The feasibility of early pulmonary rehabilitation and activity after COPD exacerbations: external pilot randomised controlled trial, qualitative case study and exploratory economic evaluation

Epidemiology

Chronic obstructive pulmonary disease (COPD) affects > 3 million people in the UK and 210 million worldwide. 1,2 Many patients remain undiagnosed and attribute the symptoms to ageing or other medical conditions. 3 Up to 2 million people in the UK4 have no formal diagnosis of the condition and are unaware of the implications until the disease is in an advanced stage. Presentation typically occurs in the over-50s, with equal prevalence in men and women when adjusting for confounders such as smoking history. 5 Although some therapies reduce the rate of progression, many clinicians focus on relief of symptoms from the outset. Acute exacerbation of COPD (AECOPD) is defined as a sustained worsening of the patient’s symptoms from his or her usual stable state that is beyond normal day-to-day variation and is acute in onset. 1 It is typified by more marked breathlessness and a mucopurulent cough. Exacerbations frequently require hospital admission and AECOPD is the second most common reason for emergency hospital admission in the UK. 6

Aetiology

Acute exacerbation of COPD may be triggered by viral or bacterial infection but is also associated with greater background sputum production, a history of gastro-oesophageal reflux disease and varying underlying immunological responses. In those patients who experience frequent exacerbations, disease progression is associated more with exacerbation frequency than smoking status. 5,7,8

Pathology

During exacerbations of COPD, patients generally develop increased cough with change in sputum viscosity or colour, breathlessness, wheeze and fatigue. Airway inflammation and excess mucus production cause gas trapping and impaired gas exchange, resulting in hypoxaemia and tissue hypoxia. Decreased appetite, in addition to excess protein loss through increased sputum production, can result in nutritional deficiency. Systemic inflammation, tissue hypoxia and nutritional deficiency can cause changes to cardiac, cerebrovasculature and skeletal muscle function,9,10 elevating the risk of cardiovascular disease and skeletal muscle wasting. 11

Prognosis

Acute exacerbation of COPD is associated with accelerated disease progression and increased mortality;12 patients with frequent episodes have a more rapid decline in lung function13 and quality of life14 and decreased exercise performance. 15 Quadriceps force and muscle mass may fall by 5–10% between the third and eighth days of hospitalisation. 16 Loss of muscle function is, in turn, associated with poor exercise tolerance;17,18 quadriceps muscle strength and mass may be predictors of mortality, independent of change in lung function. 19,20 Muscle weakness is caused by physical inactivity,21 use of oral corticosteroids,22 systemic inflammation,16 negative nutritional balance,23,24 increased resting metabolism,23 hypoxia23,24 and hypercapnia. 25–27

Many patients with COPD underestimate the severity of their disease and do not present until the pathology is advanced. 28 As a result, AECOPD requiring hospitalisation may be the first presentation of their condition. Such patients have a much poorer prognosis and are more likely to have one or more comorbidities. 29 Exacerbation frequency is a significant prognostic factor determining future outcome. Previous exacerbations predict risk of future exacerbations and there is a direct relationship between exacerbation severity, exacerbation frequency and mortality. In addition, patients with frequent exacerbations are more likely to experience functional limitations and disabling symptoms, have more frequent contact with health-care professionals, have more comorbidities and have greater associated health-care utilisation and costs. 30

Burden

It is estimated that the direct costs to the NHS in England associated with COPD care are in excess of £800M per year,1 with AECOPD being the second most common reason for emergency admissions in the UK. 6 Reducing hospital admissions and improving patient outcomes following COPD exacerbations has been a major focus for the NHS. 31 Although quality of life and functional improvements have been shown following interventions in stable COPD patients, such assessment is poorly reported immediately following AECOPD. Some studies have shown a trend towards a reduction in mortality but not readmission frequency when patients receive enhanced support at home following hospitalisation, although the evidence is not consistent. 32 Therefore, the true benefit to patients of interventions during this time is difficult to estimate.

Pharmacological management and multidisciplinary care

During hospitalisation with AECOPD, patients will receive medical interventions such as bronchodilators, corticosteroids and antibiotics. Although physiotherapy is available, the focus is primarily on airway clearance and functional assessment rather than active rehabilitation. Aside from smoking cessation, few interventions have demonstrated reductions in exacerbation frequency, disease progression or mortality. 33 Many patients will receive a review post discharge, predominantly conducted by nursing staff. There is no standard model of exercise following discharge from hospital and the offer of active rehabilitation following AECOPD is not universal, although it is recommended. 34 For those who are offered active rehabilitation, the programme will often begin at least 1 month after the exacerbation. 32 The provision of in-home physiotherapy is uncommon and not nationally recommended at present.

Pulmonary rehabilitation

Pulmonary rehabilitation (PR) is defined as an interdisciplinary programme of care for patients with chronic respiratory impairment that is individually tailored and designed to optimise each patient’s physical and social performance and autonomy. Programmes consist of individualised exercise programmes and education. 35 Traditionally in the UK, PR following AECOPD is provided in a community setting; however, PR may also be provided in hospital.

Structured PR programmes following AECOPD should incorporate interval training and continuous exercise to increase exercise capacity while improving symptoms and quality of life. 34,36–40 The British Thoracic Society (BTS) guidelines34 recommend, at a minimum, twice-weekly supervised sessions over 6–12 weeks commencing within 1 month of hospital discharge. PR aims to restore the patient to the highest possible level of independent function through increased physical activity and the provision of education about their disease, different treatment options and coping strategies. 41 Systematic reviews have shown large and important clinical effects for those who adhere to PR programmes in terms of quality of life and daily functioning. 36,42 Despite the established benefits and widespread availability of PR, many patients are reluctant to attend because of misconceptions about the nature of the exercise training, social isolation or transportation difficulties. A UK audit found that < 10% of all hospital discharges for AECOPD complete early post-hospitalisation rehabilitation43 and completion of the full programme can be especially poor in patients with a recent hospitalisation and those receiving long-term oxygen therapy. 44,45

Theoretical basis for pulmonary rehabilitation

In healthy people, physical fitness improves when exercise is undertaken 3–5 days per week at an intensity above 40–85% of the oxygen uptake reserve (difference between resting and peak oxygen uptake) for > 20 minutes (or at lower intensity, preferably for 30 minutes), continuously or in intervals. 46 These principles are no different when applied to people with COPD; however, interventions delivered during PR must recognise and adapt to the limitations to exercise caused by the disease and physical training must be specific to an individual’s requirements. For exercise programmes to be effective, the training load must exceed the loads normally faced during activities of daily living to improve both ventilator capacity and peripheral muscle strength, with training loads being increased as improvement occurs. 37,45,46 PR programmes usually combine exercises that stimulate both the cardiovascular system and the peripheral muscles to reverse the peripheral muscle weakness that is seen in COPD. 47

Targeting individual peripheral muscle groups also helps to minimise the impact of exercise on the respiratory system, allowing patients to undertake exercise without intolerable increases in their respiratory symptoms. Resistance training has been demonstrated to be an effective method of exercise training in COPD. 48–52

In moderate COPD, aerobic training results in significant physiological effects. 53

Intervention methods and materials

In the UK, PR is generally offered in hospital outpatient or community settings, with a minimum of two supervised sessions per week;34 this frequency is not demonstrably optimal but is based on published studies that encompass two supervised sessions and either a third supervised or a formalised unsupervised PR session. 42,54 Programme duration is variable across Europe and globally. On cost grounds most UK programmes last between 6 and 8 weeks; there is debate over the efficacy of programmes lasting for < 6 weeks. 34 Educational components of PR are integral, appearing in every aspect of PR and in discrete educational sessions. They aim to support lifestyle change, behaviour change and self-management to promote decision-making and self-efficacy.

Introduction

The efficacy of PR following AECOPD is established, with current national guidelines recommending post-exacerbation rehabilitation commencing within 1 month of hospital discharge. 34 However, the detrimental effects of AECOPD on physical fitness and skeletal muscle function occur rapidly during the inpatient phase,16,55 suggesting that a rehabilitation intervention delivered at the time of the acute illness might have a role in preserving muscle strength and maintaining physical function.

Small-scale trials of PR during AECOPD suggest intervention feasibility and effectiveness. 56,57 However, a large trial (n = 389) in which patients were randomised to a 6-week rehabilitation programme, starting in hospital and continuing after discharge, failed to demonstrate significant improvements in muscle strength, function or quality of life. 58

Theoretical basis of early pulmonary rehabilitation

Acute exacerbation of COPD contributes to disease progression and has a significant systemic impact. Severe exacerbations are characterised by a hospital admission, during which significant skeletal muscle weakness has been observed,16,21 along with a negative protein balance. 59 Several mechanisms, including the presence of systemic inflammation,16 a negative nutritional balance,23 administration of oral corticosteroids22 and physical inactivity,21 may contribute to this acute muscle weakness. A number of potential interventions for maintaining muscle strength and physical function during and immediately after AECOPD, in a hospital, home or community setting, have been proposed. These interventions differ from traditional PR as they must be well tolerated by highly symptomatic patients with increased ventilatory limitation and respiratory symptoms.

Evidence for effectiveness

Rationale

The National Institute for Health Research (NIHR) commissioning brief [HTA 13/24; see https://njl-admin.nihr.ac.uk/document/download/2009846 (accessed 14 December 2017)] requested a ‘feasibility study’ in ‘early pulmonary rehabilitation after an exacerbation of COPD’, to be started as soon as possible following an acute exacerbation. The brief requested that the control group receive usual care and that the study focus on the ability to recruit, randomise and deliver the intervention, leading us to interpret the brief as requiring an external pilot RCT, sometimes defined as ‘a version of the main study run in miniature to determine whether the components of the main study can all work together’. 80 In other words, we interpreted the brief primarily as requesting a study to understand the feasibility of carrying out a definitive trial. However, as the diffusion of innovations in health services is often difficult,81 and there was no clear front-running intervention to assess in a large-scale evaluation, we also assessed the feasibility of delivering early pulmonary rehabilitation (EPR) in NHS secondary care (in-hospital) and community (post-discharge) settings.

Primary objective

The primary objective was to assess the feasibility of carrying out a definitive RCT that would test the hypothesis that, compared with current practice, early initiation of PR is more clinically effective and cost-effective in AECOPD.

Secondary objectives

1. Carry out an external pilot RCT to determine:

   • the availability of eligible patients and the likely rates of participant recruitment and attrition

   • whether or not data of acceptable quality can be collected

   • whether or not the research interventions can be delivered per protocol

   • key design features including the best primary end point and sample size for the main trial.

2. Carry out fully integrated qualitative research to determine, in line with the Medical Research Council (MRC) framework:82

   • potential barriers to recruiting participating centres in the main trial

   • reasons for patient refusal of consent and obtain data on whether or not the baseline characteristics and adherence to routine treatment of non-recruited patients differ from those of consenting participants

   • the reasons for participant attrition

   • the acceptability of the research and intervention procedures to participants and health professionals.

3. Carry out a fully integrated health economic analysis and modelling to:

   • identify key drivers of NHS and social care costs

   • pilot data collection strategies in advance of the definitive trial

   • quantify the potential benefit of running the definitive trial.

Trial design

This was a parallel-group, randomised pilot 2 × 2 factorial trial (with equal allocation ratio for each of the four groups) comparing hospital EPR, home EPR, both hospital EPR and home EPR and usual care alone (delayed community-based group rehabilitation).

Participants

Participants had been admitted to collaborating centres with an acute onset and sustained increase in symptoms from a stable state outside the expected range of variation. Typical symptoms are were ‘worsening breathlessness, cough, and increased sputum production and change in sputum colour. The change in these symptoms often necessitates a change in medication’. 34

Settings and locations where the data were collected

Sheffield Teaching Hospitals (STH) NHS Foundation Trust (NHSFT) was the trial sponsor and the trust was the clinical co-ordinating centre. Co-ordination of the trial was undertaken by the Clinical Trials Research Unit (CTRU). Patients were approached and recruited at STHNHSFT and Aintree University Hospital (AUH) NHSFT, the two participating centres. Research interventions were delivered by hospital trust employees in participating hospitals (hospital EPR) and in homes in their service catchments (home EPR).

Both hospitals are large teaching hospitals: STHNHSFT serves a population of 550,000 in Sheffield and had 1194 COPD admissions in 2016; AUHNSFT serves a population of around 330,000 in North Liverpool, South Sefton and Kirkby and had 1486 COPD admissions in 2016.

Screening, assessment of eligibility and consent

Eligible patients were identified by physicians, physiotherapists and research nurses, who provided the participant information sheet and discussed the trial with patients. Patients were given a 1-hour cooling-off period to consider trial participation. Patients had to be entered into the trial within 48 hours of admission to allow enough time to complete the baseline assessments and start the in-hospital intervention. The cooling-off period therefore had to be kept short; however, if patients required more time to make a decision, this was accommodated within the 48-hour window when possible. Written informed consent was obtained from every participant by the Principal Investigator or a suitably trained delegate (physician, physiotherapist or research nurse). An eligibility form was completed for all participants.

Randomisation

Following consent and the baseline assessments, participants were allocated in equal proportions to one of four groups:

1. both hospital EPR and home EPR

2. hospital EPR but no home EPR

3. no hospital EPR but home EPR

4. usual care (neither hospital EPR nor home EPR).

All participants were invited to attend group rehabilitation as per usual care guidelines. However, those allocated to home EPR were to receive four fewer sessions to account for the four sessions delivered in the home.

The randomisation list was created using a computer-generated pseudo-random list, stratified by centre, with random permuted blocks of varying sizes; it was hosted by the Sheffield CTRU in accordance with standard operating procedures and held on a secure server. Access to the allocation sequence was restricted to those with authorisation. The sequence was concealed until recruitment, data collection and the analyses were complete. Randomisation was completed by the physician or physiotherapist (or central team member when they were not available), who then made the arrangements for the appropriate intervention(s) to be delivered. Participants were informed of their allocation in two stages: (1) they were told about hospital EPR following randomisation and (2) they were told about home EPR at discharge. Trial physiotherapists delivered both hospital EPR and home EPR (as well as aspects of usual care). Outcome data were collected by research nurses or other clinical staff, who were blind to the group allocation of the participants when possible; if blinded staff were not available or became unblinded, we still collected these data and recorded that the assessor was unblinded.

Interventions

Following randomisation the research physiotherapy team received a randomisation e-mail and arranged for the allocated treatment to be delivered. To arrange the home visit, the team delivering home EPR communicated with the hospital team to arrange the first home visit within 72 hours of discharge. Other visits were arranged directly with the participants. The interventions are described in the following sections according to guidance from Hoffmann et al. 87

Feasibility criteria

Clinical outcomes

Clinical outcome assessments are provided by time point in Table 1, with more detail provided in the following sections.

Sample size

The study was an external pilot trial intended to explore the feasibility of conducting a future definitive trial. The sample size for a feasibility study should be adequate to estimate the uncertain critical parameters [standard deviations (SDs) for continuous outcomes; consent rates, event rates and attrition rates for binary outcomes] needed to inform the design of the full RCT with sufficient precision. A sample size of 60 patients with 3-month outcome data (76 randomised with 20% dropout) would allow a SD to be estimated to within a precision of approximately ±19% of its true underlying value with 95% confidence. This estimate was synthesised with SDs observed in other published studies and ongoing trials to provide a robust estimate for use in the sample size calculation for the full trial.

Blinding

The care team and the participants were not blinded to the interventions. The research nurses (or clinicians) who collected the outcome data were blinded to treatment allocation although they recorded if they became unblinded for individual patients during the trial.

Statistical methods

Data are presented according to relevant guidance. 84,85 Analyses were performed using SAS 9.4 (SAS Institute Inc., Cary, NC, USA).

Optimisation of prescription

Fidelity is usually interpreted as the consistent delivery of intervention components. 112 However, physiotherapy interventions do not always benefit from consistency as they involve the revision of treatment plans to account for changing and uncertain experiences. 113 Evidence suggests that the fidelity to the treatment theory is more important114–116 than consistent delivery of the intervention. We therefore evaluated treatment optimisation to the patients’ needs and capabilities.

Decisions used to assess session optimisation are shown in Figure 4 (hospital EPR) and Figure 5 (home EPR). Assessment was undertaken by two co-applicant physiotherapists. Assessment was conducted on all participants allocated to hospital EPR (assessed by MC, hospital physiotherapist) and home EPR (assessed by CO’C, community physiotherapist). Decision rules were agreed with the TMG.

FIGURE 4.

Decisions used to assess clinical optimisation of hospital EPR for each session.

FIGURE 5.

Decisions used to assess clinical optimisation of home EPR for each session.

Adherence to the prescription

Physiotherapists who delivered the hospital EPR intervention recorded the number of rotations undertaken at each session and those who delivered the home EPR intervention recorded the exercises undertaken at each session. Physiotherapists recorded any AEs that occurred and, when possible, the reasons for not taking part in a session (e.g. patient choice, illness or at another appointment).

Assessment of overall treatment optimisation

An overall assessment of optimisation for each intervention by participant was obtained using the assessment of each session. The following rules were used:

1. count the number of exercises completed (1 for completed, 0 otherwise) to provide a value for each session (0–8)

2. count the number of exercises optimised (1 for optimised, 0 otherwise) to provide a value for each session (0–8)

3. for each session, divide the number of exercises optimised by the number of exercises completed to give a ‘single optimised’ score

4. if fewer than five exercises were completed no optimised score is calculated

5. the previous steps are carried out at the ‘session level’; the mean score from all sessions is used as the overall score for each participant.

Interviewer characteristics

Interviews were conducted by DH, a male graduate anthropologist and researcher, KB, a female graduate psychologist and researcher, and CO’C, a female physiotherapist and clinical research graduate.

Relationship with participants

No relationship was established with any participant prior to or outside the interviews; some participants were interviewed twice by the same researcher. Participants were fully informed of the purpose of the interview at the time of consent; this was checked at the start of each interview.

Theoretical and thematic framework

Participant selection

Setting

Semistructured interviews took place between 15 December 2015 and 11 August 2016 for participants and between 29 April 2016 and 26 August 2016 for physiotherapists and research nurses. Participants chose the setting for data collection: most were interviewed by telephone, with six interviewed in their home. In general, interviews were conducted in quiet and private settings to reduce distractions; one interview was conducted with the participant’s daughter present. Two physiotherapists were interviewed in person, one at their place of work and one at the university; other health professionals were interviewed by telephone.

Data collection

In addition to the a priori themes identified in Table 3, semistructured interview guides for participants contained questions about the acceptability of intervention and research protocols.

The interview guide for health professionals adapted questions suggested by the NPT developers. 134 Interview guides were not piloted; eight participants were interviewed at two time points but otherwise no repeat interviews were undertaken and transcripts were not returned to participants for correction.

All interviews were recorded on an encrypted digital recorder and fully transcribed, with transcriptions anonymised; notes were not taken by the interviewers. Participant interviews at 7 days lasted a median of 16 (range 4–77) minutes and at 90 days lasted a median of 14.5 (range 5–32) minutes, with the duration of the interviews related to the participants’ length of responses to the questions. Physiotherapist interviews lasted a median of 52 (range 44–81) minutes. Although we did not interview eight participants from each group, a total of 27 participant interviews were carried out in total and new content did not seem to appear in later interviews. There was, for the most part, consistency in messages across the groups.

Data analysis

CO’C, KB and DH independently coded samples of transcripts before conferring with each other and the study patient representatives to agree the working coding tree. The coding frameworks are provided in Figure 6. Umbrella themes and subthemes of a priori interest were identified deductively through recourse to the three frameworks: ICF, TDF and NPT (see Theory). The comprehensive core set of the ICF for obstructive pulmonary disease130 includes 17 functions, with a brief core set of 14 functions, and interview transcripts were coded according to these functions. The TDF consists of 14 domains used in behaviour change psychology; each domain of the TDF maps to a COM-B (capability, opportunity, motivation and behaviour) component. 136,140 Skills, although considered a single domain in the TDF, maps to both physical and psychological capability and so appears twice in Figure 6 and Table 3. To understand the perspectives of physiotherapists delivering the two interventions, we used NPT to assess how complex interventions are implemented and maintained in routine practice. NPT categorises the various tasks involved in embedding interventions into services as sense-making, relational, operational and appraisal; each construct contains four subthemes. 132,141 Implementation failure can arise if there are problems in any of these categories.

FIGURE 6.

Coding frameworks.

To assess the conflicting attitudes of participants to EPR and group PR (usual care), the following intervention dimension categories were not derived from a framework but were derived inductively by identifying the differences between EPR and usual care: timing of PR, location of PR, flexibility of visits, one-to-one therapy and introduction to PR (without committing to a PR group programme). These were reported alongside the data for group PR attendance.

Analysis of participant themes took place in NVivo 11 (QSR International, Warrington, UK). Quotations are presented to illustrate the themes. Narrative vignettes about selected individual participant cases are provided for illustrative purposes; they express maximum variation in participant characteristics and findings and combine triangulated data from different embedded units of analysis. A cross-case synthesis142 is provided aggregating findings across a series of individual cases to explore similarities and differences in the findings.

The findings were not discussed with participants for feedback.

Overview

An economic evaluation was undertaken to compare the potential incremental cost per quality-adjusted life-year (QALY) of three interventions (home EPR, hospital EPR and home EPR and hospital EPR) compared with usual care over a 90-day trial time horizon. Because of the nature of a pilot study, these cost-effectiveness results are provided to (1) suggest if there is the potential for the interventions to be cost-effective, which could be assessed further in a larger trial; (2) assess the uncertainty around the cost and effect (QALY) estimates and the incremental cost-effectiveness ratios (ICERs) produced; and (3) quantify the expected value of perfect information (EVPI) from obtaining more information from a larger study. The perspective of the economic evaluation was that of the NHS and social care.

Measurement of health status and effectiveness

The EQ-5D-5L data were collected at baseline, pre discharge (before patients were discharged from their index hospital admission; the timing of this varied between patients) and 30 and 90 days post discharge. The EQ-5D-5L was assigned a preference weight based on the UK tariff scores produced by Devlin et al. 143 to calculate QALYs. This is the method recommended by NICE144 for economic evaluation and preference-weighted measurement.

Measurement of resource use

There were three main sources of resource use information: (1) physiotherapist-recorded time, relating to how much time physiotherapists spent on a hospital or home EPR session; data recorded by physiotherapists on case report forms included whether or not a planned session was started and completed and the session time; (2) a person-reported modified Client Service Receipt Inventory (CSRI) collected at both 30 days (describing patients’ resource use between baseline and 30 days) and 90 days (describing patients’ resource use between 30 days’ and 90 days’ follow-up) post discharge; and (3) hospital ward notes, which included the dates that patients were admitted and discharged from hospital for the index hospital admission and any readmissions within the 90-day trial period. Only events or hospital stays that occurred within the 90-day trial period were included in the analysis. The types of resource use parameters from a NHS perspective included in the CSRI and used in this analysis were the number of contacts with a general practitioner (GP) (in the surgery, in the home or by telephone), physiotherapist, occupational therapist, social worker, home care worker and health visitor.

Unit costs

The price year was set to 2015/16. For the interventions, three types of resource use were identified for which a unit cost needed to be sourced: (1) the physiotherapists’ time (hospital EPR and home EPR), (2) the bike purchased for the hospital EPR sessions (hospital EPR only) and (3) a booklet describing the exercises that the patient should carry out in-between physiotherapist visits in the home (home EPR only).

Other resources for which a unit cost was sourced included GP contacts, therapy services and hospital visits. All unit costs are presented in Appendix 3; a summarised description of these unit costs is provided below.

The unit cost of the physiotherapists’ time was based on a mid-point band wage estimate;145 additional associated staff costs (employee national insurance, salary on-costs, overheads and capital overheads) were included based on the calculations used in Curtis and Burns. 146

The cost of the bike used for hospital EPR (MOTOmed viva 2;147 RECK-Technik, Betzenweiler, Germany), including value-added tax (VAT), was £5520. The equivalent yearly cost of the bike was estimated using a annuitisation procedure. 148 For this 90-day study, the equivalent cost of the bike was £295, which is attributed to the intervention cost for those people within the hospital EPR arm and the hospital EPR and home EPR arm.

The home EPR costs consisted of the physiotherapists’ time to deliver the intervention, physiotherapists’ travel time to the home of participants and back to their base of operation and the printing of a 20-page A4 information booklet (it was assumed that each patient could keep the booklet even after the intervention was over).

For GP contacts and therapy services, unit costs were sourced from Curtis and Burns. 146 Hospital visit unit costs were sourced from the NHS reference costs for 2014/15. 149

Economic analysis

The health economic analysis was restricted to those in the ITT population who completed the study. The rationale for basing the economic analysis on this group is that the model-based economic evaluation is based on the observed data from this patient group (because of the small sample size in this pilot study, multiple imputation to account for missing data was not deemed appropriate) and therefore those who withdrew or who were lost to follow-up would automatically not contribute data at specific time points of interest (in particular, at 90 days’ follow-up for the EQ-5D-5L and CSRI). The average cost per patient was calculated by combining the resource use estimates with unit costs.

It was assumed for the purpose of this analysis that all participants across all trial arms received usual care. The incremental intervention cost was therefore estimated to consist of only the cost of the intervention, omitting the cost of usual care; this is because, when using this assumption within this incremental analysis, the cost of usual care cancels out across all trial arms. However, not all participants received usual care in the intervention arms, with some receiving only the intervention without usual care [see Chapter 3, Attendance at group pulmonary rehabilitation (usual care)]. This means that we may have overestimated the cost within the intervention arms compared with the usual care arm (thus underestimating the potential cost-effectiveness, as the effect on health status is still captured in the same way using the EQ-5D-5L). However, this assumption avoids the need to estimate the cost of usual care directly and avoids complicating the exact cost estimation of the interventions with and without usual care between individuals, within and between trial arms.

The economic evaluation for this pilot study is based on an economic decision tree model over a 90-day time horizon, with incremental cost-effectiveness based on the intervention compared with usual care (the model is presented in Figure 7). The rationale for using an economic model rather than a trial-based evaluation was to utilise all of the observed data from the study and then parameterise the uncertainty around the observed data to conduct a probabilistic sensitivity analysis over 10,000 iterations to generate cost-effectiveness acceptability curves (CEACs) and to subsequently conduct an EVPI analysis150 (the EVPI analysis is described in more detail in Expected value of perfect information analysis). A one-way sensitivity analysis could also then be conducted to determine the key outcome drivers in the model and partial EVPI could be conducted to assess the value of collecting more data around specific model parameters. The information from the economic model and subsequent EVPI analysis can be used to inform the design of data collection methods and the potential economic value of a future larger trial. ICERs and cost-effectiveness planes are reported and health status, resource use and cost profiles are described as part of the analysis to provide descriptive statistics around the changes in health and resource use of this patient group between trial arms; missing data are also a focus of these descriptive statistics to assess how well the data collection methods worked in this pilot study.

FIGURE 7.

Illustration of the decision tree model used for the analysis. See Economic model: decision tree for description of terms.

Quality-adjusted life-years were calculated using the area under the curve (AUC) method. 148 The EQ-5D-5L data collected pre discharge were not used in the calculation because of the variable time points at which these data were collected (however, the EQ-5D-5L tariff scores at this time point are reported for descriptive purposes). There are two ways to quantify the AUC based on when the EQ-5D-5L data were collected: (1) use the data collected at baseline and 90 days’ follow-up and (2) use the data collected at baseline and 30 days’ and 90 days’ follow-up. The benefit of using option (1) is that it requires completion of the EQ-5D-5L at only two time points rather than three and therefore there is the potential for fewer missing data points. The benefit of using option (2) is that data from three time points should enable a better health status profile to be elicited over the trial period by accounting for an intermediate change in health status at 30 days rather than assuming a linear change between baseline and 90 days. Given the design of the economic model, option (2) is preferred to option (1) to best reflect the change in health over the 90-day period rather than focusing on the issues around missing data. However, the difference in QALYs using the two methods was assessed using a t-test for two unpaired samples and assuming unequal variance because of the likely difference in sample sizes and variability caused by the inclusion of patients who did not complete the EQ-5D-5L at 30 days.

The statistical analysis was conducted using Stata^® version 14 (StataCorp LP, College Station, TX, USA), the economic decision tree model was built and the analysis performed in TreeAge Pro 2015 (TreeAge Software, Inc., Williamstown, MA, USA), the EVPI analysis was conducted using the Sheffield Accelerated Value of Information Tool151 and all figures for the economic analysis (e.g. cost-effectiveness planes and CEACs) were generated within Microsoft Excel^® 2010 (Microsoft Corporation, Redmond, WA, USA).

Rationale

We used different methods in this study and compared their findings using a formal framework, to investigate different components of the research question, to increase the confidence of patients and professionals in our findings and to provide a platform for their feedback.

Design

The qualitative and quantitative methods were used concurrently, with priority granted to neither. We used a modified triangulation protocol152 to relate the qualitative and quantitative findings (methodological triangulation of data sets) in five steps: sorting, convergence coding, convergence assessment, completeness assessment and feedback (resource constraints meant that researcher comparison was not undertaken).

Each data set was reviewed to identify whether or not examples of intervention logic model components (Figures 2 and 3) were present (‘sorting’). We summarised similarities for each logic model component in a matrix using the following codes: ‘agreement’ – coherent interpretation of the data sets; ‘partial agreement’ – some disagreement between data sets; ‘silence’ – the logic mode component is covered by only one data set; and ‘dissonance’ – disagreement between data sets or sources (‘convergence coding’). The level of agreement between data sets was quantified (‘convergence assessment’) and differences in their contribution to the evaluation were highlighted (‘completeness comparison’). The triangulated results were shared with team members and the Trial Steering Committee (TSC) at a face-to-face meeting on 9 November 2016 and further by e-mail communication and discussion to enable ‘feedback’ to be provided, so that points of disagreement could be discussed and any changes in interpretation could be made.

Implementation summary

The trial was due to start recruitment at the beginning of September but it did not start in STHNHSFT until 28 September 2015 and in AUNNHSFT until 14 October 2015; to keep to the 7-month recruitment window, we extended recruitment to the end of April 2016. There were a number of issues with regard to recruitment, data collection and delivery of the interventions, which are discussed below. Follow-up was completed in August 2016.

NHS treatment costs

Excess treatment costs (ETCs) for delivery of the hospital EPR intervention were provided by the two acute hospital trusts: STHNHSFT in Sheffield and AUHNHSFT in Liverpool. ETCs for home EPR were agreed by Sheffield Clinical Commissioning Group (CCG) in Sheffield. In Liverpool ETCs needed to be agreed by three CCGs but they were only agreed by two, which restricted the population that we could recruit from.

Problems with the delivery of the intervention

Problems with recruitment and data collection

Research and service support costs provided funding for research nurse time; this was provided at both sites by a team of research nurses who also worked on other studies, with a dedicated research nurse for the trial. Having a dedicated research nurse who could fully understand and give time to the study was imperative. There were times when the research nurse resource was reduced because of staff sickness and staff not being replaced when leaving their posts, which affected recruitment and data collection, although this cannot be fully appreciated from the non-recruited patient data as these instances may not have been recorded.

Recruitment to the trial

Of 449 patients screened for the trial, 58 were recruited to the trial. Target and actual recruitment figures are shown in Figure 8 and recruitment data by the two centres by month are shown in Table 4. The two centres (Sheffield and Aintree) recruited for approximately 7 months each, giving 14 centre-months of recruitment. It is clear that recruitment started slowly and did not meet the target of 76 but it ended on a rate similar to the target rate. The overall recruitment rate was 4.1 participants per centre per month compared with the target rate of 5.4 participants per centre per month. The overall recruitment rate for Sheffield was 6.4 participants per month and for Aintree it was 1.8 participants per month.

FIGURE 8.

Recruitment graph.

In the last 3 months of recruitment, the target recruitment rate was exceeded, although it should be noted that the majority of participants were recruited from the Sheffield centre. The recruitment rate in this period for Sheffield was 11.6 participants per month and for Aintree was 2.3 participants per month.

As discussed in Implementation of the interventions and the trial, in the first few months of recruitment the TMG realised that the LOS exclusion criterion was the main factor affecting recruitment so the LOS criterion was reduced from 5 days to 48 hours (implemented at sites in November 2015). This did improve recruitment at both sites (recruitment rate from December 2015 to April 2016 for Sheffield was 8.4 participants per month and for Aintree was 2.6 participants per month) but was still an issue and so this criterion was removed completely (implemented at sites in January 2016), which improved recruitment in Sheffield.

Figure 9 presents the participant flow for the trial, with the follow-up figures presented overall and for the 6MWD outcome (at relevant time points). Further details of questionnaire completion for all participant-reported measures can be found in Table 10.

FIGURE 9.

Participant flow. a, n = the number of participants with some data collected at this time point. Numbers of participants completing the 6MWD outcome at associated time points are also provided.

Table 5 provides the reasons for non-recruitment to the trial, which are also categorised according to Kanarek et al. 93 The table shows that the eligibility criteria were the main barrier to recruitment, with 233 patients being ineligible; 106 eligible patients did not want to take part in the study. As mentioned previously, the criterion relating to LOS was removed during the trial, but no other criteria could be removed as they related to patient safety with regard to the interventions. Outside the eligibility criteria, the most common reason for non-participation was patient choice relating to feeling too ill to take part.

Protocol non-compliance

There were 11 cases of non-compliance reported in the trial related to trial documentation, two related to the recruitment window, five related to data collection and one related to eligibility. Non-compliance related to recruitment documentation included instances of using the incorrect consent form or information sheet during consent, although all participants were fully informed of the trial prior to participation. In relation to the recruitment window, one person was recruited outside the 48-hour window and one could not be randomised because of passing the recruitment window. Data collection non-compliance related to follow-up being completed over the telephone and an investigator collecting an outcome (prior to amending the protocol to allow this). Eligibility non-compliance related to an interview participant who did not receive any intervention. None of the reported cases of non-compliance led to a participant being withdrawn. Six cases were reported in Aintree (0.5 per participant) and 13 were reported in Sheffield (0.3 per participant). Cases of non-compliance related to the activity diaries and treatment delivery are reported in Intervention non-compliance as these are all feasibility outcomes.

In addition to the reports of non-compliance, instances of unblinding of the outcome assessors were recorded. In total, there were eight instances in which research nurses were unblinded, relating to seven of the 13 participants recruited in Aintree and one of the 45 participants recruited in Sheffield. Unblinding occurred because of recording in patient notes (n = 2), patient reports (n = 2), a lack of blinded staff (n = 1), the presence of the bike on the ward (n = 1) and a database error (n = 2).

Losses and exclusions after randomisation

Table 6 presents the number of participants who withdrew from the trial. The main reason for withdrawal of participant consent was because the participant felt too ill to continue (n = 7); this was followed by no longer having time (n = 3), being unwilling to complete the follow-up (n = 3), being unhappy with intervention allocation (n = 2), because of the presence of comorbidities (n = 1) and because of personal/family issues (n = 1), with one further participant withdrawing with no reason being given. More than one reason may have been given by participants. One participant was withdrawn by the investigator after receiving a new diagnosis of lung cancer.

Number of missing values/incomplete cases

Delivery and receipt of the interventions

Delivery of the intervention is detailed for each intervention separately in the following sections; optimisation of the interventions is presented in Chapter 4. The expected number of sessions was calculated for each participant allocated to the interventions and session attendance is detailed along with reasons for sessions not taking place.

Feasibility of recruiting participating centres

The chief investigator discussed the proposal for a full trial with a number of respiratory consultants in the UK, with eight centres identified as being willing to take part in a large-scale RCT.

Decision on the primary end point and sample size for a full-scale trial of early pulmonary rehabilitation in patients following acute exacerbation of chronic obstructive pulmonary disease

The candidate clinical primary outcome was 6MWD at 90 days. However, this measure had a low completion rate in the pilot trial, with the outcome being available for only 21 out of 58 (36%) participants. In hindsight, COPD readmission or exacerbation may have been an easier outcome to collect as it could have been taken from patient notes.

Readmission was a more reliable measure than exacerbation as the former was based on hospital data whereas the latter was based on patient self-reported events. COPD readmission was used as a continuous variable (mean number of COPD readmissions) rather than a categorical variable (i.e. proportion of participants having at least one COPD readmission during follow-up) to reduce the sample size estimation. It was unclear how to use time to readmission to define the period at risk (i.e. how to count days of hospitalisation) and whether to count only the first or all readmissions. Moreover, much of the literature on pilot trials is focused on continuous and categorical outcomes154 rather than incidence and time-to-event outcomes.

Overall, there were 34 readmissions in 48 patients within the 90-day follow-up period, giving a mean of 0.7 readmissions with a SD of 1.2 (see Table 24). To compute the sample size for a full-scale trial, the SD of the mean number of readmissions was used from the pilot trial. However, it should be noted that there is a considerable degree of uncertainty in this estimate of variability because:

• the outcome is likely to be skewed, hence SD may not be the best statistic for summarising dispersion

• the estimate is based on 48 participants.

The upper 95% confidence limit of the SD is 1.5. We decided to use 1.2 as the likely SD. Assuming a variability (SD) of 1.2, Table 14 provides the likely sample size required based on various target differences in the mean number of readmissions between groups, assuming a two-sided significance level of 5% and 90% power.

Key pilot trial outcomes

Tables 15–17 present the data for the 6WMD, EQ-5D-5L and activity monitor outcomes as these are considered the key outcomes of the pilot trial. The numbers of participants included in the 6MWD analysis are lower than anticipated, with only 21 individuals completing the 6MWD at 90 days. Because of the small numbers reported in this pilot trial, the CIs are large and we cannot draw conclusions. The results for the 6MWD outcome without the interaction term are reported in Appendix 4.

Because of issues with the timing of discharge and patient choice, it was not always possible to ensure that participants left hospital with an activity monitor. Eleven participants did not receive an activity monitor because of the timing of discharge, three participants did not want to wear the activity monitor, four participants withdrew before being given an activity monitor and one activity monitor was lost. Of the participants who received an activity monitor (n = 39), 21 (54%) provided 10 hours of data for 5 days, which were used in the analysis.

Other clinical outcomes

The LCADL, CAT and MRC Dyspnoea Scale scores are reported in Appendices 5–7 respectively.

Attendance at group pulmonary rehabilitation (usual care)

Table 18 shows the numbers of participants who had attended group PR or an alternative to PR at the 90-day follow-up point. The mean number of sessions of group PR attended for these individuals was 2.8 (SD 2.35), with five people attending the minimum of one session and one person attending eight sessions. The reasons for not attending group PR and for stopping group PR are shown in Table 19; the main reason for non-attendance at group PR (when a reason was identified) was that it was scheduled to start outside the 90-day trial period.

Baseline-adjusted analyses

Similar results were found for baseline-adjusted analyses of CAT (30 and 90 days) and LCADL (30 and 90 days) scores. These results are provided in Appendix 8.

Sensitivity analyses

Similar results were observed in the sensitivity analyses using no interaction, a time effect and multiple imputation. The results of the sensitivity analyses are provided in Appendix 9.

It was not possible to calculate a coefficient for Aintree (hospital: standard of care vs. experimental) when evaluating the centre (analyses constrained in each centre) and the interaction term because of the low number of participants (five participants overall).

Because of the low number of participants and high number of variables included, sensitivity analysis of baseline imbalances was not conducted. For instance, analyses could not be run using Stata version 14 when the centre effect was included (low number of participants in Aintree); hence, it was not appropriate to add additional variables in the models.

Physiotherapist effect was not analysed as there were too few patients per physiotherapist.

Adverse events at the participant level

Most of the participants randomised experienced at least one AE, with the lowest number of AEs reported in the group allocated to both hospital EPR and home EPR (Table 20).

Adverse events at the event level

Patients admitted to hospital with AECOPD often have significant comorbidities in addition to a high risk of COPD exacerbation and readmission. The AEs observed within the study were predominantly related to the respiratory and cardiovascular systems and would be consistent with those expected for COPD patients more generally (Table 21).

There were no unexpected AEs related to the interventions. In particular, there did not appear to be evidence of significant musculoskeletal events reported in association with the exercise-based intervention arms.

Serious adverse events at the participant level

A lower proportion of patients experiencing at least one SAE was observed in the group receiving both home EPR and hospital EPR than in the other arms (Table 22). No events related to the interventions or deaths were observed. The assessment of expectedness was problematic as readmission with COPD is related to and expected in COPD, but it is thought to be expected in only about one-third of patients; therefore, some individuals completing the assessment of expectedness events as unexpected.

Serious adverse events at the event level

The majority of SAEs were respiratory in nature and represented worsening or exacerbation of the patients’ underlying COPD (Table 23). The rate of subsequent readmission was similar to that expected for this cohort, based on their PEARL prediction score (see Post hoc analysis).

Despite the high incidence of reported exacerbations in the home EPR group, the readmission rate was subsequently lower, suggesting a combination of over-reporting and perhaps more effective intervention of other health-care professionals (e.g. GP calls).

Chronic obstructive pulmonary disease-related adverse events (including readmissions)

Chronic obstructive pulmonary disease readmission

Proportions

The proportions of participants experiencing at least one COPD readmission are presented by randomised group in Table 24. Forty-eight participants are included in this table: 44 participants who completed the study, two participants who withdrew after 90 days from randomisation and two participants who experienced the event before withdrawal. A lower proportion of events was observed in the group undergoing home EPR. However, care should be used when interpreting these results because the:

• frequency of events is low and a positive result could have been seen by chance

• study was not designed to detect a minimum clinically important difference for this outcome

• choice of the denominator (and therefore how proportions are calculated) could be disputed.

TABLE 24 - Proportions of patients experiencing at least one COPD readmission and number of COPD readmissions by randomised group

Statistic	Overall	Usual care	Home EPR	Hospital EPR	Hospital and home EPR	Treatment 1 effect: hospital EPR: usual care vs. experimental	Treatment 2 effect: home EPR: usual care vs. experimental	Interaction term
Completers, n	48	10	15	12	11	1.044 (95% CI 0.201 to 5.426)	0.385 (95% CI 0.064 to 2.296)	1.006 (95% CI 0.075 to 13.571)
Patients with events, n (%)	18 (38)	5 (50)	4 (27)	6 (50)	3 (27)
Patients with no events, n (%)	30 (63)	5 (50)	11 (73)	6 (50)	8 (73)
Readmissions, n	34	10	10	9	5
Mean (SD) number of readmissions per participant	0.7 (1.20)	1.0 (1.33)	0.7 (1.59)	0.8 (0.87)	0.5 (0.82)

Mean number of events per participant

The mean number of COPD readmissions per participant is reported in Table 24. This post hoc analysis was run to evaluate the SD of this outcome to see whether or not this would be a suitable primary outcome for the main trial. The SD ranges from 0.80 (both treatments) to 1.6 (home EPR).

Time to event

The Kaplan–Meier curve for the time to first COPD readmission is reported in Figure 10. Overall, 38% of participants experienced at least one readmission related to COPD.

FIGURE 10.

Kaplan–Meier plot of time to first readmission. Participants who are not readmitted are censored at day 90.

PEARL prediction scores

The mean (SD) PEARL prediction score was 3.0 (2.2) whereas the median [interquartile range (IQR)] score was 3 (1–5). These statistics were calculated for 48 participants who completed the study, underwent a readmission before withdrawal or withdrew after 90 days from randomisation. An increase in PEARL prediction score was associated with an increase in readmission (OR 1.57, 94% CI 1.17 to 2.11).

Chronic obstructive pulmonary disease readmission bed-days

This analysis included 48 participants (i.e. the same number of participants who experienced a COPD readmission). The median (IQR) LOS in days was 1.5 (0.0–15.0) for the usual care group, 0.0 (0.0–1.0) for the home EPR group, 0.5 (0.0–7.5) for the hospital EPR group and 0.0 (0.0–5.0) for the group receiving both interventions. The SD ranged from 6.0 (both interventions) to 13.6 (hospital EPR). A similar LOS was observed across the arms.

Summary of hospital early pulmonary rehabilitation

Resistance training during AECOPD has been shown to be effective at counteracting the catabolic effects of myostatin, improving quadriceps strength. 56 During each intervention session physiotherapists were able to increase or decrease the load based on patient symptoms. Breathlessness was measured during and after the intervention using the Borg score and participants were asked to rate the difficulty of the intervention on a scale of 1 (very easy) to 5 (very difficult) after each session. In 13 of the 131 sessions (9.9%) the load was considered to be set too low for the participant; although the load was set at 80% of the 2RM, participants reported the intervention to be very easy (difficulty rated at < 2) and had very little increase in breathlessness (Borg score of < 2). In 12 of the 131 sessions (9.2%) the equipment used was felt to be a limiting factor. The MOTOmed viva 2 cycle ergometer had a maximum resistance of 20 kg and some individuals were able to reach this maximal level. In total, 106 of the 131 sessions (80.9%) were considered to be optimised.

Summary of home early pulmonary rehabilitation

Problems associated with the Borg 0–10 breathlessness score and the lack of use of the Borg RPE score led to lower than expected optimisation scores (see Chapter 8, Evidence of feasibility, Intervention). Only 5 out of 18 people achieved an optimisation score of ≥ 50%; this is partly because of missing data. The results were shared and discussed with staff delivering the intervention and the results were reported by these staff to be predominantly related to inadequacies in study documentation and also a time lapse between training in the study protocol and delivery of the intervention.

In contrast to the optimisation data, adherence data were more positive: 69 sessions were attended out of a possible 96 and 64 of these sessions were fully completed, as recorded on the trial data entry forms; however, because of issues with the return of paper diaries it could be shown only that 58 of the 96 sessions included at least six of the core exercises. Recurrent exacerbation was the core reason for non-completion of the sessions. Only 10 out of 24 participants received all four sessions. The reasons for non-attendance are shown in Table 12.

Impairment and activity limitation

Participants in the study had difficulty moving around (d455) because of their respiration functions (b440) and exercise tolerance functions (b455):

Well yeah, yeah, can’t do it [exercise] now as much as I’d liked to do obviously with COPD but I do what I can and that’s it.

P1/092

Yeah, but it’s breathlessness that really . . . that stops you from doing anything more active?

Yes it does.

P1/107

This had an effect on carrying out their daily routine (d230) and doing housework (d640):

I was hoovering before I could go through it no problem but now I just can’t I’ve gotta stop and start, stop and start all the time.

P1/101

In general, participants said that walking (d450) was their way of getting exercise, although some participants struggled to walk much because of their breathlessness:

Well, I mean, I can’t walk a long way but I like to walk generally [mm] but every so often I have to stop you know what I mean [yeah] but I’m just across the road I’ve got fields and what have you, I usually go across there [yeah] you know.

P1/178

In addition to the functions in the brief core set, participants clearly had motivation issues related to energy and drive functions (b130) and some of the participants mentioned comorbidities that also limited their activity, for example pain in other body parts (b2801) and their weight (b530).

Er no energy, I’m always drained [yeah] so it’s always a struggle. Even the warming up exercises it takes it out of me [right] so, yeah.

P2/092

I’ve got arthritis in both me knees, so I don’t go very far.

P1/128

Participation restriction

Participants were often frustrated at not being able to be as active as they used to be. Some stated that they became breathless doing housework (d640) and participants often struggled on the stairs, causing some to have moved their bedroom downstairs. In Sheffield the majority of participants stated that the hills in the city made it very difficult to walk around (d450) and could affect their breathing (b440):

because I can walk onto end of road but then it’s an hill [yeah] I know it’s, it’s only like 10, 10 yards up and then I walk on flat again and then walk down same again 10 yards back to my house but that going up 10 yards then coming down other end 10 yards it, that kills me.

P1/098

Personal factors

Individuals found that their breathing was the main factor limiting their daily activity and in general they lacked the confidence to undertake activity because of their breathlessness (b440):

Yeah you don’t know what to don’t know what to do, you know when you have trouble breathing [yeah] you don’t know if you should take it easy or try and motivate yourself a bit more to be quite honest.

P1/223

Environmental factors

A number of participants said that they used oxygen to aid their breathing and some participants said that carrying oxygen limited what they could do and where they could go:

But I can’t you know with the oxygen and I’ve got my stuff upstairs, how can we go? So, I mean we don’t drive so I’d be you know it’s a bit awkward to try and get anywhere.

With the oxygen?

Yeah, because I have to have that beep out machine on at night and I’ve got my oxygen generator upstairs and this is my mobile one. So really we’re a bit lumbered.

P1/220

A few participants mentioned that they keep antibiotics for self-administration (e110) to control their COPD and avoid exacerbations:

Yeah because I’ve only been into the hospital once in what about year I think [mmhm] whereas really because I self-administer antibiotics and stuff for myself, I have them here, so if I know that I need them I’ll take them.

P1/161

Only one person mentioned products for daily living, having had a bathroom put in downstairs (e115), with no participants mentioning technology. Eighteen of the participants interviewed mentioned that family (e310) and friends (e320) help them with shopping or housework or act as motivators to carry out activity.

Acceptability of hospital early pulmonary rehabilitation

Of the participants who attended any hospital EPR sessions, all of them found hospital EPR to be more or less acceptable (Table 29), with three participants finding it very highly acceptable and three finding it highly acceptable:

In the hospital I was doing it three times a day [yeah] you do it with your arms and your legs, and you sit in a chair and do it [yeah] I says to her, I wish I could bring that home.

P1/059

Some commented on the fact that carrying out the exercises made them feel more capable and more confident to go home following their AECOPD:

How would you feel about doing exercises in the hospital?

Yes, yeah . . . because of being part of the recovery programme . . . they’d indicate also that I am well enough to actually go home erm, and I cater for myself alright.

P1/124

One hospital EPR participant declined all of the sessions because of feeling too unwell and another hospital EPR participant completed only one of four available sessions because of not finding the intervention acceptable. One individual did not have the chance to complete the sessions because of being discharged but indicated that hospital EPR would be highly acceptable. One participant received no sessions because of an administration error and so did not discuss acceptability. Three people appeared to show some acceptability of hospital EPR and this was because they felt unwell and thought it may have been too early to carry out the exercises:

It was a bit strenuous, yeah, yeah . . . Um, a little bit, er, a little bit heavy. You know I mean I done it but I found it a little bit, you know . . . I just, I just couldn’t seem to getting meself better.

P2/233

Participants who were not allocated to hospital EPR (n = 15) were mixed in their opinions: eight felt that it would be acceptable, one said that they felt that their bad knees would prohibit engagement and the remaining six thought that it was too early to begin exercising following an exacerbation, although one thought that they might have still tried the exercises:

Oh yeah if you go in with an exacerbation you just wanna get on the meds and feel a bit better before you do anything.

P2/092

Acceptability of home early pulmonary rehabilitation

Those who received home EPR generally found the intervention to be acceptable (Table 30), with four participants expressing a high level of acceptability and five indicating a high level of acceptability:

But the physio I had at home when they came to me after I’d been in hospital [yeah] for the erm, that was fantastic!

P1/161

Some participants found it a little too soon after leaving hospital to receive the intervention and others did not attend all of their sessions. One participant found home EPR to be unacceptable and had no sessions; another indicated little or no acceptability, declining the final session because of poor health and the perceived burden of health-care visitors:

Yeah, they were alright be nah I can’t be bothered.

P1/150

When participants indicated some acceptability they found the sessions to be acceptable when they completed them but found them difficult, felt tired after the sessions or felt that the intervention was provided too soon following discharge:

Well, I thought they might have been a bit early really because I was, I think she came the day after I came out of hospital, and I really wasn’t feeling up to it much for the first ones.

P2/176

Interview participants who were not allocated to home EPR (n = 11) said that they would find it acceptable, with one exception, who preferred not to wait in for visitors.

Participant views of the interventions: capability, opportunity, motivation and behaviour analysis

Data on participant capability, opportunity and motivation to attend group PR (usual care) are presented in Table 31; the same data are presented in Table 32 in terms of the study interventions. Further data are available in Appendix 11.

Participants were less likely to attend PR (Table 31) if they (1) did not believe that the exercises were beneficial (reflective motivation; beliefs about consequences); (2) had low self-efficacy or did not feel that they had the capability to exercise (reflective motivation; beliefs about capabilities); (3) did not enjoy exercise (automatic; reinforcement); or (4) were not usually active (psychological capability; memory, attention and decision processes). Those with no access to transport (physical opportunity) were also less likely to attend group PR (see Appendix 11).

Although physical capability is a barrier to attending group PR, participants receiving EPR generally had the skills to perform the intervention (see Tables 31 and 32). Social opportunity was a greater barrier to group PR than to the EPR intervention and tended to be related to reluctance to exercise in a group. Physical opportunity to receive the EPR intervention was also much greater than for attendance at group PR. Those people who did not have the opportunity to receive hospital EPR (see Table 32) were discharged from hospital before treatment could be commenced or completed.

Intervention dimension categories are mapped to the COM-B theoretical framework and TDF using case study vignettes (see Case study vignettes). Hospital EPR and home EPR overcame a number of the barriers to group PR reported by participants who were unable or unwilling to attend a group programme. However, limitations of hospital EPR and home EPR were also evident.

The following sections describe the barriers and enablers associated with the intervention functions of EPR in relation to the COM-B theoretical framework.

Approach

In general, participants found that the way that they were approached was acceptable and did not mind being approached when they were feeling acutely unwell. However, one of the non-participants interviewed did not take part because of how they felt in hospital and four of the participants did think that it was a bit too early in their admission to be approached, even though they still took part in the trial:

Do you remember how you found out? How you were approached?

No, I don’t [no] I can’t remember at all.

Were you quite poorly at the time?

Yeah.

So I was gonna ask questions about the information provided but you, do you, you don’t remember?

No, no.

No, so do you think it was probably just that you felt a bit poorly at the time that you . . .

. . . I think it must have been.

P1/157

Related to the approach was the information provided when introducing participants to the trial; 12 of the interviewed participants indicated that there was a lot of information to take in when they were not feeling well:

Just that when you’re in hospital, you’ve not been very well, you’ve not been able to breathe and your mind’s all over the place to be honest and I don’t think you take everything in at the time.

No I think and those information sheets we give you are quite long aren’t they.

Yeah, yeah. I told them and then afterwards I thought oh crikey did I fill that question in right.

P1/223

Randomisation

Of the 14 participants who discussed randomisation, half were happy with this part of the trial, being happy with their allocation, not having a preference or showing an understanding of the process. Some individuals seemed not to fully understand randomisation but were happy with being in the trial and with the group that they were allocated to or they had no preference. Two people expressed their disappointment when they were not allocated to the interventions, with one person withdrawing from the trial during the interview for this reason and one person expressing that they would have liked to use the equipment. One person was not happy with being allocated to both of the interventions, but still completed them both:

Yeah and how did you feel that you could have got the hospital EPR, you could have got in-home and that was sort of done randomly did that worry you at all?

I was disappointed [yeah] yeah disappointed that I didn’t get any of the equipment.

P1/161

Six-minute walk distance

Most of the participants interviewed did not have any issues with carrying out the walk test but two did not complete it while in hospital and some missed their follow-up appointments so could not complete the test at these time points. A few participants were worried prior to the walk test but felt that it was fine to complete and two participants felt good after doing the test as they felt more capable and thought that the test was good at showing them that they could do more than they thought:

Then I thought oh god will I be able to walk properly [yeah] you panic, you know, you’re not sure what you can do. I think I surprised myself how I could walk [yeah] once I got going. I don’t know whether it was because somebody else was there and you got a bit more confidence, I don’t know but once I started walking I surprised myself I think.

So do you initially, you didn’t feel very, maybe not very safe but you were happy to do it.

Yeah a bit insecure really I think, nerves.

P1/223

One person expressed that they did not believe that the test was a good measure, especially for individuals who need to walk on hills most of the time:

How you felt about the walk test, how did you find doing that in hospital?

It’s just, I think the, I think that’s about my only quibble, it’s fine doing it [yeah] but unfortunately it’s not a representative of walking in general, particularly if you live in [city name].

No that is true [laughs].

It’s all on the flat [it’s all flat] and I can walk on the flat erm, particularly on a hospital ward or a corridor [mmm] I can walk a bit and for quite a long time no problem at all [yeah]. If I walk out of my house, by the time I’ve gone 15 metres I’m out of breath if I go in one direction and I’m out of breath if I go in the other direction because it’s on a hill.

P1/226

Activity monitor

Of those who received the activity monitor only one person expressed that it bothered them, with most participants not noticing it very much:

How have you found it wearing the monitor?

Err [pause] a little bit uncomfy [yeah?] but it were only for a week so I didn’t mind.

Yeah? So did you, have you worn it quite a lot of the time, you haven’t taken it off much?

Er, just for a shower [yeah?] or a bath yeah.

Oh brilliant, so yeah I had to wear one for a week before we started it off [laughs].

I think it’s that Velcro.

Yeah the strap.

That snaps you sometimes.

P1/223

However, because of the timing of discharge, not all participants received the activity monitor (see Chapter 3, Key pilot trial outcomes).

Questionnaires

Some participants thought that the questionnaires were too long, especially when they were feeling acutely ill, but all generally saw the point of the questions. A few individuals indicated a little confusion over the questions, particularly the Borg scale (clinical measure of effort).

Burden

Three individuals expressed that the follow-up visits, intervention visits or questionnaires were burdensome as they had a number of other health visitors or people asking them questions while in hospital:

And what did you think about the number of questions you were asked?

There was a lot. A lot, yeah and I didn’t think there was all these questions and that because there were other people that was in the ward that was asking me questions as well and then I went down there it was all questions, just a lot, the whole leaflet was more than I expected to be honest with you.

Do you think it was too much?

A little bit because I think because I was getting better as well I didn’t want to be annoyed with it as well.

P1/101

Making sense of the intervention

Staff distinguished the hospital EPR intervention from usual care by its novel equipment, strength training component and greater protocolisation (‘differentiation’). It was also delivered earlier, even than in existing ESD schemes. The home EPR intervention was similar to current practice in Sheffield, but was delivered earlier – within 72 hours compared with 2 weeks; staff said that it had the potential to include sicker people who had been in hospital for longer. At Aintree the in-home service was completely novel. For both experimental interventions, higher-grade staff [Agenda for Change (AfC) bands 5–7), normally confined to assessment, took on the supervision and monitoring of exercise, typically delegated to assistants (AfC band 3 or 4). This meant better continuity of care. Physiotherapists could describe the aims of the interventions (‘individual specification’). They shared a sense of those aims with their colleagues if not always the participants (‘communal specification’), for whom they framed information, to increase engagement:

I feel some of the patients needed that additional information from us . . . they’d had all the paperwork . . . but there were some patients that needed that reinforcing during our visits.

Physiotherapist 103, AfC band 6, home EPR

The purpose is to rehabilitate them quicker and get them on a pathway of self-management. That’s what the professionals and patients see. But I think the professionals also see the bigger picture, which is, if we get these patients on this pathway of exercise and education and self-management that ultimately it’s going to lead to fewer exacerbations and hospital admissions . . . it’s cost saving. For the patient the reason is quality of life, they don’t wanna be continually having antibiotics and they don’t want numerous hospital admissions. So I think the ultimate goal is the same but for slightly different reasons.

Physiotherapist 104, AfC band 7, home EPR

I don’t think the patients fully understood what was expected of them . . . I think as soon as you mention the word exercise, patients don’t want to know.

Physiotherapist 107, AfC band 6, hospital EPR

Most physiotherapists constructed potential value for the hospital EPR intervention (‘internalisation’) in terms of maintaining strength and function without impacting on the cardiovascular system, thereby improving readmission and survival rates. They also believed that the intervention improved knowledge of the disease process, self-management and the benefits of exercise, instilling confidence to exercise. One dissenting voice doubted that the hospital EPR intervention would deliver functional gains, but acknowledged that there was a strong link between exercising and chest clearance (see Operational work around the intervention, ‘skill-set workability’). Further benefits were identified for home EPR, including improvements in general post-discharge support and case management. Professional satisfaction arose from the resulting improvements in communication between the in-hospital teams and the community teams.

Relational work around the intervention

Agenda for Change band 7 physiotherapists drove the interventions forward (‘initiation’). As well as engaging other physiotherapists and assistants to implement the interventions (‘enrolment’), they persuaded nurses to identify patients who could benefit. Many patients felt too ill or insufficiently motivated to engage with hospital EPR, unless approached by a consultant rather than an allied health professional. Resistance to the hospital EPR intervention focused around competing priorities and insufficient resources; there were concerns about both hospital EPR and home EPR interventions around the selection of sufficiently well patients (a problem of ‘legitimation’), although senior physiotherapists were less concerned than junior staff:

The sooner we can start them the better I think really . . . I think it’s beneficial because I think it helps alleviate some of their fears as well.

Physiotherapist 108, AfC band 7, hospital EPR and home EPR

Physiotherapists continued to support the interventions (‘activation’) but stressed that ‘staffing was really, really tight’ (physiotherapist 102, AfC band 6, home EPR). Patients were often discharged before hospital EPR was complete or were unavailable concurrently with the physiotherapists or bike because of family visits/clinical consultations. Patients sometimes declined in-home sessions because of competing priorities but, on the whole, remained well engaged:

That’s reflected in the diaries as well . . . they had taken the time to do the exercises in their own time, when I wasn’t there supervising them and actually document that.

Physiotherapist 108, AfC band 7, referring to home EPR

Operational work around the intervention

Staff translated or adapted instructions from manuals and training to enact intervention procedures (‘interactional workability’). For the in-hospital intervention, there was mixed satisfaction with the training and instructions; many described the bike as complicated and there being a learning curve. Some were unclear how to gauge whether or not patients could tolerate higher levels of exercise:

So the protocol that I was taught and work was to set your two repetition max and then figure out your 80% load and that is the working load; however, for the more able people they would be able to manage . . . the two repetition max for the actual intervention, and [investigator’s name] encouraged us to do that whereas, myself, I was unsure so I stuck to the protocol to make sure that nothing changed.

Physiotherapist 101, AfC band 5, hospital EPR

Logistically, hospital EPR was difficult to implement because of the competing demands on the patients’ time, the availability of the bike and positioning the equipment on the ward:

You’ve got to plan it around lunches and breakfasts and things . . . if you’ve got two patients in 1 day it’s trying to fit them around.

Physiotherapy assistant 105, AfC band 3, hospital EPR

Just things like the charging unit and where the plug sockets were . . . you were moving furniture out of the way or you had cables kind of obstructing the patient . . . you might get some patients really extending and leaning forward in the chair then you might get others that are laid back and it’s in their face so they aren’t getting a full extension . . . there wasn’t any [guidance on] how far it should be from the patient . . . it’s quite pivotal for what muscles you’re gonna be engaging.

Physiotherapy assistant 106, AfC band 3, hospital EPR

Physiotherapists reported knowledge work to maintain confidence in the intervention and each other (‘relational integration’). In-hospital staff expressed concern about the face validity of the Borg score to gauge breathlessness, before, during and after bike use. They also questioned aspects of the intervention protocol, such as the load ceiling being too low for some patients and the absence of a cool-down component:

I think that understanding of Borg and peak exertion that was like the epitome for me . . . getting them to recall back to their most breathlessness during that session, was quite difficult.

Physiotherapy assistant 106, AfC band 3, hospital EPR

At Aintree, where home EPR delivery was novel, the physiotherapists were wary of rehabilitating acutely unwell patients without a crash team being available. The Sheffield in-home team had to communicate with each other more, to ensure per-protocol delivery and intervention optimisation. They had to repeat tests, required to tailor the intervention, the results of which were unavailable from the hospital team. Although communication between the in-hospital and the community teams could be improved, community teams appreciated the increased interaction generated by the intervention. In-hospital and community teams acknowledged that respiratory consultants had to be involved to sustain implementation of either intervention:

It was good for us to be able to link up with other disciplines like the GPs and the district nurses and the consultants and the physios in the hospital . . . the link with the people at the hospital was absolutely paramount . . . we can’t do it without them.

Physiotherapist 104, AfC band 7, home EPR

Physiotherapists and their assistants were generally confident in their expertise while enacting it (‘skill-set workability’). Some comments (see Approach) revealed a poor understanding of the background theory of how exercise affects physiology immediately after an exacerbation, which could be addressed in training. Although hospital EPR could be passed on to physiotherapy assistants after the first session, continuation of home EPR could not be delegated:

I did have some tricky ones . . . there was only really one chap who I might have been able to pass on . . . on my fourth visit he was actually quite poorly and I had to get the GP to come and see him . . . it probably did need a qualified member of staff on nearly all of the visits.

Physiotherapist 102, AfC band 6, home EPR

‘Contextual integration’ of the new interventions was incomplete. Although NHS trusts were enthusiastic to participate in the research, ETCs were insufficient to create new posts or backfill old ones for the purposes of delivering the study interventions. Delivery and documentation of hospital EPR was constrained by staffing levels, the numbers trained to deliver/document the intervention and regular clinical work. Fitting the three short daily sessions around meal times and consultations was a challenge. When the staffing level was good, the ward caseload could be covered while hospital EPR was delivered. However, patients were spread out over the hospital and had to be tracked, from assessment units to respiratory or outline wards, meaning that the single available bike had to be constantly transported around the hospital. Continuity of delivery at weekends was ‘difficult’ and reliant on voluntary additional hours. Interviewees thought that hospital EPR could be integrated into usual care if a full-scale trial demonstrated efficacy.

For home EPR, staffing levels and the timescales for participant contact, post discharge, were a constraint on the ability of staff to integrate the intervention; the service was fragile in the face of leave or sickness. Interviewees believed that the service could be sustained or scaled up only if commissioners, consultant physicians, inpatient physiotherapists, therapy assistants and nurses could be engaged and extra physiotherapist capacity could be funded. To plan and deliver the service, extra hours were worked. The four home EPR visits required qualified physiotherapists, whereas Sheffield’s ESD service, which aims to see AECOPD patients within 2 weeks of discharge, delegates most of that work to physiotherapy assistants. Most of the Aintree districts did not have a community physiotherapy team. As procedures allowed little flexibility or patient choice in the timing of visits, visits were sometimes declined. Routine client visits had to be cancelled or moved to accommodate the experimental service, causing an increase in waiting times, from a typical 6–8 weeks.

At the time it was very stressful because we hadn’t got the staffing for it and it was quite labour intensive compared to what we normally do . . . It was the routine home rehabs, that haven’t had a recent hospital admission, but generally can’t get to a group because they’re housebound, or they’ve got transport issues . . . those were the ones that were neglected. And we don’t have targets for when we should see those in. And we’re still trying to catch up the waiting list for those at the moment.

Physiotherapist 104, AfC band 7, home EPR

The difficulties are delivering it at weekends and bank holidays. We obviously have a completely different set of priorities at a weekend . . . it’s not a true 7-day service . . . but certainly during the week it definitely fits and is deliverable.

Physiotherapist 109, AfC band 7, hospital EPR

Appraisal of the intervention

The extent to which physiotherapists were able to access the results of the interventions (‘systematisation’) was limited. Some believed that they could observe minor differences in function as a result of the interventions. However, evidencing gains in strength or mobility or a reduction/delay in readmissions would require a collaborative routine data collection effort by consultants, nursing staff and therapists:

Certainly with some of these chronic patients, uhm, yeah there was a strong link between exercising and chest clearance.

Physiotherapist 107, AfC band 6, hospital EPR

We’re very bad at following patients up . . . we don’t do any follow-up; we need to look at 6 months post rehab, are they still engaging with physical activity. And I’m sure most of them aren’t, because we see them come back again and again through rehab, and we don’t really want that . . . So you know, long-term stats and data is so important, and we certainly don’t have it at the moment.

Physiotherapist 104, AfC band 7, home EPR

There was little opportunity for ‘communal appraisal’; individually, resourcing aside (see previous section), physiotherapists assessed the effects of the rehabilitation programmes on their workplace in positive terms (‘individual appraisal’). They reported that many patients enjoyed the higher intensity and frequency of in-hospital rehabilitation; one assistant reported that patients thanked her for pushing them to exercise, because they felt the benefits afterwards. On the other hand, patients in hospital who were more unwell and those with longer, more complicated hospital stays or comorbidities at home could find PR too much, depending on how their chest was feeling:

I loved the fact that I had more contact . . . we’d normally see them for one or two visits and then pass them on and we’d never find out how they got on or whether they ended up coming to group rehab . . . So the continuity was really nice . . . four sessions over 2 weeks, you could really see them improving. Or if they weren’t improving, because you’d see them regularly, you could do something to help them, they weren’t on their own, you could get the GP to review them, I could ring [the consultant] which was absolutely invaluable.

Physiotherapist 104, AfC band 7, home EPR

I think a number of patients were either too breathless or too generally unwell to undertake the intervention but I think those that felt well enough, benefitted from it and could see the benefit . . . It’s basically understood that hospital admission is not good for patients’ functional mobility and that that needs to be addressed at some point but I’m not sure hospital is the right place to address it.

Physiotherapist 109, AfC band 7, hospital EPR

Physiotherapists reported having the confidence to make and document adjustments to procedures, when necessary, and proposed modifications to treatment protocols (‘reconfiguration’):

I did any adaptations I did document it and if I was unsure as to whether it would still fit into the protocol I could get in touch with [the trial physiotherapists] and just run that through them. So I didn’t feel concerned about those adaptations that I’d made.

Physiotherapist 103, AfC band 6, home EPR

For the in-hospital intervention this included allowing the use of higher settings on the bike, when patients’ exercise capacity permitted, with an additional aerobic component. For home EPR, physiotherapists suggested more of an emphasis on goal setting, that documentation follow patients from the ward into the community and that the community team carry out their own walk tests rather than relying on the in-hospital team to perform and document functional tests.

Making sense of the trial procedures

Health professionals who were involved in other trials (most were not) could distinguish the trial from others (‘differentiation’):

Quite often we get given trials to do and they’re very prescriptive and they’re very inflexible, to the patients’ detriment . . . what I’m trying to say is, it’s more patient focused, definitely than all of the other trials that I work on . . . some of them are probably more, as you say, dangerous to patients.

Research nurse

Physiotherapists understood the feasibility purpose of the study (‘individual specification’). Collectively, health professionals agreed on that purpose with each other (‘communal specification’) but were careful to frame details to often frail inpatients in ways that would not discourage recruitment overly:

You had to be very clear, it’s not running upstairs, it’s not doing anything that’s beyond what is your capacity. . . . the explanation was crucial.

Research nurse

Physiotherapists believed that, by enabling a large-scale trial, our study could increase understanding of the harm-to-benefit ratio of early rehabilitation58 and whether or not it should be part of the patient pathway (‘internalisation’). Many patients who were approached about the trial did not construct value for it; when they did so, it included benefits for other patients:

Of those that were eligible and said no . . . some of them just can’t see any further than the fact that they can’t breathe.

Research nurse

Very highly valued . . . patients were saying you know, if this is going to help other people then that’s what I want to do it for . . . I think they know how distressing it is for themselves so they obviously don’t want other people to feel like that.

Physiotherapist 102, AfC band 6, home EPR

Relational work around trial procedures

The importance of respiratory consultants and senior physiotherapists to drive the research forward was noted (‘initiation’). With persuasion and appropriate support for participant recruitment from research nurses, physiotherapists were happy to support implementation of the research (‘enrolment’), but this was made easier by contextual knowledge. In particular, involving staff in the medical assessment unit (MAU) helped in the early identification of patients. Comorbidities in those who were eligible, together with the 48-hour recruitment window, prevented higher levels of participant recruitment:

I’m pretty lucky because I came from MAU where a lot of these people were recruited . . . I knew the way that they were working, that we had a new computer system rolled out last year . . . And it saved me a lot of time because [someone with access] was able to look through and see quite quickly who would be eligible . . . in outpatients I had to take biscuits and things, because they’re really, really busy, and you’re asking them to do something that isn’t part of their job . . . there’s no real benefit for them . . . I made quite a lot of small talk.

Research nurse

Health professionals bought into the research, because of interest or belief in the treatment hypothesis and because of the involvement of consultants (‘legitimisation’). Keeping the research nurses and physiotherapists engaged in supporting the research was challenging (‘activation’). Continued engagement with research procedures was problematic for patients, many of whom were too ill to attend further study visits:

And yeah, it required a lot more flexible input than a lot of other studies . . . knowing when people are being discharged, and communicating with the wards.

Research nurse

We did have you know quite a few that wouldn’t turn up for their appointments because, they were re-exacerbating . . . I can understand why they didn’t want to come back when they’re feeling rotten.

Research nurse

Operational work around the trial procedures

Health professionals generally found the training and protocol to be straightforward and delivered the research per protocol (‘interactional workability’). However, ensuring the correct allocation of study interventions and on-schedule assessments was challenging. Assessors were sometimes unblinded by information in hospital notes/GP summaries or encounters with hospital interventionists working with participants. Research nurses had difficulty conducing ultrasonography (rectus femoris muscle cross-sectional area), which was later abandoned, and programming activity monitors because of restrictions in NHS wi-fi access when using project laptops. Almost all physiotherapists lost confidence in what the Borg scores actually represented (‘relational integration’) and research nurses were concerned that they might miss in-hospital AEs, although this should have been picked up in the discharge summaries. In-home physiotherapists had to communicate more, to ensure that participants were seen per protocol. Hand-offs occurred between senior physiotherapists rather than between physiotherapists and assistants, as was normal practice. In Sheffield, staff relied on SystmOne (an electronic patient record for community services) for scheduling visits and determining their content. Getting paperwork to clinicians in multisite trusts was challenging. Health professionals felt that delivery of the trial was appropriate to their role (‘skill-set workability’). In general, the research was adequately supported by host organisations (‘contextual integration’), although local research nurse staffing at one centre was unable to meet the demand for some of the trial because of leave and staff turnover. The fact that the research nurse services were essentially unavailable at weekends impacted on the ability of research nurses to recruit participants and made some data collection components difficult to undertake. Physiotherapists also felt that more capacity was needed and that study implementation was barely robust to staff being on leave. In-home staff, especially, described study involvement as ‘stressful’ and expressed relief at the closure of the trial:

A few weeks ago, I was the only person in the team that could administer the intervention, and I think we had two patients randomised for the exercise trial and I had another ward to run as well.

Physiotherapist 101, AfC band 5, hospital EPR

Having that short window and the four sessions to come within a certain time frame . . . the effect that that has on . . . our overall caseload and our working week.

Physiotherapist 103, AfC band 6, home EPR

The amount of SAEs and the amount of contact you try and get people to come back or to get people to answer the phones for the 7-day follow-up. A lot of it is small but bitty jobs that if you don’t do them it just makes everything a lot harder. And it works quite differently to a lot of our other trials, which are either tied in with clinic visits or are regimented to a particular day.

Research nurse

Appraisal work around the trial procedures

As already noted, those involved with the trial sometimes found the procedures burdensome (‘communal appraisal’), although some also commented ‘that it definitely got easier over time’ (physiotherapist 103, AfC band 6, home EPR). At the same time, the opportunity to be able to carry out a more systematic follow-up of patients was appreciated (‘systematisation’) as was the opportunity that the full-scale trial presented to test the hypothesis and participate in research (‘individual appraisal’):

I think probably for a period it did create some unhappiness and stress within the team [laughs]. But I think, if I’m speaking from my own perspective then I’m quite pleased that I’ve been involved . . . it was quite rewarding actually following the patients, whereas normally I would just be doing that sort of assessment and screening and passing them on.

Physiotherapist 103, AfC band 6, home EPR

Staff proposed changes to the study procedures (‘reconfiguration’), which included simplifying and integrating the AEs form; improving intersite communication/documentation and co-ordination between research nurses and physiotherapists; implementing procedures to identify patients and track them once recruited, including using routine data systems; simplifying the information sheet and adding pictures; removing unsuccessful procedures, such as the use of predicted LOS as an eligibility criterion and the research nurse-led measurement of the rectus femoris muscle cross-sectional area; providing more training in the outcome assessments; recording oxygen use; and providing training to research nurses on the 6MWD test and the circumstances under which participants should not complete it.

Summary of key results from the cost-effectiveness analysis

A summary of the key cost-effectiveness statistics is presented in Table 37. Mean baseline utility values were highest in the trial arm receiving both hospital and home EPR (0.74, SD 0.17), followed by hospital EPR (0.65, SD 0.27), usual care (0.58, SD 0.19) and home EPR (0.44, SD 0.32) (see Table 74), suggesting that those in trial arm receiving both interventions had the better health status at baseline. In contrast, between baseline and 90 days, only the home and usual care trial arms showed a mean increase in utility values (see Table 74). Mean QALYs were lower and mean costs were higher in the usual care trial arm relative to any other intervention trial arm; however, neither costs nor QALYs were statistically significantly different in any of the intervention trial arms relative to usual care. The home intervention produced the lowest mean cost saving at £145 [95% credible interval (CrI): cost saving of £3519 up to an additional cost of £3067] and the trial arm receiving both interventions produced the highest mean cost saving of £1203 (95% CrI: cost saving of £4467 up to an additional cost of £1842). In comparison, the hospital intervention produced the lowest QALY gain of 0.062 (95% CrI: a QALY loss of 0.082 up to a QALY gain of 0.092), but the trial arm receiving both interventions also produced the highest QALY gain of 0.0353 (95% CrI: a QALY loss of 0.042 up to a QALY gain of 0.104). The mean QALY gain produced by being in the trial arm receiving both interventions is equivalent to almost an extra 13 days in perfect health compared with the usual care group. Each of the three interventions therefore dominate usual care (they cost less and are more effective) at the mean point estimate.

However, it is important to note the uncertainty surrounding these cost and QALY point estimates and the small sample size associated with the pilot study used to inform this analysis. For example, the 95% CrIs indicate that for both interventions compared with usual care there is a 95% chance that the intervention could lead to anywhere between a QALY loss of 0.042 and a QALY gain of 0.104 and between a reduction in cost to the NHS of £4467 and an additional cost of £1842 per person. The cost-effectiveness planes, which indicate the difference in costs and QALYs of the home (Figure 11), hospital (Figure 12) and both (Figure 13) interventions, visually show the uncertainty around these point estimates and the dispersion of the ICERs. The CEACs (Figure 14; results are summarised in Table 37) suggest that, compared with usual care, the probabilities of the interventions being cost-effective at a willingness-to-pay (WTP) threshold of £20,000 and £30,000 per QALY gained are 58% and 59%, respectively, for home EPR, 73% and 71%, respectively, for hospital EPR and 87% and 88%, respectively, for both interventions.

FIGURE 11.

Cost-effectiveness plane for the home intervention compared with usual care.

FIGURE 12.

Cost-effectiveness plane for the hospital intervention compared with usual care.

FIGURE 13.

Cost-effectiveness plane for both intervention compared with usual care.

FIGURE 14.

Cost-effectiveness acceptability curves for each of three interventions compared with usual care. Note: this graph demonstrates the probability of cost-effectiveness at a range of decision-maker ceiling WTP values for each of the three interventions compared with usual care.

Interestingly, when the home and hospital interventions are compared against usual care, the home intervention has a higher probability of being more effective than the hospital intervention (62% vs. 56% respectively), but the hospital intervention has a higher probability of being cost saving than the home intervention (73% vs. 53% respectively) (see Table 37).

Expected value of perfect information analysis

The results presented in Table 38 are a selection of the key results from the EVPI analysis. The EVPI analysis compares the decision uncertainty around investing in a particular intervention relative to usual care at a £20,000 WTP per QALY threshold and therefore the decision uncertainty is assessed on an intervention-by-intervention compared with usual care basis.

Based on the cost-effectiveness results, the most interesting EVPI results are associated with the analysis around being allocated to both interventions compared with usual care. In this instance, the EVPI is £123 per person per year, which, if we assume that a hypothetical cohort of 100,000 people would be affected by the intervention, results in an overall EVPI of £12.3M per year (this is the upper-bound value of investing in a future trial comparing both interventions with usual care assuming that the period of time that this decision would be relevant for is 1 year and that the affected population consists of 100,000 people; the overall EVPI increases if the assumed effected population or time period of interest increases, as shown in Table 38). The reason that the EVPI is lower than for the other two interventions is because the probability of this intervention being cost-effective is much higher (there is less uncertainty around the benefit of this intervention) and therefore reducing uncertainty in any of the input parameters for this analysis is less likely to change the outcome than for the other two interventions. The partial EVPI of collecting more QALY information for the both interventions group is zero whereas there is a small (£2 per person) partial EVPI of collecting more QALY information from the usual care group (the relatively higher partial EVPI for the parameters associated with the usual care group suggests that it is the uncertainty around the outcomes in this group that is contributing particularly highly to the decision uncertainty). In contrast, there is more uncertainty around the hospital costs of the both interventions group (£23 per person) than the usual care group (£13 per person).

Sensitivity analysis and key cost and effectiveness drivers

Based on the partial EVPI analysis, there is a suggestion that the QALYs and hospital costs contribute particularly highly to the decision uncertainty. The effect on costs and outcomes can be explored by performing one-way sensitivity analysis by altering the QALYs and hospital costs to the upper and lower bounds of the 95% CIs around the mean point estimates. For example, in the base-case analysis, all interventions dominated usual care based on the mean point estimates, which were used to determine the ICERs; therefore, this one-way sensitivity analysis focused on (1) increasing the mean QALYs to the upper bound of the 95% CI and reducing mean hospital costs to the lower bound of the 95% CI for usual care (making usual care less costly and more effective within the bounds of the 95% CI) and (2) increasing the mean hospital cost to the upper bound of the 95% CIs and reducing the mean QALYs to the lower bound of the 95% CIs for each of the intervention arms (making the interventions more costly and less effective within the bounds of the 95% CIs). All other values were kept equal in the model. The results from this sensitivity analysis are presented in Table 39 for a change in mean hospital costs and Table 40 for a change in mean QALYs.

To summarise the results in Tables 39 and 40, the change in mean hospital costs resulted in the ICERs for the home and hospital interventions being substantially higher than the £30,000 WTP threshold when the comparison group was usual care; however, the both interventions trial arm remained almost potentially cost-effective at a £30,000 WTP threshold relative to usual care, with an ICER of £30,963 when the mean hospital cost associated with using both interventions was increased to £5528. The change in mean QALY values meant that the home and hospital interventions resulted in a mean QALY loss and produced an ICER representing cost savings per QALY forgone (rather than the more common cost per QALY gained associated with ICERs); however, when the mean QALY value for the both interventions trial arm was reduced to a mean value of 0.1403 (from 0.1740), using both interventions still dominated usual care in this sensitivity analysis. These results are indicative of how the mean cost-effectiveness estimates based on ICERs can change as a result of a change in hospital costs and QALY values in the model and also of how the both interventions trial arm remained relatively or almost cost-effective at the mean point estimates under these scenarios compared with the other interventions assessed relative to usual care.

Summary of the triangulation of data for hospital early pulmonary rehabilitation

Figure 15 provides a summary of the key points relating to the implementation of hospital EPR. Participants were capable and willing to undertake the exercises while in hospital, but it was difficult to deliver three sessions on 5 consecutive days, particularly during the evenings and at weekends, because of staff availability. Discharge was the main barrier to delivery of the hospital EPR intervention, as patients were often discharged within 5 days of recruitment. Some of the participants found the intervention too easy, so for these patients the exercises were suboptimal, although in general the intervention was optimised for individual patients. Optimisation may have been improved by providing enhanced training in relation to the cycle ergometer, although the workload could not be increased for some patients because of the limitations of the equipment.

FIGURE 15.

Hospital EPR summary of triangulation.

Summary of the triangulation of data for home early pulmonary rehabilitation

Figure 16 summaries the triangulation of data relating to the implementation of home EPR in this trial. Although the NHS trusts were engaged, because of changes in funding since the agreement of ETCs, this support did not translate into increased capacity to deliver home EPR in Sheffield. This in turn affected the delivery of the intervention (by pausing recruitment) and the delivery of existing services.

FIGURE 16.

Home EPR summary of triangulation.

As with hospital EPR, participants were capable and willing to take part in the intervention sessions. The participants liked the flexibility and convenience of the sessions taking place in the home at a time that suited them and 90% of the sessions that were started were completed. Both participants and physiotherapists found the activity diaries useful and participants appreciated the case management by physiotherapists, although the quantitative data suggest that case management was not successful.

The home EPR sessions were not optimised according to the guidelines for PR in stable patients, although this may not be a suitable assessment, and it was thought that senior physiotherapists were required to deliver the intervention in this group of acutely ill patients. Consultant leadership was thought to be important for both interventions and a link to a consultant was thought to be necessary for physiotherapists delivering home EPR.

Feasibility outcomes

Clinical outcomes

The primary outcome assessment was undertaken in fewer than half of the participants. The small number of participants involved did suggest better outcomes in the home EPR group and the group allocated to both hospital and home EPR.

Secondary clinical outcomes for this study were extensive, with several being exploratory. The small numbers in the pilot trial do not allow us to draw inferences from the data and the focus was on safety data, especially in light of recent evidence. 58

No deaths were recorded in any of the intervention arms during the 90-day follow-up period, which, given the severity of illness of the patients recruited (see Chapter 3, Baseline data, and previous studies32,58,156,157), is perhaps surprising.

Attendance at group PR (usual care) was poor for the participants in our trial at the 90-day follow-up period. This was partly a result of the length of the follow-up as the majority of participants had these sessions scheduled outside the follow-up period.

The observation of a potential reduction in exacerbation rates in those patients who received home EPR may be a promising outcome for both patients and health-care providers. Exacerbation frequency is a strong predictor of adverse outcomes and poorer prognosis, quality of life and function and contributes to the risk of hospital admission and mortality. 32,45,58 The rates in our trial are lower than those found in another study. 158

Readmission to hospital within 30 days is a crude indicator of rate of recovery following AECOPD, but is considered a significant marker of the effectiveness of clinical management strategies within the NHS. 38 In the small number of participants in this trial, we observed a slight reduction in readmission to hospital in those who received home EPR, in line with a recent review. 159 This outcome is dependent on the severity of patients included and can be predicted utilising tools such as the PEARL score. 109 Our analysis suggests that the overall readmission rates were similar to those predicted by the PEARL score and are representative of the wider COPD population.

After seeing the results, COPD readmission was chosen as the candidate primary outcome. Among the motivations for this were:

• Readmission is a clinically important outcome for both patients and the health-care system.

• As a safety outcome, readmission data could potentially be collected from withdrawn participants (with the appropriate permissions), resulting in a low proportion of missing outcome data.

• Readmission is an ‘objective’ outcome as it can be retrieved from medical records and is not influenced by reporting.

Among the challenges found were:

• Specification to be chosen (continuous, categorical, time to event). Continuous data were chosen as this could lead to a smaller sample size estimation. It was unclear how to define an incidence/survival outcome (and there is less literature on pilot studies using this kind of data).

• Estimation of the SD. Given the skewed outcome it could be debated whether or not choosing a continuous outcome is an efficient way of computing the sample size. A potential solution is to evaluate the outcome on the log scale (adding 1) when performing statistical inference and then report transformed and untransformed estimates.

• Denominator and participants to include. This issue will be mitigated in the main trial as we will collect data from patient notes on all patients.