitation unit.

TABLE 5 - Respondents providing multidisciplinary assessments in various clinical areas

Trusts providing a multidisciplinary assessment in acute care for older people who are frail in the following clinical areas	n (%)
Orthopaedic department	53 (88)
Wards specialising in OPM	55 (92)
Admission ward	44 (73)
Stroke team	47 (78)
Inpatient medical wards	32 (53)
ED	26 (43)
Emergency assessment unit/decision unit	21 (35)
Inpatient surgical wards	20 (33)
Hospital consultation service	22 (37)
Oncology department	7 (12)
Other	3 (5)

An assessment of non-response bias for service-level responses using the same criteria used to analyse trust-level responses did not identify systematic differences between responding and non-responding trusts.

Admission avoidance

A key issue in the provision of effective acute inpatient services is a close relationship between acute and community-based services. Accordingly, we asked questions about the provision of a community admission avoidance service. Most (79%) of the 58 responding trusts indicated that they offered such provision. Approximately half (48%) of these services were provided by a consultant specialist in OPM, who was part of the team in 25 services, available to the team in 3 services and had no involvement in 32 services.

Post-acute care

Forty-five trusts (78%) reported that they work with or provide a post-acute service. Most (n = 34) of these services were based entirely in the patients’ home and nine also had access to community beds in which to provide post-acute services. Thirty-four trusts (59%) reported that their post-acute service was provided by a consultant specialist in OPM. The types of practitioner with overall responsibility for the service are shown in Figure 2.

FIGURE 2.

Types of practitioners with overall responsibility for the service.

The service-level questionnaire

Response rate

Thirty-six (62%) of the 58 trusts that provided a completed trust questionnaire also provided completed service questionnaire(s). We received one response from Wales (14.3% of Welsh trusts), four responses from Scotland (28.5% of trusts), three responses from Northern Ireland (60% of trusts) and 28 (18.8% of trusts) responses from England. Between them, these 36 trusts returned service questionnaires describing 121 separate services – an average of 3.4 services per trust (Table 6).

TABLE 6 - Responses to the service questionnaires

Country	Number (%) of trusts
	Contacted	Returned service questionnaire(s)	Number of service questionnaires received
Scotland	14	4 (28.5)	16
Northern Ireland	5	3 (60.0)	12
Wales	7	1 (14.3)	2
England	149	28 (18.8.)	91
Total	175	36 (20.5)	121

Working with, or providing, community-based services

About half of the services provided community admission avoidance, mostly with input from a specialist in OPM. Most respondents worked with, or provided, post-acute care services, again mostly provided with or by a consultant specialist in OPM (Table 7).

TABLE 7 - Community admission avoidance services (n = 121)

Do you work with or provide	Response, n (%)
	Yes	No
Community services
A community admission avoidance service?	48 (40)	56 (51)
With input from consultant specialist in OPM?	45 (37)	11 (9)
Post-acute services
Post-acute care services	85 (70)	36 (30)
Provided by a consultant specialist in OPM?	45 (55)	18 (23)
Bed based or home based?
Bed only	10 (8)
Home only	7 (6)
Both bed and home based	68 (56)

The survey identified a range of provision of multidisciplinary assessment and management across inpatient care settings with some areas (such as orthopaedics, OPM and stroke) being more completely provided with appropriately skill-mixed MDTs than others (such as surgical and oncology departments).

Identifying patients for multidisciplinary assessment and care

In general, clinicians appear to prefer their clinical assessment processes to standardised clinical instruments and measuring tools to identify patients for multidisciplinary care. Most services [108/121 (89%)] relied on clinical assessment processes to identify patients for multidisciplinary assessment. Of these, about half (n = 61) identified processes amounting to clinical screening by an experienced health-care professional and some (n = 20) services described MDT discussion as the main clinical assessment processes used to select patients. The use of a clinical screening tool or triage method was reported by 50 services (41%). In free-text entries these services generally identified their admission processes, locally developed screening procedures and triage processes (some specifying frailty specific triage) to identify patients who will receive multidisciplinary treatment. Admission criteria were applied by 55 services (45%). Fifty-three services (44%) had a minimum age requirement and 28 services (23%) had explicit exclusion criteria.

However, those services that reported the use of screening tools to identify patients for multidisciplinary assessment generally identified their admission and referral processes and locally developed screening procedures as the tools in the free-text descriptions [rather than specific, recognised or validated clinical instruments, represented graphically by the word cloud below (Figure 3)].

FIGURE 3.

Word cloud created from free-text entries describing the clinical assessment processes used to identify patients for multidisciplinary assessment.

Identifying frailty

A total of 32 (26%) services used a standardised clinical method, instrument or measuring scale to identify patients who are frail. However, there was little consistency between services, among which a wide range of recognised assessment scales was used for this purpose (Figure 4).

FIGURE 4.

Tools used to identify frailty (n = 32). aCGA, abbreviated Comprehensive Geriatric Assessment; ISAR, Identification of Seniors at Risk; TUSS, timed unsupported stand.

Staffing and assessment at the ‘front door’

Clearly, we must be cautious about generalisation, particularly when it comes to considering detailed survey responses from relatively small numbers of specialised types of service. However, with that caveat in mind, we can identify some important dimensions of the way in which services are being staffed and delivered.

For example, it seems reasonable to say that the services generally reported being staffed by a consultant specialist, who attended regularly and was supported by junior and staff grade doctors (Figure 5). However, when asked if older people who are frail are assessed by a specialist in OPM in the ED (at the front door), only nine of the responding trusts described 16 services in which such ‘front door’ assessment is available. When available, these assessments were typically performed during the first 4–12 hours of admission.

FIGURE 5.

Staffing available in multidisciplinary treatment services.

Components of assessment

The components of the multidimensional assessments performed were similar across a range of services. More than 90% of services reported that the assessment of cognitive functioning, delirium and dementia were routinely performed, along with assessment of activities of daily living, mobility, falls and falls risk, medications, nutrition, continence and skin integrity (Figure 6). Assessment of hydration as a routine was reported in 102 services (84%) and end-of-life-care needs in 96 services (79%). Psychiatric assessment was reported in 83 (69%) of returns and 80 services (66%) reported assessing for depression. Sensory loss was the least frequently reported assessment, being reported by 73 services (60%). These data are tabulated in Table 8.

FIGURE 6.

Assessments that are routinely performed using a tool or scale (n = 121). ADL, activities of daily living; EOL, end of life.

TABLE 8 - Assessments that are routinely performed using a tool or scale (n = 121)

ADL, activities of daily living; EOL, end of life.

Assessment	Performance, n (%)
	Routinely performed	Performed using a tool/scale
ADL	110 (91)	57 (47)
Cognitive functioning	116 (96)	111 (92)
Continence	107 (88)	31 (26)
Delirium	115 (95)	89 (74)
Dementia	111 (92)	97 (80)
Depression	80 (66)	55 (45)
EOL care	96 (79)	40 (33)
Falls	113 (93)	53 (44)
Mobility	112 (93)	48 (40)
Falls risk	114 (94)	64 (53)
Medicines	111 (92)	20 (17)
Nutrition	112 (93)	103 (85)
Hydration	102 (84)	19 (16)
Pain	112 (93)	69 (57)
Psychiatric	83 (69)	31 (26)
Sensory loss	73 (60)	12 (10)
Skin integrity	114 (94)	106 (88)

Monitoring change

The majority of services [n = 100 (83%)] had processes in place to identify the development of delirium and 89 (74%) had processes in place to identify falls. Other risk factors for adverse outcomes, such as incontinence [n = 81 (67%)], depression [n = 62 (51%)] and functional decline [n = 53 (44%)], were less consistently reported.

Prevention of in-hospital deterioration

The prevention of deterioration, and of complications of hospitalisation, is particularly important for older people who are frail. Falls, urinary complications (such as incontinence), deconditioning, demotivation and depression all increase the risk of adverse outcomes from an inpatient hospital stay (Table 9).

TABLE 9 - Processes to actively prevent deterioration (n = 82)

Processes in place to actively prevent deterioration	Response, n (%)
	Yes	No
Physical functioning	80 (66)	41 (34)
Continence	46 (38)	75 (62)
Delirium	62 (51)	59 (49)
Depression	30 (25)	91 (75)
Falls	93 (77)	28 (23)
Other	7 (6)	114 (94)

Trusts are expected to have a falls prevention policy, so it is not surprising that 88% of services reported having services in place to actively prevent falls. However, only 31% of services reported processes to prevent deterioration in depression and only 40% reported processes to prevent deterioration in continence, illustrating the continued ‘Cinderella’ status of continence and mental health services in this context.

Results broken down by service type

A total of 121 services returned a survey questionnaire and responses are now summarised by service type. Some of the tables containing these data are relatively large (e.g. 10 rows by 6 or 9 columns) and contain large numbers of empty cells, so, for clarity, data are not presented. Narrative summaries (with quantitative data where appropriate) are presented here.

Identifying patients for multidisciplinary treatment

Most services (68–100%) reported that clinical assessment processes were used to identify patients for multidisciplinary treatment. The use of screening tools or standard triage methods was less frequent, generally being reported by around half or fewer services. About half [12/22 (52%)] of OPM wards reported the use of screening tools or standard triage methods.

Admission and exclusion criteria

No service reported the use of a maximum age criterion. The use of a minimum age criterion to identify patients for multidisciplinary treatment was seen most frequently among admission wards [11/17 (65%)].

Consultant reviews and assessments

The majority of services reported that patients were usually reviewed by a consultant. The rates for consultant review were > 90% in 7 out of 10 services (data not presented).

In contrast, a review of patients identified as frail ‘at the front door’ of the hospital was reported by 16 services, across nine different trusts. This included about one-quarter of admission wards and EDs. Three services reported that these assessments were usually carried out within 4 hours, eight reported that they take place within 12 hours and five reported that they take place within 24 hours.

Most services reported that they did not use a standardised clinical method, instrument or scale to identify frail patients. The use of these methods was distributed fairly evenly between the service types.

Staff who regularly work in the multidisciplinary teams

Most admissions wards (14/17), orthopaedics wards (20/22), OPM wards (22/23) and stroke teams (16/18) regularly worked with a specialist in OPM. All admissions wards and OPM wards regularly worked with junior doctors. Orthopaedics (12/22) and OPM wards (11/22) reported working regularly with staff grade doctors. General practitioners (GPs) were not reported as regular members of the team and neither were oncologists (except in the oncology teams).

Similar patterns were seen for nurses, nurse specialists or both, and pharmacy staff. Very few of the services had psychologists or old age psychiatrists as regular members of the team.

Allied health-care staff who were almost universally included in the regular MDT were physiotherapists, occupational therapists, therapy and health-care assistants.

Approximately one-third of teams reported social work staff as regular members, and very few charity, faith support or voluntary sector workers performed a regular function in the MDTs.

Staff who are available to the multidisciplinary teams

Most services reported having access to consultant specialists, including in oncology (62%), surgery (58%) and other medical specialties (67%). GPs were generally not reported as being available to the team, except in oncology where both services reported GP availability (see Report Supplementary Material 3).

Generally, teams reported availability of psychiatric nurses, old age psychiatrists and psychologists, rather than regular team membership (see Report Supplementary Material 3).

Similarly, availability (rather than regular team membership) was generally reported for podiatry and radiology staff (see Report Supplementary Material 3). The same can be said for speech and language therapists (SALTs) and dietitians, but not for stroke teams, all of which are usually part of the core team.

Trusts reported that teams also had the following staff made available to them: Faith Support Workers (Admissions Ward, Orthopaedics, OPM Ward and Stroke Team), Social Work Assistants (Admissions Ward, OPM Ward and Stroke Team) and Social Workers (Orthopaedics, OPM Ward). Orthopaedics most commonly had more Voluntary Sector staff available to teams.

Including social workers and assistants as regular or available team members, social work availability was generally high (71–100%) in inpatient services, but less so at the ‘front door services’ such as ED and Decision Units (see Report Supplementary Material 3).

Similarly, Faith Support services were almost universally available in inpatient care settings, but less so in the receiving services (see Report Supplementary Material 3).

Discussion

In this survey we sought to understand how CGA is delivered across a range of acute hospital wards and settings in the UK.

A series of questions were asked about the provision of multidisciplinary assessment and care for older people admitted to hospital, across a range of inpatient settings. This allowed identification of the components of CGA, the types of patients and clinical problems, and, importantly, the extent to which frailty is assessed and older people with frailty targeted for receipt of multidisciplinary assessment and care. Information was also sought about the provision of/relationship with the community-based assessment and post-acute services, which are essential components of integrated systems of care.

The trust-level responses are used to describe a range of available services. The service descriptions are then used to illustrate the make-up, assessment processes and resources available among the responding services. This exercise has provided an overview of service types, which is complemented by a rich source of service descriptions (from > 120 services) and which contains important details about the services that commonly provide CGA in hospital settings.

One of the key messages emerging from this survey is that although some services appear to provide multidisciplinary assessment and management routinely (wards specialising in OPM and orthopaedic wards) or commonly (stroke and admissions wards), the practice is less firmly embedded in other parts of the hospital system (e.g. medical and emergency units) and not usually found in surgical and oncology services.

With regard to selecting patients to receive multidisciplinary assessment and management, it would appear that the professionals performing this task tend to prefer their usual clinical assessment processes, using clinical and professional judgement and discussion with members of the MDT as a way to select patients who may benefit. Relatively few teams relied on standardised scales or clinical assessment tools.

Furthermore, and despite much current interest and activity related to the topic of frailty, at the time of our survey (2016), we did not find a clear consensus about the assessment/measurement of frailty and the tools required to carry it out.

More than half (56%) of the trusts that we approached trusts agreed to participate and, following multiple rounds of contact and encouragement of trust staff, with network, college and specialist society support, 60 trusts provided trust-level information and 36 trusts provided detailed descriptions of > 120 services.

Although this is a large repository of service descriptions, it cannot be said to be comprehensive, or representative, on account of the limited proportion of UK trusts that participated in the survey. This concern is mitigated somewhat by the existence of the NHS Benchmarking survey, which was carried out at the same time, which found similar results where it asked similar questions. Over a similar time period to this survey, NHS Benchmarking was conducting its regular survey of older people’s care in acute settings. 103 NHS Benchmarking worked with 45 NHS trusts and UHBs, 15 of which also responded to our survey. Between NHS Benchmarking and our survey, we therefore covered 88 NHS trusts across the UK during a similar period of data collection. This is over half of trusts so it is worth spending some time exploring the similarities and differences between the two surveys. In summary, it would appear that when we asked about the same thing, we obtained similar results.

For example, NHS Benchmarking found that 77% of trusts/UHBs delivered CGA on care of older people wards. This decreased to 42% on other specialty wards. Our survey adds a layer of detail around the same statistic, finding high levels of CGA delivery on OPM wards (92%), orthopaedic wards (88%) and stroke wards (77%), whereas the lower levels of CGA delivery were found on 53% of inpatient medical wards and 33% of inpatient surgical wards.

Similarly, NHS Benchmarking found that 40% trusts said that they had a dedicated geriatric team in the accident and emergency (A&E) department. Forty-three per cent of our respondents stated that frail older people in the A&E were assessed by a geriatrician.

When it came to the delivery of frailty-specific care there were some clear similarities in our findings. NHS Benchmarking found that about 30% of responding trusts used a standardised clinical method, instrument or measuring scale. This is close to the 29% figure that we found when we asked, ‘Does your trust use a standardised clinical method, instrument or measuring scale to identify patients who are frail?’ and we are able to add the important qualifying detail that there is not (as yet) a clear preference or consensus about which of the many available tools and scales to use for this purpose.

In conclusion, we believe that this is the biggest, and most detailed, survey of hospital CGA services ever completed in the UK. This survey has identified a range of provisions of multidisciplinary assessment and care across inpatient care settings, with some areas (e.g. OPM, orthopaedics, stroke) being more completely provided with appropriately skill-mixed MDTs than others (such as inpatient medicine and surgery).

Generally, clinicians appear to prefer their clinical assessment processes to standardised clinical instruments and measuring tools to identify patients to be beneficiaries of multidisciplinary assessment and care.

Furthermore, it would appear that, as yet, the formal use of frailty as an identifying or stratifying characteristic of patients is patchy and non-standardised.

In an emerging landscape of hospital-wide CGA service, we cannot say that patients with frailty are being identified, assessed and managed systematically in relation to their frailty, or targeted with CGA services. There are areas of inpatient hospital care in which these features are well developed, or beginning to develop, and others in which they are not often found.

Developing Hospital Episode Statistics-based proxies for frailty and creating tools to help understand the need for Comprehensive Geriatric Assessment: introduction

In this strand of work, we wanted to develop proxy measures and tools that would provide a better understanding of the needs of frail older people within a given a population. In the UK there has been a long tradition of population-based measures and the concept of needs assessment104,105 as the basis for informing decisions about where to invest/disinvest in care services. Unfortunately, comprehensive data sets that identify frailty in an individual do not exist,106 so the approach has had to be a little more tangential than previous methods employed.

There are reasonably good descriptors of the underlying population and information about health service activities contained in the fairly comprehensive Hospital Episodes Statistics (HES) data sets that record information on patients’ attendances at hospital. 107 Although these patient-level data sets have limited clinical detail, they have been fairly consistently recorded in all acute NHS hospitals in England for many years. For this analysis, we wanted to exploit this history of good data collection to look in detail at the population of older people, and, where possible, to look for frailty. The following strands of work are pragmatic and exploratory. We have sought to combine the outputs into tools that can be used for planning CGA.

Our aims were to:

• stratify local populations to identify the number of people who may benefit from CGA

• apply a series of health system performance measures at area and provider level that relate to the care of frail older people

• develop simple interactive tools to compare patients’ assessed potential benefits and costs.

We used linked population-level data sets to see if we could assess the number of frail older people who may benefit from CGA. This approach built on the idea of there being a resident population of older people in a local authority (LA) area, who could then then be split into mutually exclusive groups or segments, so that we could identify a set of individuals with high needs and form a potential target population for CGA. For the sake of consistency, and acknowledging multiple caveats about the definition and measurement of the frailty syndrome, we called this group ‘frail older people who may benefit from CGA’. For this group, we identified a set of performance metrics, which we might reasonably expect to be influenced by the nature and quality of the inpatient care experienced (particularly the use of CGA), for example annualised numbers of emergency admissions.

We tried three approaches to describing the extent of frailty within a population:

1. population segmentation based on historic patterns of hospital use

2. clinical diagnoses linked with frailty – using a predefined list of diagnoses

3. clustering of clinical diagnoses to create a new approximation to frailty.

We also looked at the direct costs of CGA, which are described in Chapter 7. Information from these various analyses was combined into Microsoft Excel^® (Microsoft Corporation, Redmond, WA, USA) tools.

Population segmentation approach

Population segmentation has been proposed for some time108 and in recent years has been a more practical application owing to the availability of large person-level data sets. 109,110 There are, however, no standard segmentation approaches that have been applied to look at care use by older people.

In this approach, we looked at patterns of past hospital activity to see if we could categorise groups of older people according to their previous use of hospital care. By implication we assume that this will help to indicate something about their future needs. At one end of the spectrum we have older people who are fit and healthy and have never used hospital services, and at the other end there are those with multiple chronic conditions who require multiple acute interventions. It is among this latter group that we believe ‘frail’ older people are more likely to be identified.

Testing the population segmentation approach in three local authorities

Our initial work used hospital admissions data linked over time for all residents in three LAs – Leicester, Nottingham and Southampton – as these would be used for more detailed clinical linkage work described in Chapter 6. Data were extracted from the HES by NHS Digital and held at the Nuffield Trust under permissions granted by NHS Digital based on the Nuffield Trust fulfilling the information governance requirements needed to analyse pseudonymised person-level hospital data, and linked Office for National Statistics (ONS) mortality data.

Hospital Episode Statistics provides information on individual episodes of hospital care for patients. Each record contains a pseudonymised Hospital Episode Statistics Identifier (HESID) that allows episodes of care for the same person to be tracked over time. We are also able to identify the LA area in which the person normally resides. This information contained within HES enables us to identify all the people aged ≥ 75 years in 2008 living in the LAs of the cities of Leicester, Nottingham or Southampton and then track their hospital use anywhere in the country between 2008 and 2012. Using the census population estimates for each LA, we could then estimate the proportion of the older people in these areas that had used the hospital services and also by implication estimate how many older people had not (Figure 7).

FIGURE 7.

Outline of how individual histories were constructed from linked patient records. H, history.

We recognise that there are limitations to understanding frailty just from past hospital records. There may be people within a population who are perfectly well and not engaging with health services but who experience an event that alters their health status and health service use. These newly emergent cases are of interest as they may be less well identified through studies of past hospital records.

Table 10 summarises the number of people in our study cohort who had any hospital contact (outpatient attendance, inpatient admission or ED visit) during the 4-year study period. It shows that after 1 year, the proportion of the estimated population in each area that were aged ≥ 75 years with at least one hospital encounter was between 62% and 76%, whereas after 4 years between 91% and 100% of people aged ≥ 75 years had had a hospital encounter.

TABLE 10 - What proportion of over 75-year-olds had any hospital contact over 4 years?

Area	Population aged ≥ 75 years (n)	Numbers (%) with hospital event
		After 1 year	After 2 years	After 3 years	After 4 years
Leicester Unitary Authority	17,559	12,122 (69.0)	14,867 (84.7)	16,327 (93.0)	17,171 (97.8)
Nottingham Unitary Authority	17,850	13,628 (76.3)	16,312 (91.4)	17,587 (98.5)	18,266 (102.3)
Southampton Unitary Authority	16,160	9953 (61.6)	12,494 (77.3)	13,775 (85.2)	14,621 (90.5)

In Nottingham the number of individuals who had had at least one hospital encounter (as measured by the number of unique HESIDs) exceeded the estimates of the resident population. This could arise for a number of reasons:

• flaws in the HESID (e.g. people had been assigned more than one HESID, basic patient details had been recorded incorrectly at the hospital or people had moved to a new house and the HESID had not been updated)

• underestimates of the census population,111 perhaps as a result of people living in care homes.

It is also worth noting that there appear to be quite large differences between the three LAs, with Southampton having consistently a smaller proportion of older people identified as hospital users.

Our first attempt was, therefore, to group the population into categories that roughly correspond with an increasing scale of hospital utilisation over a period of time. We were also interested in trying to separate out isolated acute hospital encounters and other manageable chronic diseases from indicators of longer-term deterioration and instability similar to the Bridges to Health Mode. 111

For this analysis, the key elements of health-care use that we could identify at patient level were derived from HES and therefore limited to emergency hospital admissions, elective hospital admissions, ED attendance and outpatient appointments.

The groups are mutually exclusive so a person can fall into only one group at any one time. The groups are also ‘hierarchical’, with an order of precedence loosely based on intensity of hospital use. For example, if someone has an outpatient appointment and an elective admission, they are grouped into the latter category. Death has not been accounted for when assigning individuals to these categories.

Table 11 shows the number (across all the three LAs) and proportion of cases (compared with the population of all those aged ≥ 75 years in those areas) in each category according to whether data over 1 year or 4 years are considered. For example, after 1 year, 28.5% of people aged ≥ 75 years had had only outpatient contacts, but after 4 years this had fallen to 12.9% as other forms of hospital contact became more likely. After 4 years the number of people aged ≥ 75 years who have multiple ED attendances but without an admission is very small (1.3%), whereas the largest group is for those with multiple emergency admissions. After 4 years, almost 37% of the older population have had more than one emergency admission. This reinforces the importance of emergency hospital use among older people as a factor in shaping needs for acute hospital care.

TABLE 11 - Number and proportion of the population who are aged ≥ 75 years in hierarchical population groups

Population group	Time point, n (%)
	After 1 year	After 4 years
No hospital contact (implied)	15,866 (30.8)	1511 (2.9)
Outpatient only	14,701 (28.5)	6661 (12.9)
Single ED attendance (no admissions)	1449 (2.8)	2495 (4.8)
Multiple ED attendances (no admissions)	143 (0.3)	660 (1.3)
Single elective admission	4279 (8.3)	4486 (8.7)
Single emergency admission	6499 (12.6)	7928 (15.4)
Single elective and single emergency admission	1130 (2.2)	2198 (4.3)
Multiple elective admissions	1921 (3.7)	4314 (8.4)
Multiple elective and single emergency admission	578 (1.1)	2384 (4.6)
Two emergency admissions	2968 (5.8)	7413 (14.4)
Three emergency admissions	1117 (2.2)	4322 (8.4)
Four or more emergency admissions	918 (1.8)	7197 (14.0)

Table 12 shows over 4 years the 12 hospital use categories grouped to create eight segments that are broadly consistent across LAs. For example, the proportion of the population aged ≥ 75 years who had only outpatient contact with hospitals over a 4-year period was between 12.1% and 13.6% across the different areas. In addition, the proportion of older people who had multiple emergency admissions ranged from 32.3% to 39.1%. Southampton had a larger implied proportion of older people with no hospital contact and a smaller proportion of older people who had multiple emergency admissions. Interestingly, in Southampton the proportion of older people with elective admissions was in line with the other areas. The implication is that the demand on urgent hospital care in Southampton by this age group is slightly less, but we cannot definitively say whether this was due to differences in the needs of the older people in this area or differences in the supply and organisation of care.

TABLE 12 - Proportion of the population aged ≥ 75 years in population segments after 4 years for individual LA areas

Population group	Area (%)
	Leicester	Nottingham	Southampton
No hospital contact (implied)	2.2	–2.3	9.5
Outpatient only	13.0	13.6	12.1
ED attendances (no admissions)	6.3	7.2	4.7
Single elective admission	7.9	8.7	9.7
Single emergency admission	16.9	15.1	14.0
Elective admissions and single emergency admissions	8.3	9.3	9.0
Multiple elective admissions	7.0	9.4	8.7
Multiple emergency admissions	38.3	39.1	32.3

Applying the population segmentation approach across England

A hospital use classification, similar to those above, was applied to all lower-level LA areas in England (n = 326). In order to reduce the computation burden, 2 consecutive years of data were used instead of 4. Table 13 summarises the proportions of the population aged ≥ 75 years in each segment, providing information to show the distribution across all the LAs. It shows that, on average, the highest level of utilisation (people with three or more emergency admissions) made up 3.3% of the population aged ≥ 75 years, whereas 27% of the population aged ≥ 75 years had no hospital contacts.

TABLE 13 - Proportion of cases in those aged ≥ 75 years in each segment across 327 LAs (2-year analysis)

Measure	Number of emergency admissions (%)			Other forms of hospital contacts (%)
	≥ 3	2	1	Elective admissions	ED attendances, no admissions	Outpatient	None (implied)
Average	3.3	4.8	14.4	13.7	6.4	29.9	27.5
Standard deviation	0.9	0.7	1.3	1.8	1.9	5.8	7.4
Minimum	0.0	0.0	8.0	8.1	0.0	20.2	–0
Median	3.3	4.8	14.3	13.6	6.3	28.9	28.6
Maximum	5.6	6.5	18.2	21.0	13.7	61.7	48.2

The estimated numbers of people with no hospital events came from a number of areas in which the number of people identified in the hospital activity seemed to exceed the resident population. This is because, similar to what was previously observed in Nottingham, for some LA areas the number of individuals having had at least one hospital encounter (as measured by the number of unique HESIDs) exceeded the estimates of the resident population. Most of these areas were in the north-west of England; we assume from our analysis that this geographic concentration is a local artefact of how the outpatient data are collected and HESIDs assigned.

Association with deprivation

Socioeconomic factors are a key driver of differences in observed morbidity and health service utilisation. 112 For this analysis, using the national data described above, we wanted to see if underlying socioeconomic differences between LAs were affecting the profiles across our population segments. Table 14 summarises the correlation between LA levels of deprivation using the individual elements of the Index of Multiple Deprivation (IMD)113 and our population segments based on prior activity.

TABLE 14 - Correlation between population segments at LA level and IMD subscores (values are Pearson correlation coefficients)

Characteristic	Number of emergency admissions			Other forms of hospital contacts				n
	≥ 3	2	1	Elective admissions	ED attendances, no admissions	Outpatient	None (implied)
Income	0.56	0.59	0.61	–0.02	0.42	0.04	–0.37	0.28
Employment	0.43	0.51	0.58	0.05	0.34	0.05	–0.34	0.29
Education, skills and training	0.28	0.37	0.44	0	0.19	–0.09	–0.12	0.16
Health deprivation and disability	0.48	0.57	0.64	0.03	0.34	0.08	–0.39	0.24
Crime	0.59	0.58	0.57	–0.16	0.48	0.03	–0.34	0.16
Barriers to housing and services	0.04	–0.13	–0.26	–0.04	–0.03	–0.13	0.17	–0.03
Living environment	0.11	0.03	0.01	–0.15	0.12	0.05	–0.05	0.09
Income Deprivation Affecting Children Index (IDACI)	0.58	0.59	0.6	–0.06	0.44	0.01	–0.34	0.25
Income Deprivation Affecting Older People (IDAOPI)	0.65	0.61	0.58	–0.1	0.44	0.06	–0.38	0.21

The results indicate a positive correlation between levels of deprivation and the proportion of people in the higher utilisation population segments. The correlation is particularly marked for the subset of deprivation scores in relation to ‘Income Deprivation Affecting Older People’. The exception is for the elective admissions segment for which there is no relationship. If it is assumed that deprivation is associated with greater health needs, this exception may reflect the inverse care law proposed by Tudor Hart. 114

Across the sub-elements of the IMD, the scores relating to ‘Barriers to Housing and Services’ tended to show no or a negative relationship with frailer older people. This element of the IMD tries to measure the physical and financial accessibility of housing and local services. It includes two subdomains, namely ‘geographical barriers’, which relate to the physical proximity of specific local services, and ‘wider barriers’, which include issues relating to access to housing such as affordability, homelessness and overcrowding. It seems that LAs that score highly on these domains do not have a distinctive pattern across our older population’s segments.

Similarly, the ‘Living Environment’ domain, measuring the quality of an individual’s immediate surroundings (both the ‘indoors’ living environment, which measures the quality of housing, and the ‘outdoors’ living environment, which contains two measures about air quality and road traffic accidents) tended to show no significant correlations with the older population’s segments.

There is a significant relationship between area-level deprivation as measured by combined IMD scores and a larger proportion of numerous emergency admissions for an individual in this older age band. Figure 8 plots the proportion of people aged ≥ 75 years in a LA who had three or more emergency admissions against the IMD score of that area (R² of 0.26). This association is not surprising and it is driven by two factors115 that cannot be completely separated:

1. the impact of deprivation on health status in general – and so the association with needs for care and frailty (demand side factors)

2. accessibility of care services and patterns of care (supply).

This existence of this relationship means that, in exploring the different ways that health services responded to needs in the population, it would be appropriate to look at the proportion of cases in high utilisation segments when standardised for deprivation.

FIGURE 8.

Proportion of population who are aged ≥ 75 years and have had more than three emergency admissions in 2 consecutive years by IMD for English LA areas.

Figure 8 and Table 15 show the areas in which the observed proportion of people in the segment with the highest level of hospital utilisation in the ≥ 75-year-olds group are less, or more, than expected given the level of deprivation. This relatively crude analysis is a useful way of exploring what differences are more likely to arise from patterns of service provision. It is worth noting that at both ends of this scale there is a distinct geographic clustering, with neighbouring LAs appearing closer together (e.g. Great Yarmouth and Waveney; or Watford, Three Rivers and North Hertfordshire) (Table 16). The most likely explanation for this is that it indicates local provider effects – either of capacity to admit and the availability of alternatives.

TABLE 15 - Areas where the proportion of people who are aged ≥ 75 years with three or more emergency admissions in 2 years was much less than ‘expected’ when standardising for area-level deprivation

LA	Number of people aged ≥ 75 years	IMD score	Proportion of older people who have had ≥ 3 emergency admissions (%)		Observed/expected
			Observed	Expected
Isles of Scilly	249	12.0	0.0	2.98	0.00
Wyre Forest	9848	22.1	1.6	3.58	0.46
Malvern Hills	9497	16.1	1.5	3.22	0.46
Forest of Dean	8298	17.0	1.6	3.28	0.47
Isle of Wight	16,419	23.1	1.9	3.65	0.51
North East Lincolnshire	14,357	30.9	2.1	4.11	0.51
Waveney	14,177	25.1	2.0	3.77	0.54
Cornwall	56,591	23.8	2.1	3.69	0.58
Allerdale	9779	22.6	2.1	3.61	0.58
West Lindsey	8939	19.2	2.0	3.41	0.58
Medway	17,802	22.3	2.1	3.60	0.59
East Hampshire	11,551	8.6	1.6	2.78	0.59
North Norfolk	15,199	21.3	2.1	3.54	0.59
Great Yarmouth	9973	32.4	2.5	4.20	0.60
Havant	13,405	21.2	2.1	3.53	0.60
South Hams	9777	14.1	1.9	3.10	0.60
Wychavon	12,638	16.0	2.0	3.22	0.61
West Somerset	5188	23.3	2.3	3.66	0.62
East Lindsey	16,556	28.9	2.5	3.99	0.62
Stroud	10,917	10.9	1.8	2.91	0.63
Rutland	4068	9.6	1.8	2.84	0.63
Plymouth	20,838	26.6	2.5	3.86	0.64

TABLE 16 - Areas where the proportion of people who are aged ≥ 75 years with three or more emergency admissions over 2 years was much more than ‘expected’ when standardising for area-level deprivation

LA	Number of people aged ≥ 75 years	IMD	Proportion of older people who have had ≥ 3 emergency admissions (%)		Observed/expected
			Observed	Expected
North Hertfordshire	11,347	11.6	4.0	2.96	1.35
Stockport	26,132	19.1	4.6	3.41	1.35
Solihull	20,318	17.2	4.4	3.29	1.35
Merton	11,624	14.9	4.3	3.16	1.37
Hounslow	13,284	22.5	5.0	3.61	1.37
Wandsworth	13,079	18.3	4.6	3.36	1.37
North East Derbyshire	10,059	16.8	4.5	3.27	1.38
Corby	3967	25.8	5.3	3.81	1.38
Watford	5864	15.7	4.5	3.20	1.39
Hammersmith and Fulham	7841	24.4	5.2	3.72	1.41
Chesterfield	9447	25.3	5.5	3.78	1.45
North Tyneside	17,992	21.3	5.1	3.54	1.45
Tamworth	5194	20.3	5.0	3.48	1.45
Three Rivers	7854	10.0	4.2	2.86	1.45
Hillingdon	18,407	18.1	4.9	3.35	1.48
Slough	6329	22.9	5.4	3.63	1.48
Sutton	14,125	14.6	4.8	3.13	1.52
Rushmoor	5646	15.1	4.9	3.17	1.53

Identifying possible frailty based on predefined list of diagnoses

This section presents a refinement of the population segmentation approach presented above and relies on using a set of clinical diagnostic codes predefined as being linked to frailty. The specific codes were selected by a team at the University Hospitals of Leicester, which included OPM specialists, in collaboration with a clinical coder and a public health doctor (see Appendix 4). The approach is more traditional and links well with conceptual models of frailty based on the identification of ‘geriatric’ or frailty syndromes,122,123 which takes a pragmatic approach to identifying people who may be frail based on the presence or absence of specific information – usually in coded discharge data. This approach has the advantage of conceptual simplicity and seeks to maximise the use of recorded information but it is very dependent on the consistency and accuracy of recording diagnostic codes.

Data and methods

The analysis was based on a subset of 532,608 HES records of people aged ≥ 75 years in 2008 (see above). Four years (1 April 2004 to 31 March 2008) of prior hospital activity history were used to split the subset of HES records into two cohorts, namely those with:

1. a history of frailty (based on presence of the suggested frailty codes in diagnostic history)

2. no history of frailty.

The full list of diagnostic conditions used to identify a history of frailty, including dementia, senility and falls, is presented in Table 17. The records were then assigned to population segments, similar to those outlined in previous sections, based on subsequent hospital use over the period 2008–12.

TABLE 17 - Conditions responsible for identifying a person as frail as of April 2008 baseline

Condition	Individuals
	Number	Percentage of all in the study
Unspecified protein–energy malnutrition	989	0.2
Dementia/Alzheimer’s disease	128,101	24.1
Faecal incontinence	11,236	2.1
Difficulty in walking	37,194	7.0
Unspecified urinary incontinence	31,484	5.9
Somnolence, stupor and coma	6672	1.3
Other symptoms and signs involving cognitive functions and awareness	99,741	18.7
Very low level of personal hygiene	358	0.1
Senility	102,785	19.3
Falls	345,388	64.8
Problem related to life-management difficulty	17,229	3.2
Problems related to care-provider dependency	43,786	8.2
Dependence on wheelchair	7396	1.4

A summary of the cohort is presented in Table 18 and indicates that those with a frailty flag were more likely to be older and female.

TABLE 18 - Summary of study individuals (aged ≥ 75 years)

Note that these figures include only those people with a diagnosis history prior to April 2008.

Frailty	Individuals		Average age (years)	Percentage female	Proportion of patients dying within 4 years from 1 April 2008
	Number	Percentage of all
No prior frailty code	468,695	87.2	81.3	58.0	23.7
Prior frailty code	63,913	12.8	84.9	71.4	54.0

Figure 9 shows the proportion of patients surviving over the 4 years among those identified as frail compared with those identified as non-frail and indicates that these frailty flags are associated with lower survival rates. It also indicates that this differential effect continues to grow during the four years studied.

FIGURE 9.

Comparing survival rate in patients with and without frailty code.

When looking at differential hospital utilisation, using average hospital use per patient over this period may be misleading owing to changing denominators, as many of the patients will have died during the analysis period. Therefore, Table 19 summarises differences using ‘days at risk’ as the denominator. Days at risk are defined as the number of days between 1 April and the date of death if the patient died, or the difference in days between 1 April 2008 and 31 March 2012 if the patient was still alive at the end of the analysis period. Days at risk are then converted into years at risk to aid interpretation.

TABLE 19 - Utilisation per year at risk for patients with and without frailty flag

OP, outpatient.

Utilisation per year at risk	No prior frailty code	Prior frailty code
Total number of admissions per year at risk	0.7	1
Total number of bed-days per year at risk	4.6	11.4
Total number of OP appointments per year at risk	2.9	2.5
Total number of AE attendances per year at risk	0.3	0.5

The results indicate that the presence of the frailty flag is associated with a higher level of admissions (1.0 vs. 0.7 admissions) and a much higher number of bed-days (11.4 vs. 4.6 bed-days) in 1 year. Conversely, patients with the frailty flag appear to have slightly fewer outpatient attendances. Together, these values give an indication of the additional complexity of care associated with these frailty diagnoses.

Clustering of diagnoses to create a new approximation to frailty

Cluster analysis is a promising approach for identifying subgroups of complex patients with distinct patterns of comorbid conditions or outcomes. It is an analytical technique used to classify a set of observations into smaller and more homogeneous groups. Crucially, the observations are grouped based on their own similarity to each other across multiple characteristics, so that similar patients should end up in the same cluster. 126

In a retrospective cohort study, we ran a cluster analysis on pseudonymised patient-level hospital data. The main objective of this analysis was to explore the use of clustering on longitudinal hospital data to identify a cohort of people aged ≥ 75 years, admitted to hospital over a 2-year period, who could potentially be classed as ‘frail older people’, and to characterise their patterns of hospital use, outcomes and costs. Gilbert et al. 127 provide a more detailed description of this analysis.

The hypothesis for this piece of analysis was that one or more of the groups produced would be more likely to experience adverse outcomes and have frequent interactions with hospitals and would be more likely to have one or more of the predefined flags for frailty. Further examination of such groups could potentially be used to produce other candidate markers of frailty in secondary care. This could help inform population profiling and build towards the development of a near-patient risk-assessment tool for predicting outcomes of older people admitted to acute care settings.

In order to reduce computational requirements of the analysis, the sample size was reduced by randomly selecting a smaller population from three geographical areas. The analysis used a cohort of 22,139 people, which was created by randomly selecting 80% of patients aged ≥ 75 years who were discharged from hospitals (excluding mental health and community health providers) in the Leicester, Nottingham and Southampton LA areas between 1 April 2013 and 31 March 2015 after elective, non-elective and day-case admissions.

Three different sets of variables were used to perform the clustering: number of bed-days, costs and International Classification of Diseases, Tenth Edition (ICD-10) diagnoses. 128 The number of bed-days per patient was calculated as the total number of days spent in hospital over the 2-year period.

The clustering method identified six separate clusters, which are presented in Table 20. There were some distinctive patterns in the characteristics of patients in each cluster. Some of the clusters had strong links to particular diagnoses, for example we labelled some clusters as ‘Chronic heart problems’, ‘Cancer & lung disease’ and ‘Acute heart problems’.

Cluster number 5 was more mixed and was linked with ‘Frailty’. This group of patients had the highest level of resource use in terms of bed-days and costs and the highest mortality rate, with 48% of patients dying during the 2 years compared with 28% of all patients. Patients in this cluster were generally older (mean age 84 years) and more frequently female (60.4% of the group). The most frequent diagnostic codes recorded in cluster 5 referred to cognitive disorders and confusion. Other diagnoses such as falls, abnormalities of gait/mobility, urinary incontinence or decubitus ulcer were also at least twice as prevalent when compared with all patients. Such diagnoses are typically referred to as ‘geriatric’ syndromes. 123,129

This approach has the advantage that it uses data that are readily available, which makes it possible to look at generalisable relationships between diagnoses variables and utilisation in a pragmatic way in terms of cost and time. However, the clinical detail on individual patients is limited to ICD-10 diagnoses recorded in hospital (for those patients who are admitted) and depends on the accuracy and completeness of coding. Although this method is not as affected by this as methods that work from a predefined list of diagnoses (see Appendix 4) because it relies on a greater degree of consistency in coding practice, it would not highlight any coding problems, therefore hindering the ability to address these issues.

The following section looks at the ‘Frailty’ cluster in more detail and uses it to construct a more generalisable hospital frailty risk score.

Deriving patient frailty risk score from clusters

In this part of the project we used the results of the cluster analysis described in the previous section to interrogate HES to create a patient frailty risk score, which gives an indication of the likelihood of patients being frail. This can then be used to look at the prevalence of frailty within individual hospitals. The aim was to create a method that could be easily used by hospitals without having to carry out computationally intensive clustering methods, which are not easily replicable.

Results

The frailty risk scores for patients aged ≥ 75 years admitted to England acute trusts in 2014/15 ranged between 0 and 99 but were heavily skewed to the right: < 10% of patients had a score > 20; and < 5% had a score > 25. The proportion of patients with poor outcomes increased with increasing values of the score, but the association with mortality flattened out above a score of 15, and the association with having a long stay in hospital reduced above a score of 30 (Figure 10).

FIGURE 10.

Relationship of frailty risk score to outcomes. (a) 30-day mortality; (b) long length of stay; and (c) emergency re-admission.

Three categories of increasing frailty risk were defined using thresholds of 5 and 15 points; 40.7% of the patient cohort were categorised as being at ‘low risk’ (score < 5); 37.8% as ‘intermediate risk’ (score range 5–15); and 21.5% as ‘high risk’ (score > 15). Across the categories, the mean Charlson Comorbidity Index (CCI) score increased from 2.0 to 4.5 and the proportion of patients with three or more past admissions increased from 8.0% to 49.6%. The proportions with at least one frailty syndrome increased from 22.9% to 94.7%, with the most striking gradient for cognitive impairment increasing from 4.6% to 67.0% (Table 21).

TABLE 21 - Characteristics of patients by frailty risk category in national validation cohort with an emergency admission between 1st April 2014 and 31st March 2015 (n = 1,023,337)

Being in the least affluent quintile of the national ranking of areas according to the IMD.

Past admissions to hospital over 2 years, including elective and emergency admissions.

Frailty syndromes based on ICD-10 codes. 128 Cognitive impairment covers delirium (ICD-10 code F05), dementia (F00-F04, R41) and senility (R54).

Characteristics	Frailty risk category (score range)
	Low risk (< 5)	Intermediate risk (5–15)	High risk (> 15)
Number of patients	434,952	383,871	204,514
Percentage of patients	42.5	37.5	20.0
Frailty risk score, mean (SD)	2.1 (1.5)	9.2 (2.8)	23.3 (7.4)
Age (years), mean (SD)	82.6 (5.6)	84.8 (5.9)	86.1 (5.9)
Percentage of patients who were female	54.0	59.0	61.5
CCI, mean (SD)	1.9 (2.1)	3.0 (2.5)	4.5 (2.7)
Percentage of patients from a deprived areaa	16.7%	18.2%	19.7%
Number with missing IMD score	5210	3162	1375
Number of past admissions, mean (SD)b	1.3 (2.7)	1.9 (3.2)	3.3 (3.6)
0	50.9%	36.5%	13.9%
1	22.0%	23.5%	18.6%
2	12.4%	15.1%	18.3%
≥ 3	14.7%	25.0%	49.2%
At least one frailty syndrome (%)c	22.9	66.8	94.5
Anxiety/depression	5.1	9.5	17.2
Functional dependence	0.4	2.9	12.6
Falls and fractures	12.5	34.7	58.4
Incontinence	0.7	3.9	13.5
Mobility problems	1.1	9.3	29.6
Pressure ulcers	0.6	4.7	15.9
Cognitive impairmentc	4.7	29.1	66.2

The risk of poor outcomes also increased with increasing patient frailty risk. Across the three categories, 30-day mortality increased from 5.8% to 13.6% and the proportion with a long hospital stay (> 10 days) rose from 10.8% to 44.5% (Table 22). After adjustment for age, gender, deprivation, CCI and admission history, patients in the high frailty risk category (score > 20) had a 70% increased chance of dying in hospital compared with those in the low-risk category. The adjusted odds of staying in hospital for longer than 10 days was six times higher, and the adjusted odds of an emergency re-admission were increased by 50%.

TABLE 22 - Relationship between frailty risk category and outcomes (n = 1,013,590a)

Excludes patients with a missing IMD score.

Adjusted for patients’ age, gender, socioeconomic status, admission history and CCI and hospital variation.

Emergency re-admission within 30 days of discharge excluding patients who died in hospital in the first emergency admission (n = 920,629).

Outcome, frailty risk (score)	Percentage of patients	Crude odds ratio	Adjusted odds ratio (95% CI)b
30-day mortality
Low risk (< 5)	5.8	1.00	1.00
Intermediate risk (5–15)	11.4	2.09	1.65 (1.62 to 1.68)
High risk (> 15)	13.6	2.56	1.71 (1.68 to 1.75)
Length of stay > 10 days
Low risk (< 5)	11.7	1.00	1.00
Intermediate risk (5–15)	30.8	3.36	3.29 (3.25 to 3.30)
High risk (> 15)	44.4	6.03	6.01 (5.92 to 6.10)
Emergency re-admission within 30 daysc
Low risk (< 5)	13.9	1.00	1.00
Intermediate risk (5–15)	17.5	1.32	1.23 (1.22 to 1.25)
High risk (> 15)	21.4	1.69	1.48 (1.46 to 1.50)

On its own, the frailty risk score discriminated weakly between individuals with different outcomes. The c-statistics were 0.60 for 30-day mortality, 0.68 for a long stay and 0.56 for 30-day re-admission. Inclusion of patients’ other characteristics (age, gender, deprivation, admission history, comorbidity) improved discrimination to 0.69 for mortality, 0.79 for long length of stay and 0.61 for re-admission.

We also tested variants of the frailty risk score that were based on past data only; this would be equivalent to the data available to a clinician seeing a patient in ED. Among 630,079 patients with at least one past admission over 2 years, the rank of their ‘historic’ frailty risk score was similar to that of their current score (ICC 0.88, 95% CI 0.86 to 0.89). The historic risk scores tended to be lower than the current risk scores in absolute terms (mean score 6.9 vs. 11.0) because current risk scores capture additional current diagnoses and fewer patients were categorised as being at high risk. Historic frailty risk scores of patients aged ≥ 75 years predicted outcomes, but the sizes of associations were smaller and model discrimination was also worse, with c-statistics ranging from 0.54 to 0.60.

Introduction

In Chapter 4 we described work to create a hospital frailty risk score that uses routinely collected inpatient data from the HES. 131 In this chapter, we examine agreement between the derived score using HES data and other frailty scores derived from more detailed clinical data sets.

There are a number of different ways that frailty can be conceptualised and an even greater number of ways that it has been measured. 132 No single standardised measure of risk or frailty exists, and classifications of people as frail or non-frail can vary depending on which tool is used. The two most prominent measurement approaches, the phenotype model (Fried)133 and the cumulative deficit model (Rockwood),133 show overlap in their identification of frailty,134 although agreement in practice is only moderate. 3 The Fried and Rockwood instruments have different approaches to defining frailty and are seen as complementary.

The clinical measures developed by Fried and Rockwood tend to rely on some general clinical observations and tests, such as gait speed and grip strength. A number of attempts have been made to create frailty measures from routinely collected health-care data, including hospital122 and primary care data. 123 Crucial to appreciating the role of each of these measures is the way in which they are intended to be used. This can range from clinical assessment and decision-making tools for clinicians to population-level applications looking at patterns of resource use and need across whole organisations or areas.

The information required to calculate different measures is also important to the feasibility of using them in different settings and for different purposes. For some instruments, measurement is tied into routine clinical practice and may be documented in medical notes. For others, additional measurement and data are required that are not routinely collected. For population-level analyses – hospital level, community wide or even national – the costs of collecting or collating new information can become challenging and there is likely to be value in measures that are accessible on a large scale even though they may be less precise (hence the interest in scales derived from electronically stored patient records). Electronic records can also help in moving information to where it is needed, for example in the various attempts to link patients’ prior electronic health records in ways that can be accessed by clinicians while assessing patients. A hospital frailty risk score that exploits the power of linking patient records over time could make a useful contribution to frailty metrics.

Ideally, to validate a new frailty risk score, we would compare it with a ‘gold standard’ or ideal measurement of the same construct. The challenge surrounding frailty measurement is that this is not straightforward owing to both data and methodological issues. In this chapter, we compare HES frailty risk scores described in Chapter 4 and assess how they compare with a range of clinical measures of frailty. The HES-based risk scores were pragmatic, in the sense that they were derived from routinely used ICD-10 diagnostic codes, and they also exploit the ability to link episodes of care over time. The question was whether or not there was any relationship between the HES score and the frailty metrics derived from clinical data sets.

To form a baseline expectation of how well the measures ought to agree, we examined prior work on the agreement between different clinical frailty scales. Aguayo et al. 135 analysed 35 different frailty scores using data from the English Longitudinal Study of Ageing, using a variety of statistical methods for rating frailty with the scores and estimating agreement between these ratings. Table 23 summarises median chance-adjusted agreement (kappa) between each individual measure and all of the others, set out using the descriptors to judge the magnitude of agreement proposed by Landis and Koch. 136 The highest median kappa coefficient was 0.5, representing moderate agreement, rather than substantial or near perfect agreement (kappa > 0.6). In addition, agreement would be likely to be higher in this study than in ours because the top 20% of each scale was used as a cut-off point to identify frailty rather than using clinically defined cut-off points that may categorise varying proportions of people as frail.

Creating linked data sets

For the analysis, we used three clinical data sets that were linked to HES. These data sets had different data collections, and the cohorts represented populations with varying levels of frailty in different settings. Not all of the variables needed to construct each frailty scales were available in each of the data sets.

• Leicester/Nottingham hospital cohort: the first cohort examined was from the Acute Medical Unit at Queen’s Medical Centre in Nottingham or Leicester Royal Infirmary. This included only older people (≥ 70 years) who were admitted to either hospital. All of these patients have an index admission for the frailty risk score, as they were all admitted to hospital. However, not all have a historic frailty risk score excluding the index admission, as for some they did not have any admissions in the 2 years prior to their current admission. This cohort represents a typical population of older people with an acute hospital admission and expectation of a relatively short treatment episode before discharge. 137

• Newcastle 85+ cohort: people within this cohort were recruited in the community from GP patient registers. This cohort therefore includes a mix of people with no active health problems, as well as some requiring hospital inputs. As a result, this means that the historic HES frailty risk score only can be calculated, because there is no index admission. This cohort had the lowest level of frailty, which is expected, given that the other two cohorts comprise only individuals who had been admitted to hospital. 138

• Southampton hospital cohort: the third cohort included women admitted to an acute Care of the Elderly ward at Southampton University Hospital. These patients generally had high care needs and were admitted directly by specialist in OPM and were expected to have prolonged hospital stays. According to all the measures of frailty in the data set, this cohort has the highest number of frail individuals. 139

Data linkage

In order to calculate different frailty scores for the same person, including scores derived from HES data and from the cohort study data, we linked these local data sets to national HES data (Figure 11). The clinical data sets contained sensitive information about individuals for which they had given consent for it to be used for research purposes. The HES data set held by the Nuffield Trust is pseudonymised – each person has unique HESID but this does not make it possible to identify individuals. The key to the linkage is in the more sensitive data on HES records that is held by NHS Digital and that only NHS Digital is able to link to other data sets. The process of agreeing how the data were to be linked followed a model used by the Nuffield Trust in many previous evaluative studies looking at longitudinal changes in hospital utilisation for patients receiving specific services or interventions. 140 It is worth highlighting that the process in the present study took considerably longer than it in previous studies – > 3 years – a sign that concern over data security has made this type of work increasingly difficult. The steps required were as follows:

• Agree ethics permissions through the local Research Ethics Committee (REC) and get Confidentiality Advisory Group (CAG) approval to undertake such linkage. Ethics approval was provided by Essex National Research Ethics Service Committee (East of England) (reference number 15/EE/024). In this case we explicitly sought permissions to reuse the consent given by patients in the original studies. We were required to include additional material on the hospital and university websites that ran the studies to explain how the data were to be reused.

• Agree which fields were to be used for the purposes of linkage, arrange for the custodians of each data set to create two files and agree with NHS Digital the linkage methodology to be adopted.

• Each data set created a Microsoft Excel file containing the required field. These were sent to NHS Digital using secure transfer arrangements.

• NHS Digital linked the records in the Microsoft Excel files to HES data and identified the specific subset of HESIDs to share with the Nuffield Trust, which were shared using secure transfer arrangements.

• The Nuffield Trust extracted HES records for specific patients and linked these with the pseudonymised local research data sets.

FIGURE 11.

Outline data flows for linkage.

Construction of frailty measures

Our main analysis used five different frailty measures but not all of these were calculated for all three data sets because each of them contained different variables. Table 24 displays the frailty measures available for each data set.

The hospital frailty risk score that we developed using the HES data set in England was described in Chapter 4. In brief, the HES score is created using each patient’s 2-year history from, and including, an index admission – this is referred to in tables as the HES Current score. A historic version of the score was also created, which excludes the index admission from the calculation – referred to in tables as the HES Historic score. For the community-based Newcastle 85+ cohort, only the historic score could be calculated.

The Fried Phenotype model defines frailty as a clinical syndrome,134 where three or more of a set of criteria diagnoses an individual as frail. These are unintentional weight loss, self-reported exhaustion, grip strength, slow walking speed and low physical activity. In practice, the way that each of these five constructs are measured can vary. Appendix 5 describes the individual variables used to construct the Fried score in each of the data sets. The Leicester/Nottingham data included measures of weight loss, grip strength, self-reported energy levels, gait speed and self-reported physical activity. For the present analysis, we recreated the Fried scores from these variables and tested the results against those from the original study. 138 The Rothman Score in Leicester/Nottingham and Southampton was very similar to the Fried score – the main difference being the addition of a simple cognition score (Mini-Mental State Examination) (see Appendix 5).

In the Southampton cohort, the frailty scores were developed from slightly different variables, covering change in weight, grip strength and self-reported energy, plus some elements of the Barthel score141 covering mobility and physical activity.

In the Newcastle 85+ cohort, the Fried score and Rockwood Index were calculated by the original study team142 – sections of their report are included in Appendices 6 and 7 and show which variables were used. The scores were based on a combination of questionnaire statements and some physical observations, including grip strength, walking speed and weight loss.

The Rockwood deficit index is constructed from a list of 35 deficits, including some of the same individual variables that contribute to the Fried score. 9 To calculate the score, an individual’s total number of deficits is divided by the total possible deficits to obtain a continuous score between 0 and 1. A list of the variables used to construct the Rockwood Index is included in Appendix 7.

Methodological approach

To test agreement between the HES scores and each one of the frailty scores, we used the linked data sets to identify subsamples of cohort members for whom we had complete data on the scores being compared.

We described the linear association between the continuous versions of the HES hospital frailty risk scores and Rockwood index using Pearson correlation coefficients.

For the purpose of calculating agreement in the identification of patients as frail or non-frail, we created categories using pragmatic thresholds based on previously suggested cut-off points. For the Fried score, people with two or fewer of the Fried criteria, often labelled ‘pre-frail’ or ‘non-frail’ categories, were categorised as ‘non-frail’ and those meeting three to five were categorised as ‘frail’. The Rockwood Index is designed to measure frailty as a gradable state rather than as present or absent, but we applied a threshold of 0.25 to distinguish the frail and non-frail, as suggested by Rockwood for practical use. 143 For the HES scores, those with a score of < 5 were categorised as non-frail (the ‘low risk’ group referred to in Chapter 4) and those with a score > 5 were categorised as frail (including the ‘intermediate risk’ and ‘high risk’ groups referred to in Chapter 4).

We assessed agreement between how the different measures identify patients as frail or non-frail using kappa coefficients. This examines the proportion of cases on which two frailty measures agree in the frailty classification, corrected for chance agreement. Benchmarks for describing the strength of agreement using kappa coefficients have been suggested as 0.00–0.20 poor or slight, 0.21–0.40 fair, 0.41–0.60 moderate, 0.61–0.80 substantial and 0.81–1.00 near perfect. 137 However, agreement tends to be only fair to moderate between established clinical frailty scores. 136

We also explored the predictive value of the HES frailty risk score in detecting frailty (as defined by the Rockwood Index cut-off point of 0.25) using a range of thresholds.

Results

Discussion

Although some patients are identified as frail whatever measure is employed, there are some identified as frail using one measure, but as non-frail using another. We found that the level of agreement between existing clinical frailty measures tends to be moderate. Similarly, we found moderate agreement between these clinical measures and the HES frailty score that was constructed as part of this study. For participants in the Nottingham/Leicester and Newcastle studies, the kappa coefficients indicated a fair degree of agreement in the identification of frailty by the different scores, more than chance and in line with other frailty scores. Agreement between scores based on the clinical frailty measures was only slightly higher. It is also important to recognise that the clinical frailty scores were derived from the same data source collected at the same time and using many of the same variables. This means that some degree of agreement was inevitable, especially between Fried and Rothman scales, but even for these measures agreement ranged from 0.61 (Nottingham/Leicester) to only 0.41 (Southampton). The HES data showed a very different set of information relating to the same patient, collected and recorded in a very different way to the clinical measures, with the frailty score constructed quite differently. The analysis reported in this chapter suggests that the HES frailty measure offers a distinct, complementary approach to assessing the frailty of people admitted to hospital.

In terms of practical implications, it is important to remember that the different frailty scales may be useful for different purposes. As Cesari et al. 145 have discussed, the frailty phenotype categorically describes a condition implicating risk for subsequent events (most specifically, disability), which can help clinicians decide on the possible need of changes to care and/or interventions. The frailty index conceptualises frailty as a gradable condition, rather than present or absent, which may be useful for other purposes. Identifying this group of patients would allow targeted screening for frailty syndromes and the delivery of frailty attuned approaches to care. Examples include CGA, prevention of delirium and functional deterioration and identification of end-of-life care needs on a hospital-wide basis. Additional work should explore the impact of the score on clinical decision-making, in particular that it does not have the perverse effect of increasing therapeutic nihilism. The HES frailty score has complementary value as a pragmatic way of assessing frailty by drawing on existing administrative information. This score can be used as either:

• A way to synthesise electronic information from previous hospital activity and support clinical decision-making over whether or not an individual patient needs more specialist assessment and/or care. Ultimately, the test of this is in the application of the measure to daily clinical practice – something that is beyond the remit of this study.

• A way to describe the relative levels of frailty in populations between areas and organisations over time in order to help planning and commission of appropriate health services. The utility of the scale in this respect is therefore based partly on the extent to which it corresponds to or augments clinical perspectives of frailty (as seen in agreement studies here), but also in the ability to differentiate on patient outcome and utilisation (discussed in Chapter 6).

The use of pre-existing administrative data sets for research purposes is an important area in that it makes larger study sizes possible and can draw on information about the use of health and care services that is an important adjunct to clinical information. However, the use of administrative data poses many challenges in terms of accessing data, ensuring that data are handled securely and in achieving record linkage across data sets, and ensuring consistency of data definitions and collection. We note that the process of securing the data linkage required for this validation work was considerable and had adverse knock-on effects on the way in which we were able to process the data during the project.

Summary

This chapter describes the impact of frailty on survival and hospital use and its associated costs over 2 years among two cohorts of patients admitted to hospitals in Leicester, Nottingham and Southampton and from participants aged ≥ 85 in a community-based study in Newcastle. Through gaining a better understanding of the impact of frailty, this work intends to identify potential benefits of improved assessment and management of frail older people through CGA, including end-of-life care.

This work involved the innovative linkage of detailed clinical data sets, developed for other research purposes, with routine records of hospitalisation and death. The linkage across data sets and for individual patients over time enables us to look at events before and after an index episode of care and, therefore, to better describe the lasting and longer-term consequences of frailty. In addition, linkage of the Newcastle 85+ cohort study data to their past and future administrative records provides new insights into how frailty is linked to survival and health-care use among the oldest of old people.

Introduction

Frailty has been associated with a range of short-, medium- and long-term adverse outcomes in older people. People identified as frail in community-dwelling populations in several countries, including France, Portugal, the Republic of Korea and the USA, have been found to have increased mortality and hospital admissions. Clegg et al. 106 summarised the evidence from four large cohort studies linking frailty to higher rates of adverse outcomes, including hospitalisation.

In previous community-based cohort studies, information about contact with health-care providers, such as hospital admissions, has relied on self-reported data at designated follow-up points. In contrast, the present work has made novel use of linked hospital records. In addition, none of the studies has focused specifically on health-care use and survival among the oldest old (≥ 85 years), which we have been able to do by accessing data on the Newcastle 85+ cohort study.

The present work also adds to the literature on the impacts of frailty following a hospital admission. Understanding the impact of frailty in these settings is important, because, for some people, hospitalisation is associated with an increased risk of harms over and above the presenting clinical condition, such as risk of delirium and deconditioning. Older people with frailty are at an even greater risk. Previous studies of hospital-based populations of older people have tended to focus on short- and medium-term outcomes at 30, 90 or 180 days. For example, in our Leicester/Nottingham cohort of people aged ≥ 70 years, various measures of frailty have previously been associated with significantly higher mortality and re-admission at 90 days. Another hospital study found that frailty was strongly associated with delirium among older inpatients aged ≥ 75 years, which was in turn a strong predictor of worse survival: median survival was 88 days among patients with delirium compared with 359 days among patients without delirium. In the present work, we have been able to link hospital and death records over 2 years to focus on long-term outcomes for hospitalised older adults. This could help to understand potential long-term benefits of interventions designed to improve care for frail older adults.

In addition, we have focused on the total number of days spent in hospital (‘bed-days’) as a useful measure of the intensity of long-term hospital use. People who are frail are known to have longer lengths of stay in hospital, as well as more frequent hospital admissions, and, therefore, bed-days give a more accurate representation of their overall hospital use. In a survey of older people, most of those participating said that they would prefer to have their care provided outside hospital and bed-days are also the main determinant of health-care costs among older people. Other studies have focused on this broad measure of hospital use for conditions such as heart failure.

Finally, we have explored the interaction between frailty and hospital use at the end of life. Previous work has shown that hospital use increases sharply over the last few weeks of life, and the pattern of use could depend on frailty. We also wanted to understand how far any differences in intensity of hospital use by frailty status were linked to increased use at the end of life.

Data

Methods

Kaplan–Meier survival curves were plotted for each frailty scale with the survival function, together with 95% CIs presented as shaded areas. Cox regression was used to produce unadjusted and adjusted hazard ratios describing relative hazards between people identified as frail and non-frail by the frailty scales. Models were estimated that adjusted for age and gender only, as well as models that, in addition, adjusted for past number of admissions (0, 1, 2 or ≥ 3) and CCI score (0, 1, 2 or ≥ 3). We recognise that the final set of results adjusting for admission history could represent ‘over-adjustment’ in the sense that more past admissions could be indicative of frailty, and mutually reinforcing rather than independent. Nevertheless, these results are presented alongside the other results in order to help understand the extent to which frailty relates to long-term outcomes even if patients have very similar histories otherwise.

Hospital use was modelled in two different ways to describe variation in the rate of use over 2 years, and also variation in the rate of use over a person’s remaining lifetime. The first of these measures is of particular interest from the service cost perspective, but the second may be more important from the individual perspective. The difference between the two measures is marked among the patients coming towards the end of their life.

To describe the rate of hospital use over 2 years, we used the total number of bed-days as the dependent variable in a negative binomial regression model and included the frailty measure as a binary independent variable. We used negative binomial regression because the distribution of bed-days was very positively skewed, with a small number of patients having very long periods in hospital. Models were adjusted using the same set of variables as for the survival analysis. To describe the rate of hospital use over the course of a person’s remaining life, we re-estimated the models including survival time (in days) as an offset.

The cost of inpatient hospital use over the 2-year follow-up period was also described and modelled. A detailed description of the method used to calculate inpatient costs is available in Appendix 8. In brief, all hospital inpatient activity was costed by grouping clinically similar conditions and treatments with similar costs into Healthcare Resource Groups (HRGs) using the NHS Digital grouper software. 146 The most recent mandatory national tariff prices available (2016/17)147 were then used to attach prices to the activity data. If national tariffs were unavailable, the 2013/14 national reference costs were used, adjusted for Market Force Factors, averaged nationally and then uplifted for tariff inflation. 148 If neither of these sources provided costs for a particular HRG, specialty average costs were calculated from the tariff costed spells and applied back to spells that were still uncosted (on the basis of the dominant spell treatment function). Excess per diem costs were added for each extra day spent in hospital for patients with an exceptionally long length of stay for their HRG.

The mean cost per person was presented by frailty status along with the unadjusted and adjusted difference in costs between the frail and non-frail group. Adjusted costs were produced using negative binomial regression with total cost as the dependent variable and the frailty measures as the independent variable. Two adjusted differences were calculated, one adjusting for age and gender, and one additionally adjusting for CCI score and admission history as with the hospital use models.

To gain a better understanding of the interplay between frailty, hospital use and survival, the hospital use models were also fitted on the restricted sample of people who survived the period. We used visual methods to explore hospital use in the 90 days before death by frailty status. The relative contribution of end-of-life hospital use was captured with the proportion of days in hospital over the last 90 days of life over the number of days in hospital over the 2 years before death. This was calculated by frailty status for those that died in the 2-year follow-up period.

Results

Table 32 summarises baseline characteristics of the three cohorts. Leicester/Nottingham participants were on average younger than Southampton and Newcastle cohort participants. All of the Southampton participants were female, compared with just over half (58.4%) in the Nottingham/Leicester and Newcastle (39.4%) cohorts. The Newcastle cohort were the least likely to have past hospital admissions, with 1.6 on average per person over a 2-year period, whereas the Southampton cohort were the most likely to have past hospital admissions, with 6.2 on average per person over 2 years.

There were marked differences in the levels of frailty in each cohort, regardless of which measure was used. In particular, although between 19% and 44% of the Leicester/Nottingham hospital cohort were categorised as frail, prevalence in the Southampton hospital cohort was much higher, with 48–81% defined as frail. The low prevalence of frailty based on grip strength in the Leicester/Nottingham cohort is slightly misleading, because a sample-based cut-off point was used to define frailty, but frailty based on the other measures also tended to be lower. In contrast, for the community-based Newcastle 85+ cohort, between 20% and 30% were categorised as frail based on the Fried, Rockwood or historic HES hospital frailty risk scores.

Please find additional data in Report Supplementary Material 4.

Survival analysis

The proportion of people who were still alive at the end of the 2-year follow-up period varied from 57% in the Southampton cohort to 76% in the Leicester/Nottingham cohort and > 80% in the Newcastle 85+ cohort.

Figure 12 shows examples of survival curves contrasting survival patterns in frail and non-frail groups in the two hospital-based cohorts using the Fried frailty scale. Other survival curves are presented in Appendix 8. In the Leicester/Nottingham cohort there were clear differences between the frail and non-frail groups, which become significant after around 250 days (8 months). These differences were also present for the Newcastle cohort, although there was greater statistical uncertainty around the estimates for those identified as frail. The smaller sample sizes in Southampton meant that, although median survival for the two groups was significant between the frail and non-frail groups, there was much greater statistical uncertainty around the survival estimates for each group.

FIGURE 12.

Survival over 2 years by Fried frailty status in the three cohorts. (a) Leicester/Nottingham; (b) Southampton; and (c) Newcastle.

Table 33 summarises the results of the survival analysis by frailty status. In the Leicester/Nottingham and Newcastle cohorts, the proportion of participants who survived the 2-year follow-up period was consistently lower for those classified as frail than for those classified as non-frail across all the frailty scales. In the Southampton cohort, women who were identified as frail using the Fried or Rothman scales did have significantly lower survival, but neither grip strength nor the HES-based frailty ratings identified patients with lower survival.

TABLE 33 - Two-year survival by frailty status in the three cohorts

**p-value for relative difference in hazards of < 0.05.

Southampton analysis adjusted for age only (women-only cohort) and Newcastle analysis adjusted for gender only (all 85-year-olds at baseline).

Adjusted for gender, age, number of past admissions and CCI.

Frailty scale (sample size)	2-year survival (%)		Hazard ratio (standard error)
	Non-Frail	Frail	Unadjusted	Adjusted for age and gendera	Fully adjustedb
Leicester/Nottingham hospital cohort
Fried (584)	83	63	2.51** (0.18)	2.01** (0.19)	1.72** (0.19)
Rothman (594)	83	63	2.50** (0.18)	2.17** (0.18)	1.77** (0.19)
Grip strength (805)	79	65	1.66** (0.16)	1.48** (0.16)	1.34 (0.16)
Rockwood (569)	84	66	2.31** (0.18)	2.08** (0.18)	1.69** (0.19)
HES historic (599)	78	64	1.73** (0.16)	1.60** (0.16)	1.30 (0.17)
HES current (825)	84	65	2.44** (0.15)	2.10** (0.15)	1.62** (0.16)
Southampton hospital cohort
Fried (140)	79	54	2.63** (0.43)	2.32** (0.43)	2.63** (0.45)
Rothman (192)	69	47	1.96** (0.23)	1.62** (0.24)	1.60** (0.25)
Grip strength (246)	68	55	1.55 (0.28)	1.35 (0.28)	1.25 (0.29)
HES historic (187)	52	60	0.78 (0.22)	0.70 (0.22)	0.76 (0.26)
HES current (246)	57	58	1.04 (0.21)	1.08 (0.21)	1.11 (0.24)
Newcastle 85+ cohort
Fried (538)	90	77	2.37** (0.25)	2.80** (0.26)	2.04** (0.32)
Grip strength (711)	88	82	1.61** (0.23)	1.55 (0.23)	1.60 (0.32)
Rockwood (726)	89	76	2.60** (0.18)	2.86** (0.19)	1.92** (0.26)
HES historic (418)	85	67	2.36** (0.23)	2.41** (0.23)	1.49 (0.25)

In the Leicester/Nottingham and Newcastle cohorts, the hazard of death was 1.6 to 2.6 times higher in the frail than in the non-frail groups depending on the sample and measure used. Adjusting additionally for differences in age and gender, the hazard ratios were between 1.5 and 2.1. Adjusting for admission history and CCI reduced the ratios to between 1.3 and 2.0, and, for the HES historic frailty risk score, the relative differences in hazards were no longer statistically significant, largely due to its association with admission history. The hazard ratios for the full three category frailty scales are presented in Table 33.

End-of-life care

In order to explore further the relationship between resource use and survival among frail patients, we specifically looked at hospital use in the period before death. This analysis was limited to the Rockwood frailty score for the Leicester/Nottingham and Newcastle cohorts to ensure sufficient sample sizes for subgroup analyses.

In the last 90 days alive there was no significant difference in the mean time spent in hospital between those classified as frail or not by Rockwood in either cohort (p-value > 0.3). The proportion of the Leicester/Nottingham and Newcastle cohorts in hospital on any given day over the last 90 days alive by frailty status is presented in Figure 13. The patterns of hospital use are similar among people who were categorised as frail and non-frail at baseline. Around half of all time spent in hospital over the last 2 years of life occurred in the last 90 days for those classified as non-frail by Rockwood in the Leicester/Nottingham (42%) and Newcastle (57%) cohorts. For those classified as frail by Rockwood, only one-third of hospital use over the last 2 years of life was focused in the last 90 days: 32% for Leicester/Nottingham and 34% for Newcastle. For the Newcastle cohort, those classified as non-frail had a higher proportion of time spent in hospital occur in the last 90 days of life (57%) than those classified as frail (p-value = 0.01).

FIGURE 13.

Hospital use over the last 90 days of life among patients in the (a) Leicester/Nottingham (n = 123) and (b) Newcastle (n = 119) cohorts who died over a 2-year period.

The results show that the proportion of participants that are in hospital increases towards death for both frail and non-frail people. The number of bed-days in the weeks before death seems to be largely unaffected by frailty status. Those classified as frail in the Newcastle cohort were, however, more likely to spend time in hospital at other time periods over the last 2 years of life.

Discussion

This chapter has described the impact of frailty on long-term survival, hospital use and hospital costs for three cohorts; a cohort of patients admitted to short-stay acute units in Leicester or Nottingham, a cohort of female patients admitted to a long-stay geriatric ward and a community-based study of participants aged ≥ 85 years in Newcastle. Frailty was found to be associated with these long-term outcomes but the results differed by cohort, highlighting the challenges of identifying frailty in all populations.

The Leicester/Nottingham cohort was the generally healthier of the two hospital-based cohorts. Between 25% and 40% of participants were characterised as frail, depending on the frailty rating score used. In this cohort, frailty defined by the clinical and HES-based scores was consistently associated with worse survival and higher hospital use and associated costs.

Between 50% and 80% of the Southampton hospital-based cohort were characterised as frail and there was a mixed relationship between frailty and long-term outcomes. Frailty identified by the clinical frailty scores was linked to worse survival, but the HES frailty risk scores did not discriminate. Shorter survival times for those identified as frail by the clinical frailty scores meant that no difference was found in hospital use between the two groups unless survival time was taken into account. The HES frailty risk score found higher hospital use for those classified as frail.

The community-based Newcastle 85+ cohort could be described as the healthiest group, having all survived to 85 years of age and with recruitment outside hospital. Between 20% and 70% of the cohort were classified as frail. Those classified as frail by the clinical frailty scores were more likely to die over 2 years and to have higher hospital use. Those identified as frail by Rockwood also had slightly increased hospital costs. The historic HES score identified some evidence for increased mortality and hospital costs in the frail group.

The mixed results by cohort may in part reflect the nature of the different settings but these differences offer lessons to be learnt. The clinical frailty scales were less able to discriminate in the overall frail group of women in the Southampton cohort and to create large subgroup sizes for analysis. The impact of frailty on long-term outcomes should be investigated in a larger sample of frail individuals to determine whether it was the suitability of the frailty scales for this population or the sample size that meant that a consistent association between frailty and the long-term outcomes could not be found. The Newcastle cohort could be considered a ‘healthy elite’ to have survived to this age in relatively good health. The absence of a current admission meant that only a historic HES score could be calculated, and, with 1.6 past hospital admissions on average, the information available to calculate this score was limited. In this cohort the relationship between past and future hospital use acting through the HES score may have limited the impact of frailty on the outcome. The clinical and HES-based scores were most effective at discriminating in the Leicester/Nottingham cohort. This is promising for the potential use of the HES-based scores as the Leicester/Nottingham participants were recruited in a setting similar to that which may benefit most from this type of score.

These findings are important in helping to understand the impact that frailty has on long-term survival and health-care resource use as well as gaining some understanding of the appropriateness of the varying frailty scales for different populations.

Introduction

This chapter has two main purposes. First, it describes work done to estimate the hospital staffing costs of providing CGA in an acute care setting. A consensus-based approach was used, drawing on knowledge captured via a costing workshop that brought together practitioners with expertise and direct experience of commissioning, providing and regulating hospital services for older people. The results of the costing exercise were used to develop an interactive Microsoft Excel-based costing tool to help local commissioners and providers estimate the costs of providing hospital-based CGA.

Second, the chapter also describes work done to create an interactive needs tool to help commissioners and providers describe frailty and hospital activity within their older local populations aged ≥ 75 years. Indicators have been developed and populated with data for each LA and NHS acute trust in England to describe populations, hospital costs (see Appendix 8) and hospital activity using the HES and ONS data sets. These are delivered with HES-based measures of frailty and hospital utilisation, as described in Chapter 4, to provide a range of local estimates of the number and proportion of older people who could need CGA at both LA and NHS acute trust level. For NHS acute trusts, additional indicators on patient outcomes (mortality and emergency re-admission) have also been included.

As well as a description of the work done to create the costing and needs tools, this chapter also includes a brief description of their contents.

Costing the care of frail older people and developing a costing tool

Relatively few studies about frail older people examine the impact of their management on hospital costs8 and there are no studies that have looked specifically at the hospital staffing costs of undertaking CGA. However, for organisations wishing to introduce CGA, it is important to consider the cost implications in helping to plan, shape and develop this service. Those studies that do consider resource consequences use length of stay as a proxy for hospital costs, mainly because it is relatively easy to get data on length of stay at person level. Although length of stay is a major determinant of costs, it is not the only component. Nor does length of stay provide much valuable information to inform service and personnel planning.

There are two challenges in estimating the costs associated with CGA. First, given that CGA is an ‘added extra’ to the package of care provided to frail older patients in hospital, it is necessary to identify who and what is involved in delivering this additional support, over and above what would be considered as usual care for such patients. Second, there are no routine costing data that can be used to determine these additional costs.

Given that that the additional costs associated with CGA cannot be determined from routine data, we adopted the following strategy to estimate staffing costs. We first conducted a number of interviews and site visits, which are described below. Our original intention was to see whether or not we could develop a questionnaire that would itemise the key cost elements of CGA including direct staff cost, training cost, consumables costs and some recognition of any overhead costs (e.g. information technology, premises). We would then use this questionnaire to survey other sites about the costs of their CGA services.

Following the interviews and site visits, it became apparent that a standardised survey questionnaire could not be developed, given the vagaries of defining CGA, the range of patients who each service dealt with, and the difficulty of instructing respondents to provide estimates of only the additional costs of CGA over and above usual care. Instead we decided to rely on expert opinion to estimate the costs associated with providing usual care and CGA to two typical patients, described using vignettes. Other studies of health-care costs have also used vignettes to ensure standardisation. 149 This involved bringing experts together in a workshop, which we describe below.

Site visits

A series of interviews and sites visits in November 2015 and February 2016 were conducted with staff working directly in hospital services providing older people’s care. A detailed description is provided of one of these visits to Royal United Hospital Bath emergency frailty services to illustrate the type of information that was gathered.

Visits to Royal United Hospital Bath: February 2016

Two members of the Nuffield Trust team shadowed and interviewed different members of the MDT who provide emergency and short-stay medical care to older people presenting to the hospital. The aim was to identify key members of staff involved in CGA, the time involved in conducting the initial assessment and follow-up care and any extra resources used as a part of CGA that do not form part of usual care.

The visit started from a post-take ward round with a consultant geriatrician to review patients who had been newly admitted to the Geriatric Medical Admission Unit and to decide whether or not they needed care in other wards. After that, the second ward round was on a short-stay ward for older people – the Assessment & Comprehensive Evaluation Older Persons’ Unit (ACE OPU). At ACE OPU, there was a consultant geriatrician-led ward round each morning.

The primary aim of an ACE OPU ward is to facilitate rapid clinical assessment (i.e. CGA), to support early discharge and to reduce time in hospital. The key observation from this visit was how closely the nursing and medical team worked with allied health professionals (including physiotherapists, occupational therapists and pharmacists) and social workers to plan effective discharge. During the visit, it became apparent that there was a great deal of complexity around the patient journey. Figure 14 is a simple diagram of patient pathways. Because the visit was arranged in an informal capacity, this is an account of the author’s observations and gives a general idea and may not represent a full and accurate account.

FIGURE 14.

Schematic of patient pathways used to inform costing of CGA. GMU, Geriatric Medical Admission Unit.

It also became evident how important a role the team working with the geriatrician had in proactively making arrangements to plan a safe discharge for patients. In this hospital, the initial structured assessment part of the CGA was typically conducted by a physiotherapist but was followed up and reviewed by the geriatrician. The team at the ACE OPU ward explained that they defined both this initial assessment, the review and the follow-up actions as CGA.

Visits were made to two other hospitals (Leicester Royal Infirmary and St Helier Hospital, Epsom) to shadow and interview staff providing emergency care to older people. During the site visits and via further correspondence with key contacts, we attempted to estimate the staff time and costs involved in CGA. Two common themes emerged: the importance of identifying frail older people who may benefit from CGA using frailty assessment tools in the ED and the role of the MDT working proactively with a geriatrician. However, the names and structures of these services varied, as did the patient pathways and the staff involved in delivering care. From this work, three key issues emerged that highlighted the difficulties of defining and costing CGA:

1. There was a problem of developing a consistent notion of which elements of care were specifically associated with CGA.

2. There was a shortage of basic information about resource use, including the personnel involved and diagnostic tests performed.

3. There was the challenge of how to present ‘average’ scenarios – given the variability in patients’ care requirements and their personalised care packages.

Reflecting on these issues, it was decided to abandon the idea of developing a survey questionnaire and instead to adopt an alternative consensus-based approach to assess the hospital staffing costs associated with CGA.

Workshop

The consensus-based approach involved a workshop that brought experts in the field of older people’s care together to collaborate, discuss and debate the costing implications of CGA.

The workshop was attended by 24 people from 15 different organisations. We recognised that collaboration with multiple health service organisations would be critical in developing an informed consensus, and we invited a range of experts, including public health professionals, members of government agencies, academics and front-line staff working in hospitals. Organisations represented included:

• PPI representatives

• NHS Interim Management and Support (IMAS)

• College of Occupational Therapists

• Association of Directors of Adult Social Services (ADASS)

• Systematic Care for Older People in Elective Surgery (SCOPES) – Nottingham

• Health Education East Midlands

• NHS Benchmarking Network

• Future Hospital Programme Manager

• Guys and St Thomas’s NHS trust

• Salford Royal Hospital

• Brighton and Sussex Medical Schools

• Royal Sussex County Hospital

• University Hospitals of Leicester NHS Trust

• The Nuffield Trust

• University of York.

The main objective of the workshop was to identify the major components of the clinical team involved in CGA and how it differed from a non-CGA environment. We deliberately focused our efforts on looking at staff inputs, as this was felt to be the area in which differences between CGA and non-CGA care were the most likely.

To create context for the workshop conversations, we asked participants working in small groups to undertake two exercises. The first exercise (Box 1) was designed to get an idea of typical patient pathways. The second exercise (Box 2) was designed to identify the additional costs of CGA based on two vignettes. The idea behind these exercises was to highlight the variability that exists in different CGA and non-CGA models of care. For both exercises, attendees were split into four groups comprising six experts. The discussions were documented and shared with the wider group.

BOX 1 - Exercise 1: mapping the patient pathway

For this first exercise, we would like you to imagine you are walking the patient pathway for a typical patient treated by you/your team.

Start from the point when they arrive at hospital and end at the point when they leave your team’s care.

If you see patients admitted via a wide range of pathways, please describe a typical pathway to start with. If you have time, please create more than one!

Write on and arrange the coloured cards to show which departments and wards patients can move between, and which staff are involved in their care.

Please write on the plain card to show whether or not you think that patients receive CGA.

Please nominate someone on your table to describe the pathway for us.

Results from this first exercise revealed that one of the key features of the CGA staffing model is having a co-ordinating MDT involved in the care of older patients under the care of a geriatric consultant. It was also highlighted that screening for frailty may take place in ED or frailty units at the front end of hospitals. Patients can be assessed by nurses or therapists, pharmacists, etc., and other professionals will be brought in to deal with other aspects of care, as was observed on the visit to Royal United Hospital Bath. Participants at the workshop highlighted how quickly acute issues get handled in EDs. Patients become hospitalised quickly and can lose physical function overnight, so staff believe that spending resources at an early stage and in a concentrated way is required, prioritising proactive instead of reactive care. Staff believed that this would to be cheaper in the long run and improve outcomes.

The second exercise provided information about two patients deemed to be fairly typical of people with either severe frailty or marginal frailty, as set out in Box 2. These two vignettes were designed to help workshop participants create profiles describing the mix of staff involved in CGA and the extra time that different members of staff devote to undertake a CGA relative to usual care in the absence of CGA for each of these patients.

BOX 2 - Exercise 2: estimating staff time involved in CGA

Two patients are described in the cards on the table.

Severe frailty case study

Jaspreet, 78-year-old lady, who weighs 50 kg (BMI of 20 kg/m²).

Found on the floor in her sitting room by community carers this morning.

Conscious but confused, not in pain. On floor for at least 3 hours.

Jaspreet cannot remember falling. Carers report three falls in the past year, but none have required hospital treatment.

Jaspreet was in hospital 4 months ago with pneumonia.

Also: type 2 diabetes mellitus on metformin, atrial fibrillation on Apixaban and Bisoprolol, Co-codamol for osteoarthritis and thyroxine for hypothyroidism.

She has no allergies. Carers report that Jaspreet has some memory problems, but she does not have a diagnosis or receive any treatment. She has lost 5 kg in the past 6 months or so.

Jaspreet has a walking frame but does not use it around the house, preferring to hold on to walls/chairs, etc. and she does not tend to go out very often, as she feels unsteady on her feet and her sight is deteriorating.

Jaspreet lives alone in a large house but uses only three rooms on the ground floor. Carers come once a day for medication and to help her dress in the morning. The carers report a recent decline in the condition of her home.

Marginal frailty case study

Raymond, aged 76 years, weighs 68 kg.

Severe periumbilical abdominal pain throughout the night and getting worse. He has never felt anything like this before.

Raymond has prostate cancer under a ‘watchful waiting’ programme.

Right knee replaced 10 years ago, as a result of osteoarthritis.

He has hypertension and so takes 240 mg of verapamil and 20 mg of Lisinopril.

He is allergic to penicillin. Raymond is an ex-serviceman. Raymond is fully mobile normally, but his pain worsens when he tries to straighten up to walk.

Raymond was really reluctant to come with us as his partner is in a wheelchair and Raymond is currently his sole carer.

Raymond does all of the work around the house and normally copes well.

For this second exercise, we would like you to estimate the amount of time that each member of the MDT spends treating each of these two example patients.

Fill out a separate sheet for each of the two patients.

List each member of staff typically involved in the care of this patient in your hospital.

For each member of staff:

Describe what is typically involved in the initial assessment and ongoing management of this patient’s care.

Estimate the total time in minutes that would typically be spent on their care.

Please complete the additional column, giving the time in minutes that would ideally be spent on care of this patient.

Please go through the same exercise but imagine a ‘usual care’ scenario in the absence of CGA.

BMI, body mass index.

Table 38 shows the broad estimates of additional staff time associated with providing CGA to marginally frail and severely frail patients over and above the costs of usual care. These estimates are the averages based on information collated from the four groups comprising six experts. The additional staff time corresponds to extra time required in the care of frail older people receiving CGA. For example, a Consultant Specialist in OPM might contribute 20 minutes towards CGA for someone with marginal frailty and 30 minutes for someone with severe frailty, which would not have been contributed in a usual care scenario. Note that the estimates for additional nursing time are zero. This is because workshop participants believed that each care team consists of full-time equivalent nurses that have a full caseload based on the unit/bed layout, making it too complex to disentangle and quantify additional nurse time associated with CGA.

TABLE 38 - Estimates of additional staff time (in minutes) spent on patient assessments per patient in acute hospital setting

Staff	Frailty staffing levels
	Marginal	Severe
Consultant specialist in OPM	20	30
Acute consultant	20	0
Junior doctor	0	30
Nurse	0	0
Occupational therapist	60	120
Pharmacist	20	20
Physiotherapist	45	60
Psychiatric nurse	0	30
Old age psychiatrist	0	30
Social work/community care worker	0	0
SALT	0	20
Primary care co-ordinator	0	0
Discharge co-ordinator	30	30
Dietitian	0	30
Other: meaningful activities facilitator	10	40

Table 39 converts staff time into cost estimates. The estimates suggest that the additional staffing cost of performing a CGA amounts to £90 for someone with marginal frailty and £172 for someone with severe frailty. It is worth noting that these costs elements are in the context of overall hospital costs for an average inpatient visit for a frail patient of the order of £2000–£3000. Hence, CGA represents a small increase in costs of providing care, amounting to around 4.5% to 5.7%.

TABLE 39 - Estimates of additional staff time costs associated with patient assessments (per patient)

WTE, whole-time equivalent.

Staff cost for new patients	WTE requirement for (£)
	Marginal frail patient	Severe frail patient
Consultant specialist in OPM	18.14	27.20
Acute consultant	18.14	0.00
Junior doctor	0.00	7.97
Nurses	0.00	0.00
Occupational therapist	21.15	42.29
Pharmacist	7.05	7.05
Physiotherapist	15.86	21.15
Psychiatric nurse	0.00	5.70
Old age psychiatrists	0.00	27.20
Social/community care worker	0.00	0.00
SALT	0.00	7.09
Primary care co-ordinator	0.00	0.00
Discharge co-ordinator	8.19	8.19
Dietitians	0.00	10.57
Other: meaningful activities facilitator	1.90	7.60
Total	90.41	172.00

We used the information provided during the workshop to construct a simple Microsoft Excel costing tool, allowing users to vary staffing input and wage rates according to their local circumstances. As baseline estimates, we populated the tool with the information about personnel inputs provided by the workshop experts, costed using NHS pay rates.

In terms of the factors that lead to differences in resource use between frail people who do or do not receive CGA, the workshop identified four key elements:

1. range of staff involved in their assessment

2. grade and salary levels of staff involved

3. time spent on assessment

4. length of stay.

The Microsoft Excel costing tool allows users to input different values for each of these elements, tailored to the characteristics of an individual patient and the particular circumstances of the care team.

As anticipated, we found that CGA entails increased costs of care, but the marginal increase is not substantial, amounting to < 6% of the overall cost of hospital care, even for those with severe frailty. Moreover, these slightly increased costs might be offset by downstream savings. Workshop participants suggested that investment in a CGA might be associated with a reduction in length of stay, reduced complications of treatment or lower re-admission rates. These savings might be secured through timely and safe discharge or transfer, an early identification of a problem or the need for rehabilitation interventions that could arrest functional decline for patients.

Developing a tool to assist commissioners and providers ascertain needs of Comprehensive Geriatric Assessment

To assist commissioners and providers of hospital care in England, we developed a Microsoft Excel needs tool to present descriptive information about the population of older people at both the LA and provider (NHS Acute hospital Trust) level. By making this sort of descriptive information more accessible, we hoped to help local providers and commissioners of care in England get a better picture of local needs and health-care service use patterns in their area. As it aggregates information on the number of patients who could benefit from the implementation of CGA, it was hoped that the tool could assist in planning and arguing for wider use of CGA.

Data and methods

We chose to use a Microsoft Excel spreadsheet as the optimal way to store a large volume of basic information. Although it is not the most sophisticated way to manage or present information, it has some key advantages:

• it allows the user to use relatively large data sets for all areas/providers but also allows the user to select their own data set from those that are pre-loaded

• it is flexible in the ways in which data can be added and manipulated

• Microsoft Excel is readily available software that all organisations and most individuals can access at no extra cost (see Report Supplementary Material 1).

In deciding which indicators, comparators and additional information to include in the tool, the team used an iterative approach. A first attempt was made based on the experience of the analytics team. This was shared with clinical colleagues and discussed within the patient and public forum. The changes suggested included a reordering of key information, more information at provider level, greater use of graphic display versus tables and additional explanatory notes. Following these discussions a number of amendments were made and a final version was produced.

The data are presented at either LA level for population-based needs or NHS trust level for provider-based needs. The LA level has the advantage that it provides more stable population denominators than other geographies, such as Clinical Commissioning Groups (CCGs), the boundaries of which are prone to change and which are therefore less certain in the way population estimates are constructed. The LA level also means that the data can link to LA commissioners and health and well-being boards when the issues relate to the wider care needs for the whole population. Provider analysis was based on NHS Acute hospital trusts per NHS Digital.

Please find additional data in Report Supplementary Material 5.

The LA indicators are organised in four themes, describing population, hospital activity, hospital financial costs and frailty within hospital. The provider indicators are also organised in four themes, describing hospital activity, hospital financial costs, frailty within hospital and patient outcomes.

Users of the tool can select any LA or NHS acute trust from a drop-down list, and the spreadsheet instantly populates the relevant indicators with data for that selected organisations (Figure 15).

FIGURE 15.

Screenshot of a LA indicator from the Microsoft Excel needs tool.

Some indicators have been indirectly standardised, based on age, gender and IMD, to allow comparison to national averages. Population estimates were derived from ONS population estimates for 2014. Information on population deprivation was derived from the IMD74 and reported for LA area and provider. In order to comply with information governance requirements, all small numbers (< 5) have been removed to prevent the identification of individuals.

For both LAs and NHS acute trusts, there are a set of indicators on the levels of frailty within hospitals, which shows the distribution of frailty risk scores (see Chapter 4) for different patient groups. These provide LAs and NHS acute trusts with comparisons of patients with intermediate or higher risk of frailty to patients with low risk of frailty, along with national averages. These indicators were selected to help local commissioners and providers make the case for and plan the introduction of CGA use within a LA or provider.

Introduction

Frail older patients with complex needs admitted to acute care should receive holistic care as encompassed in multidisciplinary assessment and management, regardless of their destination in the hospital. The strongest evidence is currently for discrete, ward-based services, as opposed to peripatetic teams providing assessment and advice. Liaison services that simply offer advice, rather than actively direct patient care, are not as effective.

The task for WS 4 was to bring together and synthesise evidence, including that produced by WS 1 and WS 2, craft a model of best practice in CGA, and produce a range of tools to disseminate it. The aim was to design a whole-hospital approach to improving care for frail older patients and to improve existing practice in multiple areas of the hospital in which CGA is not part of the clinical tradition (such as surgical and oncology units), but in which frail older people with a wide range of problems may be treated.

The toolkits were expected to consist of a range of resources including, but not limited to, evidence summaries, best practice guidelines for commissioning and service delivery, assessment tools to assist in identifying those most likely to benefit, case studies, benchmarking tools for service delivery, and links to generic resources on organisational change. The design of the toolkit was to deploy evidence-based approaches to knowledge translation to secure maximum uptake. In their effort, the design team aimed to draw on the expertise of the study’s External Stakeholder Group, gathering their insights on the content and framing of the toolkit(s), and harnessing their networks and influence for change.

In this chapter, we describe the process of toolkit design, focusing principally on the design and development of a service-level (clinician-oriented) toolkit, and the fit of this with other ‘levels’ of toolkit that were also developed as part of the study. We start initial efforts to ‘make sense’ of the existing evidence, then describe the cycles of consultation and consensus building that took place with the External Stakeholder Group between January and May 2016, and conclude with the process of identifying and negotiating access to the pilot sites in which the service-level toolkit was then trialled (see Chapter 9).

Making sense of the evidence

The initial source of evidence was an output from an earlier WS within the study, namely an internal review paper entitled ‘How best to deliver Comprehensive Geriatric Assessment (HoW-CGA): an Umbrella Review’ (see Chapter 2). The review summarised the available high-level evidence about clinical and organisational features of CGA, including the definition of CGA as a multidimensional, multidisciplinary process which identifies medical, social and functional needs, and the development of an integrated/co-ordinated care plan to meet those needs. Other key pieces of evidence included in the review and relevant to the design of the toolkits showed CGA to be more effective than approaches that did not include holistic assessment and management, and ‘ward-based’ (i.e. integrated) CGA to be more effective than ‘liaison (team)-based’ (i.e. peripatetic) CGA (see Chapter 2). This definition of CGA also implied features of CGA further detailed in the British Geriatrics Society’s good practice guide on Comprehensive Assessment of the Frail Older Patient:151

• CGA as multidimensional – an assessment to be comprehensive needed to cover five domains of a patient’s health and well-being, including

  1. medical

  2. mental health

  3. functional capacity

  4. social circumstances

  5. environmental domains.

• CGA as multidisciplinary – care assessment and management needed to be provided by a multidisciplinary approach delivered through MDTs. MDTs typically include

  • physicians

  • nurses

  • social workers

  • occupational therapists

  • physiotherapists

  • discharge co-ordinators

  • other relevant practitioners such as pharmacists.

• CGA supported by care planning and co-ordination – effective assessments need to be translated into care plans that deliver in a co-ordinated way.

The umbrella review and other sources of evidence, such as the Good Practice Guide152 and The Silver Book121 were specific about high-level features of the CGA process. They were, however, less specific about meso- and micro-organisational elements of CGA. At high level, the evidence suggested that to ensure comprehensive care and effective management of frail older patients hospital wide, acute services needed to move beyond the confines of peripatetic consultations with geriatricians, or confining patients to OPM wards. There was a need to delineate a future ‘third way’ model to mainstreaming key service-level competencies required to deliver high-quality care and management of frail older patients across wards and services.

Building on this starting point, early on, the team sought further sources of evidence about the following two themes:

1. how can CGA be best organised in an acute setting

2. how can CGA be best implemented in an acute setting.

In order to collect more in-depth and detailed knowledge relevant to these two themes, the team extended the scope of the evidence base to include:

• published case studies on implementing CGA in acute settings that were not specialised in older people’s care,152,153 of the kind sometimes characterised as ‘grey literature’ (usually non-peer reviewed, often not accessible via mainstream academic databases)

• knowledge of ‘good practice’ from expert focus groups (namely the study’s External Stakeholder Group, which consisted of representatives of all professions relevant to CGA), particularly where this was not available in the published or grey literature

• individual interviews with stakeholders with practical experience in designing, developing and delivering CGA in acute settings.

The last two sources were grouped under the term ‘consultations’ with the External Stakeholder Group and with others they identified as having relevant perspectives on this issue. Because the evidence base was characterised by relative agreement in relation to the macro-level organisation of CGA (that it is most effective if fully integrated in acute settings rather than offered on a peripatetic basis), and offered little information on the meso- and micro-level specifics of how to organise and deliver CGA, this consultation process did not incorporate a Delphi consensus-building activity, because there was no obvious subject of disagreement that required consensus development.

In November 2015 the team received a favourable University Ethics Committee opinion for interviews and focus groups with stakeholders, including clinical academics from a range of specialities, non-clinical academics and other experts in care of older people, namely managers and members of the study’s PPI group. In total, in undertaking this consultation work, the team organised two focus groups at two time points (four focus groups in total) involving 19 experts, and conducted six individual interviews with clinicians (five members of the External Stakeholder Group plus one other). These consultation exercises were supplemented by more informal, ad hoc communication with these stakeholders and others, such as clinicians in potential pilot sites and clinical and managerial colleagues of the team and the stakeholder group, who provided further insights into the challenges and opportunities that might be involved in mainstreaming CGA in settings that do not specialise in the care of older people. In addition, a further two focus groups were convened with the External Stakeholder Group in late 2017, after the piloting and evaluation described in Chapter 9 had been completed, to assist in analysing and interpreting findings from this work.

The initial two focus groups were organised in November 2015. Members of the External Stakeholder Group were invited to discuss their experiences with CGA. What we learned from the focus groups was confirmatory in relation to evidence from literature and novel in highlighting some issues for further exploration. For example, participants indicated that, in their experience, and confirming the view of the literature (see Chapter 2), providing advice on a peripatetic basis is a less effective way to deliver CGA to areas that were not specialised in older people’s care than shared care arrangements:

Liaison is very advisory whereas embedded [CGA] gives people a mandate to make decisions about people and follow them through autonomously without going back to the person who may have their name at the end of the bed. So if somebody has fallen over and broken their neck of femur, and they are under Mr Campbell, if you are liaison you might go in and just write a list to do a, b, c and d and you go back 3 days later. But if you are embedded you go in, you would write your list and then you would follow it through and you would have a team working with you.

Geriatrician

You go to a surgical ward where there’s different priorities . . . I can remember as a young registrar writing 15 recommendations in the notes [of an] elderly frail person and going back and finding that none of them had been taken up. So my senior took me aside and said, well, perhaps just try one at a time [laughs] But it is, it is very, very difficult, because these are not your patients, not your ward, you can [only] make recommendations.

Geriatrician

Participants also agreed that it was important to improve patient identification and, in surgery, the idea of a pre-operative initiation of CGA resonated well. Discussing the idea of developing a toolkit, participants suggested focusing on key nodes in the CGA process (e.g. identifying potentially frail patients most likely to benefit from CGA, including specific frailty screening tools), highlighting benefits of implementing CGA to target services (i.e. process outcomes), grounding the toolkit in accepted best practice guidance such as The Silver Book121 and breaking down the process and listing actions and skills needed in each stage of CGA. Participants also suggested that targeting commissioners and managers may improve understanding of the need for innovative services (and thus prompt encouragement, incentivisation and management interest at the system level), whereas targeting practitioners and those setting up CGA locally may provide enhancement to improvement initiatives (thus improving the resource base and probability of success of ‘bottom-up’, clinician-led work to improve the quality of care for frail older people). Thus, they began to identify the need for multiple ‘levels’ of intervention through any toolkit, a need that the team sought to address in toolkit design.

A picture started to emerge from the literature and consultations of local improvement initiatives that were likely to be voluntary rather than mandated efforts targeting individual services or pathways within hospitals. Innovating acute care for frail older people could suffer from a disconnection between evidence on the one hand and the prevailing way in which acute care for frail older patients is improved. Best practice knowledge, guidance and vision had been developed by the learned societies, and there seemed to be an appetite for improvement among practitioners. But, on the other hand, improvement had been slow and isolated owing to a dominant mode of innovation in which efforts to improve care in settings not specialised in older people’s care were developed primarily by lone enthusiasts with little infrastructural support – often specialists in OPM teaming up with non-specialists and mobilising networks of support both within and without individual services:

I thought I would try and change acute medicine from within. So I was actually appointed as an acute medic and . . . I spent about 2 or 3 years working on a variety of iterations of service to try meet the aim of improving outcomes [for frail older patients].

Physician

I think the challenges are probably the same everywhere. First of all it’s identifying a workforce who can deliver CGA, [you] have got the education and training to do it but then also [you] have [to have] the appetite to be doing it outside of traditional settings. And that does require quite a lot of perseverance and enthusiasm. It’s not just normal geriatric medicine . . . The challenge of trying to establish geriatric medicine in other settings . . . is hard work, and it is hard to win people over. You do have to end up spending far more time on that rather than being based on a geriatric medicine ward.

Geriatrician

Correspondingly, to overcome the fallibilities of such approaches, participants stressed that for CGA to be effective it needed to be ‘embedded’, or integrated, into the operational and clinical workings of a service. This mainly concerned the issue of control over the process of identifying potentially frail patients, over treatment where clinical decisions were made in conjunction with specialists (e.g. surgeons or oncology specialists), and how these were integrated into the care provided. To some, it appeared that the importance of integrating CGA into existing service provision had to an extent been recognised and that there was an opportunity in part because the range of people involved in medical decision-making had extended substantially in recent years. This brought advantages of multidisciplinary input into care (particularly of more complex populations, such as frail older people) and also meant that a ‘multidisciplinary mindset’ was increasingly valued and recognised by clinicians.

Expanding the scope of the toolkit design

At this point, both literature and consultations were inconclusive as regards who should be the target audience of the toolkit(s) or what should the toolkit(s) aim to achieve. The team needed to decide whether to target those seeking to realise change in practice at the ‘sharp end’ of care, most likely to be physicians in relatively senior posts, who may along the way communicate with other parties (e.g. management, clinical leads), or to focus their efforts on hospital management. A toolkit for the former audience could aim to assist their improvement efforts, and provide material that ‘lone enthusiasts’ might be able to use to embed their work. A counter-argument suggested that the toolkit should be aimed at management and commissioners to improve their understanding of care for frail older patients and generate their buy-in. Those at the sharp end, the argument implied, already have access to many of the basics that could be set out in a toolkit, for example in the form of guidelines and principles have already been published by various professional and learned societies. Therefore, an individualised approach to improving local services may not be aided by a generic toolkit.

To overcome some of these issues, in consultation with the stakeholders, the team proposed to expand the toolkit design simultaneously in two directions:

1. to adopt a multilevel approach to change that will recognise that there was a range of actors in different positions whose conjoint positive action may create synergy and ensure change more effectively

2. to identify and describe the set of key service-level competencies needed to deliver holistic assessment and management of frail older patients in a way that could be mainstreamed locally, in acute settings not specialised in the care of older people. These could then be turned into a self-assessment tool, and be made an integral part of a service-level toolkit.

The multilevel approach sought to move beyond the ‘lone enthusiast’ mode of innovation and, instead, adopt a multilayered approach. This approach stemmed from the view, strongly expressed in the consultations, that a whole-system approach would be more likely to be effective in tackling what is ultimately a whole-system problem. Changing service provision, and the demands associated with it, cannot be effectively achieved without the co-ordinated efforts of a range of actors at different levels. This, of course, is also in keeping with current research-based recommendations for achieving change,154–157 which identify the complementarity between approaches that focus on the bottom-up efforts and practical knowledge of clinicians and top-down expectations communicated through incentives, directives and policy recommendations.

Although six possible levels were identified in the course of the consultations, the team agreed to focus on four levels, in relation to which the study offered the greatest opportunity for influence. It was felt that educational and workforce aspects of change were beyond the abilities of the current project (Figure 16).

FIGURE 16.

Multilevel approach to improving acute care for frail older patients identified in consultation process. Service-level tools.

1. Strategic (regional) level: relevant strategic players, such as system resilience groups and leads of Sustainability and Transformation Plans (STPs), with CEOs of local provider and commissioning organisations as their members, should be made aware of problems in care that arise from suboptimal provision for older people, such as higher-than-expected volumes of attendance and high admission rates, length of stay, re-admission rates and institutionalisation. The purpose of a toolkit oriented towards this audience is to identify the potential solution offered by initiatives to improve care for frail older patients, and prompting system leaders to take action (e.g. to include service development in strategic planning, delegating responsibility for implementation to operational arms).

2. Operational (acute trust) level: managers are to be presented with convincing evidence of the problem, for example national reports from Royal Colleges, data from the NHS Benchmarking audit on acute care for older people, and the tools developed in the course of this study to present projected need and comparative outcomes between areas (see Chapter 7). This will help to identify opportunities for improved quality of care, so that divisional and service managers prompt service development across the hospital and provide support to improvement teams within their trusts.

3. Service level: activity will support clinician-led improvement teams that are seeking to implement CGA in services within hospitals. The toolkit will outline a journey of service innovation and guide local teams about what they should address and how they might address it on the way. It will help to build a case to persuade colleagues and relevant others about clinical benefits; it will include a self-assessment tool to help identify what competencies are already in place and what need development; it will bring together operational and implementation knowledge, and assist local improvement teams to communicate with others to increase buy-in. It will also offer a selection of clinical, improvement and evaluation tools.

4. Patient and carer level: this audience will be empowered to take a more active role in their care, if they wish to do so. Patients and carers may influence the ways acute services are provided locally and will be targeted by a specific intervention (e.g. an information leaflet or a video) to increase awareness about high-quality care for frail older people. In turn, patient and carers will be able to demand this care from their services. This was to be a toolkit that our PPI group would have an active role in developing (see Appendix 1).

Given the time frame of the current project, and the potential for early formative evaluation in pilot sites, the team decided to first focus on developing the third of these levels, the ‘service-level’ tools oriented towards clinicians seeking to improve quality and put CGA into practice in acute care services that are not specialised in older people’s care. As improvement requires a concerted effort from of a range of actors, not only those ‘on the ground’ but also actors located outside individual services, these service-level tools were subsequently supplemented by parts of the toolkit aimed at the other three levels (levels 1, 2 and 4; see Appendix 9). These aimed to target other audiences at the levels of strategic and operational management, and be a tool to help patients and carers to take an active role in their care and ensure that clinicians and their colleagues ‘on the ground’ were not improving care in isolation.

After a number of iterations, the first candidate version of the service-level tools for testing was finalised ahead of a stakeholder meeting that reviewed it in May 2016 and provided feedback via written comments and through the last two focus groups. The final version of the service-level tools, which was subsequently piloted in two hospitals (see Chapter 9), was submitted as part of an amended HRA ethics application for qualitative evaluation of the piloting process on 13 June 2016.

The service-level tools aimed to bring together a combination of clinical and implementation knowledge, together with concrete examples of the practical experiences of leaders in improving care for frail older people. It was structured into five ‘chapters’:

1. using data to identify problem and convince others about the solution

2. self-assessment

3. team approach to change

4. barriers to implementation

5. review, expansion and sustainability.

Four of these (chapters 1 and 3–5) were dedicated to the wider theme of service development, and its specific application in relation to the realisation of CGA. The second chapter presented a self-assessment tool, through which services could identify the competencies to which they already had access, and which they would need to develop, in order to deliver CGA. Each of the four service development chapters aimed to provide best evidence about key challenges in improving health care, resources proven to help in overcoming these challenges (referring to a great deal to well-established quality-improvement sources such as the NHS Change Model, the NHS Scotland Quality Improvement Hub and the Welsh 1000 Lives Plus), experiences of those who have successfully improved care for frail older people in acute settings across the country, and evidence-based guidance on the challenges involved in overcoming barriers to quality improvement (see Report Supplementary Material 2). 158

A key aspect of the service-level tools was the self-assessment tool (see Chapter 2), which was developed by the team based on existing guidance on the skillsets required for CGA to give improvement teams the opportunity to identify service-level competencies that were already in place and those in need of further development. It also offered recommendations about what to do when gaps in service provision were identified. Developed through the consultations with the professional bodies represented by stakeholder group members over the previous nine months (which included representation from all of the professional disciplines identified as contributing to CGA), it presented the structure and process involved in CGA, and a standard against which all services could assess current care provision, and identify areas for development. In this way, it sought to challenge local innovators not only to improve their service provision, but also to improve it to an extent they may not have imagined or deemed necessary, i.e. by providing an essentially gold standard model of CGA (and the competencies required) towards which to aim. Stakeholders involved in the consultation process indicated that in many areas, variations on the liaison model of CGA were under consideration. By presenting the competencies and changes required for integrated CGA, the tool sought to provide an alternative that might assist local initiatives to develop CGA in services not specialised in older people’s care in following the current best evidence for effective CGA (see Chapter 2).

Cycles of consultation and consensus building

The team found that there was little consensus in the literature and guidance about what exactly the competencies required to deliver CGA are, beyond the high-level specification of CGA’s structure and process. For example, the Silver Book contains a list of skills required in CGA. 153 However, the skills are generic and do not cover clinical practices across the five domains of CGA. To add value to the toolkit, and make self-assessment practical and concrete in terms of the gap between current and ‘best’ practice and what was needed to close this gap, the team sought to identify questions about service ‘competencies’ that improvement teams might ask themselves to assess whether or not they were delivering holistic assessment and management of frail older patients, and to map out gaps in provision. The self-assessment thus remained a distinct task for development, which required an iterative approach and consensus-building process involving a range of clinical professions – medical, mental health, social work, nursing, occupational therapy and physiotherapy. In the absence of a clear evidence base on this question, securing input from the stakeholder group – which included all these disciplines and, furthermore, the members of which had been nominated by their respective professional associations – provided the best way to build a toolkit that was both useful and grounded in the current thinking on good practice and on the practical steps involved in delivering beneficial CGA.

In the first instance, the team turned to stakeholder group members and asked them to consider the more specific level of competencies and/or interventions that embody high-quality care for frail older patients. It was stressed that these would need to go beyond the high-level specification of CGA (e.g. ‘CGA should include assessment of medical domain or functional capacity’), and identify exactly what ‘doing CGA’ implied in terms of the skill sets of practitioners who would be able to undertake effective assessment and management of frail older patients. The stakeholders were asked to suggest questions that they would ask if they wanted to check whether a service was assessing and managing frail older patients well. More specifically, based on their experience and reflecting their organisation, they were also asked what 3–5 questions they would ask to establish whether or not practitioners were doing the right thing in terms of the assessment and management of frail older patients.

The first call was circulated to External Stakeholder Group members in April 2016. This identified a good number of questions for the medical, mental health and functional domains, but largely left the others unpopulated. To obtain service competencies for social and environmental domains, and to validate the first draft of the self-assessment tool, the team extended the call to a network of clinicians across the East Midlands, beyond the study’s External Stakeholder Group. In total, a further 33 practitioners were targeted through this call in May 2016. During the second round of consultations, the team achieved a good representation of service-level competencies across all five domains of CGA.

Also in May 2016, the team organised the second set of two focus groups with members of the extended stakeholder group, who were invited to discuss the first draft of the toolkit. The group discussion looked at the multilevel approach, the structure and the content of the toolkit, and the shaping up of the self-assessment. Participants confirmed that the development was going in the right direction. They felt that the multilevel approach was a useful way of enhancing the success of improvement:

I thought it was a sensible approach to direct the different things you are publishing at the people that it is relevant to at different levels . . . I think it’s really great that you have got all four. So you may already have top level support, you may have people on the bottom and therefore you might need your patients and operational bit instead. So I think nationally you need all [those] bits but then locally people will have to decide what bits they need and they might pull various elements depending on what their local needs and people are.

Geriatrician

Overall I think it is a good length, appropriately pitched and the approach is positive.

Oncology specialist

I think this is coming together well. The design is clear, it takes you through the process easily. It looks as if it should be a very workable tool.

Medical oncologist

Overall, this looks really good and is easy to follow.

Pharmacist

Participants discussed what other groups and organisations might also be considered in use of the first and second levels (i.e. the strategic and operational tools, such as STP Boards and acute care commissioners at the regional level, and CEOs, Medical Directors, Director of Nursing, directorate leads and chief operating officers at the operational level).

The team received a wealth of comments about the service-level toolkit. Participants liked the prospect of a web-based toolkit, which was generally regarded as more user-friendly and accessible than a printed toolkit, and easier to update. The idea of dipping in and out of individual ‘chapters’ on implementation was also welcomed. Personal experiences mentioned in the toolkit in the form of case studies received more mixed feedback. Although used as illuminations rather than key reference points throughout the text, some felt that the comments were too ‘anecdotal’ and not ‘hard enough’, yielding only limited power to convince stakeholders that implementing a new service would give rise to tangible benefits. Others felt positive that the toolkit included the personal experiences of those who had already achieved positive improvement results. Overall, participants liked the combination of improvement knowledge applied to CGA, including details reflecting the realities of ‘improvement as practised’ identified in the social scientific literature on health-care improvement, such as a mention of ‘corridor discussions’, and the toolkit’s focus on sustainability in its fifth ‘chapter’. In their view, more could have been said in the toolkit on which outcomes were most useful in bringing CGA to the attention of colleagues and managers. Some would also have liked an extended methodology chapter. Participants discussed the point at which frailty should be identified and where the assessment process starts. Who was likely to be setting up the service, and the extent to which the toolkit was ‘doctor biased’ were also debated. Questions were asked about the possible role of other professions and groups in improvement teams, and leaving the decision to local sites. Although some suggested that the label ‘CGA’ had become unhelpful because of its perceived association with geriatric expertise, overall, participants felt that it was important to keep the label as it had become an established and widely understood term referring to a well-structured, known and validated model of care.

Stakeholders, and later participants in the piloting process (see Chapter 9), spent a great deal of time discussing the contents of the self-assessment chapter. In their view, the self-assessment was a particularly crucial part of the tool:

I thought it was a nice document that was easy to read and gave you a good focus on where to start your self-assessment [which] is the nuts and bolts of it, and to move on to what you want within your environment, and what a CGA will hopefully give to your environment. So in a sentence I thought it was very good.

Anaesthetist

They also stressed the importance of having all those CGA process items together in front of those doing the self-assessment:

The service competencies made a lot of sense. [They] integrated what we require. Because you’ve got to remember that most people have not ever in this environment seen a CGA . . . I went down all these, and there wasn’t a lot of things that we don’t do, which is good. But putting them together in this structure was very useful. So I think having the service competencies in the self-assessment tool is very useful.

Anaesthetist

Participants stressed the need to include a mention of person-centred care, and items such as pressure damage assessment, continence and the prominence of discharge-planning. Some felt that the functional section seemed to be falls dominated.

Negotiating test sites

In February 2016 the team began preliminary discussions with a number of prospective pilot sites about trialling the service-level tools. Ultimately, these gave rise to an agreement to pilot the service-level toolkit and formatively evaluate the efforts of clinical teams to introduce integrated CGA into their services, in three services within two sites.

In one site in the East Midlands, local partners included a surgeon, who was also a head of colorectal cancer surgery, and a clinical nurse specialist in surgery. They had already begun thinking about improving surgery for older patients. Their ideas included setting up a monthly clinic to refer frail older patients for a specialist OPM input to improve decision-making and optimise patients. A second service in the same trust, in the field of vascular surgery, also began to make to make use of the toolkit sometime later, and limited formative evaluation work also took place here (see Chapter 9 for more details).

In the second hospital in the north-east of England, the main contact was an anaesthetist who was also a clinical lead for pre-operative assessment. In his words, there was enormous interest in ‘frail elderly patients’ in perioperative services and in the Royal College of Anaesthetists. This service dealt with high numbers of older patients, of whom > 40% were cancer patients. In surgery, pre-operative optimisation was seen as an important task, allowing patients to be put on the surgical pathway. Local surgeons had already been liaising with specialists in OPM to initiate social and functional interventions. Their wish was to ‘have something to make this fast and effective’. Their initial idea was to introduce a screening tool to identify a population of frail older patients who will benefit from further input and have a list of interventions to improve functional outcomes. The opportunity to pilot the toolkit was thus a timely one.

In May 2016, the finalised version of the toolkit was circulated to the prospective test sites for early feedback. Local contacts were reminded that the toolkit was aimed at clinicians from services that were not specifically oriented to the needs of older people and should allow services to identify gaps in competencies and plan changes to ensure delivery of high-quality CGA (through the self-assessment tool), and support services on an improvement journey (through resources blending clinical and improvement knowledge). They were asked to share reflections on the early draft with the research team, bearing in mind that it was a work in progress.

After initial negotiations, the initial two services in the East Midlands and the north-east agreed to test the service-level toolkit. In these two sites the toolkit was seen as a potentially useful aid to local improvement thinking, and that was timely in that it coincided with their own windows of opportunity for service development. The East Midlands service had started a project through which they hoped to develop a holistic, streamlined, one-stop ‘preparation for surgery’ service for colorectal cancer patients. A multidisciplinary collaboration had been established between a cancer specialist, a consultant anaesthetist in charge of a high-risk colorectal anaesthetic clinic, and a surgeon. As many of their colorectal cancer patients were frail older patients who need more time to prepare for surgery, the idea of improving the process was seen as beneficial for patients and service alike.

Submission to the Research Ethics Committee and beyond

The team finalised and submitted the toolkit, along with documents relating to the formative evaluation in the pilot sites (observation guide, interview guide and Information leaflets together with a statement of activities of participating organisations), to the NHS East of England REC on 12 June 2016 for approval (as a substantive amendment to the study’s original ethics application). Favourable opinion was received on 1 July 2016. Pilot sites were informed about the success and, following research and development approvals, both services began to welcome the project team from October 2016 for initial interviews and observations as the formative evaluation commenced (see Chapter 9).

Following the completion of the formative evaluation work covered in Chapter 9, the External Stakeholder Group was reconvened for two final focus groups, building on the four carried out earlier in the process. They reviewed emergent findings from the fieldwork and considered the service-level toolkit (alongside the draft strategic, operational and patient- and carer-level tools) in light of these findings. They made further recommendations for further development of the tools and made suggestions for further work to refine and disseminate them, including use by frailty improvement networks and the British Geriatrics Society (see www.bgs.org.uk/resources/hospital-wide-comprehensive-geriatric-assessment-how-cga-overview). In addition, we are in discussion with NHS Digital about embedding the frailty risk score into NHS systems. Level 4 is insufficiently developed and the risks and benefits of its use unclear, so we do not feel that it is yet in a state to be widely disseminated (see Appendix 9).

Introduction

From October 2016 to September 2017, the research team piloted the so-called ‘service-level’ components of the toolkit in three sites in two NHS trusts, identified during the development of the toolkit as described in Chapter 8. Alongside this, the team conducted a qualitative study to evaluate formatively the pilots to further inform the development of the toolkit. This chapter highlights key findings from this evaluation. It provides an overview of the contexts of the services, their clinical foci, organisation, plans for change and anticipated use of the toolkit. It should be noted that the work described in this chapter is not an evaluation of the service-level toolkit per se; rather, it seeks to examine the opportunities and challenges involved in putting CGA into practice in acute settings that do not specialise in the care of older people. It summarises the key issues involved in this process and highlights important differences in stakeholders’ perceptions of the role of geriatricians in mainstreaming CGA. In discussing our findings, we draw on normalisation process theory to facilitate greater understanding of work to implement CGA in the wider context of improving health-care organisations.

Methods

Case study selection was partly purposive and partly based on opportunistic sampling. As described in Chapter 8, during the development of the service-level tools, the team engaged with a number of sites that had expressed an interest in making use of the toolkit as part of their own improvement plans. The services ultimately included were drawn from this sampling frame, with a view to putting the toolkit into use in services that varied in potentially important ways, for example clinical focus and professional leadership. 160 It is noteworthy, however, that all three services were actively seeking to improve the quality of their care for older people and were known to the investigators. As such, they arguably represent relatively receptive contexts for efforts to incorporate CGA in services that do not specialise in care of older people, with some willingness and enthusiasm for change.

We were principally interested in the work involved in local efforts to improve quality, how the toolkit fitted into these activities, and what gaps clinicians identified between their own aspirations and the changes put forward in the toolkit. We were interested in the process, challenges and opportunities involved in integrating more holistic care for older people, based on the principles of CGA, into areas that had their own pressures, existing processes, structures and norms, and potentially divergent expectations of the advantages that CGA might bring.

Data collection in the sites included ethnographic observations and interviews. Our observations mainly focused on the ‘backstage’161 work, whereby the new intervention (both the content of CGA and the toolkit as an artefact implicated in efforts to implement it) and concomitant changes in processes, systems and professional relationships were introduced and negotiated. We observed meetings where care improvements were discussed, colleagues and staff engaged, progress evaluated and next steps planned. Some ‘frontstage’ observational work of clinicians at work was also included. This involved clinicians testing new organisational arrangements (e.g. preparation for surgery clinic and handling new triaging tools) and discussing interdisciplinary aspects of clinical work with colleagues within and beyond their own professions. We did not observe any clinical encounters with patients. Topic guides used in the interviews, and as prompts for observations, drew on Normalisation Process Theory162,163 as a means of ensuring that our data collection was attentive to key conceptual issues involved in the introduction of new approaches into existing clinical and organisational environments, grouped under four generative mechanisms: coherence, cognitive participation, collective action and reflexive monitoring. ‘Coherence’ refers to the extent to which an innovation is understood as meaningful, achievable and desirable by participants. ‘Cognitive participation’ refers to the enrolment of participants and tools necessary to deliver the innovation. ‘Collective action’ is the work that brings the innovation into use and ‘reflexive monitoring’ is the ongoing process of adjusting the innovation to keep it in place. 164

Several visits were made to sites 1 and 2 (located in the East Midlands and the north-east of England, respectively) during the evaluation period between October 2016 and September 2017. A third site (in the same hospital as the first, but with a different focus – vascular surgery rather than cancer) was activated in the last 4 months of the pilot, and further limited data collection was undertaken here. In total we undertook 28 hours of observational work and 52 interviews across the three services (Table 40). Interviews targeted clinicians acting as leads in improving care, and their close collaborators from other clinical specialisms and professions including surgeons, anaesthetists, geriatricians, nurses, occupational therapists, physiotherapists and health-care assistants.

Interviews and ethnographic field notes were fully transcribed and analysed integratively using the constant comparative method. 165 Again, themes from normalisation process theory were used as ‘sensitising concepts’ to guide our analysis and ensure coverage of key issues known to be involved in the process of integrating new working practices into clinical environments,165 but we present our analysis and findings below under empirically rather than conceptually derived headings.

Findings

At sites 1 and 2, change was instigated by various clinicians involved in perioperative care. At site 1, the improvement lead was a colorectal surgeon working in close collaboration with an anaesthetist. At site 2, it was an anaesthetist – surgeons were not an active part of the improvement effort in this site. Leads in both sites developed improvement ideas prior to their encounter with the CGA toolkit. Initially, their thinking revolved around preparation for surgery, or what is known as pre-operative assessment, and its key question: should a patient be having an operation? Site 3 was a trust-wide vascular surgery service. The service lead here, a surgeon, was keen to try aspects of CGA to ensure that ‘the right person has the right operation’.

The findings presented here summarise the sites’ initial plans for improving care for frail older surgical patients, key challenges faced by clinicians, their use of the CGA toolkit, how their improvement projects related to other care pathways, and the different understandings of the role geriatricians should take in improving care. They suggest important challenges that will be vital to address in future efforts to ‘mainstream’ CGA, not least divergent opinions about whether such mainstreaming is ultimately achievable and sustainable without ongoing active involvement from specialist geriatricians.

Discussion

The evaluation of pilot sites’ efforts to incorporate CGA into their work summarised in this chapter shows that sites made only limited progress in implementing CGA during the study period. At the end of the pilot, participants across sites were in many ways still at the beginning of their work to identify patients and plan change. In part, this was because of the sheer volume of work involved in such an important change to the practices involved in the three sites, as well as the interaction with existing procedures, policies and norms, such as the cancer pathway and associated timelines, which loomed large in the thinking of many of those involved, which were not always synergistic. 170 It is noteworthy, however, that the seemingly ‘receptive contexts’ of the three sites, in which clinicians all shared an explicit, shared aim to incorporate CGA, and where there was a good history of multidisciplinary collaboration, did not guarantee smooth incorporation of the principles of CGA into routine practice. Indeed, at times, the apparent fit of the ambitions of CGA with wider ongoing work towards improving the quality of care and creating a more holistic approach to individual patients could actually work to marginalise CGA, with its particular, specific ambitions and (perceived) narrower beneficiary group.

Our data collection and analysis were guided in part by Normalisation Process Theory, using it as a sensitising tool to think through the challenges of change and improvement. 165 The four generative mechanisms articulated by the theory offer a matrix for investigating the collective work that may or may not lead to embedding a complex intervention in clinical practice. 170 The matrix of normalisation process theory provides a reminder that what may seem like modest results after 11 months of ‘mainstreaming’ CGA involved, in fact, a lot of action and practical interaction, much of which might risk being lost or written off if it is not explicitly accounted for. Across the sites, all four generative mechanisms of normalisation were widely represented. In many situations, more than one mechanism was active at a time. For example, collective action, cognitive participation and reflexive monitoring all propelled team meetings that typically enrolled participants and tools to assess, monitor and review, both formally and informally, the opportunities and challenges of implementing CGA in local services. Clinicians also engaged in making the innovation meaningful to themselves and others (coherence) in ongoing routine interactions with colleagues.

However, the success of these efforts was mixed, and, again, normalisation process theory offers some clues as to why this was the case. In particular, the competing definitions of the role of geriatricians that were evident in the sites highlights an important dynamic of negotiation, whereby the sense-making work of clinicians produced more than one viable way of constructing an appropriate vision of what ‘successful implementation’ might look like. As those differing sets of meanings continued to develop without effective reconciliation, fostering coherence and cognitive participation required further collective action and, in turn, further joint input into meaning-making. 106,165 As of the end of the evaluation period, this process was incomplete.

Context: acute hospital care for older people

Older people accessing acute hospital care often present with a combination of cognitive impairment, multiple comorbidities, polypharmacy and functional impairment. The accumulation of such deficits is one means of assessing frailty,171 an independent predictor of falls, delirium, disability, re-admission and care home admission. 106,134,172 Traditional medical intervention (diagnosis and medical treatment) is necessary but insufficient to optimise health outcomes for older people with these problems. A more holistic care model is required, a widely accepted definition of which is a multidimensional, multidisciplinary process that identifies medical, social and functional needs, and the development of an integrated/co-ordinated care plan to meet those needs, which conventionally is abbreviated to ‘CGA’, as in this report3 (see Chapter 2). Much of the evidence for CGA indicates that it is most effective when applied in discrete ward settings or specialised units. The aim of this research was to ascertain how best to identify frailty in acute hospital settings and to develop tools to help clinicians, service managers and commissioners deliver CGA for older people with frailty who are not on discrete wards.

Evidence base

Our literature review summarised the extensive evidence base for CGA and helped us to characterise new and emerging services that are beginning to deliver CGA on a hospital-wide basis. We confirmed a widely used definition of CGA (see above) and found that the trials and evidence syntheses regarding the use of CGA in acute hospital care to date have not usually selected older people for intervention using frailty risk tools as now defined. Rather, according to the conventions of their times, the early trials mostly selected people on the basis of age, with many studies using lower cut-off points of 60 or 65 years. In addition, they have tended to focus on clinical and operational outcomes that matter to clinicians and researchers rather than outcomes that matter to patients.

We identified a number of types of service that are beginning to deliver CGA where it has not traditionally been found, such as oncology, emergency medicine and perioperative care. However, the evidence base in these areas is still rather nascent and does not provide clear insights as regards the optimal methods to spread CGA across the whole hospital.

Evidence of variation in current practice

Our survey was one of the largest surveys of CGA in acute hospitals to date, yielding a rich database that described > 120 hospital CGA services. The findings are consistent with reports from NHS Benchmarking, in that where we asked similar questions in a different, but overlapping, sample of acute trusts, the responses were similar.

The key messages were:

• The use of frailty or other risk stratification tools was uncommon (30%).

• Multidisciplinary assessment and management was routine in wards specialising in OPM and orthogeriatrics wards, common in stroke and admissions wards, but less firmly embedded in other parts of the hospital system (such as medical and emergency units) and not usually found in surgical and oncology services.

• Clinical teams are mainly using informal as opposed to formal assessment processes (e.g. the use of structured tools to assess cognition or function).

Overall the survey describes an emerging landscape of hospital-wide CGA: we cannot say that patients with frailty are being identified, assessed and managed systematically, or targeted with CGA services.

Population segmentation

There is a growing body of evidence that shows the link between measures of frailty and patient outcomes and the cost of care, and a growing acceptance that the notion of frailty may be a better predictor of what happens to patients than chronological age. If we accept the idea that frailty is a helpful construct that can inform choices about care and treatment for individual patients, then, ideally, the information systems that we use to support clinicians should be able to capture the data required to measure frailty in ways that are consistent and robust. Such information can potentially have value at three levels:

1. micro-level – support to clinical decision-making; there is little consistency in the tools used to assess patients prior to CGA

2. meso-level – helping to develop the right fit of services within a given provider (e.g. where do we need to prioritise work on better CGA; what workforce is needed?)

3. macro-level – input to planning and allocation decisions looking at service provision. This could affect decisions about resource allocation to areas or (dis)investments in new service models or major infrastructure projects.

The development of better clinical information systems is a multimillion pound industry, yet, as the Wachter review120 observed, achieving success in major information systems is challenging, takes time and is not simply a case of technical change (buying new hardware), but requires a much wider range of actions:

We believe that the NHS is poised to launch a successful national strategy to digitise the secondary care sector, and to create a digital and interoperable healthcare system . . . The experience of industry after industry has demonstrated that just installing computers without altering the work and workforce does not allow the system and its people to reach this potential; in fact, technology can sometimes get in the way. Getting it right requires a new approach, one that may appear paradoxical yet is ultimately obvious: digitising effectively is not simply about the technology, it is mostly about the people.

Wachter review120 p. 6.

Crown copyright. This publication is licensed under the Open Government Licence v3.0 except where otherwise stated. To view this licence, visit www.nationalarchives.gov.uk/doc/open-government-licence/version/3/

In this study, we were interested in understanding the needs of older people who are frail, and the potential for altering services (such as introducing CGA) to meet these needs. We have shown how it is possible to generate population segments that identify subsets of the older population with distinctive patterns of hospital use. Although information based only on hospital events is limited, most of the population aged ≥ 75 years will have some encounter with a hospital that provides some basic information about the makeup of the local population. Over 30% of the ≥ 75-year-old population will have multiple emergency admissions in a 4-year period; this will be higher in more deprived areas. Through the lens of acute hospital care use, we have shown that it is possible to use a predefined list of diagnostic codes associated with frailty to identify people retrospectively after hospital admission. We subsequently showed that patients with a frailty flag attached are more likely in the future to have a hospital admission and much more likely to have greater overall bed use, with associated high costs.

Risk stratification

Using cluster analysis, we have been able to derive frailty risk scores for patients aged ≥ 75 years and group these into ordinal categories of risk that are linked to poorer outcomes for patients. In a national cohort of more than 1 million patients aged ≥ 75 years, those with high frailty risk scores had a 70% higher chance of inpatient mortality, six times the odds of a prolonged stay and a 50% increased odds of emergency re-admission within 30 days compared with those with low frailty risk scores. The hospital frailty risk scores correlated moderately with formal frailty rating scales.

Separate analyses with the clinical frailty rating scales confirmed that they can all identify cohorts of older people at increased risk of future hospital use, costs and mortality. This was also the case for our hospital frailty risk score when fully adjusted for other baseline predictors. The limitations of these frailty measures for use in all population types was, however, highlighted.

Both the population segmentation and risk stratification approaches can be applied to national data sets for provider- and area-based comparative analysis. Tools aimed at the hospital organisation or commissioning roles to show up similarities and differences in distribution of frail older people. Although this information can be useful, there are a number of challenges in exploiting these data to the full, which include:

• technically linking and bringing together sometimes disparate data sources

• access and information governance around sensitive information

• having skills to make sense of the data

• developing new data-capture technologies that allow for more clinical detail, better descriptors of patient/population health status and real-time analysis.

Costing of Comprehensive Geriatric Assessment

We attempted to describe the costs of CGA so that they could be incorporated into the interactive planning tool containing data on frailty prevalence and service utilisation. The direct costing exercise proved difficult to undertake, primarily because of the difficulty distinguishing the costs of CGA as a process of care from usual care. Instead, we undertook a costing exercise using clinical vignettes, which estimated the cost of CGA to be around £90–172 per patient, over and above usual care costs of £2000–3000 per admission. Although this estimate is subject to multiple caveats, it indicates that, broadly, CGA is unlikely to add a significant financial burden over and above usual care – reflecting the clinical model of CGA, which is about changes to the process of care, supplemented by some additional expertise, rather than a whole new structure.

Hospital-wide Comprehensive Geriatric Assessment toolkits

The implementation WS of this research had the bold ambition of incorporating CGA as a process of care into non-geriatric services, in this example, oncology and surgery. The evaluation of pilot sites’ efforts to incorporate CGA into their work showed only limited progress during the study period. In part, this was because of the sheer volume of work involved in such an important change, as well as the interaction with existing procedures, policies and norms – such as the cancer pathway and associated timelines. It is noteworthy, however, that the seemingly ‘receptive contexts’ of the three sites, in which clinicians all shared an explicit aim to incorporate CGA, and where there was a good history of multidisciplinary collaboration, did not guarantee smooth incorporation of the principles of CGA into routine practice. Indeed, at times, the apparent fit of the ambitions of CGA with wider ongoing work towards improving the quality of care and creating a more holistic approach to individual patients could actually work to marginalise CGA, with its particular, specific ambitions and (perceived) narrower beneficiary group. In particular, the competing definitions of the role of geriatricians that were evident in the sites highlights an important dynamic of negotiation, whereby the sense-making work of clinicians produced more than one viable way of constructing an appropriate vision of what ‘successful implementation’ might look like.

Clinical care for older people with frailty

The technology of CGA is not new, and its effectiveness over the ≥ 50 years since its inception4 has been well evidenced. Given the rapidly increasing ageing population, with the associated increase in frailty shown in our study and others, the poor outcomes experienced by older people with frailty in a range of settings (not least acute care), it is timely to ask why CGA has not become more widely adopted already on a hospital-wide basis and (for some of us) difficult to understand why not. Although teaching about ageing and medicine in old age is included in most UK medical undergraduate curricula,173 it does not receive the prominence it merits on the basis of population demography. Similarly, post-graduate training in multidisciplinary care and management for older people is relatively underdeveloped. Recently, novel ways of integrating multidisciplinary approaches tailored to the needs of older people into settings and specialities where they have not traditionally been found, such as emergency medicine, are starting to develop,174 and become visible through (for example) our survey and the work of NHS Benchmarking.

Medical care for older people has not traditionally been high profile, often being described as a ‘Cinderella’ specialty. Perhaps this relates to the juxtaposition of old age and the end of life in many people’s minds, evoking thoughts of futility and decrepitude, and the associated difficulties this brings in choosing and using a consistent and inclusive language for the discussion (see Chapter 1). We prefer the view that ‘The measure of a society is found in how they treat their weakest and most helpless citizens’, a sentiment that was shared by (and variously attributed to) many of the twentieth century’s most distinguished leaders (Mahatma Gandhi, US Vice President Hubert Humphrey, US Presidents Jimmy Carter and Harry Truman, Winston Churchill and Pope John Paul II, among others). This view regards later life (and all of its uncertainties and challenges, including disability and death)175 as a valued part of life’s rich tapestry, with a wide range of lived experiences including healthy active components and variability in the rapidity and nature of the inevitable terminal decline. 176,177 Society at large does not yet fully embrace disability, death and dying, so it is unsurprising that health-care practitioners, who enter their professions to ‘save lives’, are no different. If we can move society into an understanding of the limitations of modern medicine and clinical care, then we can start to have honest conversations that allow a focus on ‘care not cure’.

More specifically, older people with frailty pose a challenge to clinicians’ diagnostic and therapeutic approaches: the clinical presentations associated with frailty may be due to a plethora of conditions,177 not easily identifiable because of the lack of ‘classic symptoms’. Therapeutic strategies are often confounded by comorbidities (e.g. certain treatments for urinary incontinence can cause confusion). This gives rise to clinical uncertainty,178 which is characterised by risk (indeterminacy of future outcomes), ambiguity (limitations in the reliability, credibility, or adequacy of information) and complexity (features of information that make it difficult to comprehend). 179 This broader perspective provides some insights as regards the apparent limitations of the toolkit approach that was adopted in this programme of work. The limited progress of highly trained surgeons and oncologists in developing active ownership of caring for frail older people is better understood given their relative lack of training in older people’s care, which is compounded by the complexity frailty presents and which is in contrast to the relative certainty and precision that characterises their practice.

Implications for health services

The finding that surgical teams felt a strong need to include geriatricians in direct clinical care, as well as service design, is important. There are insufficient geriatricians to manage all frail older people now, and this gap will widen as population ageing continues to increase. We have the opportunity to test the HoW-CGA toolkit in other specialised services (renal dialysis, chemotherapy, emergency cranial neurosurgery, interventional cardiology, complex spinal surgery and adult critical care) as part of work commissioned by NHS England, which started in 2019. For some of these specialist areas (neurosurgery, interventional cardiology, complex spinal surgery), it would be surprising if the outcomes will be vastly different from those seen so far. Others might be more attuned to interacting with frailty (renal haemodialysis and critical care); it will be important to evaluate this further iteration of the HoW-CGA toolkit.

Aside from specialist services, many frail older people receive their acute hospital care in general medical settings. Although clinicians in this setting typically possess many of the competencies to care for older people with frailty, our survey findings would suggest that there is significant variability in the delivery of frailty attuned care. Interventions aimed at enhancing the clinical skills, and particularly the organisational elements of care in this setting, are likely to be more feasible and potentially to have a greater impact. For example, in a hospital admitting 1000 older people per month, around 200 would be classified as severely frail, in whom the application of CGA might result in 12 more people being alive and in their own home, and 40 fewer people being admitted to long-term care.

There are important messages in this study for educational bodies, such as Health Education England and the General Medical Council (which regulates undergraduate education), which may wish to reflect on current and future training programmes (see dissemination plans below).

Apart from education and training, there are other mechanisms by which health services research can influence practice, for example outcome measurement. PROMs can be a powerful mechanism through which to change practice and focus care on outcomes of greatest importance to patients. 180 At the individual patient level, PROMs can drive improvements in diagnosis, communication and prioritisation of needs. 181 At the population level, PROMs can be used for research, benchmarking and feedback to providers to inform service improvements. 182 There is no existing, evidence-based outcome measure for older people with urgent care needs183,184 nor any systematic process to embed such an outcome measure. However, an outcome measure that leads to a focus on ‘what matters to the patient’ rather than what is the matter with the patient (cf. 62-day cancer target) could be very powerful.

Medicine for older people and broader care models

For the past few years the NHS has been looking at different ways to deliver care. The Five Year Forward View (NHS England) has led to a series of ‘vanguard’ projects that are seeking to redefine historical organisational boundaries. 185 Before these, there were pleas for models of integrated care that sought to look at patient care across traditional historical divides – whether from primary to secondary care or between health and social care. 186,187 Underpinning these ideas was a recognition of the growing importance of managing long-term care needs and of getting better at preventative management. 188,189 Part of the story for many of these was the need for better multidisciplinary assessment of individual care needs. These are just the sort of assessments we observed in CGA – so there is a sense that CGA should fit into emerging models of care for older people. The NIHR has recently published a themed review on this topic. 190

Summary

In summary, we have confirmed that CGA is an effective, and quite possibly cost-effective, care model for older people with frailty. We have evidenced significant variation in practice across the country, which implies that there is potential for improvement. The development and validation of the hospital frailty risk tool is powerful, as for the first time it can potentially make frailty and its consequences routinely and systematically visible, at the individual patient, service, hospital and commissioning levels. There are parallels with the community-based eFI123 and how identification of frailty in primary care settings is already starting to influence care provision for older people (e.g. mandatory medication reviews for the most frail). There is potential to develop a similar momentum in secondary and integrated care using (for example) large-scale improvement networks to reduce unwarranted variation, and to further develop the uptake of CGA on a hospital-wide basis, for example through the Acute Frailty Network. 191,192

The HoW-CGA toolkit can be used to activate health services to respond at multiple levels (strategic, operational and clinical) as well as, subject to further evaluation, to activate patients, carers and the public. Describing the limitations of the HoW-CGA toolkit has allowed us to highlight some of the avenues to address these issues, providing a road map to make hospital care fit for frail older people now and in the future.

Conclusions

Older people in acute hospitals are at high risk of poor outcomes, which can be improved through the delivery of specialist geriatric care in dedicated ward areas. The optimal method to deliver such care across the whole hospital is unclear. Current service provision is patchy, poorly standardised and, in surgical and oncology settings, does not usually involve teams specialised in older people’s care.

Most older people will come into contact with acute hospitals over a 4-year period, with more than one-third having an emergency admission. It is possible to use predefined lists of diagnostic codes associated with frailty to identify people retrospectively following hospital admission. Patients with frailty are more likely to have a hospital admission and much more likely to have greater overall bed use.

A frailty risk score derived from routine data was tested in more than 1 million patients. Those with high frailty risk had 70% higher odds of inpatient mortality, six times the odds of a prolonged stay and 50% increased odds of emergency re-admission within 30 days than patients without features of frailty. Although predictive at the group level, the ability of the risk categories to discriminate between individuals with different outcomes was low.

Clinical toolkits designed to empower non-geriatric teams to deliver CGA were received with initial enthusiasm but did not fully achieve their stated aims as a result of the need for an extended period of service development and geriatrician support, and competing priorities.

Further research is needed. A case for the further development and evaluation of tools to assist in the delivery of CGA on a hospital-wide basis can be made. Further trials would be justified and should be stratified by frailty, use PROMs and collect sufficient economic data to determine cost-effectiveness. RCTs of hospital-wide CGA must almost inevitably use the hospital as the unit of randomisation (or else it is not hospital-wide) and this pushes the limits of the feasibility of the methodology. Such trials will need careful process evaluations embedded within them in line with current research frameworks for the evaluation of complex interventions. 101

Organisations

• British Geriatrics Society (BGS).

• HoW-CGA PPI network.

• Emergency Care Intensive Support Team (ECIST).

• Acute Frailty Network, NHS Elect.

• College of Occupational Therapist (COT).

• Royal College of Nursing (RCN).

• Association of Directors of Adult Social Services (ADASS).

• Systematic Care for Older People in Elective Surgery (SCOPES), Nottingham University Hospitals.

• Royal College of Physicians (RCP).

• Royal College of Emergency Medicine (RCEM).

• Physiotherapists working with older people AGILE, Chartered Society of Physiotherapists.

• Royal Pharmaceutical Society (RPS).

• Society for Acute Medicine (SAM).

• Royal College of Surgeons (RCS).

• Association for Cancer Physicians.

• Royal College of Anaesthetists (RCA).

• Ageing Speciality Group, NIHR.

External Stakeholder Group

• British Geriatrics Society: Gill Turner, Jo Preston, Shane O’Hanlon, David Oliver, Adam Gordon, Rob Morris, Laura Daunt, Andrew Williams, Jugdeep Dhesi, Sarah Turpin, Matt Thomas, Recia Atkins, Colin Nee, Jo Gough, Geraint Collingridge and Aamer Ali.

• Lay representative, PPI Network: George Wood.

• Emergency Care Intensive Support Team: Russell Emeny and Jack Hawkins.

• Pip Logan, Occupational Therapist; Professor of Rehabilitation Research, College of Occupational Therapists. British Association of Occupational Therapists.

• Helen Lyndon, Nurse, Consultant Older People, Clinical Lead Frailty, RCN, NHS England.

• Grainne Siggins, ADASS.

• Geraint Lewis, Chief Data Officer, NHS England.

• Raymond Jankowski, Head of Healthcare Public Health, Programme Improvement and Delivery, Health and Wellbeing, Public Health England.

• SCOPES team members: Heather McCormack, Elizabeth Bailey, Sarah O’Leary, Emma Clarke, Samantha Sheriff, Angela Smith.

• Barry Evans, Health Education East Midlands and BGS.

• NHS Benchmarking Network: Debbie Hibbert and Leigh Jenkins.

• Royal College of Physicians: Ian Bullock, David Hunt, Susan Latchem, Aimee Protheroe & Mark Temple (Future Hospital Programme).

• Helen Dobbin.

• Jane Brothers.

• Gordana Babic-Illman, Guy’s and St Thomas’ hospital.

• Royal College of Emergency Medicine (RCEM): Jay Banerjee.

• Louise McGregor, AGILE.

• Heather Smith, The Royal Pharmaceutical Society.

• Society for Acute Medicine: Mark Holland, Lauren Wentworth, Olivier Gaillemin, John Soong.

• Royal College of Surgeons: David Ward.

• Royal College of Psychiatrists: Hari Subramaniam.

• Adonika Brown.

• Anne Thomas, Association for Cancer Physicians.

• The Royal College of Radiologists (RCR): Sally Appleyard.

• Lisa Barrott, Royal Sussex County Hospital.

• Chris Snowden, Royal College of Anaesthetists.

• Andrew Clegg, University of Leeds.

Study Steering Committee

• John Gladman (chairperson), Nottingham University.

• Marcel Olde-Rikkert, Raboud UMC, Nijmegen.

• Anette Ranhoff, University of Bergen.

• Gillian Parker, Social Policy, University of York.

• Nigel Rice, Health Economics, University of York.

• Sasha Shepherd, Health Services Research, Oxford University.

• Tom Bryden and Lynne Corner, Voice North.

Hospital-Wide Comprehensive Geriatric Assessment patient and public involvement network

• Anna-Teresa Lefort.

• Anthony Locke.

• Brian Todd.

• Cynthia Conrad.

• Deb Tanner.

• Diana Robinson.

• Donald Robertson.

• Edmund Brooks.

• Elizabeth Sclater.

• George Wood.

• Gerry Bennison.

• Gerry Freedman.

• Harbhajan Singh.

• Heather Redshaw.

• Helen Edwards.

• Irene Richards.

• Jacqui Gath.

• Jennifer Bostock.

• Jeremy Dearling.

• Jim Harding.

• Joya Morris.

• Julian Paul Ash.

• Margaret Hence.

• Margaret Ogden.

• Marguerite Beard-Gould.

• Maureen Godfrey.

Contributions of authors

Professor Simon Paul Conroy had overall responsibility for preparing, editing and finalising the report.

Dr Martin Bardsley led the population modelling (Chapter 4).

Dr Paul Smith undertook the initial cluster analysis and development of the hospital frailty risk score.

Dr Jenny Neuburger led the development and validation of the hospital frailty risk score.

Ms Eilís Keeble led on the validation of the hospital frailty risk score.

Ms Sandeepa Arora led on the costing of CGA.

Mr Joshua Kraindler led on the development of the level 2 toolkit and supported the validation of the hospital frailty risk score.

Dr Cono Ariti led on the initial cluster analysis.

Dr Chris Sherlaw-Johnson contributed throughout on the big data analyses.

Professor Andrew Street contributed throughout on the big data analyses and supervised the economic analysis.

Professor Helen Roberts supported throughout the project, in particular on the survey work and development of the HoW-CGA toolkit.

Dr Sheila Kennedy led on the PPI elements of the project.

Professor Graham Martin led on the development of the HoW-CGA toolkit with Dr Kocman.

Ms Kay Phelps supported all aspects of the study including interviewing and interpretation of the interviews.

Ms Emma Regen supported all aspects of the study including interviewing and interpretation of the interviews.

Dr David Kocman led on the development of the HoW-CGA toolkit with Professor Martin.

Dr Patricia McCue led on the literature review and survey work.

Dr Elizabeth Fisher further developed and validated the level 2 toolkit.

Professor Stuart Parker provided strategic leadership to the whole project and led on the survey work and literature review with Dr Patricia McCue.

Publications

Gladman JRF, Conroy SP, Ranhoff AH, Gordon AL. New horizons in the implementation and research of comprehensive geriatric assessment: knowing, doing and the ’know-do’ gap. Age Ageing 2016;45:194–200.

Conroy SP, Turpin S. New horizons: urgent care for older people with frailty. Age Ageing 2016;45:577–84.

Parker SG, McLeod A, McCue P, Phelps K, Bardsley M, Roberts HC, et al. New horizons in Comprehensive Geriatric Assessment. Age Ageing 2017;46:713–21.

Parker SG, McCue P, Phelps K, McCleod A, Arora S, Nockels K, et al. What is Comprehensive Geriatric Assessment (CGA)? An umbrella review. Age Ageing 2018;47:149–55.

Kocman D, Regen E, Phelps K, Martin G, Parker S, Gilbert T, et al. Can comprehensive geriatric assessment be delivered without the need for geriatricians? A formative evaluation in two perioperative surgical settings [published online ahead of print April 22 2019]. Age Ageing 2019.

Data-sharing statement

All data requests should be submitted to the corresponding author for consideration. Access to available anonymised data may be granted following review.

Patient data

This work uses data provided by patients and collected by the NHS as part of their care and support. Using patient data is 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.

Patient and public involvement in the project: who, where, how many and doing what?

More than 50 older people living in England and Scotland contributed to the project as PPI research volunteers. The volunteers ranged in age from the mid-50s to the early 90s, with both genders equally represented. Participants work, or used to work, in a broad range of occupations and professions, with many from health and social care backgrounds.

Some were members of established NIHR PPI research groups based in their local area (e.g. Leicester and Rutland PPI NIHR Ageing Specialty Group; Voice North in Newcastle); more than 30 were individuals with varying amounts of experience of PPI who were recruited directly via the PPI leads doing the following:

1. posting about the project and requesting PPI volunteers on the NIHR People in Research website (see www.peopleinresearch.org)

2. distributing the project description and PPI recruitment poster/flyer at the NIHR INVOLVE conference, 26–27 November 2014, Birmingham

3. Making e-mail contact with Sheffield-based community action groups involved in campaigning around a medical condition or ageing-related topic, with potentially interested individuals suggested by the Age UK (London, UK) engagement team and researchers involved in the New Dynamics of Ageing research programme

4. making a personal invitation to contacts known through the PPI lead’s previous research involving PPI volunteers and other work with older people

5. engaging with the NIHR Ageing speciality group, with a specific aim of establishing aligned PPI groups in region.

Around 50 volunteers from around the UK participated in the project (see Table 12).

The types and frequency of involvement

Between them, the volunteers were asked to contribute to 19 different research tasks and activities organised during the project. Individual volunteers were involved during the project from once up to 12 times (Table 41). The main reasons for the variation were the individual’s specific knowledge and skill-set and/or relevance of her/his personal or professional experiences to the task, along with the volunteer’s interest and availability at the time of the involvement opportunity. In addition, some volunteers contributed more often as they got involved in the project much earlier than others and, finally, reflecting the success in recruiting so many volunteers from so many areas of the country (Figure 17), some inevitably live in places less accessible or further from the project sites used for PPI meetings (mainly at the University of Sheffield, the PPI lead’s base, with a few in Leicester and a couple in London and Newcastle).

FIGURE 17.

Locations of PPI volunteers.

Some project volunteers’ reflections on their patient and public involvement experiences

The following outlines one participant’s involvement with the CGA research project:

I got involved because I was interested in improving the experiences of older people in hospital. During the project I commented on some draft documents being prepared for an ethics meeting. I also attended two meetings about the statistical work being done by researchers at the Nuffield Trust. It was illuminating to learn that in some areas of some hospitals a Comprehensive Assessment is being carried out with older people, but it is not being done in others. At the meetings I attended we were equally concerned about providing the best quality services to older people as this aspect often seems to be put aside by the NHS – probably due to limited resources. Learning that the patients, assessments and care being studied in this project were in relation to what is being done in hospitals, and because of my past experience of work in social care, I mentioned that it is important to recognise that some people may be able to have their health and life enhanced by co-operation between the NHS and Local Authority social care departments, although this is rarely mentioned. Since attending the meetings I tried to find out if my local Hospital Trust carries out the Comprehensive Geriatric Assessment but met with some difficulty and still am awaiting a reply. Perhaps encouraging some organisations like the National Pensioners Convention to seek out this information nationally would be of benefit to getting it taken up.

The following outlines one participant’s involvement with reviewing documents for the Research Ethics Committee and attending meetings with Nuffield Trust researchers:

The PPI information sheet and consultation sheet were fairly standard and explanatory about the need to understand how the frail and elderly can be treated in specialist services. However, they raised issues about not being able to erase data if people leave the study. The data should be anonymous and stored securely. The lay summary was much harder to understand in relation to the use of large data sets and the role of experts in analysing the linkages of data because the research methodology is assumed to be scientific and thus outside the realm of a lay person’s understanding. It would have been better to explain the value of large data sets in lay terms and what linkages in the data the researchers were looking for that would explain the value of this research more precisely and explicitly. It struck me that the lay summary was a work in progress and I would like to have seen the final version based on the view of lay members’ feedback. I found the meetings at the Nuffield Trust most challenging and problematic as to the validity and reliability of using routine hospital records in establishing what ‘good care’ looks like for a health population with specialist complex needs, especially with regard to the coding of data and inevitable variations in the quality of the data from different hospitals. In my experience, people with dementia or learning difficulties are not always identified (show up in routine hospital records). They have particular needs that need to be recorded and analysed before one can say what ‘good’ looks like for this group of patients. I would very much like to know the results of this work being done by the Nuffield Trust, when it has been completed. Will the results lead to other NHS specialists such as surgeons and oncologists treating older people to routinely identify those with frailty and establish how best to treat and care for them?

The following outlines one participant’s involvement with co-producing a conference poster:

For the Gerontology Society biennial meeting at Brighton on October 15th 2015, the PPI work stream lead suggested a poster about the project and its PPI should be produced as a group endeavour. Those involved were from Glasgow, North and South Yorkshire and East Anglia, which complicated some of the practical and decision-making aspects of the process. Firstly we identified a few basic principles: the objective of the poster, the target audience – lay, professional, or both – the main message, what other information to incorporate and how. We had many points of view and questions and the amount of information to consider was considerable. How should we decide what to include? We recognised that the colour scheme, font style and size were important considerations especially for people with impaired vision or colour blindness. Image and font quality were also important, so all graphics used were of high resolution. With the impressive number and wide distribution of the people engaged in the project and number of project sites we decided a map would be a good image to use but found it difficult to agree which was best. The work stream lead suggested inserting smiley faces to indicate volunteers’ locations.

The following outlines one participant’s involvement in the External Stakeholder Group:

I was asked by the PPI work stream lead to be the patient/public representative on the national External Stakeholders Group (the Group) of the project. The Group consisted of representatives of many different professions/organisations, along with members of the research team. Being the only lay person in the Group could have been quite daunting and challenging, but I soon recognised (and have reflected on) how warmly I was received by the Group. The Group leaders always made me feel valued – this was important in terms of encouraging my participation. Four meetings were held between July 2015 and March 2017. There were also occasions when responses were required to questions and pieces of work between meetings and I was very much aware of my responsibility to respond as the sole lay person in the Group. As a lay person you can sometimes bring, depending on your own experiences, a different perspective to such a Group. You can also bring ‘validation’ in relation to a particular approach being described or considered. As I have also witnessed on other research projects, real value can come when researcher, professional and lay views are shared and developed together, as happened in this Group. A question that might arise is whether, when you are the only lay person involved in this type of group, you should be consulting with other lay people. I took the view ‘not’ because of the dynamic nature of the Group, but nonetheless it is a legitimate question that perhaps warrants discussion.

Patient and public involvement evaluation

At regular intervals throughout the project, a questionnaire about the experiences, impact and outcomes of PPI activities was e-mailed to the project team and PPI volunteers who had been actively involved (as individuals) in PPI related to the project. The PPI evaluation form was adapted by the work stream lead from a pre-existing pro forma produced by the NIHR Greater Manchester Primary Care Patient Safety Translational Research Centre.

The questions asked were the same for both the PPI volunteers (VRs) and professional researchers (PRs), but the number of recipients varied for the eight rounds of the evaluation, depending on each individual’s experience of PPI since the previous round. A total of 119 forms were sent to the PPI volunteers, with 64 forms completed and returned. A total of 84 forms were sent to the professional researchers, with 35 completed and returned. The average response rate for the PPI volunteers was 60% (ranging from 20% to 100%); the average response rate for the professionals was 36% (ranging from 33% to 58%).

Results

Nearly all the volunteers said that they wanted to continue their involvement in the project and be involved in further PPI activities:

The feedback from me and others will have made the questionnaire more user friendly, and possibly more focused. I find it hard to say what impact it has for me, as I am involved in PPI in a variety of ways: generally speaking, I am pleased to have the opportunity to use my brains and communication skills.

VR 19

A large number of the volunteers’ responses and a few of the professionals wrote about the positive impact of PPI on the researcher. An impact commonly identified by the volunteers was about the enjoyment they derived from undertaking the task. Several volunteers and professionals wrote about the impact of PPI on the research process and richness of ideas incorporated with some extending the scope of its impact to potential improvements in the NHS.

Discussion

Overall, many positive comments about experiences of PPI and its benefits to the research were provided by the respondents. Several professional researchers expressed how much they both value and appreciate the volunteers’ input into the project. Most of the volunteers wrote about how they have enjoyed being involved in the project and felt their views and opinions were valued.

However, some concerns and a few criticisms from a small number of the volunteers indicate a need for improvements, especially with regard to volunteers being provided with more feedback on their input, more information prior to meetings and more information about the progress of the project as a whole.

Patient and public involvement research network in Ageing Research

The PPI research volunteers involved in the HoW-CGA project were contacted by the PPI lead in late April 2017 asking them to inform her whether or not they would like to remain involved in NIHR Age and Ageing-related health research, for example by contributing to the development of local and or national research project ideas, co-writing funding applications and/or the delivery of funded projects.

The mechanism agreed for this process is that the PPI lead would link the interested PPI volunteers to the lead for the NIHR CRN ASG and from there each volunteer would be connected to their geographically relevant LCRN lead in the ASG network.

As of 1 June 2017, nine of the volunteers had informed the PPI lead of their wish to stay involved in NIHR Ageing Research.

Conclusion

More than 50 older people, as PPI research volunteers from many different locations in England and Scotland, were involved in the HoW-CGA project.

A few were involved from the start, when project ideas and approaches were first being developed, whereas others got involved later and assisted the delivery of the numerous tasks and activities associated with the four research work streams and the evaluation of the PPI work stream. Between them, the volunteers undertook or commented on 19 research-related tasks, activities and plans (at home or in group meetings), with one also being involved in the Study Steering Committee and another in the External Stakeholder Group.

The value of PPI to the project cannot be quantified easily, but it is very clear from comments made at project meetings, by e-mail and on the evaluation forms, that the researchers and managers of the project as a whole have positive views of its contribution.

Cochrane Database of Systematic Reviews

Date range searched: January 2005 to February 2017.

Date searched: 22 October 2015.

EMBASE

Date range searched: January 2005 to February 2017.

Date searched: 22 October 2015.