Developing a multidisciplinary rehabilitation package following hip fracture and testing in a randomised feasibility study: Fracture in the Elderly Multidisciplinary Rehabilitation (FEMuR)

Phase I: developing the intervention

1. To undertake a realist review to identify the important components of a multidisciplinary rehabilitation programme following surgical treatment for hip fracture in older people and to understand the mechanism, context and outcome of successful interventions.

2. To assess the current provision of rehabilitation programmes following hip fracture surgery in the NHS throughout the UK.

3. To assess the views of patients, their carers and health professionals in multidisciplinary rehabilitation teams on the rehabilitation that they received or provided following surgical repair of a proximal hip fracture; how programmes could be improved; and the findings from the realist review and survey.

4. To design a rehabilitation programme based on the findings from the realist review, survey and focus groups.

Phase II: feasibility study

1. To assess the feasibility of a future definitive RCT by assessing the number of eligible patients, monitoring recruitment and retention rates and exploring the willingness of patients to be randomised and the willingness of patients and carers to complete process and outcome measures.

2. To produce means and standard deviations (SDs) of the quantitative measures so that effect sizes can be calculated for planning the future RCT.

3. To assess the acceptability of, and compliance with, the rehabilitation programme among patients, carers and clinicians and the fidelity of its delivery and to identify any adverse events (AEs).

4. To explore the methodological issues associated with conducting an economic evaluation alongside a future RCT.

Rationale for the review

This realist review, along with a national UK survey of current rehabilitation practice and focus groups with patients, carers and multidisciplinary rehabilitation teams, was performed to inform the development of an enhanced rehabilitation programme following proximal femoral fracture.

Objectives and focus of the review

The main objective of this review was to identify the important components of a multidisciplinary rehabilitation programme following surgical treatment for hip fracture in older people, in particular to distil and understand the evidence relating to how successful interventions work, in which setting and context, for which outcome and in which group of patients.

Research questions

1. What community-based multidisciplinary rehabilitation programmes have been developed and what were their main aims (intended outcomes)?

2. What were the mechanisms by which community-based rehabilitation of hip fracture patients is believed to result in its intended outcomes?

3. What are the identified contexts that determine whether different mechanisms yield intended outcomes?

Given the evidence in response to questions 1–3 we also drew conclusions regarding the following questions.

1. In what circumstances are the rehabilitation programmes likely to be clinically effective and cost-effective if implemented in the NHS?

2. In what circumstances and with which combination of mechanisms and contexts are the rehabilitation programmes likely to generate unintended effects or costs?

Rationale for using realist synthesis

A realist review was undertaken to identify suitable components for an enhanced multidisciplinary rehabilitation programme following proximal femoral fracture. Such rehabilitation programmes are complex interventions because they are multifaceted and interact in complex ways with many contextual factors39 (see Appendix 1). Compared with systematic reviews, realist reviews aim to build a deeper understanding of the mechanisms behind an intervention and to identify ‘what works, for whom, in what circumstances and why’. 39,40 Whereas conventional systematic reviews judge the overall effectiveness of an intervention and pay less attention to context, realist reviews attempt to explain mechanisms by which interventions produce different patterns of outcomes according to different contextual factors (see Appendix 2). The realist review was conducted by a researcher experienced in large-scale systematic reviews, traditional and network meta-analyses, large-scale database analyses and mixed-method process evaluations of policy or intervention trials, supported by team members with expertise in realist review and realist evaluation methodology.

Realist reviews use a theory-driven approach with a philosophy of realism and adopt an explanatory rather than a judgemental approach to evidence synthesis. 40 They seek to produce more transferable findings by taking into account, for example, the heterogeneous nature of rehabilitation programmes and the heterogeneous hip fracture population. The findings are then formulated into statements, the ‘programme theories’, which are propositions for how a programme is considered to produce intended outcomes. They can be generated from various sources of evidence such as the literature, discussions with experts and, as in our study, a survey of current practice (see Chapter 4) and focus groups with patients, carers and rehabilitation professionals (see Chapter 5). During the review process, these intermediate theories are tested, rejected or developed into the final programme theories to make recommendations for future practice, policy and research. We used the guidelines developed by the RAMESES (Realist And Meta-narrative Evidence Syntheses: Evolving Standards) collaboration41,42 (see Appendix 3) to report our methods and findings.

Extending the realist review to include any economic evidence allows the consideration of behavioural economic theories relating to factors such as welfare judgements,43 expected utility gains44 and choice architecture. 45 Additional costs may be accrued when modifying the setting in which the rehabilitation takes place (e.g. home based vs. hospital based) or the delivery team responsible for the rehabilitation programme (e.g. multidisciplinary vs. a single practitioner). The intervention itself could accrue additional costs, for example through additional training required by practitioners, additional time required by practitioners to deliver the rehabilitation programme and additional technology or equipment required for the rehabilitation programme (e.g. instruction packs for exercises). However, we recognised that the literature may not be rich enough to provide understanding of all behavioural economic factors in this field.

Scoping the literature

A scoping search of the literature was carried out in MEDLINE, EMBASE and PubMed for relevant systematic reviews concerning multidisciplinary rehabilitation following hip fracture and stroke and in the frail elderly using the broad search terms ‘rehabilitation’, ‘frail’, ‘elderly’, ‘stroke’, ‘hip/femur fracture’. The reviews identified3,5,9,10,14,17,21–38,46–60 and their reference lists were the starting point for identifying both the implicit and the explicit theories behind the success or failure of rehabilitation programmes or their components. Existing UK and international guidelines were also searched for additional contributions to theory development.

Immersion in the literature to develop initial theory areas

Initial immersion in the rehabilitation literature sought to identify an initial list of relevant intermediate programme theories. We scanned relevant primary studies and other linked papers with a strong theoretical content identified from the reference lists of the included reviews. This process helped to map out important areas and research gaps in the literature, resulting in a list of unanswered questions under different domains related to receivers (patients), deliverers (health-care and rehabilitation teams), programmes (rehabilitation) and settings or systems (hospital, community, etc.) used to deliver such rehabilitation programmes (see Appendix 4).

Developing and refining the intermediate programme theories in interactive workshops

These lists of questions were formulated into statements (see Appendix 5) to signify how the different domains of a programme interact and might affect all of the agencies (stakeholders) involved. These intermediate programme theories were refined during discussions between members of the evaluation team and with other researchers engaged in similar realist evaluations (at two realist evaluation workshops in the School of Healthcare Sciences, Bangor University, convened by one of the senior researchers, JR-M). To keep track of these emerging programme theories a table was constructed in which the theories could be recorded, cross-referenced and commented on. Feedback from the workshops was integrated into this table.

The list of questions enabled the building of context, mechanism and outcome (CMO) configurations that formed the basis of the development of the final programme theories of how complex programmes (systems) work in certain contexts to produce intended (or unintended) outcomes. The initial list of these CMO configurations is presented in Appendix 6; again, this was refined iteratively in team meetings.

Feedback from patient/carer interviews and the health professional survey

Results from the survey of health professionals (see Chapter 4) and focus groups with patients, carers and rehabilitation professionals (see Chapter 5) were also used to refine these programme theories. These refined theories were incorporated into the review as it progressed. Findings from the health professional survey that contributed to theory development included the importance of tailoring, the importance of feedback mechanisms and variation in the delivery of rehabilitation in different areas based on the availability of staff and facilities (see Chapter 4). The focus groups with patients and their carers highlighted unmet information needs with regard to the process of recovery, the availability of services that patients are entitled to access but which they are not necessarily aware of and geographical variation in the provision of services (see Chapter 5).

Developing programme theories

As already described, the summary of findings from our initial immersion in the literature, feedback from meetings and workshops (from experts in health psychology, rehabilitation and implementation research) and the findings from the patient/carer focus groups and health professional survey were integrated into our candidate programme theories. The emergent list of intermediate working theories was used as the basis for the development of bespoke data extraction forms.

Developing bespoke data extraction forms

Two sets of bespoke data extraction forms were developed using a Microsoft Access^® database (2013; Microsoft Corporation, Redmond, WA, USA) to extract data from both comparative studies (RCTs/quasi-RCTs/non-RCTs, comparative cohort and case–control studies) and non-comparative studies (qualitative studies involving patients or health professionals, service evaluations, routinely collected database studies). The data extraction form for comparative studies (see Tables 37 and 38) was designed to collect data from each study on study characteristics (design, sample type, sample size), the intervention/programme and the control, process details (fidelity of the intervention, dosage), contextual factors in the study setting, outcomes collected and theories or mechanisms postulated by the authors to explain the results. The data extraction form for non-comparative studies (see Table 39) was designed to collect data on study characteristics, research methods, the theoretical approach, the sample type, the intervention/programme and the method of analysis as well as evidence to test the programme theories.

The forms were used in two stages to extract data from included studies and test the intermediate and final programme theories. The first set of forms was used to populate the initial themes with evidence from effective (or ineffective) components of rehabilitation programmes and how these interacted with outcomes in given contexts. These themes were then refined into statements, which led to the development of intermediate programme theories. The second set of forms was used to test these theories and adjudicate between competing theories (see Table 40).

Literature search

The literature search strategy used in the NICE guideline review of multidisciplinary rehabilitation programmes for hip fracture7 was adapted to encompass all of the theory areas of the first phase of the review process. No filters for study design were applied so that all study designs such as RCTs and non-RCTs and observational, economic and qualitative studies could be included. Full details of the search strategies for the major electronic databases are reported in Appendix 8.

The following databases were searched from inception to February 2013 for published, semi published and grey literature:

• MEDLINE

• MEDLINE In-Process & Other Non-Indexed Citations

• OLDMEDLINE

• EMBASE

• Cumulative Index to Nursing and Allied Health Literature

• Allied and Complementary Medicine Database

• British Nursing Index

• Health Management Information Consortium

• PsycINFO

• Cochrane Central Register of Controlled Trials

• Database of Abstracts of Reviews of Effects

• Cochrane Database of Systematic Reviews

• Health Technology Assessment database

• NHS Economic Evaluation Database

• Science Citation Index

• Social Science Citation Index

• Index to Scientific & Technical Proceedings

• Physiotherapy Evidence Database

• Biosciences Information Service

• System for Information on Grey Literature in Europe

• ProQuest Dissertations & Theses database.

Identified references were deduplicated and transferred to bibliographic software (EndNote X5; Thomson Reuters, CA, USA) to facilitate assessment for inclusion and the categorisation of relevant studies. Multiple publications arising from the same study were identified, grouped together and represented by a single reference.

Realist review involves iterative and purposive literature searching39,41 and so citations were tracked (forwards and backwards) and internet search engines, such as Google Scholar (Google Inc., Mountain View, CA, USA), and individual publisher websites were used to identify additional evidence as the review progressed and new ideas emerged. The reference lists of previous systematic reviews and included studies were also screened to identify relevant studies. Using this method, no attempt was made to include every relevant study but materials were retrieved purposively to answer specific questions or test-specific theories. The process stopped when sufficient evidence had been collected to answer these questions or test the theories. Conversely, if a new question arose, it triggered further literature searching to answer the question posed and to determine its fit within existing theory or whether or not a new theory needed to be formulated.

Screening of references for relevance

A working definition of multidisciplinary rehabilitation to be used for screening sources of evidence (Table 1) was adapted from a review of intermediate care services;61,62 the working definition in this review had been adapted, in turn, from Godfrey et al. 63

This definition was used when screening the titles and abstracts of identified studies in the EndNote library for relevance. Screening was carried out independently by separate reviewers and discrepancies were resolved after discussion. In addition, potentially relevant studies were categorised according to study type: systematic review, RCT or non-RCT, observational study, economic evaluation or qualitative study. There were no language restrictions and non-English publications were translated whenever possible using Google Translate (Google Inc., Mountain View, CA, USA) or by other research colleagues who could speak the relevant language.

Participants of interest were elderly adults with proximal hip fracture. The intervention of interest was multidisciplinary rehabilitation following proximal hip fracture. The outcomes of interest were mortality, pain, functional status, quality of life, health utility, health service use, costs and patients’ experiences.

Literature identified in the initial search was screened in two stages for both behavioural economic evidence and evidence of economic evaluation (e.g. cost analysis, cost-effectiveness analysis, cost–benefit analysis, cost–utility analysis). Screening for economic studies at the title and abstract stage was conducted by the four main reviewers. Potential economic studies identified in the initial search were then screened by two experienced health economists, who excluded studies based on the following criteria:

• clearly falls outside the definition of multidisciplinary rehabilitation for hip fracture (see Table 1)

• clearly is not an economic evaluation or comparative cost study or does not include behavioural economic theory

• does not involve services users who belong to our service user group of interest.

The detailed screening process for economic evidence and the study flow chart are presented in Chapter 3.

Conceptual categorisation of screened relevant references

Potentially relevant references were conceptually categorised as ‘rich’, ‘thick’ or ‘thin’ based on the criteria described by Ritzer64 and Roen et al. 65 and as used in a recent review of intermediate care. 61 This process made the database manageable and enabled information to be gleaned from the most appropriate studies for theory building and testing. A detailed description of the criteria used for this purpose is provided in Appendix 9.

Inclusion and exclusion of studies

Selection and appraisal of documents

After the initial screening and conceptual categorisation of the references in the EndNote library, potentially relevant studies were exported into a separate library for full document retrieval. Study inclusion criteria were applied to these retrieved documents by two reviewers independently and conflicts were resolved by discussion or after consulting a third reviewer. A list of all studies to be included was prepared for data extraction.

Data extraction

Data were extracted by one reviewer and checked for accuracy by a second. Inconsistencies or disagreements were resolved by mutual discussion and checking against the source study.

Quality assessment

Study quality was assessed using the Mixed Methods Appraisal Tool for mixed studies reviews,68 which can be used across different study designs (qualitative studies, trials, observational studies). The purpose of appraising the ‘quality’ of studies was to assist in the judgement of the relevance and rigour of different evidence from a ‘fitness for purpose’ perspective as opposed to scoring the studies for acceptance or rejection.

Data synthesis

The data from the quantitative, comparative effectiveness and qualitative studies were synthesised separately.

The data from the effectiveness studies were exported into structured tables to show the strength and direction of the treatment effects. Outcomes reported in the included studies were broadly categorised into four domains: physical/physiological, psychological, health service utilisation and AEs. These were subcategorised further under the following headings (the outcome measure instruments used are listed in Appendices 10 and 11):

• physical/physiological

  • ADL

  • composite scores

    • favourable clinical outcome

    • functional recovery

  • exercise behaviour

  • quality of life

  • function

    • physical function

    • mobility

    • functional recovery

    • balance

  • physiological measurements/muscle strength

• psychosocial

  • patient satisfaction

  • carer satisfaction

  • cognitive function/dementia

  • depression

  • fear of falling

  • psychological morbidity

  • self-efficacy/falls efficacy

  • socialisation

  • social support

• health service use

  • physical/occupational therapy sessions

  • discharge destination/new nursing home admissions

  • falls and hospital readmissions

  • health-care utilisation

  • length of hospital stay

  • severity of illness/disease burden

• AEs

  • malnourishment

  • morbidity rate

  • mortality rate

  • pain

  • rate of (repeat) falls.

The rehabilitation settings where the programmes were delivered varied from the acute hospital setting to the community setting and were categorised as below:

• acute hospital

  • inpatient

    • specialised orthopaedic ward

    • specialised orthogeriatric ward

  • outpatient

    • general outpatient rehabilitation unit

    • specialised orthogeriatric outpatient rehabilitation unit

  • rehabilitation unit

    • general elderly rehabilitation unit

    • specialised orthogeriatric rehabilitation unit

• community

  • place of residence

  • nursing, care or residential home

  • specialised nursing home rehabilitation unit

  • community hospital

  • community rehabilitation centre.

Testing the theories with quantitative and qualitative evidence

Data from each individual study were examined in terms of the identified programme theories and the interaction between mechanisms, context and outcomes. Next, the data across the different studies were examined to detect patterns and themes for each theory in turn. Separate fields were created in the Microsoft Access database to capture these interactions as well as raw statement data from the included studies to support reviewers’ reflections. Data synthesis involved individual reflection and team discussions to question the integrity of each theory, adjudicate between competing theories, consider the same theory in comparative settings and compare the theory with actual practice. When candidate theories failed to explain the data, new theories were sought from included studies or from the wider rehabilitation literature, such as studies of rehabilitation following stroke or following inpatient admission after being unable to stand. The narrative of the review was guided by the final theories that emerged from this process. The literature analysis relating to each identified theory is presented in detail, followed by a data summary to show the relationships between data themes and the theories in the final synthesis. Extracts were taken from participant quotations (patient, carer or health professional) reported in the included qualitative studies and used as evidence to support subthemes of the main theories. This is an established method used in a recently reported review of intermediate services61 to incorporate and integrate the theoretical perspectives from qualitative evidence into quantitative evidence.

Results of the initial scoping review

The scoping search for systematic reviews and other reviews as well as guidelines relating to the rehabilitation of older frail populations identified 39 reviews, both Cochrane reviews9,10,27,33,37,49 and other traditional systematic reviews. 3,5,14,17,21–26,28–32,34–36,38,46–48,50–60 The majority of the reviews were related to hip fracture rehabilitation,3,5,9,10,14,17,21–24,26,27,29,30,32–36,49,51–57,59,60 but a few also included rehabilitation for stroke as well as for other conditions in older frail populations needing continuous care. 24,25,28,31,37,46–48,50,58 A few conceptually rich and theoretically sound primary studies from the reference lists of these reviews were also obtained. 69–73 The search also identified five sets of guidelines, from the UK [NICE,7 Scottish Intercollegiate Guidelines Network (SIGN)74], USA,75 Canada76 and Australia and New Zealand. 77

Study flow diagram for the realist review

The electronic searches identified 19,646 references, with a further 24 references identified by hand searching. Deduplication resulted in 12,278 unique references that were screened for relevance by two independent reviewers. The full texts of 610 references were obtained and, after collating multiple publications, 128 studies were included in the review12,13,69,72,78–201 (Figure 2; see Appendix 12 for the total number of references retrieved from each electronic database).

FIGURE 2.

Realist and economic review flow chart.

Study characteristics

Of the 128 primary studies included in the review, 17 were conceptually rich13,69,72,78–80,83,86,89,106,107,124,131,165,176,193,197 and 111 were conceptually thick12,81,82,84,85,87,88,90–105,108–123,125–130,132–164,166–175,177–192,194–196,198–201 (see Appendix 13). Thin sources were screened but were not included in the review for data extraction (see Appendix 14). A list of studies excluded from the review with reasons can be found in Appendix 15. Appendices 16–18 present the raw data tables describing the general characteristics of the included studies, the populations of interest, the treatment categories with characteristics of the interventions and the strengths, limitations and conclusions as presented by authors, respectively. These data are described briefly in the following sections according to the types of research methods used.

Summary of participant characteristics

The number of patients/participants included in the studies ranged from 1 to 2762. The review included a total of 22,443 patients and 97 health professionals. In total, 6282 (range 90–401) patients participated in RCTs,12,69,78,80,87,90–126,128–130,197,199 276 (range 24–95) patients participated in quasi-RCTs,83,131–133 116 (range 20–30) patients participated in non-RCTs,79,134–137 3044 (range 1–919) patients participated in historical cohort studies,13,157–165,170,173,178 7136 (range 18–946) patients participated in concurrent cohort studies,81,138–152,169,177,179,181,182,184,187,188,190–193,196,200,201 1697 (range 3–764) patients participated in controlled before-and-after studies,85,153–156,180,186,195 45 patients participated in mixed-method studies,84,89 3243 (range 130–2762) patients participated in database analyses166–168 and 521 (range 12–222) patients/health professionals participated in qualitative studies,72,82,86,88,127,174–176,183,185,189,198 with two studies involving health professionals (n = 97) rather than patients. 176,198 Two studies used administrative/work process data and did not include any patient data. 171,172

The majority of the studies included patients aged ≥ 65 years. 79,80,82–86,88,90–93,95,97,99–102,105,107,109,111,113,115,116,120,121,124–127,131,135,138–148,151,153,155–157,160,162,163,165,166,175,178–180,182–184,189–191,199,201 Six studies included adults of any age with a hip fracture and undergoing rehabilitation;78,112,167,174,185,186 two of these included carers174 or health professionals. 185 Eight studies included patients aged ≥ 50 years;87,98,103,118,173,187,188,192 18 studies included patients aged ≥ 60 years;12,89,96,104,114,117,119,122,132,133,149,154,164,193–197 19 studies included patients aged ≥ 70 years;69,72,79,88,106,108,110,111,116,120,131,136,143,148,150,161,180,183,184 and seven studies included patients aged ≥ 80 years. 13,81,123,134,137,152,178 The age of the included participants could not be determined from the study reports for four studies. 171,172,181,200 Sixteen studies included only female participants69,79,80,102,104,123,126,147,156,158,165,174,180,183,185,191 and one included only male participants. 159 The rest included participants of both genders but the majority of studies included a greater proportion of women.

The majority of studies excluded patients who had a cognitive impairment or dementia or who lacked mental capacity to give informed consent;12,13,69,72,78–100,102–109,111–118,120–122,124–138,140,142–151,153–156,158–169,171–181,183,185–201 11 studies included such patients,101,110,119,123,139,141,152,157,170,182,184 with one study stating that such patients would be included only if suitable carers ready to participate in the study were identified. 119 The majority of the studies included participants who were mobile and living independently in their own home or in a care home before their hip fracture. 12,69,72,79–86,88–90,92–114,116–119,121,123,124,126–130,132–136,138,139,143,145–149,153,156–158,161,165,166,168,169,171,175–177,179,180,183–186,191,193,194,196,197,199,201 Only seven studies included patients with a medical or psychological comorbidity;123,139,145,157,158,184,193 the majority of studies excluded such patients, especially when exercise would have been contraindicated. Only two studies included patients who had a history of a previous fracture. 175,184

The majority of the studies were carried out in English-speaking countries and involved mainly white Caucasian populations. Three Swedish studies,90,99,127 two Taiwanese studies,119,196 one German study93 and one Danish study84 included only patients who could speak, read and write in these languages, with other patients excluded.

Summary of interventional studies

Forty-eight of the studies were RCTs,12,69,78,80,87,90–126,128–130,191,197,199 with 10 from Australia,87,94,95,110,112,117,118,123,125,129 nine from the USA,69,80,91,92,107,113,121,126,191 six from the UK,12,102,109,115,128,197 four each from Sweden90,98,99,120 and Taiwan,95,105,119,122 two each from Canada,108,111 Norway,116,199 Finland,101,114 Hong Kong103,106 and Switzerland,93,96 and one each from Denmark,104 Belgium,97 Italy,78 Spain124 and Germany. 130 Only seven studies69,78,80,106,107,124,197 were categorised as being conceptually rich.

Four of the studies were quasi-RCTs83,131–133 and five were non-RCTs,79,134–137 with two each from Canada134,135 and the USA,79,83 and one each from Israel,131 Japan,137 Italy,133 South Africa136 and Taiwan. 132 Three of these studies were categorised as being conceptually rich. 79,83,131

Thirty-two of the studies were concurrent cohort studies,81,138–152,169,177,179,181,182,184,187,188,190,192–194,196,200,201 with nine from the USA,81,142,146,147,149,181,191,192,194 four each from Italy,141,143,177,182 Israel138,148,179,201 and Sweden,139,187,188,193 two each from the UK,150,200 the Netherlands,140,144 Taiwan190,196 and Germany,151,152 and one each from Norway,145 France184 and Canada. 169 None of these studies was categorised as being conceptually rich.

Eight of the studies were controlled before-and-after studies,85,153–156,180,186,195 with two each from the USA153,156 and the UK,85,155 and one each from Canada,195 Denmark,154 Sweden180 and the Netherlands. 186 None of these studies was categorised as being conceptually rich.

Thirteen of the studies were historical cohort studies,13,157–165,170,173,178 with three from the USA,159,165,173 two from the UK,13,158 and one each from Australia,161 Austria,178 Canada,157 Germany,163 Israel,162 Italy,170 Japan164 and Sweden. 160 None of these studies was categorised as being conceptually rich.

Among the non-comparative interventional studies there were two mixed-method studies,84,89 one each from the USA89 and Denmark. 84 One study from Finland166 reported a cross-sectional analysis of pre-trial data, two studies from the USA167,168 reported a hospital database analysis and another study from the USA reported longitudinal data from a survey. One study from Australia169 reported before-and-after outcome data for a cohort who underwent an intensive rehabilitation programme in the acute hospital. There were also two case report studies:13,170 one from Italy170 and one from the UK. 13 Two mixed-method studies84,89 and a case series81 were categorised as being conceptually rich.

Summary of non-interventional studies

Non-interventional studies did not use any intervention or treatment to affect the outcomes but were useful for their conceptual input to the theoretical framework and provided explanations for elements of the proposed theories. Two studies, one from the USA171 and one from Canada,172 reported service and work process restructuring. One study from Australia161 utilised hospital data on hip fracture patients 4 months post surgery who had been successfully rehabilitated into the community. These patients were divided into fallers or non-fallers after their rehabilitation. None of these studies was categorised as being conceptually rich.

Twelve of the studies were qualitative studies,72,82,86,88,127,174–176,183,185,189,198 with three each from the UK86,176,198 and USA,72,174,183 two from Sweden88,127 and one each from Australia,185 Canada,175 China189 and Taiwan. 82 Six of these studies72,86,127,175,183,189 interviewed hip fracture patients after discharge about their experiences of the whole process and the rehabilitation that they went through. Two of the studies176,198 interviewed health professionals providing rehabilitation services regarding their experiences about such provision as well as any issues encountered that might be amenable to service improvement. Three of these studies were categorised as being conceptually rich. 72,86,176

Summary of the study settings

Twenty-two of the included studies69,72,82,92,107,114,115,117,118,121,122,125–127,135,142,156,161,163,165,174,201 were conducted in the community after the patients had been discharged from the acute or community hospital to either their pre-fracture place of residence or a care home. Sixty-seven studies12,13,78–80,83,87–89,93,96–98,100–102,104,106,108,111,112,116,119,124,129,132,134,137–141,146,148,152,157–160,162,164,166–173,175–187,194,195,197,198,200 were conducted while patients were still in the acute hospital following surgery. In 39 studies the intervention started in the acute hospital but continued in the community following discharge. 81,84–86,90,91,94,95,99,103,105,109,110,113,120,123,128,130,131,133,136,143–145,147,149–151,153–155,188–193,196,199

Overview of the rehabilitation programmes

Appendix 17 summarises the interventions and comparators as described in the included studies.

Physical activity components of the rehabilitation programmes

Fifty-two of the included studies reported some form of physical intervention69,78–81,83,87,91–93,96,98,99,103,104,107–110,112–115,117,118,121,122,126,130,131,134,136,148,151,153,154,156,158,162,164,169,170,177,180–182,186,187,191,192,194,199 and seven also included a psychological component. 78–80,91,148,180,191 Twenty studies compared intensive physical exercise with less intensive physical activity or an inactive control. 69,78,80,83,91–93,104,107–110,113,114,117,118,122,126,130,169 Twenty-four studies compared supervised programmes with conventional programmes that either did not include supervision as part of the programme or included only minimal supervision to ensure patient safety. 69,80,83,87,91–93,104,107–110,112–114,126,131,136,151,153,180,186,191,199 Nine studies compared specifically tailored programmes with generic rehabilitation programmes. 69,79,80,83,99,109,113–115

Psychological components of the rehabilitation programmes

Fourteen studies reported using a psychological intervention in isolation12,106,137,141,171,197,201 or as part of a comprehensive rehabilitation programme along with physical components. 78–80,91,148,180,191 Three of these studies141,171,191 did not report any outcome data but were utilised mainly for theory explanation.

Place of rehabilitation

Twenty-six studies compared different rehabilitation settings. 81,97,101,103,112,120,125,128,131,133,135,143,145–147,149,151,152,154,160,161,179,188,195,200,201 Ten of these studies compared some form of community (own home or care home) rehabilitation with hospital-based rehabilitation,101,103,125,128,131,133,135,143,145,160 with one comparing hospital plus home rehabilitation with hospital rehabilitation only. 133 Eight studies97,112,120,146,149,151,152,200 compared hospital-based rehabilitation with usual care, no post-discharge care or rehabilitation in nursing facilities. Other studies did not compare rehabilitation settings per se but included comparisons based on patients’ characteristics, such as fallers compared with non-fallers,161 very old patients compared with younger patients201 and treated in a cognitive specialised rehabilitation unit compared with treated in a non-cognitive specialised rehabilitation unit. 152 One study compared the discharge practices of four hospitals after inpatient rehabilitation. 147

Process or system improvement

Twenty-nine studies investigated the effects of improvement or change in existing health-care rehabilitation structures. 69,83,94,95,98,100,102,105,111,116,117,119,120,123,124,127,128,132,138–140,144,150,155,157,159,163,167,184 Seventeen studies compared the development of multidisciplinary co-ordination programmes with usual care or another existing programme. 69,83,98,102,111,117,119,120,123,124,127,132,144,155,157,159,184 There was large variation in these programmes from different health-care systems, but common features included comprehensive geriatric assessment both pre and post surgery, assessment of patient needs and assignment of appropriate health-care staff to address those needs, regular multidisciplinary meetings to discuss progress and care pathways that continue into the community after discharge. Usual or conventional care varied greatly among the studies, ranging from simple control of post-operative symptoms117 to comprehensive assessment. 69,83,120

Eight studies reported on structured discharge planning from hospital to the community based on patients’ abilities, the extent of support needed and the availability of support from family or friends during the recovery period. 94,95,100,105,128,139,140,150 Six studies compared the early discharge of patients to their own home with usual discharge,94,95,100,128,140,150 one study compared early discharge to a rehabilitation unit of a community hospital with early discharge home139 and one study compared early discharge to the rehabilitation ward of a nursing home with conventional (delayed) discharge to the same ward. 105

Four studies reported the implementation of new ward protocols. 116,138,163,167 One study compared a newly commissioned orthogeriatric ward with a traditional orthopaedic ward116 and another study compared comprehensive geriatric assessment with usual care. 138 The other two studies were non-interventional improvement reports that utilised routinely collected hospital data in their analyses. 163,167 One did not report any patient-related outcomes but was useful for theory development. 163

Summary of outcomes

Outcomes data were extracted from 70 of the included studies. 12,69,78,80,81,83,87,90–93,95–97,100–103,105–111,114,115,117–124,130–133,135–140,143,144,146–149,151–158,160–162,167,169,184,192,195,199–201 Sixty-five of these studies reported physical or physiological outcomes,12,69,78,80,81,83,87,90–93,95,97,100,102,103,105–111,114,115,117–124,130–133,135–140,143,144,146–149,151–157,161,162,167,169,184,192,195,199,201 22 reported psychological or social outcomes,69,80,81,83,90,95,105,106,110,115,117–119,121,130,136–138,143,151,161,169 26 reported health service utilisation83,87,90,93,97,100–102,105,111,119,120,124,138,140,143,147,148,152,154,155,157,158,160,200,201 and 16 reported AEs93,96,97,100,101,110,111,119,124,133,137,144,148,154,155,162 as their main outcomes.

The rest of the included studies13,72,79,82,84–86,88,89,94,98,99,104,112,113,116,125–129,134,141,142,145,150,159,163–166,168,170–183,185–191,193,194,196–198 mainly contributed to theory building and explanation.

The directions of effect at various follow-up points are presented in structured tables in Appendix 19. The outcomes reported are discussed further when appropriate in the following discussion of the evidence for the final programme theories.

Study quality assessment

The results of the quality assessment are presented in Appendix 20.

Programme theory 1: improve patient engagement by tailoring the intervention according to individual needs and preferences

Proximal hip fracture patients presenting with a range of pre-fracture physical and mental functioning and a variety of comorbidities (C) need a rehabilitation programme that is tailored to individual needs (M) to achieve appropriate outcomes such as improved physical functioning, greater mobility, reduced disability and independent living (O).

Tailoring of rehabilitation activities involved the interplay of many factors encompassing the patient, the health-care professional and the environment in which the rehabilitation took place. The main theme revolved around making rehabilitation planning patient centred and contextualising what is important for patients so that this can be incorporated into their care plan, allowing a better chance of engaging patients in their recovery.

Summary of programme theory 1

A summary of programme theory 1 is provided in Box 1.

Programme theory 2: reducing the fear of falling and improving self-efficacy to exercise and carry out activities of daily living

Proximal hip fracture results in poor physical functioning, fear of falling, low mood and lack of self-efficacy (C), requiring improved quality and an increased amount of practice of physical exercises, ADL and psychological tasks (M) to gain mastery and control to improve confidence, mobility and physical functioning (O).

Summary of programme theory 2

A summary of programme theory 2 is provided in Box 2.

Programme theory 3: co-ordination of services and sectors delivering rehabilitation

The diversity of services provided by different disciplines across sectors from a variety of funders (C) requires co-ordinated provision of multidisciplinary rehabilitation programmes (M) to deliver appropriate physical, functional and psychological interventions to patients in a timely manner (O).

Summary of programme theory 3

A summary of programme theory 3 is provided in Box 3.

Summary of findings

Three programme theories arose from this realist review that appeared to put patients at the heart of rehabilitative care and improve outcomes. These were:

1. tailoring the intervention according to patients’ individual needs and preferences to improve patient engagement

2. reducing fear of falling and improving self-efficacy to exercise and perform ADL through increased quality and frequency of the practice of exercises and daily tasks under supervision

3. the co-ordination of services and sectors delivering the rehabilitation.

Strengths and limitations

This was a realist rather than a systematic review of multidisciplinary rehabilitation following hip fracture aimed at synthesising evidence to develop an enhanced rehabilitation package for such patients. To our knowledge, this is the first realist review that has attempted to build an explanatory account about the effectiveness of different components of rehabilitation interventions following hip fracture in the context of what works for whom and under what circumstances. 39,40 As such, we did not attempt to summarise all of the evidence and judge whether or not rehabilitation programmes were effective, but rather we attempted to build an explanatory account of the mechanisms behind their effectiveness and to establish which components were effective in certain circumstances and contexts.

No studies were excluded based on a particular outcome and data for all reported outcomes were extracted. Particular caution needs to be exercised when the included studies have used non-validated measures or measures have been combined to give a cumulative outcome, such as combining numbers of readmissions and deaths into a single outcome designated as ‘poor outcome’. 26 About 35 outcomes using 152 outcome measure instruments were recorded (see Appendices 10 and 11). Such a variety of outcomes in hip fracture rehabilitation studies has been reported by the majority of reviews3,21 explored during theory-building and scoping searches. This applies to both long-term3 and short-term14 disability arising as a result of hip fracture. This was recognised as a limitation of traditional review methods, as were the myriad of treatment strategies used and inconsistent reporting,9,10,23,24,47,50 which made comparisons across studies difficult and led to inconclusive results. 3,8–10,21,23–25,28,29,47,49,50,240,241 These issues also led to difficulties in replicating intervention designs in different health-care systems as well as in drawing evidence-based conclusions about best practice. 10,24 Successfully combining a few commonly reported outcomes to estimate the long-term health outcomes of rehabilitation programmes10 could help to more accurately estimate the comorbid disease burden, which may benefit from longer-term rehabilitation and falls prevention programmes, leading to health gains and reduced costs in the long term. 3

It is acknowledged that the pragmatic method of reviewing the literature and synthesising the evidence may be difficult to replicate and that another team carrying out a similar review may reach different conclusions. This review has attempted to be as inclusive as possible and incorporate evidence from multiple sources and study types, which helped to provide context to the findings of the comparative studies. The use of a specific definition of rehabilitation programmes and categorising sources of evidence according to their conceptual richness might have resulted in some relevant studies being missed or excluded. However, this was mitigated by the comprehensive electronic literature searches, complemented by manual citation tracking.

The process of identifying and formulating programme theories was challenging, particularly when some studies did not explicitly state how the intervention being tested was developed or the content and operation of services in the context of that particular intervention. This similar issue has been highlighted in a recent review of intermediate care services, which recommended that the research community provide more information about ‘how’ and ‘why’ the interventions/services being evaluated were developed and delivered. 62

Comparison with previous literature

As discussed in the previous section, a realist approach was used to tease out which components of rehabilitation programmes were useful and which were not,10,23 rather than summarising overall effectiveness as in previous systematic reviews. It also addressed some of the gaps in hip fracture rehabilitation, such as the lack of interventions that enhance self-efficacy24,169,242,243 and address postural balance and fear of falling,14,21,24,169 the exclusion of moderately cognitively impaired patients,10,16,17,24,36,57,244 the short-term provision of rehabilitation services10,244 and the lack of reporting of carer outcomes. 10

Self-efficacy to exercise has long been recognised as an important mediator for regaining confidence in daily activities and function and overcoming the fear of falling, to enable an increase in the practice of physical activities and exercise. 14,69,72,227,245 But such potential has not been realised fully in rehabilitation interventions, especially once patients have been discharged from the acute hospital. 157,169,242,246 Targeted information provision and education,246 skills enhancement,227 addressing and accommodating user needs and views214 and social and peer support are important to improve motivation and engagement and help patients regain control and confidence after the medical and psychological crisis presented by hip fracture. 227 A rehabilitation intervention model addressing both psychological and physical needs may be an answer to this dilemma, as such strategies in isolation have minimal positive effects on outcomes. 10,73,91 This has been successfully shown in the rehabilitation of patients with stroke247 and other conditions. 61,62,214 This review identified that mutual aid and social support,246 combined with the development of practical skills through supervised practice and exercise,243,248 have the potential to improve quality of life and reduce the rates of rehospitalisation and long-term institutionalisation. This is mediated by building confidence, cognitive understanding and practical skills.

Recent drives to both improve patient outcomes following hip fracture and reduce costs through effective rehabilitation programmes have highlighted the importance of tailoring specific rehabilitation packages according to individual patient needs rather than following set programmes that do not take into account patients’ circumstances. 227 Such tailoring activity involves multidisciplinary team decision-making in consultation with the patient and his or her carers about the appropriate timing and place of such activities. The timing of different interventions is important, as some intervention activities appear to produce better outcomes if introduced at certain time points and in appropriate settings. For example:

• coaching may have a minimal effect if started immediately after surgery in hospital91 but has been shown to enhance patients’ self-efficacy, skills and mastery to perform exercises independently if provided in the community soon after discharge36,69,72,73,244,245,249

• occupational therapy and interventions for fear of falling in the acute hospital have a minimal effect99,245 but appear to be effective when provided at patients’ usual place of residence,244,245,250 probably because patients start ambulating either independently or with help and can appreciate that the occupational therapy supervision aims to make the environment safe for them to practise exercises and ADL28

• supervised physical training and earlier mobilisation in rehabilitation programmes during the acute hospital stay,36 and strength and progressive resistance training later in the programmes, either during or after discharge, appear to improve outcomes,25,36,57,251 as earlier commencement of intensive physical therapy may act as a hindrance and there is a danger of patients becoming disengaged or dropping out of programmes9 because of the challenging nature of the tasks

• falls prevention programme training combined with supervised exercise to improve self-efficacy and regain confidence,36 but only after patients have restarted walking. 5,9

Similarly, no single pathway provides the answers to all patient issues because of the complexity of comorbidities with which hip fracture patients can present. 8,29,49,240 Geriatric orthopaedic rehabilitation units are likely not to be cost-effective but can be beneficial for frailer patients to reduce complication rates, readmissions, nursing home placements and in-hospital deaths; however, they showed little effect on the rate of death post discharge, in both the short term and the long term. 29 Similarly, geriatric hip fracture programmes and early supported discharge can be cost-effective if offered to suitable patients, as they appear to reduce the length of hospital stay in such groups. 47,252 If they are offered to all patients without distinction, however, they will simply transfer the cost from the hospital to the community, as some patients may need prolonged care in the community. 140 Multidisciplinary co-ordination would play an important role in such instances to determine the appropriate skills mix of rehabilitation professionals during a patient’s inpatient rehabilitation and discharge planning8,49,240 as well as rehabilitation beyond the acute care setting and into the community. 252 Such rehabilitation programmes would need to be flexible with a mix of components to respond and adapt to individual patient needs and requirements.

Implications for practice, policy and research

In the early post-operative period, patients’ needs and wishes should be used to tailor the rehabilitation package (the timing of delivery of certain components of the programme, the skills mix of rehabilitation professionals needed, the appropriate place and support element built into the programme) to allow it to produce positive outcomes and be cost-effective in the long term.

Rehabilitation programmes need to take into account the longer-term needs of patients to reduce the detrimental sequelae that follow, for example fear of falling again leading to movement restriction, reduced participation in physical activities and ADL, reduced function leading to disability and dependence and psychological issues leading to isolation and reduced quality of life.

Rehabilitation programmes should have elements of support, supervision and coaching to help patients regain lost confidence and realise the importance of the enhanced practice of physical activities and ADL for improving function and reducing disability.

Realist review methodology provides a flexible and pragmatic way of developing complex intervention such as rehabilitation programmes for hip fractures through an explanatory building exercise and teasing out individual and effective components of such programmes.

The programme theories that emerged from this realist review were combined with the findings of the survey of therapy professionals (see Chapter 4) and the patient/carer focus groups (see Chapter 5) to develop the rehabilitation package (see Chapter 6) to be tested in a feasibility study (see Chapter 7).

Data extraction

Data were extracted by one reviewer and checked for accuracy by a second. Inconsistencies and disagreements were resolved by discussion after checking against the source study.

Quality assessment

Study quality was assessed using the checklist for economic evaluations of Drummond et al. 66

Evidence synthesis

The economic outcome data were exported into structured tables. Unfortunately, there was insufficient homogeneity between studies to allow for meta-analysis and therefore a narrative review of the findings is presented.

The method of interrogation used to appraise the included papers was to explore the study setting (country), year of publication, perspective of measurement for cost and benefits, type of study, comparator, range of costs and benefits measured, source of unit costs, year of unit costs and currency. We also assessed whether discounting of costs or benefits took place, whether or not sensitivity analysis was undertaken, the extent to which the results were compared with the results of other studies and, finally, whether or not the results were argued to be generalisable to other settings.

Cost-effectiveness studies

Three cost-effectiveness analyses were identified in the review.

Olsson et al. 260 conducted a cost-effectiveness analysis of an individualised hip fracture rehabilitation pathway, the integrated care pathway (n = 56), compared with usual care (n = 56) in a Swedish hospital, with a focus on motivation, prerequisites for rehabilitation and early first ambulation. Patients aged ≥ 65 years, admitted for a hip fracture and living independently, participated in the study. In the analysis, a hospital perspective was used and only direct hospital costs were considered. Costs were collected over an 18-month period during 2003–5 from the hospital’s financial database (cost year 2004). Costs were not discounted as the study follow-up period was < 1 year and loss in productivity costs were not calculated because of the age of the participants (mean age 84 years). Costs were converted from Swedish crowns to euros using the annual average exchange rate for 2004 (9.1268 SEK per €). Direct hospital costs included treatment costs (e.g. surgery, post-operative care) and hotel costs (salaries, drugs, food, accommodation, administration). The effects of the integrated care pathway and usual care were assessed by comparing participants’ self-estimated pre-fracture ADL level with their actual ADL level at discharge using a tool developed by Katz et al. 269 This tool used a hierarchical scale with steps ranging from A (completely independent) to G (completely dependent). The integrated care pathway produced 36 successfully rehabilitated participants at a cost of €534,249, whereas usual care produced 27 successfully rehabilitated participants at a cost of €861,532. Olsson et al. 260 calculated an ICER of €14,840 per successfully rehabilitated participant (classed as an ADL level of A–C) and a failure rate of 25% in the integrated care pathway group, and an ICER of €31,908 per successfully rehabilitated participant and a failure rate of 45% in the usual care group. No sensitivity analysis was reported. The authors concluded that the integrated care pathway was less costly and more effective than usual care, and that the application of patient-centred care appears to enhance both rehabilitation outcomes and cost-effectiveness. They stated that it was difficult to compare costs and outcomes for a group of patients such as hip fracture patients because of individual variability within the sample and the differing organisational structures of health care between countries. Concerning the generalisability of the results, they acknowledged that the sample size was small but argued that the statistically significant improvements in levels of ADL were likely to remain statistically significant in a larger sample.

Kramer et al. 261 conducted a cost-effectiveness analysis of different rehabilitation settings with 518 randomly selected elderly hip fracture patients and 485 stroke patients admitted between November 1991 and February 1994 from 92 hospital-based units across 17 states in the USA. The patients were randomly assigned to the following different rehabilitation settings: (1) traditional nursing home, (2) subacute nursing home and (3) rehabilitation hospital. A hospital perspective was used and only direct hospital costs were considered. Service use information was collected from admission until 6 months post admission from Medicaid forms. Costs were presented in US dollars (cost year not reported). Costs were not discounted as the study follow-up period was < 1 year. Direct health service costs were considered, such as Medicare-reimbursed costs. The outcomes of treatment in these settings were assessed in terms of the recovery of patients to premorbid levels of ADL 6 months post admission. This study assessed five ADL: bathing, dressing, toileting, transferring out of bed and walking 20 feet. Analysis was conducted for completers only. Hip fracture patients admitted to rehabilitation hospitals did not differ from patients admitted to nursing homes in terms of returning to the community [adjusted odds ratio (OR) 1.3, 95% CI 0.6 to 2.6] or the number of ADL recovered to premorbid levels (difference 0.09 ADL, 95% CI 0.27 to 0.44 ADL). There was no difference in return to the community between patients admitted to subacute nursing homes and patients admitted to traditional nursing homes (adjusted OR 1.6, 95% CI 0.7 to 3.6). Medicare costs were greater (p < 0.001) for rehabilitation hospital patients than for subacute nursing home patients. The costs of subacute nursing home patients were greater (p = 0.009) than those for traditional nursing home patients. No sensitivity analysis was conducted. The authors concluded that enhanced health outcomes were obtained at higher costs. Enhanced outcomes were obtained with higher costs for elderly patients treated for stroke in rehabilitation hospitals. However, patients treated for hip fracture did not achieve enhanced outcomes. The authors did not state if the results of the study were generalisable to other settings and did not compare the results with the results of other studies.

Cameron et al. 253 conducted a cost-effectiveness analysis of an accelerated rehabilitation intervention (n = 127) compared with conventional rehabilitation (control group) (n = 125) for elderly patients with proximal femur fracture in an Australian general hospital between 1989 and 1990. Patients in the accelerated rehabilitation group received earlier assessment of rehabilitation goals, early commencement of rehabilitation (usually within 24 hours of surgery), early mobilisation, retraining of physical independence, closer family caregiver contact and detailed discharge planning combined with elements of geriatric assessment. The control group received conventional rehabilitation (a variety of rehabilitation programmes depending on functional status following surgery). Forty-five control group participants received interdisciplinary rehabilitation (equivalent to orthogeriatric care), 23 participants received little formal rehabilitation as they had limited disability and were discharged directly home, 12 patients were discharged to a nursing home before they could return to their own home because of poor prognosis and 45 patients were transferred back to their nursing home on discharge. Costs were considered from a hospital and community perspective. Direct hospital costs included the costs of inpatient surgical and post-surgical care, readmissions, community support services and institutional care. No family costs of rehabilitation were considered. Costs post hospital discharge were calculated and included the costs of follow-up treatment, changes in residential care and home support services. Service use was collected from admission, with a 4-month post-operative follow-up. Unit costs were gathered from published data, local service providers or government subsidies for residential care. Costs were presented in Australian dollars (cost year 1990). Discounting of costs and benefits was not reported; however, the study follow-up period was < 1 year. The effects of the two rehabilitation interventions were assessed by comparing patients’ pre-fracture and post-rehabilitation physical independence, using the Barthel Index of Activities of Daily Living. 265 This index, which is a standard, validated method of measuring physical independence levels, was administered by a trained research nurse during admission and again 4 months after fracture. A recovered patient is a patient who has returned to disability-free living, defined as requiring no or limited assistance in ADLs and characteristically having a score of ≥ 80 on the index. If a patient had significant disability prior to fracture (classed as scoring < 80 on the index) then recovery was defined as regaining the premorbid level of physical independence. In the accelerated rehabilitation group, 63 of the 127 patients (49.6%) met the criteria for recovery, whereas in the conventional care group, 52 of the 125 (41.6%) patients met the criteria for recovery; this difference was not statistically significant. The total cost of accelerated rehabilitation was AUS$10,600 (this included an additional AUS$10 per patient for an extra 30 minutes of nursing support or allied health professional support per day and AUS$25 per day for physician costs) and of the conventional rehabilitation programmes was AUS$12,800. Cost-effectiveness ratios were calculated as the total direct cost per recovered patient of accelerated rehabilitation and conventional care, based on 50% of patients recovering in the accelerated rehabilitation group and 41% of patients recovering in the conventional rehabilitation group (difference of 9%, 95% CI –3% to 21%). The cost per recovered patient was AUS$21,240 in the accelerated rehabilitation group and AUS$31,190 in the conventional rehabilitation group. When only the costs after surgery were considered, the cost-effectiveness ratios produced a cost per recovered patient of AUS$15,290 in the accelerated rehabilitation group and AUS$25,250 in the conventional rehabilitation group. Sensitivity analysis was conducted modifying the duration of hospitalisation and varying the rates of recovery for the two types of rehabilitation. The sensitivity analysis demonstrated that patients in the accelerated group needed to be hospitalised for approximately 1.5–2 days fewer than patients in the conventional care group for the intervention to start generating cost savings. Given the reduction in length of stay achieved by accelerated rehabilitation, accelerated rehabilitation patients could receive treatment costing up to 40% more per bed-days than conventional care patients and still generate savings in gross costs. If the criterion for recovery changed from a Barthel Index score of 80 to a score of ≥ 90, no differences were found in the relative costs of the two rehabilitation programmes; the cost per recovered patient in each of the two rehabilitation programmes increased by approximately 20%. The authors concluded that accelerated rehabilitation was more cost-effective than conventional rehabilitation in treating proximal femur fracture. They did not state if their results were generalisable to other settings and did not compare their results with the results of other studies.

Cost–benefit analysis

Two cost–benefit analyses were identified in the review.

Ruchlin et al. 262 conducted a cost–benefit analysis of a patient education and high-intensity strength intervention to improve rehabilitation after hip fracture compared with standard post-operative care (control condition). In total, 114 patients aged > 64 years who had been admitted for a hip fracture to the New York Hospital, the Hospital for Special Surgery or the New York Hospital – Queens received either the patient education self-efficacy video and high-intensity strength intervention (n = 59) or usual care (n = 55). Costs were collected from a societal perspective and included direct medical costs and non-medical costs (e.g. costs of community-based care). Direct medical costs included the costs of outpatient and inpatient care, emergency care, acute hospital care, post-hospital rehabilitation in a long-term care facility, nursing home care, physical and occupational therapy, visiting nurse care and prescription drugs. Community care-based costs included the costs of home assistance, transportation and informal care provided by family and/or friends. Service use was collected via self-reported service utilisation questionnaires administered by telephone interview to patients. Medicare reimbursement rates were used to provide the unit costs of direct and community care. Service use data were collected until 18 months post fracture. Costs were presented in US dollars (cost year 1995). As data were collected until 18 months post fracture, a discount rate of 3% was applied. Physical role limitation, physical functioning and social functioning components of the Short Form questionnaire-36 items (SF-36)266 were assessed at baseline and 6 months’ follow-up for the intervention and control groups. The protocol stated that each patient in the intervention group should receive 16 strength training sessions; however, only 35 of the 59 intervention patients (59.3%) received any strength training. The video was seen by 42 of the intervention patients (71.2%) and 24 of the intervention patients (40.7%) had at least one contact with a peer advocate. The change between baseline and 6 months in the physical role limitation component was significantly higher for the intervention group (66.1) than for the control group (38.9) (p = 0.02). The change in physical functioning score was 46.3 for the intervention group and 38.9 for the control group (p > 0.05). The change in social functioning score was 44.2 for the intervention group and 39.4 for the control group (p > 0.05). The cost of the intervention was US$13,842 per patient (SD US$11,941) and the cost of usual care was US$17,139 (SD US$21,577) per patient. The total intervention cost was US$42,593. The total saving for the intervention compared with usual care was US$194,527 (US$66,799 in direct medical savings and US$127,725 in non-medical savings). The cost–benefit ratio was 4.57, higher than the threshold of 1.0, and the net present value of the intervention was US$151,934. Sensitivity analyses were conducted varying the unit cost applied to staff wages from the federal minimum wage (US$4.25 per hour) to market wages (US$11.50 per hour). When the 3% discount rate was maintained and the higher market wage was applied (US$11.50 per hour), the total saving for the intervention compared with usual care was US$226,661 (US$66,799 in direct medical savings and US$159,856 in non-medical savings). The cost–benefit ratio was 5.32, with a net present value of US$184,068. In addition, as part of the sensitivity analysis a 5% discount rate was applied to test the robustness of the findings. When applying the 5% discount rate to the estimates using the federal minimum wage value (US$4.25 per hour) to calculate wage costs, the total saving for the intervention compared with usual care was US$193,567 (US$66,746 in direct medical savings and US$126,809 in non-medical savings). The cost–benefit ratio was 4.54, with a net present value of US$150,974. When the 5% discount rate was applied to the estimates using the market wage value (US$11.50 per hour) to calculate wage costs, the total saving for the intervention compared with usual care was US$225,660 (US$66,746 in direct medical savings and US$158,898 in non-medical savings). The cost–benefit ratio was 5.30, with a net present value of US$183,067. The authors concluded that the intervention results in significant improvements in the physical role limitation component compared with usual care. The economic benefits of the intervention exceeded its costs. The authors did not state if the results of the study were generalisable to other settings and did not compare their results with the results of other studies.

Fordham et al. 254 conducted a cost–benefit analysis of a new joint management system compared with a single specialty orthopaedic management system (standard care). In total, 108 women aged > 65 years with a fractured neck of femur were randomly assigned to either the new joint management system (n = 50) or the single specialty orthopaedic management system (n = 58) in Huddersfield, UK. The joint management system consisted of joint geriatric and orthopaedic management including early post-admission assessment, joint weekly bedside consultations, joint consultant decision-making with regard to the rehabilitation programme and any ward/hospital transfers and discharge. Standard care patients were treated solely by an orthopaedic consultant, with advice from a geriatrician when necessary. Costs were collected from a hospital perspective and only direct hospital costs were considered. Service use was collected from hospital records using a specially designed patient profile data collection form, from admission up to and including discharge, at fixed intervals. Sources of unit costs were not reported in the paper. Costs were presented in UK pounds sterling (cost year 1985). Discounting of costs and benefits was not reported. Costs taken into consideration included the cost of the bed-days utilised and staff inputs into rehabilitation, for example physiotherapists’ and occupational therapists’ time. Additional costs for the joint management system were considered and included the costs of extra geriatrician input, medical travel and ambulance transfer. Costs were not discounted as the study follow-up period was < 1 year. The different management systems were assessed based on the average cost of the hospital stay per patient, the length of stay per patient, ADL, prognosis and place of discharge. Measures were taken at fixed intervals during the study: at admission, during the fourth week of hospitalisation and at discharge. ADL, prognosis and place of discharge were assessed against a classification system devised by the research team rather than a standardised system. The mean length of stay was 56 days for the intervention group and 44 days for the control group. The study showed no savings in terms of bed-days used. There were no statistically significant differences between groups for ADL, prognosis and place of discharge. The joint management system cost £151.20 per patient and the single specialty (control) system cost £55.10 per patient. Joint management was £96 more expensive per patient than single specialty management. In relation to total costs, this represented a 3.6% increase, which was quite small in proportion to the total hospital costs. No sensitivity analysis was reported. The authors concluded that future geriatric–orthopaedic collaborations should take particular care in assessing the impact on length of hospital stay and personal benefits to patients. They also stated that costs could have been reduced further if arrangements for joint management had been based at a single site, as travel between sites was a major source of additional costs. In this case, the costs of the joint management would have been only 2% more expensive. The authors advised caution when generalising the results of this study to other settings because of the small sample size and did not compare the results with those of other studies.

Cost-minimisation analysis

Two cost-minimisation analyses were identified in the review.

Polder et al. 263 conducted a cost-minimisation analysis of two discharge programmes in 208 patients with a fresh hip fracture between October 1996 and October 1998 in a general hospital in Rotterdam, the Netherlands. The two discharge programmes were classed as conventional discharge (n = 102) and early discharge (n = 106). Conventional discharge patients stayed in hospital longer for rehabilitation than early discharge patients. The treatment consisted of physical therapy, which was administered twice per day by the hospital’s physiotherapists under the supervision of ward physicians. Early discharge was implemented by a discharge protocol that started 5 days post-operatively. Administrative procedures were accelerated and the number of available beds on the rehabilitation ward was increased. Physiotherapists, occupational therapists and social workers were involved in the rehabilitation process and were supervised by a physician trained in geriatric medicine. Costs were considered from a societal perspective and included the costs of manpower, equipment, materials, housing and overheads, all medical costs and patient-borne costs including those borne by families (e.g. for travelling). Service use was gathered through interviews, including with family members to gather any family-borne costs, and from hospital records. The costs were divided into seven categories: (1) preadmission, (2) from admission to day 5 after hip fracture, (3) from day 6 after fracture until discharge, (4) nursing home, (5) care home, (6) own home and (7) readmission to hospital or nursing home. This resulted in a calculation of the total average cost per day for each patient. Service use was estimated for a 7-month period (3 months pre-operatively and 4 months post-operatively) and unit costs were gathered from hospital financial records. Costs were presented in euros (cost year 1998). Discounting was not applied as the costs of the interventions fell within a 1-year time horizon. The effects of the two discharge programmes were assessed by comparing patients’ functional outcome and cognitive status using the Rehabilitation Activities Profile (RAP)267 and the Mini Mental State Examination (MMSE). 268 The RAP is based on the International Classification of Impairments, Disabilities and Handicaps and measures disabilities in communication, mobility and personal care. Four months after hip fracture, the RAP score was 14.9 for the early discharge group and 14.5 for the conventional discharge group (based on a possible score of 0–36). The MMSE score was 20.6 for the early discharge group and 20.8 for the conventional discharge group (based on a possible score of 0–29). Early discharged patients stayed an average of 13.5 days less in hospital than conventionally discharged patients, although this difference was not statistically significant. Average costs during the 4 months after hip fracture were €14,281 for early discharged patients and €15,338 for conventionally discharged patients, although this difference was not statistically significant. No sensitivity analysis was reported. The authors concluded that early discharge of hip fracture patients from hospital led to a limited non-significant reduction in total costs. Although the reduction in hospital stay for the early discharge programme was larger than that reported in Australian studies by Cameron et al. ,253 Farnworth et al. 247 and Sikorsky and Senior,270 significant cost-savings were not observed. The authors advised caution when generalising the results of this study to other settings because of the small sample size.

Elliot et al. 264 conducted a cost-minimisation analysis in Christchurch, New Zealand, of geriatrician assessment (n = 61) of patients aged 65 years with a femoral fracture admitted to an acute orthopaedic ward compared with standard care (n = 57), whereby geriatrician input was given only following a request from the orthopaedic medical staff. Costs were considered from a hospital perspective, with direct hospital costs calculated (e.g. surgery, post-operative care). Service use was estimated from admission to discharge from the hospital, gathered from hospital data. Unit costs were obtained from hospital financial records. Costs were presented in New Zealand dollars (cost year 1993). Discounting was not applied as the costs fell within a 1-year time horizon. The total mean length of stay was 20.7 days (95% CI 17.4 to 23.9 days) for the intervention group and 26.3 days for the non-intervention group (95% CI 21.2 to 31.4 days). The mean length of stay in the rehabilitation ward was 12.7 days (95% CI 9.0 to 16.3 days) for the intervention group and 18.9 days (95% CI 13.6 to 24.2 days) for the non-intervention group. The proportion of patients discharged to a higher level of care (i.e. more nursing intensive) was 11% and 23% for the intervention and non-intervention groups, respectively. The length of stay, cost of care and discharge destination were compared for each group. The cost per case on the intervention ward was NZ$9400 (95% CI NZ$8300 to NZ$10,500) and on the control ward was NZ$11,500 (95% CI NZ$9900 to NZ$13,200). No sensitivity analysis was conducted. The authors concluded that the intervention led to a shorter hospital stay and reduced hospital costs, without the need for increased patient dependency. These results agree with other findings that geriatricians provide additional health benefit, either at a lower cost or at an equal cost to that of conventional care. Because of the small sample size, caution is needed when generalising the results of this study to other settings.

Quality assessment

Table 3 provides the results of the quality assessment of the included economic studies using the checklist of Drummond et al. 66

Summary of findings

Two out of the three cost-effectiveness studies found that rehabilitation pathways were cost-effective. Olsson et al. 260 demonstrated that an integrated care pathway with a focus on motivation for rehabilitation and early first ambulation was less costly and more effective than a usual care pathway. The application of patient-centred care appeared to enhance both rehabilitation outcomes and cost-effectiveness in a Swedish hospital. Cameron et al. 253 found that an accelerated rehabilitation intervention using early assessment, early mobilisation, physical independence and detailed discharge planning was more cost-effective than conventional rehabilitation in treating proximal femur fracture in an Australian general hospital. On the other hand, Kramer et al. 261 conducted a cost-effectiveness analysis of different rehabilitation settings and found that enhanced health outcomes were obtained with higher costs for patients admitted with stroke and that therefore the intervention could not be considered cost-effective. Of the two cost–benefit analyses, only one found that the benefits of the intervention outweighed the costs. Ruchlin et al. 262 found that a patient education and high-intensity strength intervention to improve rehabilitation after hip fracture resulted in significant improvements in physical role limitation compared with standard post-operative care and that these benefits exceeded the costs. Fordham et al. 254 found that the number of bed-days and the costs increased for the intervention group compared with the control group. Both Polder et al. 263 and Elliot et al. 264 revealed that the interventions under study reduced costs without reducing outcomes compared with usual care, thus demonstrating that the interventions could be lower-cost alternatives to regular practice.

Strengths and limitations

Many of these studies had small sample sizes. This is common in economic analyses as the power calculations of most studies are based on detecting differences in clinical outcomes. 271 A small sample size can lead to difficulties in detecting differences between groups and to the misinterpretation of results.

Other potential biases were also noted. In the study by Cameron et al. ,253 both the control group and the intervention group were treated on the same ward by the same practitioners, with a potential for cross-contamination of those in the control group if they overheard advice from the practitioners or received elements of the intervention by mistake. Polder et al. 263 found that health-care costs in the study shifted from hospitals to nursing homes, highlighting the importance of a detailed cost analysis based on real resource use to detect whether there are true cost-savings or just redistribution of costs. Polder et al. 263 also found that costs increased for individuals with comorbidities and impaired cognitive status and for the oldest members of the sample. Elliot et al. 264 did not report their recruitment process; however, in their discussion they highlighted the potential for bias, in particular in patients who were allocated to the treatment ward based on staff judgements that they would benefit more from the intervention than from usual care.

The studies were carried out in different countries, making cost and care pathway comparisons difficult. In particular, there was a lack of UK evidence; of the seven economic studies identified, only one came from the UK. Three methods of economic evaluation were identified in the included studies: cost-effectiveness analysis, cost–benefit analysis and cost-minimisation analysis. Each of these methods uses different techniques to value and compare the costs and effects of an intervention. Each study compared a different rehabilitation pathway, intervention or setting, making comparisons difficult, although Olsson et al. 260 and Cameron et al. 253 both championed patient-centred approaches to rehabilitation such as early assessment and early ambulation/mobilisation. There was also a lack of recent economic evaluations; the latest evaluation included in the review was published in 2009.

Another limitation noted was the lack of detail when describing how the costs of the interventions were calculated to help determine exactly what costs were included in the analyses. As shown in Table 3, the majority of studies stated that only direct hospital costs were considered. Certain studies gave examples of these costs, but this was not consistent across the included studies. Although Olsson et al. 260 showed that the integrated care pathway was cost-effective in a Swedish hospital, they did not state if the costs of the integrated care pathway, including any development costs, were included with the direct hospital costs, such as treatment and hotel costs, in the cost-effectiveness ratio. Kramer et al. 261 provided a limited explanation of how the cost data were handled and no price year was reported in the paper. A more detailed description of how costs were collected and used for the purpose of the analyses would have been useful.

Implications for practice and future research

This review is in agreement with the NICE hip fracture guidance7 and the Cochrane review by Handoll et al. 8 in finding that there is a limited evidence base for multidisciplinary hip fracture rehabilitation, particularly from UK settings. Five of the seven interventions studied were shown to be a good use of resources, using cost-effectiveness, cost–benefit or cost-minimisation analysis. However, the range of methods used and the range of interventions studied did not allow us to make firm conclusions about any one type of rehabilitation intervention. Despite this, two of the cost-effectiveness analyses showed that patient-centred approaches to rehabilitation were a good use of hospital resources. However, more economic evidence is needed, particularly from the UK.

Aim

To investigate and describe current UK NHS hip fracture physical rehabilitation service provision for patients aged ≥ 65 years in both acute and community settings who have had dynamic hip screw or hemiarthroplasty surgery for proximal hip fracture.

Objectives

To describe the variability in current practice and service provision, and obtain examples of good practice and respondents’ views on how provision could be improved.

Questionnaire design and pilot

National Institute for Health and Care Excellence guidance on hip fracture rehabilitation7 was used as the starting point for developing the survey questions because it outlines best practice recommendations based on current evidence. The particular recommendations were those pertaining to rehabilitation:

• orthogeriatric assessment

• early mobilisation and physiotherapy

• early assessment of cognition

• early identification of individual goals for multidisciplinary rehabilitation to recover mobility and independence

• facilitate return to pre-fracture residence and long-term well-being

• ongoing multidisciplinary team co-ordination and review

• liaison or integration with related services, particularly mental health, falls prevention, bone health, primary care and social services

• offer patients (or, as appropriate, carers and/or family members) information about treatment and care.

The questions were designed to provide descriptive data on how these recommendations may have been operationalised in clinical practice and service organisation. In addition, patient mood,11 self-efficacy13,272 (defined as a belief in one’s ability to organise and carry out/execute a general or specific action273) and fear of falling12 have been shown to impact on rehabilitation outcomes following hip fracture as well as elective hip surgery and so we also asked all respondents whether or not these were routinely assessed and, if so, what measures were used. We also asked respondents for their views of good practice and where improvements could be made.

Most elements of rehabilitation that this project focused on were delivered by both physiotherapists and occupational therapists across different health-care settings and so we chose to survey these professions from both the clinical and the service management perspectives. As a result, three versions of the survey were developed to address the different professional roles of potential respondents: one for physiotherapists (see Appendix 21), one for occupational therapists (see Appendix 22) and one for hip fracture centre therapy service managers (see Appendix 23). The therapist versions were further subdivided by type of health-care setting, that is, for those working in acute hospitals, community hospitals or similar community inpatient facilities and community-based teams providing services to patients in their own homes, including care/nursing homes.

By combining the answers from the different versions of the survey we sought to describe the organisation of services, including multidisciplinary working, working across acute and community settings in terms of discharge planning, liaison and signposting to supportive social groups and activities, and coverage of rehabilitation services. We also planned to be able to describe current clinical practice in terms of assessment and the content of routine rehabilitation.

Demographics

All versions of the questionnaire included questions on the location of the service and the hip fracture centres served, the respondent’s role in hip fracture rehabilitation and the profession of the respondent. In addition, the therapist versions asked for the respondent’s NHS clinical band and type of setting in which they worked (to direct them to the appropriate section of the questionnaire).

The study manager developed the questions in consultation with the chief investigator. The content of the questions and response options for the physiotherapy and occupational therapy questionnaires were developed by the study manager and the physiotherapist and occupational therapist members of the research team. Other members of the research team then commented on all of the questions to further refine them. All of the questionnaires were piloted on members of staff across one health board in Wales. This organisation has responsibility for both acute and community services and so all versions of the questionnaire could be piloted within this one organisation. The pilot was used to assess the content of the questionnaires and the functionality of the web-based survey. A few minor amendments were made following feedback from this pilot, for example adding ‘other’ as a response option for questions and correcting questions that referred to the wrong professions.

Organisation of services

The survey of managers focused on the organisation of services and included questions concerning:

• the availability of different acute and community service provision models

• the involvement of different professions in the multidisciplinary rehabilitation team both in acute centres and in the community

• discharge planning and co-ordination with community services and signposting to social support or activities

• the proportion of patients discharged from acute centres to their own homes who received rehabilitation after discharge.

A few service organisational questions were also included in the two therapist questionnaires. As for the managers, acute therapists were asked about the types of wards where patients were treated and how weekend care was organised. Community therapists were asked about their involvement in multidisciplinary team discharge planning meetings. Community hospital therapists were asked about discharge timing, that is, the length of time post-operatively before patients were usually transferred from an acute hospital.

All respondents were asked if they routinely signposted patients to formal/informal social support services or activities. They were also asked if they used a hip fracture integrated care pathway that included rehabilitation and, if so, to send it to the research team. Physiotherapists were also asked if they used patient information leaflets about exercises and, if so, to send examples to the research team.

Clinical practice

The physiotherapist and occupational therapist questionnaires addressed:

1. assessment, including its timing

2. the content of routine rehabilitation

3. the frequency and length of sessions

4. details of any home visits carried out before discharge from inpatient care and for which patients

5. whether or not they routinely assessed cognitive status, mood, self-efficacy or fear of falling, which tools were used and which profession conducted these assessments.

In addition, physiotherapists were asked on which post-operative day patients were mobilised.

Areas of good practice and those needing improvement

Open-ended questions were used to allow respondents to comment on good aspects of their service and anything that they thought could be improved. Open-ended questions were also used to give respondents the opportunity to add any further comments about aspects of rehabilitation not covered by the survey.

Data collection

The survey was open for 7 weeks from 6 August 2013 to 25 September 2013.

Strategies used to reach respondents

Analysis

Descriptive statistics were used to provide frequency (counts, percentages) data concerning current services and practice when the answer format provided predetermined response options. When the response format was open-ended, responses were coded and categorised into themes. The integrated care pathways and physiotherapy exercise sheets returned to the team were qualitatively reviewed to provide a description of commonalities and differences.

Demographics

In total, 210 respondents completed the survey, consisting of 13 managers, 57 acute inpatient physiotherapists, 29 community inpatient physiotherapists, 43 community team physiotherapists, 37 acute inpatient occupational therapists, 18 community inpatient occupational therapists and 13 community team occupational therapists. Of the 70 hospitals contacted in relation to the survey of therapy service managers, 25 agreed to take part, with 13 actually completing the survey, as shown in Appendix 24. It was our aim to recruit therapy service managers from 10% of UK hip fracture centres to which we achieved 7%. Although 24 centres agreed to take part in the survey of managers, the survey links may have been passed on within the centre, resulting in respondents answering a more relevant version of the questionnaire. An example of this is that, although we were unable to recruit any managers from Northern Ireland, we did receive completed surveys from some therapists in Northern Ireland.

Geographical spread

Respondents were geographically spread across the UK, with respondents from community hospitals and teams generally being from the surrounding areas of respondents from acute hospital teams. Of the respondents to the survey of therapy managers, two were from Wrexham and two were from Portsmouth.

Respondents’ roles/job titles

The exact job roles and titles of the different groups varied, although managers were generally principals, leads or heads of departments. Eleven of the 13 manager respondents were physiotherapists and two were occupational therapists. In all three settings (acute hospitals, community hospitals and community teams) the majority of the physiotherapist and occupational therapist respondents reported spending most of their time delivering front-line clinical care, with some also having some management responsibility.

Table 4 denotes the workload structures for physiotherapists and occupational therapists in different settings.

Service organisation

Rehabilitation in acute hospitals

For all of the responding centres, the multidisciplinary rehabilitation team in acute hospitals included physiotherapists, occupational therapists and nurses. An orthogeriatrician was involved in 11 of the centres with general geriatricians involved in the others. A social worker and pharmacist were involved in 10 of the centres. A dietitian was routinely involved in five centres and a mental health professional in four. Some respondents also mentioned the pain team.

Type of hospital ward where post-operative rehabilitation was delivered

The wards used for post-operative rehabilitation by the 57 physiotherapist and 37 occupational therapist respondents from acute hospitals are described in Figure 4; the majority of respondents worked on orthopaedic trauma wards. Twenty (54%) occupational therapist respondents routinely saw proximal hip fracture patients pre-operatively.

FIGURE 4.

Acute hospital wards used for post-operative rehabilitation by physiotherapists and occupational therapists. GORU, geriatric orthopaedic rehabilitation unit; MARU, mixed assessment and rehabilitation unit.

Frequency and length of therapy sessions

Sixteen (28%) physiotherapy respondents gave physiotherapy twice a day, 12 (21%) once or twice a day, 27 (47%) once a day, one (2%) every other day and one (2%) only two to three times a week. The length of the sessions is shown in Figure 5.

FIGURE 5.

Length of physiotherapy and occupational therapy sessions in minutes.

The number of times that occupational therapists in acute inpatient hospitals saw patients was variable. The minimum length of these sessions was 15 minutes and the maximum length was 2 hours (see Figure 5), with a mode of 30 minutes.

Rehabilitation in community hospitals

Occupational therapists from community hospitals reported that their routine rehabilitation consisted of prescribing specific equipment (such as long-handled aids) (n = 18, 100%), practising various ADL (n = 18, 100%), education about hip precautions (n = 15, 83%), providing information about falls services (n = 12, 67%), other activities to encourage independence (n = 12, 67%), falls prevention techniques (n = 15, 83%), anxiety management (n = 6, 33%) and developing self-awareness (n = 3, 17%). Eleven (61%) respondents referred some patients to a falls prevention service, 16 (89%) referred some patients to social services and 12 (67%) referred some patients to a discharge team. Seventeen (94%) respondents carried out home visits with patients prior to discharge and all made a home environment visit without the patient. Other duties included promoting patient empowerment and increased confidence in patients’ ability, teaching relaxation and pain management techniques, and providing advice to carers and patients about good moving and handling practices and joint protection.

Rehabilitation in the community

According to the occupational therapist respondents from community teams routine rehabilitation consisted of prescribing specific equipment (n = 13, 100%), practising various ADL (n = 13, 100%), education about hip precautions (n = 13, 100%), providing information about falls services (n = 12, 92%), other activities to encourage independence (n = 12, 92%), falls prevention techniques (n = 13, 100%), anxiety management (n = 6, 46%) and developing self-awareness (n = 6, 46%). Seven (54%) respondents referred some patients to a falls prevention service, 10 (77%) referred some patients to social services and one (8%) referred some patients to a discharge team. Other duties included checking medication, providing support and advice to family members, partners or carers and referring clients to intermediate care if they had complex needs.

Physiotherapy rehabilitation exercises

All physiotherapist respondents from community teams indicated that they used strengthening exercises, with 39 (91%) respondents indicating that they used progressive resistance training with some of their patients. Forty-one (95%) respondents used weight-bearing exercises, whereas only 34 (79%) used gait training. None of the physiotherapists used treadmill gait training. Other exercises used are listed in Box 4. Thirty-nine (91%) physiotherapist respondents from community teams gave their patients exercise sheets. In terms of ADL, 42 (98%) physiotherapist respondents from community teams encouraged walking and climbing stairs and 39 (91%) encouraged transferring. Other ADL addressed included personal care and meal preparation, stair climbing, accessing the community, outdoor walking, car and bus transfers, feeding the cat, walking the dog and caring for a relative. The frequency of physiotherapy sessions varied from patient to patient but was either weekly or fortnightly. The minimum length of therapy sessions was 15 minutes and the maximum was 90 minutes, with a mode of 30 minutes (Figure 6).

FIGURE 6.

Length of community physiotherapy and occupational therapy sessions.

Summary of survey findings

Survey respondents were geographically spread across the UK. Therapy service managers were principals, leads or heads of departments. Most therapy service managers had a written integrated care pathway for patients with proximal hip fracture that included their rehabilitation. In contrast, less than half of the physiotherapists and occupational therapists in all three settings stated that they had a written integrated care pathway. The initial assessment of patients varied between pathways but often included a specific falls assessment for identifying patients at risk of further falls.

Acute hospital rehabilitation consisted of strengthening exercises, the practice of ADL, education about hip precautions, the provision of information about falls services, other activities to encourage independence, the prescribing of specific equipment, falls prevention techniques, anxiety management and the development of self-awareness. Patients were mostly seen and mobilised by a physiotherapist the day after surgery and seen by an occupational therapist any time from the same day as surgery to up to 4 days later. How often a patient received physiotherapy and occupational therapy varied depending on patient need. In terms of routine assessment, acute hospitals measured cognitive status and assessed mood, self-efficacy and fear of falling. In most cases, patients were referred or signposted to social support or social activities on discharge. Patients were transferred into the community hospital variably according to patient and departmental factors.

Community hospital rehabilitation consisted of strengthening exercises, individual assessment of functional tasks, personal and domestic ADL and the environment, the prescribing of specific equipment and education about hip precautions and falls prevention. In most cases, progressive resistance training was not used. Most patients were given exercise sheets. As with the acute hospital, the frequency with which patients were seen by a therapist varied depending on patient need. Community hospitals measured patients’ cognitive status and assessed mood, self-efficacy and fear of falling. However, routine assessments using validated tools were not carried out in all centres and how often progress was assessed varied between centres.

Not all patients received community rehabilitation after acute hospital discharge. For those who did, community rehabilitation consisted of prescribing specific equipment, practising various ADL, education about hip precautions, providing information about falls services, other activities to encourage independence and falls prevention techniques. Most therapists said that they referred some patients to social services. As with community hospitals, routine assessments using validated tools were made of cognitive status, mood, self-efficacy, fear of falling, disability, mobility and balance.

Good points according to managers and therapists

The survey found that the range of rehabilitation programmes provided in all three settings had similar goals and highlighted the importance of multidisciplinary teams. Therapy service managers felt that the close collaborative work of the multidisciplinary team was one of the most positive aspects of their service, which contributed to improved discharge times and better patient care, and that having the same team throughout a patient’s hospital stay was beneficial. When asked to comment on good aspects of their service, more than half of the acute hospital physiotherapist respondents and most of the occupational therapists highlighted the benefits of having a multidisciplinary team for patient treatment. Similar numbers of respondents also highlighted that they were able to provide a timely response to patient care. Community hospital physiotherapists frequently commented on their ability to carry out thorough assessments and provide intensive input to patients, allowing them to tailor rehabilitation to individual patients’ needs. More than half of the community hospital occupational therapists indicated that they followed a successful client-centred approach. Some of the respondents also highlighted the good multidisciplinary team working. The majority of community team physiotherapists mentioned the relationship with the patients and their ability to be responsive and flexible to tailor treatment to individual patients’ needs. More than half of the community team occupational therapists felt that the amount of time that they were able to spend with patients, particularly in their own home, was a positive aspect of their service.

Areas for improvement according to managers and therapists

For therapy service managers, areas for improvement included better liaison between acute hospitals and community services, better access to rehabilitation beds and more therapy/nursing resources. More than half of the acute hospital physiotherapists wanted more resources for staff, more beds in community hospitals and more rehabilitation equipment; acute hospital occupational therapists wanted better communication within the multidisciplinary team, better external services (such as falls groups) and the provision of more staff to enable more time to be spent rehabilitating patients. The main area for improvement mentioned by more than half of the respondents from community hospitals and community teams was staffing levels. The community teams also wanted better links to hospitals.

Strengths and limitations of the survey

This is the first UK-wide survey aiming to describe rehabilitation for patients following hip fracture across acute and community settings since the introduction of NICE guidance in 20117 including recommendations for rehabilitation. A wide range of respondents was sampled in terms of profession, health-care sector and geographical spread of hip fracture centres. The link to the survey was easily circulated by e-mail to relevant potential respondents from the hip fracture centres. Because of data protection issues, the professional organisations could not provide us with their contact lists, but they did let their members know about the survey and provided the link to the survey itself. We were also able to place a notice about the survey on the National Hip Fracture Database website news section. We also encouraged people to pass on information about the survey to colleagues. As there is no register or centrally held record of physiotherapists and occupational therapists working in hip fracture rehabilitation, we were not able to establish the total number of these professionals in the UK or use such a register as a sampling frame.

We recruited therapy service managers by telephone, which enabled us to provide them with detailed information about the study as well as ask them to circulate the survey to their colleagues. Although this method did work well, with many asking to be informed of the results of the study and being happy to circulate the survey to their staff, it was time-consuming.

It is possible that our survey results may not have been entirely representative of the UK situation regarding rehabilitation for hip fracture patients because we were unable to sample settings and therapists and community service managers proportionately. However, we did sample acute hip fracture centre managers proportionately, aiming for a 10% sample of the hip fracture centres in the UK and obtaining a 7% sample. However, no acute hip fracture centre managers from Northern Ireland responded to the survey and so we were unable to obtain the perspectives of managers from Northern Ireland. We did, however, receive responses from therapists in England, Wales, Scotland and Northern Ireland. Our strategy for reaching respondents resulted in 210 respondents completing the survey, with a good geographical spread of respondents from different acute and community settings.

As with all surveys, the findings provide only a description of what was happening; causality for the variation found could not be demonstrated and may warrant further investigation.

Comparison with previous literature

Respondents in both acute and community hospital settings mostly reported that routine clinical practice followed the latest NICE7 and SIGN74 guidance. However, there was variability in the provision of services, especially in terms of what was available in the community. This variability in service provision has been reported elsewhere, for example in a report of physiotherapy services for rheumatoid arthritis. 289 Similar findings are reported for the focus groups involving hip fracture patients, described in Chapter 5.

Staffing levels were reported to be an issue that impacted on the level of service that could be provided in the community. This finding agrees with an earlier report of NHS physiotherapy waiting times and workload with regard to hip fracture and other conditions requiring rehabilitation. 290

Important psychological issues, such as fear of falling and self-efficacy, were not measured in many settings, although the realist review discussed in Chapter 2 found them to be important components of a successful rehabilitation programme. Previous qualitative research with Australian physiotherapists planning discharge home following hip fracture found that lack of confidence was seen as a barrier after hospital discharge but that this was not one of the discharge criteria. 185

Communication within teams in a particular setting was often seen as good; however, communication across boundaries between community services and acute services was more of a challenge. This has been noted previously. 176

The findings that therapy managers appreciated the role of integrated care pathways in streamlining the flow of patients rapidly through the health-care system, and that many of the health-care staff providing treatment felt that such a checklist approach devalued the human aspect of dealing with patients as individuals, agreed with the results of previous research. 176

Implications for rehabilitation programme

The variability in rehabilitation programmes could be reduced, particularly in the community, but not at the expense of tailoring programmes to individual need. There needs to be greater awareness of available resources for patients, carers and clinicians, with more standardised referral procedures and less reliance on the need for patients and their carers to push for access to services. Communication should be improved between acute hospitals and community services, and also with patients and their carers. There should be a more consistent assessment of important prognostic variables such as self-efficacy and fear of falling, perhaps by the adoption of a set of measures that could be used to evaluate the progress of patients during the course of their rehabilitation.

Aim

To describe the experiences of multidisciplinary rehabilitation programmes in the North Wales area from the perspective of health-care professionals, patients and carers.

Objectives

To assess the views of patients and their carers on:

• the rehabilitation that they received following surgical repair of a proximal hip fracture

• how the programme could be improved

• the intervention ideas emerging from the different strands of work in Phase I.

To assess the views of health professionals in the multidisciplinary rehabilitation teams on:

• the rehabilitation programmes currently available to patients following surgical repair of a proximal hip fracture

• how the programmes could be improved.

Ethics approval

The study received ethics approval from the UK NHS North Wales Research Ethics Committee – Gwynedd and Anglesey (reference number 12/WA/0355) and NHS research and development (R&D) approval from the BCUHB Internal Review Panel.

Focus groups with rehabilitation team members

Three focus groups were organised with members of multidisciplinary rehabilitation teams in the community around three North Wales hospital sites [north west (Ysbyty Gwynedd and surrounding areas), central (Ysbyty Glan Clwyd and surrounding areas) and north east (Ysbyty Maelor and surrounding areas)].

Focus groups with hip fracture patients and their carers

Three focus groups with patients and carers were originally planned, also based around the three main hospital sites within the BCUHB; however, as more than eight participants responded for one of the groups, two focus groups were run at this site.

Data collection

The focus group discussions were semistructured and run by a moderator (the study manager) and co-moderator (study researcher) using a topic guide291,292 that contained open-ended questions regarding experiences of, and perceptions and beliefs about, rehabilitation following proximal hip fracture (see Appendix 25). In the later patient focus groups we were able to start exploring and gaining feedback on initial ideas for the intervention.

As the focus groups were held in an area where two languages are spoken (Welsh and English), we allowed participants to choose which language they wanted to participate in. As a result, one patient and carer focus group was run with the use of simultaneous translation, to enable all participants to discuss their personal experiences using the language that they were most comfortable with. In this focus group a third moderator (study researcher) was present who was fluent in both languages to ensure equality of access to both English- and Welsh-speaking participants. All of the study documentation was provided in both languages.

In the focus groups with professionals, in addition to the topic guide, we developed patient scenarios to stimulate discussion about the sorts of rehabilitation programmes patients would be likely to receive (see Appendix 26).

The focus group discussions were digitally recorded and fully transcribed. The transcriptions were made in the original language and any transcripts in Welsh were translated into English for analysis.

Analysis, credibility and plausibility

The focus groups were thematically analysed using the framework approach. 293 The initial framework used for this analysis was broadly developed from the theory areas identified as important in guiding the realist review (see Chapter 2). The framework (see Appendix 27) was used to index the transcripts and was populated with summaries of data by one researcher (the study manager). This was carried out separately for the patient and carer focus group data and the professionals’ focus group data. Once summarised, each subsection of the index was transferred onto a new sheet for coding. The coding was completed for both sets of data for a particular subsection of the framework in sequence, to begin identifying similarities and differences in the perspectives and to start the process of synthesising the analysis. The researcher was able to add reflective notes as the coding progressed. At the end of each section, emerging key ideas and issues were noted for development into categories and subcategories.

Once this process was completed for all sections of the framework, the researcher developed an initial interpretation using the categories, grouping them into themes. At this point a second researcher experienced in framework thematic analysis, and who had been a co-moderator in one focus group but who was otherwise independent of the study, reviewed the charts and agreed the plausibility of the emerging themes.

A third researcher involved in the study then reviewed the initial framework, the original transcripts and the draft of the initial analysis. This researcher then made the final decisions about theme structure and content. The initial and third researchers agreed on the final analysis. Involving three researchers in the analytical process strengthened the plausibility and credibility of the findings, as the identified themes were not generated from one person’s perspective. The indexing of the transcripts and the framework charts provided an audit trail connecting the themes back to the original data.

Patient and public involvement

A patient and public involvement representative commented on all aspects of the study protocol,294 including the content of the topic guides, and was actively involved in the design of the patient-facing materials, such as the information sheets and consent forms.

Description of participants

Thirteen health-care professionals involved in delivering rehabilitation throughout the patient pathway consented to take part in the focus groups. These were:

• one physiotherapist from an acute hospital orthopaedic trauma ward

• one nurse manager of an acute hospital orthopaedic trauma ward

• three occupational therapists from three different acute hospital orthopaedic trauma wards

• one physiotherapist from a community hospital providing inpatient rehabilitation

• one physiotherapist from a community rehabilitation unit who also saw patients at home

• one physiotherapist from a community intermediate care team

• one occupational therapist from a community intermediate care team

• one occupational therapist from a local authority social services team

• two local authority domiciliary support team managers (teams that provide reablement care as part of their remit)

• one co-ordinator of a local authority leisure services centre providing the National Exercise Referral Scheme295 (a national scheme in Wales).

Thirteen hip fracture patients (11 women and two men) and four associated carers (two women and two men) consented to take part in the focus groups. Three of the carers were spouses and one was a friend. Patient and carer participant ages ranged from 60 to 90 years and participants were from both rural and urban areas of North Wales.

The numbers of participants in the patient and carer focus groups ranged from three to seven and in the health-care professional focus groups from two to six. Some participants in the health-care professional focus groups were unable to attend at short notice because of work commitments and two participants in the patient and carer focus groups were unable to attend because of illness.

Focus group themes

Four main themes emerged from the data analysis:

1. the variation in rehabilitation care provided

2. the need for information

3. facilitators of, and barriers to, rehabilitation

4. the psychosocial impact of hip fracture.

Summary of main findings

Four main themes emerged from the focus groups: (1) the variation in rehabilitation care provided, (2) the need for information, (3) the facilitators of, and barriers to, rehabilitation and (4) the psychosocial impact of hip fracture. The variation in care provided occurred partly because of the individual tailoring of treatment, but was also the result of geographical variation in the availability of resources, the complex variety of providers delivering different components of rehabilitation programme and the lack of awareness by referring clinicians of the full extent of available services. This complexity in programme provision meant that there was a strong need for more information to be provided to patients and their carers about what to expect following hip fracture and how to access all of the resources available for rehabilitation. Facilitators of, and barriers to, rehabilitation included the reliance on patients’ self-motivation for seeking out and accessing services, and also their confidence to do so, their engagement in the rehabilitation programme, which could depend on its relevance to their day-to-day activities, and their self-efficacy beliefs, as well as their pre-fracture level of functioning and comorbid conditions. As rehabilitation progressed, some flexibility in response to changing patient needs was required, such as requiring access to transport for community services when home visits were no longer required. Good co-ordination and communication between the different components of the programme were important; as different components were provided by different organisations or services, it was evident that failures in communication had the potential to cause problems at various stages along a patient’s rehabilitation journey. Finally, in addition to physical needs, falling and fracturing had an important psychosocial impact in terms of anxiety, fear of falling, loss of independence and loss of participation in previously enjoyed activities.

Strengths and limitations

Although the best efforts were made to ensure that participants in the health-care professional focus groups included representatives from across the whole multidisciplinary team involved in post-fracture care and rehabilitation, it was not possible for members from all disciplines to attend, for example pharmacists or orthogeriatricians. Therapy managers involved in the organisation of wider teams were also absent from the focus groups. Although the inclusion of such therapy managers may have provided further insights into the organisation of resources across areas, there is the potential for the presence of senior staff to limit the discussions generated from more junior members of teams. As the groups consisted generally of staff at similar levels of seniority, the inclusion of such therapy managers was avoided, and good levels of discussion were seen between focus group members. Representatives from social services from one council were able to attend one focus group, which was useful to obtain their perspectives; however, there are several other councils in this area so these perspectives may be limited, particularly in relation to the geographical variation that was apparent.

Similarly, the absence from focus groups of patients or the carers of patients who were living in residential care, or who had cognitive impairment, limits the breadth of information gathered from the patient and carer perspective. The patients who were able to attend had a broad range of functional abilities and comorbidities, allowing us to gain insight from patients with different experiences of rehabilitation. However, it is possible that those patients who declined to take part in the focus groups or who were unable to attend because of ill health may have been those who were struggling the most with their recovery.

Comparison with previous literature

There are significant similarities between the focus group findings and other published reports of qualitative research in comparable groups. Other studies have found a similar spectrum of patient participants with varying levels of post-fracture function and a range of levels of dependency on carers and available health services,296 but generally displaying strong self-motivation and ‘zest for life’. 88 A study exploring mobility levels pre and post fracture also reported a fear of falling, a lack of confidence (low self-efficacy) and reliance on others as having an impact on patients’ experiences of rehabilitation. 297 Patients and health-care professionals expressed a difficulty in predicting post-fracture self-efficacy and confidence, which has also been evidenced in a review of surgical outcomes in total hip and total knee replacements. 272 This review concluded that post-operative self-efficacy was a more consistent predictor of functional outcomes than pre-operative self-efficacy, highlighting the importance of feedback on post-operative achievements for continuing rehabilitation engagement. The patient need for information from health-care professionals and the importance of this in successful rehabilitation has also been previously identified. 298

A qualitative study of physiotherapist perceptions of rehabilitation also found that there is tailoring of care to patients’ individual needs, based on their own goals and level of support available. 185 A study into the challenges of team working in the rehabilitation of hip fracture patients also demonstrated that there are breakdowns in communication within multidisciplinary teams and issues relating to the organisation of resources and services that lead to variation in patient care. 299 In addition, this theme was prevalent in the survey findings (see Chapter 4).

The description of care across North Wales was also similar to the findings from across the UK, as presented in Chapter 4. There is an emphasis on early mobilisation after surgery and rehabilitation programmes vary according to the resources available and the individual needs of patients.

Implications for practice and future research

The goal of conducting focus groups was to gain insight into the perspectives of patients, carers and health-care professionals on the rehabilitation of patients following hip fracture and highlight areas that could be improved. This has been successfully achieved. The emergent themes encompassed a range of experiences and insights, which were used to inform the development of the study intervention (see Chapter 6) and further refine the theory areas of the realist review (see Chapter 2). In particular, they emphasised the need of patients for information following hip fracture, which led to the development of an information workbook to be used as part of the study intervention, the content of which was partially based on information requirements that were detailed in focus groups.

The variability in care provision was an important factor in both patient and professional perspectives of what constitutes a good package of rehabilitation, and the decision to provide six additional therapy sessions to intervention group participants aimed to address this issue.

On a wider scale, there are also implications in terms of informing health and social care services about areas that could be improved, particularly with regard to how to communicate information to patients and their families to support them during a time when they report feeling isolated and vulnerable. By including patients more explicitly in the decisions regarding their care, providing feedback on progress and being transparent around expectations, including the lack of service provision or potential treatment delays, it may be possible to improve patient engagement in rehabilitation and allow patients and families to regain control of their recovery.

Study objectives

1. To assess the acceptability of, and compliance with, the rehabilitation programme among patients, carers and clinicians, and identify any AEs.

2. To determine the feasibility of a future definitive RCT by assessing the number of eligible patients, monitoring recruitment and retention rates, and explore the willingness of patients to be randomised and the willingness of patients and carers to complete process and outcome measures.

3. To produce means and SDs for the quantitative measures so that effect sizes can be calculated for planning the future RCT.

Study design

Phase II encompassed the second stage of the MRC framework for assessing complex interventions,20 using a parallel-group randomised feasibility study to inform a definitive RCT comparing usual care with the enhanced rehabilitation package. It also consisted of a cohort study of all hip fracture patients (see Chapter 8). The acceptability of the new rehabilitation programme to multidisciplinary rehabilitation teams, hip fracture patients and their carers, as well as its feasibility, were assessed with further focus groups (see Chapter 9). An exploratory economic evaluation was conducted as part of the feasibility study (see Chapter 10). Patients recruited to the feasibility study were also recruited to a triangulation study that aimed to compare the quality of data on service use collected from patient-reported outcome measures with the quality of the same information extracted from patient electronic records (see Chapter 11).

Feasibility study interventions

We compared the enhanced rehabilitation intervention with usual rehabilitation care. Usual care was varied and consisted of a multidisciplinary rehabilitation delivered by the acute hospital, community hospital and community services depending on patients’ individual needs at different times during their recovery and on the availability and accessibility of services in different areas. The multidisciplinary teams delivering care and rehabilitation included orthopaedic surgeons, orthogeriatricians, nurses, physiotherapists, occupational therapists, dieticians, pharmacists, GPs and social workers. The settings for care included acute orthopaedic or orthogeriatric wards, rehabilitation units in community hospitals, rehabilitation beds in care homes, patients’ homes and care home settings.

The main aim of the intervention was to enhance usual rehabilitation by increasing the amount and quality of patients’ practice of physical exercise and ADL to improve their functional outcomes at 3-month follow-up. We also hypothesised that improving patients’ self-efficacy would increase their motivation to engage in the rehabilitation process and improve the quality and quantity of this practice.

Rehabilitation was enhanced by means of a patient-held information workbook and diary given to participants in the acute hospital and kept with them throughout the follow-up period of the study. Six additional therapist/technical instructor sessions were available to patients once they returned to their usual place of residence. These extra sessions were tailored to individuals’ needs at the discretion of the therapists allocated to deliver the extra sessions. The workbook also included information about what to expect from their recovery and information about NHS, council and voluntary sector services that they could use. This included a variety of community services such as falls preventions programmes.

The objectives of the workbook and diary and a logic model of the enhanced rehabilitation intervention are provided in Chapter 6.

Assessment of safety

Statistics

Trial management

Patient and public involvement

A patient and public involvement representative was a member of the SSC and commented on all aspects of the study protocol,294 including outcome measurement, and was actively involved in the design of patient-facing materials such as information sheets and consent forms.

Data handling

Indemnity

Bangor University had appropriate clinical trials indemnity and professional indemnity insurance in place that covered members of the research team to conduct the research as per protocol. NISCHR CRC staff were on NHS contracts and their work was appropriately insured. NHS and social services staff who worked with patients involved in the intervention were not expected to do anything that was not covered by their contract and they remained covered by NHS or social services insurance arrangements.

Changes to the protocol

After the trial began some minor changes were made to the protocol. An additional patient information summary was provided alongside the patient information sheet to ensure that patients had a good understanding of the trial before recruitment. Reasons for declining participation that had not previously been collected were added to the screening information taken from participants. The DCE was also shortened to reduce patient burden and aid completion, and a Welsh version of the HADS was made available to Welsh speakers alongside the English version in an attempt to validate the Welsh version of the measure. As a result of recruitment being slow, the recruitment period was increased from 6 to 9 months in Ysbyty Glan Clwyd and Wrexham Maelor. Recruitment was not extended in Ybsyty Gwynedd as resources were not available to continue delivery of the intervention for an additional 3 months.

Demographic characteristics

In total, 61 patients were randomised into the feasibility study and 49 patients subsequently completed the trial. Of the 61 participants, 32 were randomised into the control group and 29 were randomised into the intervention group, showing a good balance between the groups.

The characteristics of the recruited participants are shown in Table 10. Participants ranged in age from 66 to 99 years. The difference between the mean ages for the two groups was 2.9 years, which, within the 33-year age range, is a reasonably small difference.

Most of the participants (75%) were female. The proportions of men and women were similar in both groups, which would be expected as gender was used as a stratification variable.

There was an even split in both groups between participants who lived alone and those who lived with others before being admitted to hospital. The majority (79%) lived in owner-occupied properties before they were admitted to hospital. After discharge there was a small difference between those who were discharged directly to their original accommodation (27 patients) and those who were not (22 patients). This was not consistent between the two groups, with 17 (53%) in the control group discharged directly home but only 10 (34%) in the intervention group discharged directly home.

The most common type of fracture was intracapsular (44%). There was a slight difference between the two treatment groups, with equal numbers in the intervention group having intra- and extracapsular fractures but more people in the control group having an intracapsular fracture than an extracapsular fracture.

Recruitment at each of the three centres was not equal, with 34 patients recruited at Ysbyty Glan Clwyd in Denbighshire, 16 patients recruited at Wrexham Maelor and 11 patients recruited at Ysbyty Gwynedd in Bangor. Hospital site was a stratification variable within the randomisation process to ensure that the proportions from each site were similar between the groups, which worked effectively.

Mean scores for most of the outcome measures and physical function tests at baseline were similar between the groups (see Table 17); however, there was an imbalance in NEADL scores, with a higher mean score in the control group.

Participant flow

Figure 10 shows the Consolidated Standards of Reporting Trials (CONSORT) diagram, detailing the flow of participants through the trial.

FIGURE 10.

Consolidated Standards of Reporting Trials (CONSORT) diagram.

A further breakdown by recruitment centre of the number of patients screened, the number of eligible patients and the numbers recruited and retained is provided in Tables 11 and 12 and Figure 11.

FIGURE 11.

Numbers of screened, eligible, recruited and retained patients within the trial by recruitment centre.

The overall eligibility rate was just short of the 50% target. There was some variation in eligibility rate between the centres, with Ysbyty Gwynedd reaching the target, and the other two centres being within approximately 10% of the target, which can be explained by natural variation. The main reason for ineligibility was a lack of mental capacity, which accounted for 49% of ineligible patients.

The recruitment rate was within the 95% CI of the 25% target. This rate differed between centres, with Ysbyty Glan Clwyd recruiting 34% of the eligible patients and the other two centres recruiting lower percentages of the eligible patients. Patients were recruited after 193 of 266 (73%) eligible patients were approached, with 176 (91%) of these agreeing to talk to the researcher. Those who were not approached lived in areas where it was not possible to deliver the intervention or were deemed by clinical staff to be too ill to take part in the study; in addition, in some cases there were safety concerns because of lone worker policies that would have prevented the intervention being delivered.

The CONSORT diagram in Figure 10 details the reasons for non-recruitment of eligible patients to the study, with the most common reason being the perceived study burden. Information concerning the number of visits it took to recruit participants was collected for 36 patients. The majority of patients had two visits and the researchers reported that, as recruitment occurred early in patients’ recovery from surgery, many requested a return visit to discuss the study after they had been discharged (Table 13). It may be worth considering the timing of when patients are approached in a future definitive RCT.

The overall retention rate for the trial was satisfactory, but varied between centres. The retention rate in Wrexham Maelor was 100%, but in Ysbyty Gwynedd it was only 36%. This centre did encounter particular difficulties accessing staff for the trial, which might partly explain the poor retention rate observed. Nine of the patients who did not complete the study withdrew and the remaining four could not be contacted and so were lost to follow-up.

Alongside the recruitment of patient participants we also identified and recruited their carers. In total, 41 carers were identified and 31 (76%) of these agreed to participate. Six carers withdrew during the study and seven were lost to follow-up, leaving 18 (58%) carers who completed the follow-up questionnaire.

Defining usual therapy care

Data were collected from Therapy Manager manually by a researcher accessing individual patient entries, identified by name, address, date of birth and hospital ID number.

Of the 35 participants recruited at Ysbyty Glan Clwyd, all had entries on Therapy Manager relating to their post-fracture care in the acute hospital. Six of these participants withdrew from the study and no further data regarding usual care were collected. Following treatment at the acute hospital, patients were discharged to their place of residence or for further rehabilitation in a community hospital prior to going home (Figure 12).

FIGURE 12.

Rehabilitation pathways in usual care on discharge from Ysbyty Glan Clwyd following surgical repair of fracture.

Ten patients had no details recorded relating to usual care following acute hospital discharge from Ysbyty Glan Clwyd. Of the 20 patients who did have entries on Therapy Manager, four did not receive any face-to-face appointments with a health-care professional; their entries related to telephone calls to patients who were either uncontactable or who declined further treatment. The 16 patients who did receive an appointment received a median of three appointments (n = 4). The maximum number of appointments for one patient was 21. In total, these 16 patients received a total of 81 appointments, with 73 of these appointments (90%) conducted as home visits. Home visits were completed by different members of the therapy team (Figure 13), whereas 90% of outpatient appointments were conducted by a physiotherapist (10% not recorded). In general, the time between outpatient follow-up appointments was longer than the time between home visits.

FIGURE 13.

Percentage of home visits for usual care conducted by different members of the rehabilitation team.

The role of the health-care professional carrying out a home visit appeared to be dependent on individual patient need. For example, if an assessment was required then a qualified physiotherapist or occupational therapist completed the visit, with subsequent visits, following an agreed care plan, conducted by a technical instructor. It was also common for more than one therapist to attend.

Although there was no universal format for recording how appointments were used, information was available about the activities that team members completed in relation to patients. These activities were categorised as direct or indirect. For usual care, the majority of direct activities involved the practice of ADL (25%) or physical exercise (23%). The remainder of the direct activities involved telephone calls with patients, discussion of progress and assessment of mood. Indirect activities were predominantly referrals to other services (33%) or contact with other members of the multidisciplinary team (30%).

Therapy Manager also allowed health-care professionals to record qualitative data relating to their interactions with patients. It was apparent from entries in the system that information given to patients about waiting times for usual care was inconsistent. In one case a patient was told that the standard wait for an appointment was a minimum of 4 weeks; however, this patient was seen within 1 week after she contacted the relevant department. The therapist who recorded this visit stated that the waiting list was based on patient need, as was the provision of sessions. This did not seem to be clear across the service and there was confusion for both patients and staff. A number of patients also reported to the therapy staff that they were reluctant to engage in physical activity until they had been seen by a physiotherapist, even though in many cases they were told that there would be a wait of at least 4 weeks.

Adherence to and fidelity of the intervention

Twenty-nine patients were randomised to the intervention group and details of intervention session use were available for 20 of these participants. Reasons for missing data are shown in Table 14.

Of the 20 participants with information on intervention session use, the majority (n = 13) received all six sessions. Other participants did not feel that they needed all six sessions; this may have been a characteristic of the younger, healthier subpopulation recruited or may reflect the variability in usual care. The average number of sessions delivered to the intervention participants was five, four of which took place in their own home. One patient randomised to the intervention was discharged from the community hospital to a respite care home and so her intervention therapy sessions were delivered there.

A technical instructor alone conducted the majority (55%) of the intervention sessions, with 44% of sessions conducted by a physiotherapist and the remaining 1% conducted by more than one team member.

As with the usual care sessions, the content of the intervention sessions was determined according to individual patient need and was at the discretion of the therapist conducting each session. Therapists consistently completed the intervention paperwork detailing the types of activities that were carried out in sessions and the time taken for different activities in each session. Each session lasted for approximately 1 hour, with an additional hour spent travelling. Activities were categorised as direct or indirect, as with usual care. In the intervention sessions, a lower percentage of direct activities were categorised as the practice of exercises (15%) or ADL (14%) than in usual care; instead, there was a higher rate of activities such as answering questions raised by the intervention workbook, working with goal-setting diaries, giving feedback on progress and discussing the emotional needs of patients. For indirect activities, only 7% of the time was used for discussion with the wider team and 4% was used for referring to other services. This is likely to be a result of the intervention being used alongside usual care, with the referrals and multidisciplinary team discussions being covered by the usual care therapists. The remaining indirect activities for intervention therapists consisted of travel to appointments, writing notes, arranging further appointments and discussions with carers.

Ten participants returned their goal-setting diaries and workbooks to the study team following completion of their intervention sessions. All of the goal-setting diaries included input from a therapist detailing the goals that were set in the initial assessment session and five had also been updated by the patients and their carers. These participants used the diaries extensively, updating their progress on the initial goals agreed and entered by the therapist and adding new goals into the diary themselves. Three of these participants also completed the quizzes and hip fracture story sections of the workbook. One of the workbooks was also completed by a carer, who detailed the things that had been challenging in the patient’s recovery and what they were doing to overcome these problems.

Measure completion rate

There were good completion rates at baseline for all of the outcome measures (> 90%) (Table 15). Completion rates were reduced at follow-up, when there were particular problems with the GSES, SEE and grip strength test.

Completion rates were based on the final score for a measure being present, with all individual scores having to be completed for the final score to be calculated. In some cases, one question might have been omitted and, as the final score could not be calculated, this measure would be classed as incomplete.

Exploratory correlation analysis

Correlations were calculated to assess the relationships between several variables as prespecified prior to the trial data becoming available. All correlations are shown in Table 16.

Exploratory analysis of the primary outcome

Exploratory analysis was performed on the primary outcome measure, the BADL index. An ANCOVA was completed, which included the participants’ baseline score as a covariate, with the aim of achieving an effect size along with a 95% CI. 330 This resulted in an extremely low effect size of 0.29 (95% CI –0.31 to 0.89), which suggests that there was no discernible difference between the two treatments within this study. 331 All assumptions were tested and satisfied before completion of the analysis.

Exploratory analysis of the secondary outcomes

Further exploratory analysis was performed on the remaining outcome variables and physical function tests using the same methods. The baseline and follow-up scores, the differences between the groups and the effect sizes with CIs are shown in Tables 17 and 18. For the variables contained in Table 17, an ANCOVA controlling for baseline score was used to evaluate the effect sizes, and the mean differences shown in these tables have been adjusted for the baseline scores. When baseline scores were not collected, a t-test was completed on the follow-up data to obtain the effect sizes and 95% CIs (see Table 18). All analysis was completed on an intention-to-treat basis.

Many of the effect sizes were extremely small; however, when considering the NEADL scale an effect size of 0.63 was found, which is considered a medium effect size. For this measure, the intervention group therefore achieved better results than the control group.

One further variable, the 50-foot walk test, had a Cohen’s d effect size of 0.40, which is a reasonable size. This variable represents the physical function of the patient and the results show that the control group completed the 50-foot walk in a shorter time than the intervention group, which suggests that the control group had better physical function than the intervention group. Those in the control group received their physical function appointment, on average, 22 days later than those in the intervention group because of the relocation of physiotherapy services in the central area where the majority of participants were recruited from, limiting access to the gym required for carrying out the tests. This may have affected the results as control group participants had longer to recover from surgery than intervention group participants. However, there was also one outlier in the control group who affected this result and so a sensitivity analysis was completed with this outlier removed. This resulted in an effect size of 0.02 (95% CI –0.80 to 0.84), which suggested that there was no longer a difference between the groups. In the analysis, only data for the physical function tests from patients who completed the tests unaided were used. When possible, participants completed the tests as intended; however, when problems occurred they were permitted to complete the test with the aid of a walking stick, using the arms of the chair, etc. These results were recorded but were not included in the analysis.

Summary of findings

The trial methods were feasible in terms of recruitment, randomisation and follow-up, although recruitment was initially slow and the recruitment period was extended in two sites for an additional 3 months. In total, 45% of the potential participants screened were eligible, with the main reason for ineligibility being lack of mental capacity. Of the eligible population, 23% agreed to participate; the main reason for non-participation was the perceived study burden. The participation rate of identified carers was 76%. Although recruitment was challenging as patients were approached soon after a traumatic injury and subsequent surgery, the overall recruitment rate achieved was within the 95% CI of the target rate. The completion rate for the baseline questionnaires was > 90%.

The intervention and control groups were similar with respect to gender, living status, type of property, type of fracture, type of surgery and admitting hospital. After hospital admission there was a small discrepancy between the groups for those discharged directly home and those sent to a community hospital for rehabilitation. The baseline scores for the outcome measures and physical function tests were similar between the two groups. However, the NEADL score was higher in the control group.

At 3-month follow-up, the patient retention rate was 79% and the carer retention rate was 44%. The completion rate for the follow-up questionnaires varied from 64% to 100%. There were minimal differences between the two groups for most of the outcome measures, including the main outcome measure, the BADL index. In the intervention group there was a trend for a greater improvement in self-efficacy and mental health, but with small effect sizes. However, the intervention group showed a medium-sized improvement in the ability to perform ADL (as measured by the NEADL scale) relative to the control group. Paradoxically, one of the physical function tests, the 50-foot walk test, reported better function in the control group, with a medium effect size. However, this can be explained by one outlier, as without this participant the effect size was massively reduced. The improvement in the ability to perform ADL suggests that the community-based rehabilitation intervention could be beneficial in enabling participants to regain better levels of independence compared with usual care.

Strengths and limitations

This was a single-centre feasibility study conducted in one local health board at its three acute hospital sites in North Wales. It was able to assess the feasibility of the trial methods in terms of recruitment, randomisation and outcome measurement. It was not designed, or powered, to test the effectiveness or cost-effectiveness of the rehabilitation intervention.

As expected, usual care varied between patients. The provision of physiotherapy services in usual care was based on individual patient need, with variation in the number of sessions provided, how they were used and which health-care professional delivered them. There was less variation in the intervention sessions when considering which type of health-care professional conducted the sessions because of the limited number of staff involved in the delivery of the intervention. There were challenges in delivering the intervention in some areas, but there was good adherence to the planned intervention by those patients who did receive sessions, with a high level of detail provided in the completed intervention paperwork.

We were able to assess the different outcome measures to determine which would be the most suitable for a larger definitive RCT. The NEADL scale was more responsive than the BADL index for measuring the ability to perform ADL. The FES-I and SEE scale were more responsive than the GSES, but the researchers reported that many participants struggled to understand the SEE scale and so the FES-I seems the most appropriate measure of self-efficacy for the main definitive RCT. The most appropriate health economic outcome measure is discussed in Chapter 10.

The physical function test with a medium effect size, the 50-foot walk test, reported better function in the control group than in the intervention group. However, this can be explained by the outlier in the control group, as the effect size was greatly reduced by removing this outlier. This did, however, identify that those in the control group completed the 50-foot walk test later than those in the intervention group, which must be addressed within any future study.

A strength of the feasibility study was the high completion rates for the outcome measures. However, patients’ frailty, lack of manual dexterity and comorbid conditions, such as hearing or visual problems, sometimes meant that assistance was required from carers to complete the measures. This improved the completion rates but also might have introduced bias, as the carers might have influenced how the patients responded to some measures, especially the quality-of-life and mental health measures. 332

Examination of therapist notes using the Therapy Manager System highlighted the variability in usual care, which could impact on the ability to measure differences in patient outcomes between the two groups.

Originally, we had hoped to include participants who lacked mental capacity to assess if the rehabilitation intervention was feasible and acceptable for different levels of cognitive functioning. However, the NHS research ethics committee excluded participants who lacked capacity because the study was not investigating the condition causing this lack of capacity. Participants with cognitive impairment but with sufficient mental capacity were included, but the mean AMTS demonstrated a good level of cognitive functioning in the feasibility study participants. Hence, the findings might not be generalisable to patients with more severe cognitive impairment.

Comparison with previous literature

Previous studies of hip fracture rehabilitation interventions were reviewed in our realist review in Chapter 2. We identified previous studies of rehabilitation interventions that combined the promotion of physical exercise and practice of ADL with psychological interventions designed to tackle self-efficacy and fear of falling, for example the study by Resnick and Jenkins. 311 However, none of these studies was set in the UK and we are not aware of any other RCTs of similar interventions following hip fracture in the NHS.

As discussed, the overall recruitment rate of eligible patients was 23%, with the main reason for non-recruitment being the perceived study burden. Recruitment into trials, particularly those involving surgical treatments, is challenging and has been discussed previously. 333,334 Although the data show that a large number of patients would require screening in a full definitive trial, this was a relatively simple process as all patients admitted to the ward were assessed by the acute physiotherapist and the eligibility criteria for the trial would be included in such an assessment.

Phase II studies are primarily intended to test the feasibility and acceptability of the various trial methods prior to a larger definitive Phase III RCT. However, they also provide an opportunity to compare and contrast different methods, for example in this trial different patient-completed outcome measures for self-efficacy, functional ability and quality of life, and different objective measures of physical function were compared. Measuring outcomes comprehensively is important in health services research to inform improvements in clinical practice and improve patient outcomes. 335 However, reducing patient burden while completing such measures is also important for patient adherence and, hence, a core set of measurement instruments in hip fracture has been advocated. 336–338 Hip fracture populations pose particular challenges to researchers as a numerous outcomes have been reported in the literature (see Chapter 2) and deciding on a core set is difficult. 278 A recent UK consensus study has made recommendations on a core set of outcome measures338 and the current study complied with most of these recommendations, although it differed in the use of the NEADL scale to measure the ability to perform ADL. The NEADL scale is an instrument that measures patients’ actual ability to perform their day-to-day activities, rather than what they presume they can do, as is the case with the BADL index, which this study used as its primary outcome. As previously discussed, changing this will be considered for a future full RCT.

Implications for future research

This was a difficult population to recruit from, particularly because of the large number of patients with cognitive impairments and the perceived study burden. Recruitment was initially slow but the recruitment rate was acceptable and good completion rates for the different measures indicate that obtaining the required data was feasible. In future research it would be beneficial to capture the experiences of those patients who declined to take part to try and identify their needs. The intervention was generally delivered as planned and was well tolerated by participants.

Problems encountered with accessing the physiotherapy gym in some areas led to a delay in the physical function tests being conducted. Using an alternative test, the Short Physical Performance Battery (SPPB),279 could address this problem in future research as it does not require such a long distance to perform the walk test and could therefore be completed in a patient’s home. With sufficient training, this test could also potentially be conducted by researchers, further reducing problems with resource availability that are encountered when a therapist is required to complete the test.

Usual care data included notes from therapists and administration staff who had spoken to patients over the telephone. In some cases it was recorded that patients were reluctant to engage in physical activity without seeing a therapist for an initial ‘fact-finding’ session, to determine which activities were safe to perform. This theme is further explored in the focus groups presented in Chapter 9.

This feasibility study allowed us to refine the trial methods for a future definitive RCT. This is expanded on in Chapter 12.

Methods

Demographic characteristics

The final data set for the cohort analysis consisted of 400 patients. The characteristics of patients in the cohort study and in the feasibility study are compared in Table 19. The cohort population was slightly older than the feasibility study population, with a mean age difference of 4.5 years. In both data sets there was a higher proportion of women (73–75%), with similar proportions of women in both data sets.

Fracture details

Information relating to the type of fracture suffered by patients along with the type of surgery used to treat the fracture were recorded from patients’ medical records. Intracapsular fractures were more common than extracapsular fractures, with similar proportions in the two data sets. In the extracapsular group, intertrochanteric fractures were the most common type of fracture.

There was a higher percentage of unavailable information in the feasibility study than in the cohort study. This could be related to the timing of examination of the hospital records. For the cohort study the records were examined retrospectively by experienced researchers working for Health and Care Research Wales. Within the trial, the records were examined by study researchers while patients were being recruited. This occurred either while a patient was on the ward, when the hospital notes were also being used by the various clinical teams or after a patient was discharged home, when the hospital notes were not available.

Most of the cohort study population underwent hemiarthroplasty (40%) or internal fixation (38%). In the feasibility study population, a slightly higher proportion received hemiarthroplasty (47%) and a slightly lower proportion received internal fixation (28%). Within the trial, all of the patients underwent a surgical procedure, as this was one of the inclusion criterion; however, 7% of the cohort study population did not receive any surgery and would therefore have been excluded from the study during screening.

Living arrangements

The type of accommodation that patients lived in, both before and after their hospital admission, was recorded. The aim of this was to detect any change that had occurred in patients’ living conditions and related independence following their hip fracture and subsequent surgery.

Living independently prior to hip fracture surgery was an inclusion criterion for the feasibility study. Most of the cohort population (78%) lived in a private property before their admission to hospital and a further 3% lived independently in sheltered accommodation. The remaining 19% of patients admitted with hip fracture in the cohort study did not fulfil the living independently criterion for taking part in the feasibility study. In contrast, in the feasibility study, 10% of the assessed patients were ineligible because they were not living independently prior to hip fracture surgery. Some patients would have been ineligible for multiple reasons and, in particular, may not have been living independently because of a lack of capacity. As only the primary reason for ineligibility was recorded for the feasibility study, this may account for the discrepancy.

In the cohort study, the majority of patients (56%) were discharged from the acute hospital to a community hospital. Only 26% of patients were discharged directly back to their private property, with 11% discharged to other types of residential care. The same data were not available for the feasibility study population to allow a direct comparison to be made between the data sets. However, information was available at 3-month follow-up in the feasibility study on participants’ place of residence and whether or not this was the same place that they had been discharged to after their hospital stay. At 3-month follow-up, 27 patients were living in the same residence that they were discharged to and, in most cases (93%), this was a private property. The remaining 22 patients had an interim residence between discharge and their 3-month follow-up; however, details of where this was were not recorded.

Readmissions and deaths

In the cohort study, information was collected on the number of patients who had an unplanned readmission to hospital or who had died in the 3 months following their hip fracture. As shown in Table 20, 69 patients (17%) were readmitted to hospital and 68 patients (17%) died within 3 months of their hip fracture. As these numbers were very similar, additional information was collected to determine whether or not these two events occurred in the same patients. However, 117 patients either were readmitted or died during the 3 months post hip fracture, suggesting that the two events mostly occurred in separate populations.

Table 20 also shows how many times patients were readmitted in the 3-month follow-up period for those patients who were readmitted. Most patients (n = 58) were readmitted only once; however, nine patients were readmitted twice and two were readmitted three times. In the feasibility study population, only one death (2%) and two readmissions to hospital (3%) were recorded. These proportions were much lower than those found in the cohort population, suggesting that the feasibility study recruited a younger and healthier subpopulation of patients.

Complications at initial admission

Information was collected in the cohort study on any serious complications that patients encountered during their initial admission for hip fracture. These serious complications included repeat falls, repeat fractures and adverse reactions to any drugs received.

In total, 160 (40%) patients experienced at least one complication during their initial admission. Some patients experienced numerous complications during their admission, which could be numerous instances of the same complication. Data on complications are provided in Table 21, which shows that, in total, 359 complications were experienced by patients in the cohort study. Examples of incidents that would come under the other medical complications category were myocardial infarctions, urinary tract infections or deep vein thromboses.

Screening logs

The numbers of patients identified for inclusion in the cohort study are shown by centre in Table 22. The cohort study identified 462 patients in total over a 6-month period. In this same period the feasibility study screened 372 patients for eligibility. Therefore, most of the potential feasibility study participants (81%) were identified by the researchers during screening. The majority of the patients’ hospital notes (87%) were pulled from the system. The reasons for patients’ hospital notes not being pulled are shown in Table 22, with the most common reason being that the notes were not available; some of the patients had been moved between hospitals and so would have been included within the cohort for the other hospital.

Summary of findings

The anonymised cohort allowed us to assess the representativeness of the population recruited to the feasibility study in relation to the broader hip fracture population. Participants recruited to the cohort and feasibility studies were similar with respect to gender, type of hip fracture and type of surgery, but the feasibility study participants were younger and less likely to die or be readmitted to hospital. This suggests that the feasibility study recruited a healthier and younger subset of the total hip fracture population.

Strengths and limitations

One of the main strengths of the cohort study was that data were extracted and collated by experienced researchers working for Health and Care Research Wales, who have a great deal of expertise in and familiarity working with patients’ medical records. For reasons of patient confidentiality, this was an anonymised cohort and so it was not possible to match participants in the cohort with those in the feasibility study. In total, the feasibility study identified 81% of those in the cohort study. It is not known how the 19% who were not identified in the feasibility study differed, nor is it known why they were identified in the cohort study but not in the feasibility study.

A limitation that could have influenced the study was the search terms that were used to identify patients. It is possible that patients who had not been correctly coded on their admission with a hip fracture would not have been included in the cohort study. After identifying patients who fit the criteria for inclusion in the cohort study, their physical notes had to be located to capture any relevant information; this was often problematic and the research workforce was unable to locate 13% of the medical records. If hospital records were completely digitised this would avoid the problems associated with locating paper record folders.

Although not essential to the conclusions drawn, it may have been beneficial if the feasibility study had recorded the place of discharge in a more consistent manner, similar to that used in the cohort study. This would have allowed a closer comparison of the data. This could be considered in any future study.

Comparison with previous literature

The age and gender composition in the cohort study were similar to the age and gender composition in other studies from Sweden340 and the USA. 281 The mortality rate in the cohort study was higher than that reported in California, where 12.7% of patients died in the 3 months after hip fracture,341 in Sweden, where 13% of patients died in the 6 months after fracture,340 and in Northern Ireland, where 15.1% of patients died in the 12 months after fracture. 342 Readmission rates were not directly comparable but were similar: in the USA, 11.9%281 and 12.3%341 of patients were readmitted within 30 days; in Sweden, 32% of patients were readmitted within 6 months. 340

The proportion of patients identified as having non-surgical conservative management of their hip fracture appeared to be higher in our cohort (7%) than is reported elsewhere in the literature (between 2% and 5% in European studies342,343). The reason for non-surgical management was not collected in our cohort study.

There were fewer deaths and readmissions in the feasibility study population than in the cohort population. The recruitment of younger, healthier patients to a physical activity or exercise intervention study was not surprising and has been reported previously. 199

Implications for future research

Including an anonymised cohort has allowed us to consider the representativeness of the population recruited to the feasibility study. It is highly likely that a future definitive RCT of the rehabilitation intervention would recruit a younger, healthier, more motivated subgroup. One important group excluded from the feasibility study was those frail individuals who lack mental capacity. To ensure that a future RCT is more representative it will be necessary to include such groups. This is discussed further in Chapter 12.

Patient focus group participants

All patient participants and their carers were asked when initially consenting to take part in the feasibility study whether or not they also agreed to be invited to a focus group later in their recovery. All those who agreed were invited to participate in a local focus group following completion of follow-up. Participants were contacted by telephone to gauge interest and those who expressed an interest in taking part were sent an information sheet, specimen consent form, topic guide and details of the venue. Separate focus groups were conducted for those in the control group and those in the intervention group. Written informed consent was obtained just prior to starting the focus group from those who attended. We aimed to recruit approximately eight participants to each focus group.

Health-care professional focus group participants

Staff who had been involved in the screening and recruitment of participants, delivery of the intervention and conducting physical function tests were approached by study researchers to ask if they would be willing to take part in a focus group. Focus groups were arranged at the hospital site where staff were based to minimise disruption to their working day and maximise the number of staff who could attend. When staff were unable to attend focus groups, one-to-one interviews were offered. Staff who expressed an interest in participating were sent an information sheet, specimen consent form and topic guide prior to the focus group. Written informed consent was obtained just prior to starting the focus group from those who attended. For telephone interviews, verbal consent was given and recorded, and written informed consent was posted back to the study team.

Focus group procedures

All participants gave written consent to participate and agreed that their comments could be recorded, transcribed and anonymised for analysis. The discussions in the focus groups were semistructured and run by a moderator (the study manager) and co-moderator (a study researcher) with the use of a topic guide (see Appendix 31). Topic guides were specific to patient and carer or health-care professional focus groups and contained open-ended questions regarding experiences, perceptions and beliefs about the rehabilitation that they had delivered or received, including the different aspects of the enhanced programme for those patients and carers in the intervention group. The topics covered were informed by the aims and objectives of the study and the intervention logic model. In focus groups with patients and carers randomised to the intervention, ranking exercises were also used to elicit discussion on the importance to patients of the different aspects of the intervention. Focus groups were run in English, although simultaneous translation into Welsh was offered to participants. Digital recordings of the focus group discussions were made and notes were taken during the focus groups by the moderator or co-moderator. At the end of the focus groups, the recordings were downloaded onto an encrypted NWORTH laptop and subsequently downloaded and stored on the university server. Access to the data was limited to core members of the study team. Transfer of the recordings to an approved transcriber and return of the transcripts involved encrypting the recordings and uploading them to/downloading them from a secure server.

Care was taken not to record personally identifiable data in written notes. Written notes were stored in a locked cabinet in a locked room in NWORTH and were accessible to authorised team members only.

Qualitative data analysis

Transcripts of the focus groups were analysed by the study manager, who had acted as group moderator, using the framework approach to thematic analysis. 293 Data from the focus group notes and transcriptions of the discussions were transferred to Microsoft Excel^® version 2013 (Microsoft Corporation, Redmond, WA, USA) software for qualitative analysis. A framework was developed based on the programme theories, logic model and questions of feasibility and acceptability that the study was designed to answer. The transcripts were then coded and grouped together into categories. The study manager discussed, compared and named the categories that were used to populate and refine the framework with an objective researcher experienced in qualitative research who had not been present at the focus groups and not been directly involved in the study. A process of refinement of the overarching themes across the data set to more specific recurrent issues within these themes provided a more in-depth appreciation of the way in which the intervention was delivered and received. The simultaneous analysis of the transcripts and themes by two researchers prior to consultation with the chief investigator as a third reviewer ensured the plausibility and credibility of the findings.

Patients’ and carers’ experiences of post-discharge rehabilitation following a fractured neck of femur

Five overarching themes emerged from the focus groups conducted with patients and their carers. These provided an overarching framework to delineate the key issues pertinent to patients regarding their post-hospital discharge rehabilitation.

Four themes related to general patient and carer experiences of the rehabilitation process following hip fracture:

1. the impact of variation within usual care on patient experiences and expectations

2. patient expectations of post-fracture recovery and their management by health-care professionals

3. patient preferences for individualised care and support

4. the relationship between the patient and the health-care professional and its role in patient engagement in rehabilitation.

The fifth theme related specifically to patient experiences of taking part in a research study and the acceptability of the intervention.

Therapist experiences of post-discharge rehabilitation of patients following a hip fracture and delivery of the study intervention

Four main themes emerged from the health-care professional focus groups in relation to general patient care and delivery of the study intervention:

1. the impact of variation within usual care on delivery of the intervention and overall therapist experiences

2. therapist perceptions of their role within the rehabilitation of hip fracture patients

3. patient-led goal-setting to facilitate engagement in rehabilitation

4. challenges of delivering the intervention.

Summary of findings

Figure 14 provides a visual representation of the overarching themes relating to patients’ and health-care professionals’ experiences of post-discharge rehabilitation following hip fracture in this study. It shows, in the context of varied usual care, the pivotal role of the physiotherapist in managing patient needs and expectations through the use of personalised care and goal-setting. In the intervention group, these activities were well supported by the workbook and goal-setting diary.

FIGURE 14.

Overarching themes from the focus groups with patients and therapists with regard to post-hospital discharge rehabilitation following hip fracture.

The themes identified in the patient focus groups can be categorised according to three distinct levels. On a health services level we identified a concern over the large variation in usual care following discharge from hospital with a hip fracture. Patients described an ad hoc experience of home-based care and rehabilitation and, in some cases, no post-discharge physiotherapy at all. This variation in experience revealed an uncertainty and associated anxiety among patients about how they could safely promote their own recovery after they were discharged from hospital.

On a patient experience level, frustration at wanting to restore their former physical functioning coupled with a lack of clarity with regard to expectations of recovery (such as what could be considered achievable or not achievable on an individual basis) led to patients feeling demotivated and needing reassurance from a health professional. Patients stated a preference for individual therapeutic care at home on a one-to-one basis that was responsive to individual physical needs. In the absence of such support, patients felt a sense of frustration at the speed of their physical recovery, which was felt to have been poorly managed.

On an interventional level, the key recurring message with regard to community physiotherapy was the importance of regular home visits, particularly in the early phase of post-hospital recovery. These visits should provide tailored advice promoting confidence and enable patients to support their own recovery by providing reassurance about which physical activities can be safely performed, along with promoting self-efficacy by highlighting patient progress.

The themes identified in the therapist focus groups mapped onto the themes identified in the patient focus groups. For example, the successful delivery of the study intervention was affected by the highly varied nature of usual care. However, there was a disparity of views between the patient focus groups and the therapist focus groups regarding the initial contact with a therapist. For the therapists, this was merely information sharing prior to starting proper rehabilitation, including the practice of exercises and ADL. For the patients, however, this initial contact with their qualified community therapist was vital. Once at home, patients’ anxieties mainly related to the uncertainty about the trajectory of their physical recovery and which daily activities were safe to perform independently. Uncertainty about what was safe and the fear of causing further damage over-rode the impulse to begin exercising and resume normal ADL, thus becoming an obstacle in the early and critical phase of patients’ physical rehabilitation. Therapists provided reassurance about recovery and verbal guidance on strategies to support recovery in the context of patients’ homes. The provision of such evidence-based guidance on self-management and recovery optimisation was pivotal for a phased return to normal ADL, while ensuring that the return to pre-fracture activities was achieved safely. The therapists were not fully aware of the pertinence, for patients, of this early psychological support and reassurance in underpinning and directing a confident trajectory of recovery.

In summary, the therapists were providers of evidence-based guidance and provided reassurance and clarity on:

• what was available to patients to support their rehabilitation, for example physical aids and psychosocial support

• what could be realistically expected in terms of recovery, for example when to expect markedly increased mobility

• what was safely achievable on an individual basis over a given time frame.

Strengths and limitations

Although we were able to conduct one-to-one interviews with staff at one of the hospital sites (Ysbyty Gwynedd, Bangor) where there had been problems with delivery of the intervention, these staff were unable to come together and discuss these issues in a focus group because of clinical commitments. As patient participants in this site were spread over a large geographical area, it was not possible to arrange a focus group that these participants could attend. Also, because of problems with intervention delivery in this area (see Chapter 9), only one participant actually received the extra sessions, and she was unable to attend the other groups that were arranged. Patient and professional participants from Gwynedd and Anglesey were therefore under-represented in this study.

The intervention was delivered only by physiotherapists and technical instructors; there was no involvement of occupational therapists as we had originally anticipated. This was because of resource availability. As such, the focus groups included only two types of professionals, although the physiotherapists were of differing grades and had different roles within the health board. Although we attempted to recruit nurses who had been involved in screening patients for eligibility, they were unable to attend the focus groups because of clinical commitments.

These factors also impacted on the number of participants in each group, which in some cases was as few as three. Although the participants represented a range of rehabilitation experiences, it would have been beneficial to reach our target of eight participants per focus group to encourage discussion from a variety of viewpoints.

Comparison with previous literature

Comments from health-care professionals and patients regarding the variable nature of usual care were comparable to those provided in the Phase I focus groups (see Chapter 5). Similar views have also been documented in the literature. 344–347

Goal-setting and supporting patients’ self-efficacy were seen to be important in helping patients to engage with their rehabilitation. Another study on patient empowerment347 also found that empowered patients were more likely to benefit from their rehabilitation and return to previous living.

The focus groups highlighted that, without involvement from a health-care professional, patients’ expectations of their recovery may be unrealistic. A previous study has also shown that patients need detailed counselling to create realistic expectations of their recovery and enable them to prepare for the outcomes. 348

Implications for practice and future research

The focus groups supported the findings from the wider feasibility study that usual care is variable and often unpredictable, and demonstrated that this had a negative impact on patient experience. Although this was expected, they highlighted the challenges that this can cause when delivering a rehabilitation intervention in this context and emphasised the need for more thorough training of intervention therapists, with continuing support for how the intervention can best be delivered. This should be considered in a definitive RCT and the training opportunities increased. Because of the challenges involved in delivering the intervention, which related to resource allocation, it would be beneficial to use a dedicated therapist to deliver the intervention sessions. This would also help to address the problems that were encountered when intervention therapists were left to provide usual care for the same patient, causing confusion for patients and therapists.

Therapists highlighted that in some cases patients felt that they did not need all six additional therapy sessions. This is supported by adherence data (see Chapter 7) and is likely to be influenced by the variability in usual care and also the specific population recruited to the feasibility study, who were shown to be a younger and healthier subpopulation of the whole cohort (see Chapter 8). This may be of relevance to future practice and to defining the minimum dose of therapy that is optimal for this population.

Patients’ perception of the therapist’s role in their rehabilitation appeared from the focus groups to be one of a guide during their recovery process. In these self-motivated and previously independent people, there was a requirement of the therapist to support a patient-led recovery process, allowing patients to take ownership of their recovery while being reassured by a professional that they were progressing safely. From the specific patient needs that were highlighted in the focus groups, and the disparities between the role perceptions of therapists, we have produced a mnemonic that could be used by therapy managers to assist their teams in rehabilitating these types of patients (Figure 15).

FIGURE 15.

The GUIDE (Guidance, Understanding, Individualised, Direction, Expertise) tool for therapists for rehabilitating hip fracture patients.

This aims to highlight the central role of the therapist in guiding patients to set realistic and achievable goals and providing feedback on progress to support patients’ own self-efficacy. It also reminds therapists of the information needs of patients and how addressing these psychological problems underpins their physical progress.

Although there were some significant challenges involved in delivering the study intervention, it was well tolerated by participants, who found taking part in the study a positive experience overall. Therapy staff involved in the delivery of the intervention also provided several suggestions that could be implemented in a future RCT of the intervention to overcome some of the problems that they encountered. One of the issues experienced by therapists was how best to utilise the intervention sessions in the context of variable usual care. They commented that they would have welcomed more training with regard to this; the use of the GUIDE tool (see Figure 15) in training may be helpful in this regard.

Published economic evidence

The National Institute for Health and Care Excellence7 highlighted in its guidance on the management of hip fracture in adults the lack of cost-effectiveness evidence available. A Cochrane review conducted by Handoll et al. 8 concluded that there was no conclusive evidence on the effectiveness and cost-effectiveness of multidisciplinary inpatient rehabilitation following proximal femoral fracture surgery in older people. Our own systematic review of the economic evidence (see Chapter 3 for the full methods and results) identified seven published economic evaluations conducted alongside rehabilitation interventions for proximal femoral fracture in older people. 253,254,260–264 The year of publication in the included evaluations ranged from 1986 to 2009. The evaluations were conducted in Australia, New Zealand, the Netherlands, Sweden, the UK and the USA. The majority of the evaluations calculated costs from a hospital perspective, considering direct hospital costs only, such as the costs of treatment, medication and care. 254,260,261,264 Two of the three cost-effectiveness evaluations found that rehabilitation pathways were cost-effective. 253,260 Of the two cost–benefit analyses, only one found that the benefits of the intervention outweighed the costs. 262 Both cost-minimisation evaluations revealed that the interventions under study reduced costs without reducing outcomes compared with usual care, thus demonstrating that the interventions could provide lower-cost alternatives to regular practice. 263,264 The findings in Chapter 3 correspond with the findings of previous economic evaluations7,8 that there is a limited evidence base for multidisciplinary proximal femoral fracture rehabilitation, particularly from a UK perspective.

Objective of the feasibility study economic analysis

The objective of the feasibility study economic analysis was to explore the methodological issues around conducting an economic evaluation alongside a future RCT, including identifying the most efficient way of measuring patient-level costs and health benefits, and programme costs and identifying potential payer stakeholders.

The objective of the feasibility study economic analysis was divided across three work streams:

1. work stream 1 – assessing the feasibility of conducting an economic analysis from data collected in the feasibility study

2. work stream 2 – assessing the feasibility of measuring patients’ preferences for proximal femoral fracture rehabilitation services using a DCE

3. work stream 3 – scoping the potential to conduct SROI analysis in a future trial based on data gathered in the feasibility study.

These work streams are described in turn in the following sections.

Objective

The primary objective of the health economics analysis was to test whether or not it was feasible to conduct economic analysis using the data from the feasibility study. We also wished to explore the use of a generic health-related quality-of-life measure and a capability measure to explore the extent to which the two approaches could help guide commissioning decisions following a full trial and full economic evaluation.

Methods

Results

Health economics sample

In total, 61 participants were recruited in the feasibility study, of whom 12 withdrew during the course of the study, resulting in 49 sets of data available for analysis. We excluded six participants as they were missing data at either follow-up or both baseline and follow-up (one participant did not complete the EQ-5D-3L questionnaire at follow-up, one participant did not complete the CSRI at follow-up and four participants did not complete the ICECAP-O313 questionnaire at both baseline and follow-up). Therefore, the final sample for the economic complete case analysis included 43 sets of data (intervention, n = 21; control, n = 22). The six participants with missing data were excluded from the analysis because cost-effectiveness or cost–utility analysis requires complete data for both costs and outcomes. Although we were unable to perform cost-effectiveness or cost–utility analysis because of a lack of a significant effect for the primary outcome, we felt that it was important to use the feasibility data in this way to inform future economic evaluations. Therefore, we used only complete data across the three economic measures (EQ-5D, ICECAP-O and CSRI) in the final sample, as would be required by a full economic evaluation. Furthermore, as this was a feasibility study, we did not impute missing data, as stated in the statistical analysis plan.

The ICECAP-O questionnaire had a lower completion rate than the EQ-5D-3L. Three participants stated that they were unable to answer the questionnaire as they felt that the domains did not apply to them (e.g. they felt that they had no social life at the moment and so could not answer the questionnaire). One participant was not able to finish the questionnaire as one of the questions led her to think about something upsetting in her life and she became distressed. These participants were omitted from the analysis because, to calculate a capability index, all questions on the ICECAP-O questionnaire must be answered. 313

Table 24 summaries the characteristics of the economic sample at baseline; Table 25 shows the mean EQ-5D-3L utility scores, EQ-5D-3L VAS scores and ICECAP-O capability index scores at baseline and at 3-month follow-up; and Table 26 shows, for both the intervention and the control groups, the mean QALYs, change in the mean ICECAP-O capability index score and mean total service use costs (including the cost of the intervention for the intervention group) over the 3-month study period. The differences between groups in mean QALYs, capability indices and service use costs are also presented, with 95% CIs around these differences.

TABLE 24 - Baseline characteristics of the economic sample (N = 43) by group

Characteristic	Intervention (N = 21), n (%)	Control (N = 22), n (%)
Gender
Female	17 (81)	15 (68)
Male	4 (19)	7 (32)
Usually lives
Alone	11 (52)	11 (50)
With others	10 (48)	11 (50)
Accommodation type
Owner-occupied property	17 (81)	17 (77)
Privately rented property	2 (10)	2 (9)
Housing association/local authority property	2 (10)	2 (9)
Sheltered accommodation	0 (0)	1 (5)
Type of fracture
Intracapsular	8 (38)	12 (55)
Extracapsular	9 (43)	7 (32)
Not recorded in notes/notes unavailable	4 (19)	3 (14)
Type of surgery
Hemiarthroplasty	9 (43)	12 (55)
Internal fixation	10 (48)	4 (18)
Intramedullary nailing	0 (0)	1 (5)
Total hip arthroplasty	1 (5)	4 (18)
Not recorded in notes/notes unavailable	1 (5)	1 (5)
Direct discharge
Yes	10 (48)	16 (73)
No	11 (52)	6 (27)
Recruited from
Ysbyty Glan Clwyd	12 (57)	12 (55)
Wrexham Maelor	7 (33)	9 (41)
Ysbyty Gwynedd	2 (10)	1 (5)
Age at baseline (years), mean (SD); range	80.95 (6.2); 69–89	76.82 (7.5); 66–89

TABLE 25 - Mean EQ-5D-3L utility and VAS scores and ICECAP-O capability index scores at baseline and at 3-month follow-up by groupa

Scores all rounded to two decimal places.

Outcome	Baseline, mean (SD), range		3-month follow-up, mean (SD), range
	Intervention group (n = 21)	Control group (n = 22)	Intervention group (n = 21)	Control group (n = 22)
EQ-5D-3L utility score	0.50 (0.26), 0.02–1.00	0.37 (0.43), –0.48 to 1.00	0.66 (0.27), 0.09–1.00	0.60 (0.27), 0.02–1.00
EQ-5D-3L VAS score	64.43 (16.37), 35.00–100.00	55.14 (25.72), 0.00–90.00	71.10 (17.89), 18.00–100.00	68.55 (18.44), 30.00–98.00
ICECAP-O capability index score	0.82 (0.11), 0.54–0.98	0.75 (0.21), 0.35–1.00	0.84 (0.13), 0.50–1.00	0.78 (0.19), 0.19–1.00

TABLE 26 - Mean QALYs and mean change in ICECAP-O capability index score over the 3-month study period and mean total service use costs at follow-up including the cost of the intervention by groupa

All rounded to two decimal places.

Outcome	Intervention group (n = 21) (1000 bootstrapped 95% CI)	Control group (n = 22) (1000 bootstrapped 95% CI)	Difference between groups (1000 bootstrapped 95% CI)
Mean QALYs over 3 months	0.15 (0.12 to 0.17)	0.12 (0.09 to 0.15)	0.02 (–0.02 to 0.06)
Mean change in ICECAP-O capability index scores over 3 months	–0.03 (–0.08 to –0.03)	–0.03 (–0.12 to 0.07)	0.00 (–0.11 to 0.22)
Mean total service use costs at follow-up including cost of the intervention (£)	149,243.02 (119,376.32 to 186,035.64)	105,243.95 (78,934.94 to 132,971.08)	43,999.07 (4026.98 to 88,818.07)

The economic sample represented 72% of the main clinical sample. The characteristics of the two samples were the same (see Chapter 9 and Table 24), with more women recruited to the trial than men. The age of the participants was similar in both samples, with the intervention group slightly older (mean age 81.0 years in the economic sample and 80.9 years in the clinical sample) than the control group (mean age 76.8 years in the economic sample and 78.0 years in the clinical sample). The participants’ home circumstances were similar between the groups in the economic sample, with the same number of participants living alone or living with others (n = 11) in the control group and one more participant living alone (n = 11) than living with others (n = 10) in the intervention group. As in the clinical sample, the majority of participants in both groups in the economic sample owned their own home; experienced an intracapsular fracture and received hemiarthroplasty surgery, with internal fixation the second most common type of surgery; and were recruited from Ysbyty Glan Clwyd, with the lowest number of participants recruited from Ysbyty Gwynedd.

The control group had a lower EQ-5D-3L index score, EQ-5D-3L VAS score and ICECAP-O capability index score than the intervention group at baseline (see Table 25). However, both groups had improved scores at follow-up, with the maximum range being at the ceiling score of the measures (1.00 for the EQ-5D-3L and ICECAP-O indices, and 100 for the EQ-5D-3L VAS). Differences in EQ-5D-3L index scores, EQ-5D-3L VAS scores and ICECAP-O capability index scores between groups were assessed using Mann–Whitney U-tests, which found no statistically significant differences (p > 0.05).

Table 26 shows the mean QALY gains (if any) by each group over 3 months. For the control group we calculated a mean QALY gain of 0.12 (1000 bootstrapped 95% CI 0.09 to 0.15), equating to 45 days gained. The intervention group showed a slightly higher QALY gain of 0.15 (1000 bootstrapped 95% CI 0.12 to 0.17), equating to 53 days gained. The difference in QALYs between the two groups was 0.02 (1000 bootstrapped 95% CI –0.02 to 0.06), equating to 8 days gained.

The mean change in the ICECAP-O capability index scores over 3 months was –0.03 (1000 bootstrapped 95% CI –0.12 to 0.07) in the control group and –0.03 (1000 bootstrapped 95% CI –0.08 to –0.03) in the intervention group (see Table 26), indicating that both the intervention and the control groups reported lower levels of capability at follow-up than at baseline. This could be because of reduced mobility and independence while participants recovered from their fracture, resulting in lower perceived capability as measured by the ICECAP-O across the whole sample. The difference in capability index scores between the two groups was 0 (1000 bootstrapped 95% CI –0.11 to 0.22).

The intervention group reported mean total service use costs per participant at follow-up (including the cost of intervention) of £149,243.02 (1000 bootstrapped 95% CI £119,376.32 to £186,035.64), whereas the control group reported mean total service use costs at follow-up of £105,243.95 (1000 bootstrapped 95% CI £78,934.94 to £132,971.08). The difference in total service use costs at follow-up between the groups was £43,999.07 (1000 bootstrapped 95% CI £4026.98 to £88,818.07).

Mann–Whitney U-tests found no statistically significant differences between the groups in terms of QALYs gained (U = 184.00, p = 0.18) or ICECAP-O capability index scores (U = 249.00, p = 0.86). However, Mann–Whitney U-tests revealed a statistically significant difference in mean total service use costs per participant between the two groups (U = 136.00, p = 0.02).

Box 5 outlines how the service use categories – primary health-care and community services, secondary health-care services, social care services, proximal femoral fracture-specific services, voluntary/charity services and medication use – were derived from the contacts listed in the CSRI questionnaires. These service use categories are used throughout this chapter when describing participant service use and associated costs. Service use data were collected for the 3 months prior to baseline and the 3 months prior to follow-up.

BOX 5 - Service use categories used in the economic analysis

Primary health-care and community services

GP.

Practice nurse.

District nurse.

Health visitor.

Physiotherapist.

Occupational therapist.

Technical instructors/rehabilitation assistants to physiotherapists or occupational therapists.

Community pharmacist.

Secondary (hospital) health-care services

Inpatient services.

Outpatient services.

A&E services.

Day hospital.

Any other hospital services not listed in the questionnaire.

Social care services

Social worker.

Community psychiatric nurse/community mental health nurse.

Psychologist.

Community psychiatrist.

Counsellor.

Home-care worker.

Care attendant.

Proximal femoral fracture-specific services

Enablement centre for rehabilitation.

Nursing home for rehabilitation.

Falls prevention classes.

National Exercise Referral Scheme.

Voluntary/charity services

British Red Cross.

Borrowing of equipment (e.g. wheelchairs).

Medication use

All medications used, including prescribed and over-the-counter medicines.

Figure 16 shows the mean total frequency of contacts and the mean frequency of contacts with primary and secondary health-care services in the intervention group (n = 21) and the control group (n = 22) during the 3 months prior to baseline.

FIGURE 16.

Mean frequencies of primary and secondary health-care service use by participants in the intervention group and the control group, as measured by the CSRI during the 3 months prior to baseline. Frequencies are presented to one decimal place.

During the 3 months prior to baseline, participants in the control group accessed more services than participants in the intervention group. Participants accessed more primary care services than secondary care (hospital) services prior to their proximal femoral fracture. Participants were most likely to see a GP, with practice nurses the second-highest accessed health-care professionals. Participants underwent more outpatient procedures than inpatient procedures during this time period.

Figure 17 illustrates the mean costs of the service use associated with the intervention group and the control group during the 3 months prior to baseline. The mean costs of primary service use were minimal, with the costs of secondary (hospital) services accounting for the majority of the total service costs. Secondary (hospital) services accounted for 98% of the costs in the intervention group, with primary services accounting for the remaining 2%; the corresponding percentages for the control group were 99% and 1%. The majority of the hospital costs were related to multiple outpatient appointments attended by the sample.

FIGURE 17.

Mean costs of primary and secondary health-care service use by participants in the intervention group and the control group, as measured by the CSRI during the 3 months prior to baseline. Costs are rounded to the nearest pound and are presented for the cost year 2013–14.

Figure 18 shows the frequency of contacts with primary and secondary health-care services, social care services, proximal femoral fracture-specific services and voluntary/charity services, as well as medication use, in the intervention group (n = 21) and the control group (n = 22) during the 3 months prior to follow-up. Participants primarily accessed primary services and social services. There were high levels of polypharmacy within the sample, with the sample, on average, taking more than five medicines at the time of completing the CSRI. The most common medications prescribed were ramipril, omeprazole, alendronic acid, amlodipine, warfarin, calcium and vitamin D supplements. Differences in frequencies between the groups were assessed using Mann–Whitney U-tests. No statistically significant differences were found between the groups in the frequency of total health-care service use (U = 197.00, p = 0.41), primary health-care service use (U = 187.00, p = 0.28), secondary (hospital) health-care service use (U = 259.50, p = 0.48), social care service use (U = 215.00, p = 0.65), hip fracture-specific service use (U = 217.50, p = 0.61) or voluntary service use (U = 238.50, p = 0.74), or in the frequency of medication use (U = 189.50, p = 0.31).

FIGURE 18.

Mean frequencies of primary and secondary health-care, social care, proximal femoral fracture-specific and voluntary/charity service use and frequency of medication use by participants in the intervention group and control group, as measured by the CSRI during the 3 months prior to follow-up: (a) intervention group; and (b) control group. Frequencies are presented to one decimal place.

Figure 19 shows the costs of primary and secondary health-care service, social care service, proximal femoral fracture-specific service and voluntary/charity service use, as well as medication use, in the intervention group (n = 21) and the control group (n = 22) during the 3 months prior to follow-up. Service use costs were mainly accrued by secondary (hospital) health-care services, as patients underwent surgery and required time in hospital to recover. In total, 51% (n = 22) of the sample reported a longer than average inpatient stay post surgery. A longer inpatient stay was reported by 67% (n = 14) of the intervention group and 36% (n = 8) of the control group.

FIGURE 19.

Mean costs of primary and secondary health-care, social care, proximal femoral fracture-specific and voluntary/charity service use by participants in the intervention group and the control group, as measured by the CSRI during the 3 months prior to follow-up: (a) intervention group; and (b) control group. Costs are rounded to the nearest pound and are presented for the cost year 2013–14.

Secondary (hospital) health-care services accounted for 96% of the costs of service use at follow-up in the intervention group, with the remaining 4% accrued by hip fracture-specific services such as the National Exercise Referral Scheme and enablement centres. Similarly, in the control group, secondary (hospital) health-care services accounted for 95% of the costs of service use at follow-up, with the remaining 5% accrued by hip fracture-specific services.

Mann–Whitney U-tests revealed a statistically significant difference between the two groups in mean total hospital costs (U = 135.00, p = 0.02) and total service use costs per participant (U = 136.00, p = 0.02). This finding could be attributed to a larger proportion of the sample in one group than in the other experiencing a longer inpatient stay. No statistically significant differences were detected between the groups in the costs of primary health-care service use (U = 202.00, p = 0.48), social care service use (U = 208.00, p = 0.52), hip fracture-specific service use (U = 219.00, p = 0.65), voluntary service use (U = 240.50, p = 0.68) and medication use (U = 181.00, p = 0.22).

Table 27 shows that, when further exploring participant inpatient stays at baseline, the mean number of inpatient bed-days in both groups was less than the average length of stay specified by the Department of Health NHS reference costs 2013–14. 339

TABLE 27 - Mean number of inpatient bed-days reported by participants in both groups at baseline compared with the average length of stay specified in the Department of Health NHS reference costs 2013–14339

Type of ward	Intervention group (n = 21)		Control group (n = 22)		Average number of bed-days according to the Department of Health339
	Inpatient bed-days, mean	Percentage of inpatient admissions that were readmissions	Inpatient bed-days, mean	Percentage of inpatient admissions that were readmissions
Orthopaedic trauma inpatient ward	8.00	0	7.00	0	14.27
Renal inpatient ward	0.00	0	9.00	0	9.27
Medical inpatient ward	0.00	0	4.00	0	5.56

Table 28 shows similar data for inpatient stays during the 3 months prior to follow-up for both groups. More often than not, the mean number of inpatient bed-days in the two groups exceeded the average length of stay specified in the Department of Health NHS reference costs 2013–14. 339 This finding further explains the high cost of participant service use at follow-up, as higher-cost excess inpatient bed-day costs were applied in the majority of cases. Additionally, as shown in Table 28, 50% of surgical inpatient ward stays were stated by participants to be readmissions, demonstrating the complexity involved in treating this cohort of elderly patients.

TABLE 28 - Mean number of inpatient bed-days reported by participants in both groups during the 3 months prior to follow-up compared with the average length of stay specified in the Department of Health NHS reference costs 2013–14339

Type of ward	Intervention group (n = 21)		Control group (n = 22)		Average number of bed-days according to the Department of Health339
	Inpatient bed-days, mean	Percentage of inpatient admissions that were readmissions	Inpatient bed-days, mean	Percentage of inpatient admissions that were readmissions
Orthopaedic trauma inpatient ward immediately following hip fracture	15.15	5	10.47	0	14.27
Surgical inpatient ward immediately following hip fracture	15.50	50	17.33	0	14.27
Assessment inpatient ward	7.50	0	7.00	0	14.27
Rehabilitation inpatient ward	12.11	0	18.80	0	11.57
Continuing care inpatient ward	21.00	0	70.00	0	11.57
Respite inpatient ward	10.00	0	0.00	0	7.83
Medical inpatient ward	13.00	0	0.00	0	8.92

Key findings

A sample of 43 participants was used for the cost–consequences analysis.

The ICECAP-O313 questionnaire had a lower completion rate than the EQ-5D questionnaire,287 which has implications for a full-scale future trial. The cost–consequences analysis demonstrated that it cost a total of £6711.46 to deliver the intervention (£231.43 per person, based on the 29 participants randomised to the intervention group).

Both the intervention and the control groups showed improvements in the EQ-5D health utility index, EQ-5D VAS score and the ICECAP O capability index at follow-up compared with baseline. There were no statistically significant differences between the groups in any of these three measures, but this small feasibility trial was not powered to test such a difference.

The intervention group had slightly higher mean QALY gains than the control group, although this difference was not statistically significant; however, this small feasibility trial was not powered to test such a difference.

There was a statistically significant difference between the two groups in mean total hospital costs and total costs of service use per participant. This could be attributed to a larger proportion of the sample in one group than in the other experiencing a longer inpatient stay. No other statistically significant differences were detected for the other service use categories.

Although in a disaggregated cost–consequences form our economic analysis shows that the multidisciplinary intervention, workbooks and diaries were no more effective than receiving usual care, this small feasibility trial was not powered to test such a difference.

Objective

The objective of this workstream was to assess the feasibility of using a DCE to measure participants’ preferences for proximal femoral fracture rehabilitation services.

Methods

A DCE is a method for eliciting individuals’ preferences about goods and services. 349 It is based on the theory that individuals make trade-offs between choices. 349 The DCE used in this trial explored participants’ preferences for rehabilitation services. Hypothetical rehabilitation services (A and B) were presented in the form of a questionnaire given to each participant according to a set number of characteristics, each of which had different levels. These characteristics and levels created a set of alternative scenarios. In each scenario, the participant was asked which service they preferred by ticking A or B. In the DCE, the characteristics and levels were chosen based on the outcomes of the literature review and focus groups conducted in Phase I of the study (see Chapters 2 and 5).

Table 34 describes the attributes (characteristics) and levels chosen to create the alternative scenarios used in the DCE questionnaire. The different attributes and levels yielded a total of 108 hypothetical scenarios. However, to create a manageable number of scenarios for participants, we used the appropriate mixed-level orthogonal arrays360 to derive a reduced number of 16 hypothetical scenarios. For each of the 16 scenarios, participants were presented with two hypothetical service scenarios and asked to choose the preferred service, A or B. A member of the research team was present during completion of the DCE questionnaire at the 3-month follow-up to answer queries and help participants understand the questions and how to complete the questionnaire.

Analysis

Conditional logit regression techniques recommended by Ryan et al. 349 were used to assess which attributes were important to participants and the direction of participants’ preferences for rehabilitation services using Stata 10.1 (StataCorp LP, College Station, TX, USA). Non-parametric 5000 bootstrap replications using Stata 10.1 were also conducted as part of the analysis to gain stability around the coefficient estimates, to create a 95% CI.

Results

Of the 61 participants recruited to the trial, 12 withdrew. Of the remaining 49 participants, one participant completed the follow-up assessment over the telephone and therefore the DCE could not be completed as the measure cannot be administered over the telephone. One participant was given a previous version of the questionnaire that had more than 16 scenario choice sets. One participant was not available at follow-up to complete the measures. Eight participants did not receive any rehabilitation and therefore felt that the DCE questionnaire was not relevant to them and declined to complete it. Four participants struggled to understand the questionnaire and how to answer it, and therefore did not complete the measure. Two participants left questions blank; as the DCE questionnaire requires all questions to be answered for analysis to be conducted, these participants were removed from the final analysis. After removing the 12 withdrawn participants and the 17 non-completers, 32 participants remained in the DCE analysis. Ryan et al. 349 suggest that a sample size of at least 30 individuals is required to undertake DCE analysis; therefore, we analysed the results from the 32 sets of data received.

Table 35 describes the characteristics of the DCE sample. The DCE sample was similar to the main clinical and economic samples. There were more women than men in the sample and the mean age of the sample was 79.28 years. Slightly more participants in the DCE sample lived with others, whereas in the main clinical and economic samples there was an even split between those who lived alone and those who lived with others. As in the main clinical and economic samples, the majority of participants in the DCE sample owned their own home, with fewer participants living in privately rented properties, housing association/local authority properties or sheltered accommodation. Hemiarthroplasty surgery was the most common type of surgery received, followed by internal fixation. The majority of participants who completed the DCE questionnaire were recruited from Wrexham Maelor, with the fewest number recruited from Ysbyty Gwynedd.

During development of the DCE, we hypothesised that participants would prefer more time with the health-care professional and sessions to be delivered by a fully qualified physiotherapist or occupational therapist rather than a supervised unqualified assistant. We expected participants to prefer to monitor their own progress using a workbook/diary as opposed to progress monitoring being carried out by a health-care professional using medical records. We also hypothesised that participants would prefer to undertake any rehabilitation at home. Finally, we expected participants to prefer to receive information about both hospital and community additional services to support their rehabilitation. Table 36 provided the results of the conditional logit regression. The sign of the beta-coefficient indicates the direction of preference for each attribute; based on our hypotheses we expected positive beta-coefficients for all five attributes analysed.

Only two of the five coefficients (qualifications of health-care professional delivering the rehabilitation sessions and time with the health-care professional delivering the rehabilitation sessions) were statistically significant (p < 0.05), indicating that these attributes were important to the participants questioned. When these two significant results were ranked in order of importance, the qualifications of the health-care professional delivering the rehabilitation sessions was the most important. In accordance with our a priori hypotheses, the coefficients for time with the health-care professional delivering the rehabilitation sessions, rehabilitation progress monitoring and venue of rehabilitation were all positive. However, only time with the health-care professional delivering the rehabilitation sessions was statistically significant. In contrast to our a priori hypotheses, the coefficients for qualifications of the health-care professional delivering the rehabilitations sessions and additional rehabilitation service information were negative. This indicates that the participants preferred rehabilitation to be delivered by a supervised unqualified assistant rather than a fully qualified physiotherapist or occupational therapist and that participants also preferred the provision of additional information about hospital-based services only, rather than information about both hospital and community services. However, of these two attributes, only one was statistically significant: the qualifications of the health-care professional delivering the rehabilitations sessions. In the feasibility trial, physiotherapy/occupational therapy assistants delivered the majority of the five additional sessions received. Participants who had experienced the rehabilitation intervention may have had more contact with the assistants, thus building a relationship with them and preferring them to deliver rehabilitation. In addition, in usual care it is possible that physiotherapy/occupational therapy assistants conducted more of the rehabilitation sessions. There is further evidence from the Phase II focus groups (see Chapter 10) that physiotherapy/occupational therapy assistants focused more on ADL and confidence than physiotherapists/occupational therapists and that focus group attendees preferred these sorts of activities. With regard to the provision of additional information about services to support rehabilitation, following discharge from hospital, usual care in the trial was patchy and varied greatly across the region, and therefore participants were more likely to access community services such as the National Exercise Referral Scheme, leading them to prefer the provision of advice about additional hospital-based services.

Key findings

The results from this DCE analysis demonstrated that the methodology can be applied in an elderly population [mean age 79.28 years (SD 7.54 years)] following proximal femoral fracture.

Only four participants failed to complete the questionnaire because of a lack of understanding.

Two attributes were found to be important to participants: time with the health-care professional delivering the rehabilitation sessions and the qualifications of the health-care professional delivering the rehabilitation sessions.

The coefficients for three of the five attributes were positive (time with the health-care professional delivering the rehabilitation sessions, rehabilitation progress monitoring and venue of rehabilitation), as expected in our a priori hypotheses. However, in contrast to our a priori hypotheses, the coefficients for the qualifications of the health-care professional delivering the rehabilitation sessions and information about additional rehabilitation services were negative.

If DCE methods are to be used in a full-scale trial, further pilot work is required to ensure that the attributes and levels chosen for the feasibility trial are meaningful across the population of interest and to further test the questionnaire with regard to layout and wording.

Objective

The objective of this workstream was to scope the potential of conducting a SROI analysis based on the data received from the feasibility trial to inform whether or not a full SROI analysis would be possible in a full-scale RCT of the multidisciplinary rehabilitation intervention.

Methods

Using the measures administered in this feasibility trial, we assessed the completeness of the responses and the level of detail provided by the respondents to scope the potential for using these as a source of data for a larger-scale SROI analysis. SROI analysis requires data to quantify the tangible and intangible outcomes from an intervention compared with those of a control condition, assign a monetary value to those outcomes, and assess attribution and the length of effect of an intervention.

Results

For the main outcomes of the feasibility trial (see Chapter 9) and in the health economics analysis (see Work stream 1), we received a high level of complete data from the sample. At baseline the measures were approximately 90% complete, although at follow-up this reduced to approximately 80%. The CSRI,315 in particular, was completed well by the participants, with good detail recorded of the services accessed. This made calculating the costs of service use by participants in the economic sample, using national sources of costs, very manageable. However, we recognise that the CSRI questionnaire administered at follow-up was lengthy and that other methods of data collection, such as using hospital records, may improve the efficiency of data collection, in particular for hospital service use, in the future.

The current measures collected in the feasibility trial would provide data to describe what changes have occurred for two potential payer stakeholders: patients and publicly funded services such as the NHS. Measures of health-related quality of life, self-reported capability, improvement following rehabilitation, ADL, anxiety and depression, pain, self-efficacy and fear of falling describe changes for patients, whereas the health, social care and voluntary service use data collected using the CSRI describe the changes that have occurred for these services.

The benefit of using SROI analysis is that it allows wider and intangible outcomes, for example happiness, leisure time and socialising, to be explored. The measures used in the feasibility trial assessed a wide range of outcomes including fear of falling and ADL. However, we did not further explore potential wider outcomes as we wanted the number of measures used in the trial to be appropriate for the sample in question. This consideration was particularly important at the baseline assessment, which occurred very soon after the participants had experienced a proximal femoral fracture. As we did not include these wider outcome measures in the feasibility trial, we cannot say whether or not their inclusion in a full-scale RCT would allow SROI analysis to be conducted. In addition, we would need to further explore the data available for this population of interest, with particular reference to data banks and valuation techniques such as well-being valuation, to place a value on non-market goods and services. 361,362

In SROI analysis, it is essential to calculate what would have happened without the intervention, to attribute how much change the intervention is responsible for. In RCTs, the control group can provide an estimate of what would have happened without a particular intervention. However, we note that in the feasibility trial, participants assigned to the control group requested additional help and support from NHS rehabilitation services. In the trial, usual care was patchy and varied greatly across the region. The majority of control participants who requested additional help and support did so because they knew they were part of a research trial and therefore expected to receive something. A large-scale full RCT would need to ensure that those in the control group do not receive an intervention or additional services. The issue of control group compared with intervention group is key, because if the groups in the trial are not distinct then SROI analysis will not be able to determine what changes are the direct result of the intervention. Furthermore, to assist attribution, questionnaires can be administered to participants assigned to the intervention group, asking them to what extent the outcomes that they have experienced are the result of the intervention. Figure 20 provides an example of an attribution question. The appropriateness of using an attribution questionnaire for the sample in question needs to be carefully considered. As mentioned previously, participant burden when completely questionnaires was a key consideration for the research team.

FIGURE 20.

Example attribution question.

Key findings

Based on the data collected in the feasibility trial we cannot currently answer the question of whether or not a SROI analysis would be possible in a full-scale trial.

We identified the patients and publicly funded health and social care services as potential payer stakeholders in our scoping exercise.

Measures administered in the feasibility trial were acceptable to participants; however, we recognise that the CSRI315 questionnaire administered at follow-up was lengthy. In a future trial, efficiency may be improved by gaining access to hospital records to collect data about hospital services used by participants (see Chapter 12).

Further work is required to test the completeness and acceptability of additional measures such as attribution questionnaires and questionnaires to assess wider outcomes, for example happiness.

Additionally, in further work, the added benefit of conducting a SROI analysis compared with cost-effectiveness or cost–utility analyses using a multiagency perspective needs to be determined.

Summary of findings

We report a range of health economic analyses to address a number of objectives for this feasibility trial. A cost–consequences study explored the feasibility of conducting an economic evaluation in this trial, setting out disaggregated findings as the feasibility trial was not powered to test an effect in the primary outcome measure, the BADL index. Both the intervention group and the control group showed improvements in the EQ-5D health utility index, EQ-5D VAS score and ICECAP-O capability index at follow-up compared with baseline. There was no statistically significant difference between the groups for any of the measures, but this small feasibility trial was not powered to test such differences. The intervention group had slightly higher mean QALY gains than the control group, although this difference was not statistically significant. The only statistically significant differences between the groups were in mean total hospital costs and the total cost of service use per participant. However, this could be attributed to a larger proportion of the sample in one group than in the other experiencing longer inpatient stays. No other statistically significant differences were detected for the other service use categories. A DCE demonstrated that it is possible to use this method with an elderly population (mean age of 79.28 years). Only four participants failed to complete the questionnaire because of a lack of understanding. Two attributes were found to be important to participants: time with the health-care professional delivering the rehabilitation sessions, with participants preferring more time with the health professional, and the qualifications of the health-care professional delivering the rehabilitations sessions, with participants preferring supervised unqualified assistants to fully qualified physiotherapists/occupational therapists. A scoping study of the potential to conduct a SROI analysis as part of the feasibility trial identified that measures were well completed in the trial, with a small drop-off rate at follow-up (approximately 80% completion) compared with baseline (approximately 90% completion). Potential payer stakeholders included the patients and publicly funded health and social care services. Further work is needed in this area to assess the added value of conducting a SROI analysis compared with a cost-effectiveness or cost–utility analysis using a multiagency perspective.

Strengths and limitations

There is limited economic evidence relating to rehabilitation following proximal femoral fracture, particularly in a UK context. This feasibility trial and concurrent economic analysis, including a cost–consequences analysis, DCE and scoping study of SROI analysis, adds to the evidence base and addresses the particular need for more UK-based evidence, as highlighted by NICE. 7 We demonstrated that the intervention was relatively low cost, with a cost per participant of £231.43, and that measures such as the EQ-5D and ICECAP-O were acceptable to most participants. However, the ICECAP-O was subject to a lower completion rate than the EQ-5D (92% and 98%, respectively). Additionally, we demonstrated that a DCE questionnaire could be used with this elderly population. The lack of power to detect an effect in the main outcome measure (the BADL index) resulted in the use of cost–consequences analysis rather than cost–utility or cost-effectiveness analysis, which were specifically highlighted by NICE7 as being needed in the area of proximal femoral fracture. However, we used a range of methods to answer the specific objectives, including cost–consequences analysis, which is championed as a method that is particularly relevant to economic evaluations carried out alongside public health interventions. 352–354 The need to ensure that the battery of measures used in the feasibility trial was manageable for participants and reduce participant burden, particularly as the baseline assessment was conducted very shortly after participants had experienced their fracture, meant that the extent to which a SROI analysis could be conducted was limited. In the feasibility study, data were collected on outcomes such as health-related quality of life, self-reported capability, improvement following rehabilitation, ADL, anxiety and depression, pain, self-efficacy and fear of falling. However, wider and intangible outcomes, for example happiness, leisure time and socialising, were not assessed in the feasibility study. One of the main benefits of a SROI analysis is the ability to value intangible outcomes and further work is needed to assess the acceptability of such additional measures to the population included in this feasibility trial.

Comparison with previous literature

The systematic review reported in Chapter 3 did not identify any cost–consequences studies; therefore, we cannot make direct comparisons between our study and the previous literature. However, Olsson et al. 260 demonstrated using cost-effectiveness analysis in a Swedish hospital setting that an integrated care pathway, with a focus on motivation for rehabilitation and early first ambulation, was less costly and more effective than the usual care pathway for patients admitted with hip fracture. Additionally, Cameron et al. 253 found that, in an Australian general hospital setting, an accelerated rehabilitation intervention using early assessment, early mobilisation, physical independence and detailed discharge planning was more cost-effective than conventional rehabilitation in treating proximal femur fracture. These two analyses found that patient-centred approaches to rehabilitation were a good use of hospital resources. The intervention applied in this feasibility trial used personalised goal-setting and diaries to provide a record of progress during rehabilitation, utilising a similar ethos to that in the studies by Olsson et al. 260 and Cameron et al. 253 of patient-centred approaches to rehabilitation.

The mean EQ-5D health utility index scores for the intervention group [0.50 (SD 0.26)] and the control group [0.37 (SD 0.43)] were lower than the UK population mean score for men and women aged ≥ 75 years living in Wales [0.80 (SD 0.24)]. 363 At follow-up, the EQ-5D health utility index scores remained lower than the UK population mean score of 0.80 (SD 0.24), with the intervention reporting a mean score of 0.66 (SD 0.27) and the control group reporting a mean score of 0.60 (SD 0.27).

Additionally, the mean EQ-5D VAS scores for both groups were lower at baseline [intervention 64.43 (SD 16.37), control 55.14 (SD 25.72)] and follow-up [intervention 71.10 (SD 17.89), control 68.55 (SD 18.44)] than the UK population mean score for men and women aged ≥ 75 years living in Wales [82.91 (SD 12.33)].

Compared with the mean EQ-5D health utility index score reported by Davis et al. 364 for a female sample of older adults (mean age 79.3 years) with mobility impairments [0.701 (SD 0.291)], the mean EQ-5D health utility index scores in the feasibility study, which included a sample of predominantly women (women, n = 32, men, n = 11), were lower at both baseline and follow-up in both groups. The mean EQ-5D VAS scores in the feasibility study were also lower for both groups at both baseline and follow-up than the mean EQ-5D VAS score in the study by Davis et al. 364 [72.3 (SD 17.6)].

Davis et al. 364 administered the ICECAP-O to their female sample with mobility impairments, reporting a mean ICECAP-O capability index score of 0.815 (SD 0.177). In the feasibility study, the control group reported a lower mean score at both baseline and follow-up than the mean score reported by Davis et al. 364 Conversely, the intervention group in the feasibility study reported a slightly higher mean score at both baseline and follow-up than the mean score reported by Davis et al. 364

As in the feasibility study, previous DCE studies conducted by Milte et al. 365 and Laver et al. 366 demonstrated that this method can be used in a frail elderly population aged between 71 and 80 years who have experienced proximal femoral fracture, stroke and other neurological conditions. However, rather than focusing on the elements of rehabilitation as in this study, the line of inquiry in the study by Milte et al. 365 was to establish what frail older people in Australia are prepared to endure to achieve improved mobility following hip fracture. Attributes included the risk of falling and breaking another bone during rehabilitation, level of pain accepted to recover the ability to walk short distances, the level of effort needed during rehabilitation and ability to recover walking following participation in the programme. Laver et al. 366 conducted a DCE in southern Australia to explore the differences in preferences of patients, occupational therapists and other rehabilitation clinicians for the way in which rehabilitation services are provided.

Implications for practice and/or future research

The feasibility study demonstrated that measures such as the EQ-5D,287 ICECAP-O313 and CSRI315 are acceptable to participants, with an approximately 90% completion rate at baseline and an approximately 80% completion rate at follow-up. However, we note that the ICECAP-O was subject to a lower rate of completion in the economic sample. Although the measure was developed with a general population sample of older adults, we found that during the feasibility study three participants were unable to answer the questionnaire as they felt that the domains did not apply to them (e.g. they felt that they had no social life at the moment and so could not answer the questionnaire). Horwood et al. 367 conducted a think-aloud study of the face validity of the ICECAP-O with hip and knee arthroplasty patients. In the sample of 20 older individuals with hip or knee arthroplasty, a minority (n = 5) showed some confusion over, and had difficulty understanding, the questionnaire. Horwood et al. 367 commented that during the think-aloud exercise a minority of the sample interpreted the questions within their disease context, linking it directly with their joint replacement or other conditions such as arthritis. Horwood et al. 367 postulated that this context-specific answering was a result of the timing of the interview, which took place days before their joint replacement surgery. The acceptability of the ICECAP-O questionnaire in the feasibility study could also be linked with the timing of administration. This measure was administered shortly after fracture and then 3 months later. At 3 months post baseline, many participants still may not have gained mobility and, thus, may have felt dependent and frustrated that they could not currently do the things that they had previously enjoyed, such as gardening. The use of the ICECAP-O in a future RCT should be considered carefully. Although the measure explores a broader concept of well-being rather than being a purely health-related measure, the implication that participants may be answering it within a disease-specific context, and therefore find it distressing or difficult to answer, may mean that just administering the questionnaire is not sufficient. In a future RCT it may be useful to include an additional qualitative element as part of questionnaire administration to uncover how participants are interpreting the questionnaire.

Future research should also consider the value of conducting a SROI analysis compared with a cost-effectiveness or cost–utility analysis using a multiagency perspective. The feasibility study did not include the valuation of wider, intangible outcomes such as happiness, to keep the battery of measures manageable and to reduce the burden on participants. Further work is required to assess the acceptability and completeness of additional measures, such as attribution questionnaires and questionnaires, to assess wider outcomes.

Conclusion

The complete case cost–consequences analysis included a sample of 43 participants. The total cost of the intervention was £6711.46 (£231.43 per participant). There were improvements in both the intervention and the control groups in the EQ-5D health utility index, EQ-5D VAS score and ICECAP-O capability index at follow-up compared with baseline. However, these improvements were not statistically significant for any of the measures. The intervention group had slightly higher mean QALY gains than the control group, although this difference was not statistically significant. Statistically significant differences between the two groups in mean total hospital costs and total cost of service use per participant were found. In the DCE, participants indicated preferences for spending an increased amount of time with the health-care professional delivering the rehabilitation sessions and for a physiotherapy/occupational therapy assistant to deliver the rehabilitation sessions. The scoping SROI analysis showed high levels of completeness for the measures chosen for the feasibility study, although further work is needed to investigate more intangible outcomes and how these could be quantified.

In summary, although NICE7 has called for further cost-effectiveness evidence in this area, we have demonstrated that it is feasible to conduct a range of health economic analyses, including cost–consequences analysis, a DCE and a scoping SROI analysis. The intervention resulted in improvements in the economic measures used. These improvements were not statistically significant, but this small feasibility study was not powered to test such differences. The intervention adopted a patient-centred approach to rehabilitation following proximal femoral fracture, which has been highlighted by others in the field as being a cost-effective method. 253,260

Identifying patients

A patient identity check was undertaken through IDRIS (Identifying Data for Research in Information Systems), a bespoke software system designed by NWORTH and developed by the BCUHB informatics department. The identification number, surname, forename, date of birth, gender, postcode and NHS number of each of the trial participants were inputted into the IDRIS system, which resides on servers within the BCUHB network. Participant information was checked in real time against data held on the Master Patient Index at the Welsh Demographic Service to confirm (or fail to confirm) the identity of the participants. If the enquiry to the Welsh Demographic Service failed to confirm a participant’s identity, no data for that participant would be made available to the researchers. Data were extracted for the 49 participants in the study (which excluded those who had withdrawn from the main trial).

Extracting data

The following data were extracted from BCUHB systems: inpatient days and episodes, outpatient episodes and A&E department episodes. Each patient whose identity was confirmed by the IDRIS system was added to a register containing their details and participant identification number. Once all participants had been checked by IDRIS the register was closed and sent electronically to the BCUHB informatics department, remaining on BCUHB secure servers at all times. The previously agreed data records for each participant were then extracted and anonymised, and patient-identifiable information was removed, in accordance with BCUHB information governance instructions, leaving only the participant identification number to link the records to the other anonymised data held for each patient. The data were then transferred to the study researchers using a secure file transfer system.

Comparing data

Data extracted from BCUHB systems were compared with data obtained from the CSRI questionnaire for consistency using the intraclass correlation coefficient (ICC). A value on this measure of ≥ 0.7 shows good consistency between the two data sources.

Accident and emergency department attendances

In total, 60 A&E department attendances were logged in medical records compared with 43 reported in the CSRI data, which equates to under-reporting by 28%. Records for 24 of the 49 participants were found in both sources. Of the 24 participants, 21 under-reported attendances compared with BCUHB records and three over-reported attendances compared with BCUHB records. Other possible causes of over-reporting were loss of NHS records or episodes in which the patient presented directly to the acute medical unit, which was mistaken for the A&E department. An ICC model found a medium degree of agreement between the CSRI data and the BCUHB data. The single-measure ICC was 0.45 (95% CI 0.20 to 0.64; F_49,49 = 2.619; p < 0.001). This suggests that there was an approximately 45% agreement between the two data sets, but with a wide CI.

Outpatient episodes

For outpatient data, the under-reporting appeared to be less pronounced, with 204 reported episodes in the CSRI data compared with 215 reported episodes in the medical records. However, a review of the CSRI data for the two patients who were outliers (one participant reported 30 in the CSRI data, with five logged in the medical records; the other patient reported 42 in the CSRI data, with three logged in the medical records) suggested that the CSRI data represented the length in minutes of each episode, rather than the number of episodes. When these outliers were excluded, the CSRI data again suggested that patients were prone to under-reporting. Records from both sources matched for only seven participants, with 28 participants under-reporting compared with the medical records and 15 participants over-reporting compared with the medical records. The agreement between the CSRI data and the BCUHB data was low, with an ICC of 0.28 (95% CI 0.0 to 0.52; F_49,49 = 1.774; p = 0.024). This suggests that there was little similarity between the two data collection methods.

Inpatient days

For inpatient days, the data from the CSRI forms were well matched with the data obtained from the medical records. In total, 1254 inpatient days were recorded on the CSRI forms and 1274 inpatient days were logged in BCUHB records. The single-measure ICC was 0.88, representing a strong agreement between the CSRI data and the BCUHB data (F_49,49 = 15.311, p < 0.001), which suggests that the two methods reported similar results for the number of inpatient days. Only five of the individual patient records matched exactly between the two data sets, which may be because days is a more difficult unit to record accurately than episodes, as was used within the A&E attendances and outpatient episodes comparisons. After several weeks in hospital it may have been difficult for patients to recall the exact number of days that they had been admitted, or there may have been differences in the way that arrival and departure days were counted. If an error margin of 6 days was allowed to counter such small errors, 39 of the 49 patients correctly recorded their inpatient stay.

During the comparison of the two data sets three data input errors were identified, which occurred when inputting data from the hard-copy CSRI forms into the MACRO database. The errors seemed to have occurred at random and straddled both treatment groups and so the discrepancies did not generate bias. These errors were within the 2% error rate that would trigger a full data entry audit (as specified in the data management plan) and so no further action was required.

Summary

The triangulation study embedded within the Fracture in the Elderly Multidisciplinary Rehabilitation (FEMuR) study tested the methodology devised by NWORTH and BCUHB to make routinely collected data in BCUHB secondary care systems available to researchers after informed consent has been given by study participants. The methodology to provide such data was successful and it was demonstrated that it was feasible to obtain research data in this manner. Although some areas showed high levels of comparability between data obtained from medical records and data obtained from the CSRI, the two methods did not universally produce the same results. Further investigation will be required to elucidate the reasons for the discrepancies and confirm which method of data collection provides a higher level of accuracy.

Strengths and limitations

Provision of routinely collected data using electronic methods for research purposes holds considerable potential for future studies. In this study a novel methodology was applied, which meant that the process of making the data from hospital records available was slow and laborious. The data required were confidential and of a potentially sensitive nature, and each stage of the process required careful planning and liaison with stakeholders to ensure that the correct procedures and security processes were followed. In particular, liaison with the data keepers and BCUHB information governance had to progress with care and there were understandable and valid delays. Despite these challenges, the data were successfully obtained and the software and methodology are now in place. Future uses of this process will thus be much more streamlined while still ensuring that security procedures and good practice are adhered to.

Although it was possible to account for some of the discrepancies in the data by identifying data input errors or accounting for errors in patient recall, it was not possible to check the accuracy of the data collected from BCUHB systems. This is a current limitation that should be addressed in future work as it is not currently possible to confidently conclude which of the data sets provides the most accurate representation of service use.

Comparison with previous literature

There have been many studies in which routinely collected administration data have been compared with data obtained from patient self-reporting. These studies have been carried out in different settings and in different countries, thus reflecting different methods of collecting routine data and different IT systems for storage. The results reported have been contradictory and inconclusive. For example, one study reported a high concordance between self-reported and claims-based hospital episodes, but concordance for physician visits was low. 368 Factors significantly associated with bidirectional (over- and under-reporting) and unidirectional (over- or under-reporting) error patterns were detected. Caution was advised when drawing conclusions based on just one physician visit data source. Consistent with our findings, another study found that patients tended to report a lower number of visits to physicians than was recorded in the computerised medical records. 369 Similarly, a study of external sources of secondary care data in older German adults showed that survey estimates based on self-report tended to underestimate true health-care use in the older population. 370 In a prospective cohort study in Germany, self-reports regarding psychiatric inpatient and day-care use collected by telephone interviews based on the Client Socio-Demographic and Service Receipt Inventory, as well as calculated costs, were compared with data from computerised hospital records. The different data sources resulted in similar estimates at the population level; however, there were pronounced differences for outpatient visits at an individual level. 371 The accuracy of the results was heavily dependent on context. For example, GP records provided more accurate data on primary care contacts than patient reports, but less reliable information on contacts with other health services. In a further study, reliance on GP records for data related to hospital services and other community health services based outside general practice surgeries was not recommended. 372 A recently published review of studies on the Database of Instruments for Resource Use Measurement (DIRUM)315 found evidence for a good correlation between medical records and patient or carer recall. 373

Finally, the conclusions of Williams et al. 374 are still valid today:

Routine data have the potential to support health technology assessment by RCTs. . . . further work is required to improve the validity of routine data . . . . Better knowledge of the capability of local systems and access to the data held on them is also essential.

Implications for future research

The use of routinely collected data has the potential to improve the efficiency of trials and other studies. Although the methodology to make the data available has been successfully demonstrated, further investigation is required to evaluate the quality and accuracy of the data obtained and to test the validity of different data sources in different contexts.