Holistic services for people with advanced disease and chronic or refractory breathlessness: a mixed-methods evidence synthesis

Design and registration

The systematic review and meta-analyses were conducted and reported in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement,125 and adhered to guidelines from the Centre for Reviews and Dissemination and Methodological Expectations of Cochrane Intervention Reviews. The review protocol was prospectively registered with PROSPERO (reference number CRD42017057508). 126

Eligibility criteria

The following inclusion criteria were applied.

• Participants: adults experiencing breathlessness related to advanced disease (described as suffering from breathlessness, dyspnoea, shortness of breath, difficulty breathing, laboured breathing and with advanced stages of diseases with a high prevalence of breathlessness) including, but not limited to, cancer (advanced local or metastatic), chronic respiratory disease [Global initiative for chronic Obstructive Lung Disease (GOLD) stage III or IV/grade C or D], heart failure (New York Heart Association stage III or IV) or progressive neurological conditions. Studies in which ≥ 50% of participants met these definitions were included.

• Interventions: as there is no standard definition, holistic breathlessness services were defined as services that draw on multiple specialties and disciplines, encompass pharmacological and non-pharmacological interventions selected on the basis of a holistic needs assessment, enrol patients because of their breathlessness (not their diagnosis), emphasise self-management, aim to reduce the perception and impact of breathlessness and related symptoms, and are offered in outpatient, community or day hospice settings.

• Comparators: all comparators were considered for controlled studies, including no treatment, usual care, an attention control (e.g. a patient support group without a specific focus on breathlessness) or an active control (e.g. alternative service).

• Outcomes: health outcomes included breathlessness intensity; breathlessness affect and impact;2 anxiety and depression; physical functioning; health status or QoL; and survival. Economic outcomes included formal health and social care service utilisation and costs, unpaid caregiver costs including caregivers’ time off work, quality-adjusted life-years (QALYs) derived from generic QoL measures [e.g. EuroQol-5 Dimensions (EQ-5D)] and perspectives of economic analysis. Acceptability outcomes included patient flow data (uptake, adherence, completion), as well as patient and caregiver perspectives on acceptability, satisfaction and/or experience.

• Study design: RCTs with a parallel, single-stage or cross-over design, including studies using minimisation, non-randomised studies including prospective and retrospective designs, quantitative and qualitative designs to elicit patient and caregiver satisfaction and experience.

Studies were excluded if they did not specifically target patients with breathlessness, or if interventions that targeted breathlessness used only a single treatment (e.g. physical exercise). Pulmonary rehabilitation and disease-specific services (e.g. integrated respiratory care) were deemed outside the scope of this review. Interventions that exclusively targeted service providers or carers were also excluded. Narrative reviews, opinion papers, case studies and case series with fewer than five participants were excluded.

Search strategy

Quality assessment

All included studies were independently assessed by two researchers (LJB and MM) for methodological and reporting quality, using standardised checklists, as outlined below. Information to aid quality assessment was obtained from primary, secondary and protocol articles.

Data synthesis/data extraction and analysis

Data on the characteristics of the service, outcomes and study information were extracted from each paper by one researcher (SM/LJB) using a predesigned electronic data capture form. These were checked by a second researcher to ensure rigour (LJB/MM). Authors were contacted if additional information was needed for meta-analyses. To increase validity and ensure comprehensiveness, the analysis and interpretation were reviewed by members of the PAG including researchers, patient/carer representatives and clinicians.

Design

To identify patient characteristics that predict response to holistic breathlessness services, in terms of patient breathlessness mastery and distress due to breathlessness, we conducted a secondary analysis of pooled individual patient data from three RCTs. 91,94,95 These were a convenience sample of available data through project team members who were involved with, and act as, data custodians for these trials. As a result of collaborations during the trial designs, there was also a high number of variables in common across these data sets to aid pooling.

Data sources and ethics approvals

Trials included Higginson et al. ’s94 2014 trial of a 6-week service with people with advanced disease, Farquhar et al. ’s91 2014 trial of a 2-week service for people with advanced cancer and Farquhar et al. ’s95 2016 trial of a 4-week service for people with non-malignant disease. Participants in each trial had chronic or refractory breathlessness and advanced disease.

Anonymised individual patient data were obtained from the data custodians, and data sets were cleaned and harmonised. This secondary analysis of anonymised data did not require ethics approval. Each of the contributing studies followed appropriate ethics approval procedures (King’s College Hospital reference number 10/H0808/17; Cambridgeshire 2 NHS reference number 08/H0308/157).

Methods

Our study built on methods of a previous analysis of patient predictors of response to opioids. 141 The primary analysis considered an intervention response on Chronic Respiratory Disease Questionnaire (CRQ) mastery, defined by the minimal clinically important difference as an improvement of 0.5. 142 This outcome was selected as it was the primary outcome for the Higginson et al. 94 study, and a secondary outcome in the Farquhar et al. 91,95 studies.

The secondary analysis considered an intervention response on Numerical Rating Scale (NRS) distress due to breathlessness (henceforth, NRS distress), defined as an improvement of 1 point. 143 This was selected for secondary analysis as it was the primary outcome for both Farquhar et al. 91,95 studies, but not measured in the Higginson et al. 94 study.

Candidate variables (and their reference group/possible score ranges) for the primary and secondary analysis included diagnosis (reference group: COPD); predicted forced expiratory volume in 1 second (FEV₁) and baseline scores for CRQ dyspnoea; CRQ fatigue; CRQ mastery and CRQ emotional function (1 to 7); EQ-5D Utility Index (–1 to 1) and EQ-5D visual analogue scale (VAS) (0 to 100); Hospital Anxiety and Depression Scale (HADS) anxiety and HADS depression (0 to 21); and NRS average breathlessness in the last 24 hours (0 to 10; henceforth, NRS average). As it was not measured in Higginson et al. ,94 baseline NRS distress (0 to 10) could be included only in the secondary analysis. For all variables, the time point immediately pre intervention for each group was treated as ‘baseline’ (e.g. in the Higginson et al. 94 trial, fast-track participants’ baseline was week 1, wait-list participants’ baseline was week 6).

Variables deemed significantly related to response (p < 0.05) in univariate logistic regression models were included in the multivariate analyses. Multivariate analyses comprised backward stepwise logistic regression, in which variables with the largest non-significant (p > 0.05) p-value in the model were sequentially removed. Dependent variables were assessed for multicollinearity. To control for age, sex (reference group: male) and study of origin (refence group: Farquhar et al. 95), these variables were forced into the multivariate analyses.

To aid interpretation, variables that were significant predictors of response were categorised (when possible, using previously established groupings for that measure). The percentages of participants classified as responders for each baseline category were then tabulated.

Design

Nominal group techniques are commonly used for consultation with stakeholders, but have been criticised for lacking transparency, reliability and opportunities for clarification. 93 Although the Delphi technique overcomes these issues, this method can be time-consuming with multiple rounds of consultation, and the initial content can be shaped by a minority. Transparent expert consultation (TEC) methods have, therefore, been developed in response to these limitations,93,144 and have been successfully used to elicit recommendations in palliative and end-of-life care research. 145–149

Our TEC comprised a stakeholder consultation and follow-up online consensus survey. Like nominal group technique, the stakeholder workshop provided structured opportunities for expression of views by a group of experts. This activity was followed by a wider online consensus survey to rate the suggested priorities (similar to a single-round Delphi technique), enabling rapid consultation of multiple key stakeholders.

Participants

A wide range of stakeholders involved in the provision of care for people with chronic breathlessness were purposively selected and invited to participate in the stakeholder workshop. Service providers and commissioners, voluntary sector organisation representatives, patient and carer representatives, and health and social care practitioners from a range of specialties and professional groups were identified through contact lists of people and organisations that the research team had previously worked with, online literature and website searches, and via recommendations from participants.

The PAG members, workshop participants and those who expressed interested but were unable to attend the stakeholder workshop, were invited to complete the online consensus survey. Additional individuals from groups that were under-represented at the workshop (including service users) were also purposively invited to complete the consensus survey, using the methods described above. All participants were adults with capacity to give informed consent.

Ethics, recruitment and informed consent

Ethics approval for this study was obtained from the KCL research ethics committee (reference number LRS-16/17-4692).

A personal e-mail invitation to the stakeholder workshop was sent to purposively selected stakeholders, including a copy of the information sheet, the consent form and the workshop schedule (see Appendix 2). This provided an opportunity for attendees to consider the study information prior to the day of the workshop, at which time they were asked to provide written consent prior to the recorded discussions and data collection. Participants were reimbursed reasonable travel costs to attend the stakeholder workshop.

For the online survey, information regarding its purpose and how the data would be used was included in an attachment to the invitation e-mail and summarised at the start of the survey. Consent for the online survey was presumed through participation. A hard-copy postal consensus survey with a freepost return envelope was available to participants preferring that format.

Procedure

Breathlessness severity

Ten studies90,94,98–100,154–157 assessed the severity of breathlessness using one or more of the following measures: VAS, NRS or Borg scores (Table 7). For ‘best breathlessness’, two studies using VAS found a greater improvement in the intervention group than in the control (differences in median change 5.7, p = 0.03,98 and 1.0, p = 0.0290), and three studies with unspecified measures found a significant intervention effect [F(2,44) = 5.30; p = 0.009]156 or no difference (data not reported). 155,156 For ‘worst breathlessness’, one study using VAS found a greater improvement in the intervention group than in the control (difference in median change 3.5; p = 0.0590), whereas no significant differences were found by two studies using NRS (MD –0.35, 95% CI –1.71 to 1.01; p = 0.6194; MD 0.41, 95% CI –0.86 to 1.67; p = 0.53157), one study using VAS (difference in median change 3.8; p = 0.1498) and one with an unspecified measure155 (data not reported). For ‘average breathlessness’, one study using an unspecified measure found a greater improvement in the intervention group than in the control (difference in mean change 1.2155), whereas two studies using NRS did not (MD –0.33, 95% CI –1.28 to 0.62; p = 0.49;94 MD 0.65, 95% CI –0.49 to 1.80; p = 0.26). 157 One study using NRS found no effect on breathlessness on exertion (MD –0.73, 95% CI –1.69 to 0.22; p = 0.13)94 and one study using Borg scale ratings for breathlessness at rest and on exertion found no difference between groups (data not reported). 154 In line with their feasibility study results,99 a powered trial comparing one with three service contacts found no significant difference in NRS worst (MD 0.2, 95% CI −2.31 to 2.97; p = 0.83) or average (MD 0.3, 95% CI –2.00 to 2.62; p = 0.79) breathlessness. 158

Breathlessness affect

Ten studies90,91,95,98–100,155–157 assessed ‘distress due to breathlessness’ using VAS (range 0–100, higher = worse) or NRS (range 0–10, higher = worse). 91,95 Eight of these studies compared breathlessness services to usual care, three of which155,156 reported no significant difference but did not provide data. Data from five studies90,91,95,98,157 were combined in a meta-analysis (n = 324; Figure 1) that showed significantly lower NRS distress following the intervention than following the control (MD –2.30, 95% CI –4.43 to –0.16; p = 0.03). A sensitivity analysis excluding two outlier studies90,98 resulted in a reduced point estimate of effect and non-significant difference (MD –0.29, 95% CI –1.00 to 0.43; p = 0.43; I² = 0%). Similar to the finding from the preceding feasibility study,99 a RCT testing service variations found no difference on NRS ‘coping with breathlessness’ (MD –1.7, 95% CI –4.27 to 0.90; p = 0.20),100 and significantly higher NRS distress due to breathlessness following three sessions versus one session (MD 3.9, 95% CI 0.98 to 6.91; p = 0.01). 100

FIGURE 1.

The NRS distress due to breathlessness. df, degrees of freedom; IV, inverse variable. Reproduced with permission from Brighton et al. 92 This is an open access article distributed in accordance with the Creative Commons Attribution 4.0 Unported (CC BY 4.0) license, which permits others to copy, redistribute, remix, transform and build upon this work for any purpose, provided the original work is properly cited, a link to the licence is given, and indication of whether changes were made. See: https://creativecommons.org/licenses/by/4.0/.

Reproduced with permission from Brighton et al. 92 This is an open access article distributed in accordance with the Creative Commons Attribution 4.0 Unported (CC BY 4.0) license, which permits others to copy, redistribute, remix, transform and build upon this work for any purpose, provided the original work is properly cited, a link to the licence is given, and indication of whether changes were made. See: https://creativecommons.org/licenses/by/4.0/.

The quality of evidence for distress due to breathlessness was rated as very low. Owing to the lack of blinding of participants or outcome assessors, plus evidence of (or unclear information regarding) attrition bias, the evidence was downgraded for risk of bias. This evidence was also downgraded for inconsistency (owing to wide variation in effect estimates across studies), and imprecision (evidenced by wide variance in point estimates and lower 95% CIs, indicating potentially very little effect).

Four studies91,94,95,154 assessed ‘mastery over breathlessness’ using the CRQ mastery domain. 94,154 A meta-analysis of these data (n = 259; Figure 2) showed a statistically non-significant increase in mastery (range 1–7, higher = better) favouring the intervention (MD 0.23, 95% CI –0.10 to 0.55; p = 0.17). A sensitivity analysis excluding one study,154 deemed to be at a high risk of bias, increased the point estimate of effect (MD 0.30, 95% CI –0.06 to 0.66; p = 0.11). One study found significantly lower mastery scores after three service contacts than after one service contact (MD –0.6, 95% CI –1.06 to –0.11; p = 0.02). 100

FIGURE 2.

The CRQ breathlessness mastery. df, degrees of freedom; IV, inverse variable. Reproduced with permission from Brighton et al. 92 This is an open access article distributed in accordance with the Creative Commons Attribution 4.0 Unported (CC BY 4.0) license, which permits others to copy, redistribute, remix, transform and build upon this work for any purpose, provided the original work is properly cited, a link to the licence is given, and indication of whether changes were made. See: https://creativecommons.org/licenses/by/4.0/.

Reproduced with permission from Brighton et al. 92 This is an open access article distributed in accordance with the Creative Commons Attribution 4.0 Unported (CC BY 4.0) license, which permits others to copy, redistribute, remix, transform and build upon this work for any purpose, provided the original work is properly cited, a link to the licence is given, and indication of whether changes were made. See: https://creativecommons.org/licenses/by/4.0/.

One further study found improved dyspnoea-12 (range 0–36; higher = worse) scores with the intervention than with control (MD 5.19, 95% CI 0.62 to 9.75; p = 0.026). 157

We judged the quality of evidence for mastery to be low. This evidence was downgraded because we deemed the studies to have a high risk of bias due to the lack of blinding of participants or outcome assessors, plus the evidence of (or unclear information regarding) attrition bias. In addition, we downgraded this evidence for imprecision because of the wide variance in point estimates and lower 95% CIs, indicating potentially very little effect.

Psychological outcomes

Seven studies assessed anxiety and depression using HADS. 90,91,94,95,98,155,157 Data from these seven studies (n = 552, Figure 3) showed a statistically non-significant reduction in anxiety scores (range 0–21; higher = worse), favouring the intervention (MD –1.59, 95% CI –3.22 to 0.05; p = 0.06). The sensitivity analysis, excluding one study155 deemed to be at a high risk of bias, increased the point estimate (–1.85, 95% CI –3.76 to 0.06; p = 0.06). A sensitivity analysis removing one outlier study98 resulted in a reduced point estimate but statistically significant group difference (MD –0.66, 95% CI –1.23 to –0.10; p = 0.02; I² = 0%). No statistical differences in anxiety were reported when comparing one and three contacts. 99,100

FIGURE 3.

The HADS – anxiety. df, degrees of freedom; IV, inverse variable. Reproduced with permission from Brighton et al. 92 This is an open access article distributed in accordance with the Creative Commons Attribution 4.0 Unported (CC BY 4.0) license, which permits others to copy, redistribute, remix, transform and build upon this work for or any purpose, provided the original work is properly cited, a link to the licence is given, and indication of whether changes were made. See: https://creativecommons.org/licenses/by/4.0/.

Reproduced with permission from Brighton et al. 92 This is an open access article distributed in accordance with the Creative Commons Attribution 4.0 Unported (CC BY 4.0) license, which permits others to copy, redistribute, remix, transform and build upon this work for any purpose, provided the original work is properly cited, a link to the licence is given, and indication of whether changes were made. See: https://creativecommons.org/licenses/by/4.0/.

We rated the quality of evidence for anxiety to be very low. This evidence was downgraded because of the risk of bias, including the lack of blinding of participants or outcome assessors, plus the evidence of (or unclear information regarding) attrition bias. We also downgraded for inconsistency (owing to wide variation in effect estimates across studies), and imprecision (evidenced by wide variance in point estimates, and lower 95% CIs indicating potentially very little effect).

For depression, one study156 reported no difference between groups but did not provide data. Meta-analysis using the six remaining studies (n = 408, Figure 4) showed reduced depression scores (range 0–21, higher = worse) favouring the intervention (MD –1.67, 95% CI –2.52 to –0.81; p < 0.001). No statistical differences in depression were reported when comparing one and three contacts. 99,100

FIGURE 4.

The HADS – depression. df, degrees of freedom; IV, inverse variable. Reproduced with permission from Brighton et al. 92 This is an open access article distributed in accordance with the Creative Commons Attribution 4.0 Unported (CC BY 4.0) license, which permits others to copy, redistribute, remix, transform and build upon this work for any purpose, provided the original work is properly cited, a link to the licence is given, and indication of whether changes were made. See: https://creativecommons.org/licenses/by/4.0/.

Reproduced with permission from Brighton et al. 92 This is an open access article distributed in accordance with the Creative Commons Attribution 4.0 Unported (CC BY 4.0) license, which permits others to copy, redistribute, remix, transform and build upon this work for any purpose, provided the original work is properly cited, a link to the licence is given, and indication of whether changes were made. See: https://creativecommons.org/licenses/by/4.0/.

The quality of evidence for depression was judged to be moderate. This evidence was downgraded for risk of bias only, indicated by the lack of blinding of participants or outcome assessors, plus the evidence of (or unclear information regarding) attrition bias.

Three further studies reported no significant differences between the intervention and control groups in ‘psychological symptoms’: two using an unspecified measure (data not reported)155,156 and one using the Rotterdam Symptom Checklist (range 7–28, higher = worse; difference in median change –8; p = 0.21). 98 One study comparing one session with three sessions found no significant difference on CRQ emotion scores (MD –0.09, 95% CI –0.54 to 0.36; p = 0.69). 100

Physical function and health status/quality of life

Five studies90,91,94,95,98 assessed physical function. Two studies found greater improvements following the intervention than in the control participants using the Functional Capacity Scale (range 0–14; higher = better; MD for change 1.25; p < 0.02)90 and World Health Organization (WHO) Performance Status Scale (range 0–5, higher = worse; difference in median change 2, p = 0.02). 98 Three studies observed no difference in functional outcomes between groups assessed using the London Chest Activities of Daily Living Scale (MD –5, 95% CI –12.22 to 1.02; p = 0.10)94 or patient-reported number of times leaving their house (data not reported). 91,95

Seven studies91,94,95,98,100,154,157 included a measure of health status or QoL, none of which found significant differences between groups across the CRQ dyspnoea domain91,94,95,154 (including the comparison between one and three sessions100) or total score,94 EQ-5D index94 or VAS,94,157 and the Rotterdam Symptom Scale QoL domain. 98 Owing to heterogeneous measures, change from baseline and post-intervention scores, and cases of non-normally distributed data, we decided against meta-analysis for these outcomes.

Survival

Two studies reported survival data. 94,98 One found a significant difference in survival (generalised Wilcoxon score of 3.9; p = 0.048) in favour of the intervention. 94 Subgroup analysis found that the difference was driven by participants with non-cancer diagnoses. The remaining study, enrolling only patients with cancer, found no difference in survival across groups (data not reported). 98

Carer outcomes

Two studies91,95 assessed outcomes for carers of patients using the service. No significant differences were reported between groups on carer anxiety (HADS) or depression (HADS) or distress due to the patient’s breathlessness (NRS).

Economic evaluation

Evidence for cost-effectiveness of the interventions was limited and non-conclusive. Four studies of three services were eligible for inclusion,91,94,95 and all used GBP 2011/12 costs. Three studies compared the service with usual care, and one study compared offering one session with offering three sessions within a service.

The perspective of the analysis for Farquhar et al. ’s91 study of their Breathlessness Intervention Service (BIS) was societal by including costs of informal care. Among patients with advanced cancer, total health/social costs including informal care for 8 weeks prior to the baseline assessment were £6137 [standard deviation (SD) £6099] in the BIS group and £5461 (SD £6099) in the usual care group. Costs between baseline and follow-up at 2 weeks were £794 (SD £866) for BIS and £1121 (SD £1635) for usual care. Intervention costs for the BIS were £119 (SD £62). Total costs were £354 lower for the BIS than for usual care (95% CI –£1020 to £246) and incremental QALY gain was 0.0002 (95% CI –0.001 to 0.002) after controlling for baseline. The chance of the BIS having lower total costs and resulting in a greater QALY gain than usual care was 80.9% according to cost-effectiveness planes. There was a 50.9% chance of the BIS being lower than usual care for total costs, and having a greater number of QALYs.

For Farquhar et al. ’s95 study of their BIS, a NHS perspective was used. Among patients with advanced non-malignant disease, total health/social costs for 8 weeks prior to the baseline assessment were £1952 (SD £3290) in the BIS group and £3630 (SD £5588) in the usual care group. Costs between baseline and follow-up at 4 weeks were £1371 (SD £2948) for BIS and £659 (SD £1253) for usual care. Intervention costs for BIS were £156 (SD £80). After adjusting for baseline values, total costs were £799 higher for BIS (95% CI –£237 to £1904) and the BIS group gained 0.003 extra QALYs (95% CI –0.001 to 0.007) than the control group. The cost per QALY for BIS was £266,333. The chance of BIS having lower total costs and resulting in a greater QALY gain than usual care was 7% according to cost-effectiveness planes.

Higginson et al. ’s94 Breathlessness Support Service (BSS) was assessed in terms of hospital inpatient days and formal care costs for 12 weeks prior to the baseline assessment. The cost of formal care was £2911 (SD £2729) for BSS and £3709 (SD £4484) for usual care. Incremental QALY gain between baseline and follow-up at 6 weeks was 0.092 (95% CI –0.23 to 0.04). No more information related to economic evaluation was reported in this paper, but a separate paper51 analysed the same data to measure the cost of care, regardless of randomisation. The total cost for all participants was £11,507 (SD £9911), which was the sum of health-care costs [£2624 (SD £3456)], social care costs [£628 (SD £1132)] and informal care costs [£8254 (SD £8777)]. Increased breathlessness on exertion was associated with higher health-care costs and informal care costs, and having a carer was associated with higher informal care costs.

Johnson et al. ’s158 trial compared one session with three sessions of a breathlessness service from a NHS perspective, including costs of the service in each arm and other health-related resource use costs. For the three-session service, there was a non-significant reduction in overall QALYs (MD –0.006, 95% CI –0.018 to 0.006) and non-significant increase in costs. There was no evidence that the additional costs were offset by lower health-related resource use costs elsewhere. The probability of the single session being cost-effective at a threshold value of £20,000 per QALY was > 80%. 100

Referral, uptake, adherence and adverse events

Of the 11 RCTs included in the analysis, fewer than half reported numbers screened for eligibility. When available, studies reported screening between 53 and 932 people for eligibility, of whom 15% to 48.6% were deemed to be eligible for inclusion. The total number of participants randomised into each trial ranged from 22 to 156. Of the participants randomised, between 50% and 90.8% of participants completed the trial to key outcome stages. Full data including reasons for attrition are described in Table 8. No adverse events were reported.

Experiences

Across the 11 studies with qualitative data90,91,94,95,102–106,159–161 and the five studies with quantitative data99,100,103,157,161,162,168 on participants’ satisfaction, three themes emerged with regard to experiences: valued characteristics, perceived outcomes and challenges to services.

Chronic Respiratory Disease Questionnaire mastery

Of the participants for whom the CRQ mastery response could be calculated, half (97/194) were classed as responders. Variables significantly associated with a CRQ mastery response in the univariate analyses included CRQ mastery (p < 0.001), CRQ dyspnoea (p = 0.008), CRQ fatigue (p = 0.009), EQ-5D Utility Index (p = 0.007), EQ-5D VAS (p = 0.014) and NRS average breathlessness in the last 24 hours (p = 0.039) (Table 10).

Baseline CRQ mastery remained the only significant predictor of intervention response in the multivariate regression when controlling for age, sex and original study. Participants with lower baseline CRQ mastery scores were more likely to respond on this measure [odds ratio (OR) 0.57, 95% CI 0.43 to 0.74; p < 0.001] (Table 11).

Seventy-six per cent (n = 31) of participants with a baseline CRQ mastery score of ≤ 2 responded to the intervention, whereas 53% (n = 49) of those scoring 3–5 and 28% (n = 17) of those with a score of ≥ 5 responded on this measure.

Numeric Rating Scale distress due to breathlessness

Of participants for whom NRS distress response could be calculated, 81 out of 129 (62.8%) were classed as responders. The baseline NRS distress score was the only variable significantly related to distress response in the univariate analysis (p < 0.001) (Table 12).

In the multivariate model, participants with a higher NRS score at baseline were more likely to respond on this measure than those with a low score (OR 1.64, 95% CI 1.35 to 2.03; p < 0.001) (Table 13).

Eighty-nine per cent of participants (n = 41) with a ‘severe’ (≥ 7) baseline NRS distress score responded to the intervention, whereas 59% (n = 23) of those with ‘moderate’ scores (4–6) and 39% (n = 17) of those with ‘mild’ scores (≤ 3) responded on this measure.

Attendance

The workshop was attended by 37 stakeholders (40 registered and 117 were invited). Most attendees were from the UK (including Bristol, Cambridge, Exeter, Glasgow, Leicester, London, Norwich, Nottingham and York); however, two were visiting the UK from abroad (Australia and the Netherlands) and were able to attend.

Of the 37 event attendees, 33 stakeholders participated in the group work and completed response booklets (there were three groups: one with 8 participants, one with 12 and one with 13). An additional two response booklets were completed by patient representatives who were unable to attend the event but wanted to provide their input. Table 14 shows the characteristics of participants who completed the response booklets.

Prompt question booklet responses

Event summary

Throughout the stakeholder workshop, the need for improved collaboration, integrated working and standardisation were strong themes. Stakeholders acknowledged the successful components of existing practices, across multiple specialties and disciplines, which should be developed but not duplicated. Figure 5 shows the graphic recording summarising discussions throughout the workshop.

FIGURE 5.

Graphic recording of stakeholder workshop discussions. Reproduced with permission from Joel Cooper, April 2019.

Reproduced with permission from Joel Cooper, April 2019

Respondents

Those invited to the workshop, plus an additional 43 stakeholders, were invited to take part in the online consensus survey. Of the 160 invited, 72 participated (an additional two participated after receiving an invite forwarded from an originally invited participant), representing a 46% response rate. Seventy-eight invitees did not respond, six were away until after the survey close date, three e-mail addresses were invalid and one person declined. Twenty-six of those who responded to the consensus survey had completed a workshop response booklet. Table 18 shows the characteristics of participants who completed the online consensus survey.

Survey responses

Of the 34 final statements from participants included in the online survey, 10 related to clinical practice, eight to policy, and 16 to research. Figure 6 shows boxplots of online consensus survey scores. The majority (n = 20) of participants received strong agreement but low consensus, seven participants received strong agreement and high consensus, six participants received moderate agreement and low consensus, and one participant received moderate agreement and high consensus. No statements were disagreed with. Participants’ priorities for clinical practice were more often met with strong agreement and high consensus, with lowest agreement and consensus being around participants’ priorities for research.

FIGURE 6.

Boxplots of online consensus survey scores. (a) Clinical priorities; (b) policy priorities; and (c) research priorities. Reproduced with permission from Brighton et al. 172 This article is distributed under the terms of the Creative Commons Attribution-NonCommercial 4.0 License (http://www.creativecommons.org/licenses/by-nc/4.0/) which permits non-commercial use, reproduction and distribution of the work without further permission provided the original work is attributed as specified on the SAGE and Open Access pages (https://us.sagepub.com/en-us/nam/open-access-at-sage).

Reproduced with permission from Brighton et al. 172 This article is distributed under the terms of the Creative Commons Attribution-NonCommercial 4.0 License (http://www.creativecommons.org/licenses/by-nc/4.0/) which permits non-commercial use, reproduction and distribution of the work without further permission provided the original work is attributed as specified on the SAGE and Open Access pages (https://us.sagepub.com/en-us/nam/open-access-at-sage).

Systematic review

For the systematic review, registration of the protocol and the systematic and comprehensive search across multiple databases (inclusive of grey literature), with no exclusions by publication year or language, ensured that there was a high level of transparency and representativeness. The review eligibility and quality assessments were conducted independently by two researchers. Furthermore, multiple stakeholders, including researchers, clinicians and patient/carer representatives, contributed to both the analysis and the interpretation of the review data. The integration of qualitative and quantitative studies was another strength of this work and allowed for a comprehensive evaluation of service experiences and outcomes. Recipients of services reported health benefits in line with those found with established tools, again highlighting an effect on affective and psychological domains of health.

The review also had some limitations. The meta-analyses combined data from services with varying structure, delivery and recipients, and there was evidence of statistical heterogeneity. We completed sensitivity analyses in response to this, but the overall data set was moderate in size and these sensitivity analyses inevitably compromised the precision of the effect estimates. Effect estimates may also have been inflated by lack of blinding of study personnel, and low expectations from participants in studies for which a fast-track design was not employed. Moreover, we did not assess for statistical evidence of publication bias given the low number of studies, and there was some evidence of selective reporting where study authors did not provide data for statistically non-significant findings. Our estimates therefore do not include these data. In addition, challenges with inconsistent use of and/or unclear reporting of outcome measures sometimes precluded meta-analysis (e.g. for breathlessness intensity). Evidence around cost-effectiveness was also limited by a lack of studies with an economic component.

Qualitative data were predominantly drawn from two UK services91,94,95 and patients who had fully engaged with the services. Less is understood about the experiences of the carers, the patients using these services internationally and those who dropped out and who perhaps might report less benefit. Finally, although limiting our review to studies with people with advanced disease was reflective of the current evidence base, it is important to acknowledge that limiting service access based on disease severity may limit reach and not serve those with distressing breathlessness during early stages of disease. For this reason, and in the light of our findings, we advocate access to these services based on the presence of distressing chronic or refractory breathlessness, rather than disease severity.

When interpreting the data, we note that some of the study group were co-authors on included studies. These individuals were not involved in any data extraction or quality assessment of these studies, providing a level of objectivity and impartiality. However, we acknowledge that influences of motivated reasoning (e.g. confirmation biases) have the potential to influence all stages of research, from deciding research questions to presentation of findings. 173 As such, team discussions encompassing those who were co-authors (4 out of the 12 report authors) and were not co-authors on previous studies of breathlessness services were important in balancing existing expertise and new insights, in addition to continually exposing our findings to feedback from diverse audiences (see Dissemination strategy).

Responder analysis

A strength of this work was that individual patient data were pooled from three high-quality RCTs. The combined data set enabled a more powerful analysis to be undertaken than would have been possible via study-level meta-analysis. However, it did mean that the analysis was limited to variables common across the three data sets. For this reason, we were unable to test some potentially important patient factors (e.g. multimorbidity, functional status) and intervention characteristics (e.g. setting, duration, professional input), which could have been important moderators of effect. There were fewer cases available for the secondary analysis of NRS distress due to breathlessness, as this was included in only two out of the three data sets, reducing statistical power. Finally, it is unclear to what extent the primary finding reflects regression to the mean;174 however, the magnitude of the effect observed following the breathlessness services suggests that this is not the only contributing factor. The finding that response was not influenced by other patient characteristics may relate to the baseline scores of the sample. For example, baseline scores for anxiety and depression were mild across the included studies175 and, therefore, the analysis may not have tested for influences of higher levels of psychological distress. Conversely, this finding could relate to the tailored nature of the services, which should consider each patient based on their physical, psychological, social and spiritual profile.

Transparent stakeholder consultation

The diversity of stakeholders who contributed to the workshop and consensus survey was a strength of the TEC process; participants included service providers and commissioners, voluntary sector organisations, patient and carer representatives, as well as health and social care practitioners from a range of specialties and professional groups. Although this diversity of knowledge of existing research may have led to lower consensus on participants’ statements relating to research, it also led to generation of a wide variety of suggestions for areas in need of further evidence. The priorities elicited from participants are also closely linked to emerging evidence as a result of the workshop structure, where this information was presented at the start for reflection and discussion. For example, participants’ suggestions around understanding the optimal models, economic modelling and medium- to long-term effects of breathlessness services, relate directly to findings from our systematic review, while recognition and inclusion of informal carers drew from evidence presented around their role and its impact.

It is also important to note that the suggestions made by our participants may be most applicable to health and social care in UK settings, as most people in the stakeholder consultation process were based in UK universities and NHS secondary care settings. There was also a higher number of participants with roles as a researcher or doctor, but these were also the participants for whom having more than one role was most common (e.g. 69% of researchers were also health-care professionals; 47% of doctors had additional roles in research, charities and/or commissioning). A smaller number of service user representatives attended the workshop than we had anticipated. This suggests that this engagement method may be less suited to people living with chronic breathlessness and/or with caring responsibilities and may have biased the types of priorities elicited from participants. However, more than 10 out of 74 (14%) respondents to the survey identified as a patient/carer. In addition, patient/carer representatives were members of the PAG and helped with the synthesis of the suggestions generated from the workshop participants in this role. This included working closely together to ensure that the participants’ statements were clear and understandable for people with a range of professional and personal expertise.

Having a clear and structured workshop process ensured that there was efficient collection of the attendees’ views. We mitigated the risk of bias inherent in face-to-face consultation techniques, whereby some participants may contribute to discussions more than others, by providing the opportunity to submit individual written suggestions in the response booklets. These individual responses were the primary focus in generating content for the consensus survey. A full Delphi process176 or having additional consultation rounds may have provided more opportunity to refine participants’ statements, but the TEC technique utilised the multiple forms of data collected at the workshop (including the scribe notes, graphic recording, response booklets and audio-recordings, which enabled the rapid synthesis and revision of participants’ statements). Although this was substantially undertaken by two researchers (MM and LJB), the full list of original suggestions had been shared with the PAG for transparency. Crucially, individual suggestions from participants were removed or modified only in line with the reasons outlined above (e.g. duplicates, low priority) and not on the basis of controversy or creativity. This method resulted in generally high levels of agreement and consensus, particularly around participants’ priorities for clinical practice. The consensus survey response rate was limited, although a high proportion of those completing the workshop booklets participated (74%) and the overall response rate was similar to studies that have previously utilised this method. 146–148 Importantly, all key stakeholder groups were represented.

For clinical practice

To our knowledge, this evidence synthesis offers the first systematic review in this field. The bias of clinical effect towards improved psychological health is consistent with the primary focus of holistic services to support people with advanced disease to live with breathlessness rather than to take this symptom away. The wider effect on psychological health may have been achieved through management of breathlessness, but also through addressing concurrent symptoms and the receipt of expert holistic care that put the person before their disease. The effect sizes from our meta-analyses (point estimates: distress –0.57, mastery 0.25, anxiety –0.45, depression –0.55) are larger than those achieved with psychological therapies and self-management programmes in similar populations, and more comparable to pulmonary rehabilitation. 86,87. For NRS distress and HADS anxiety and depression, the MDs also represent a minimum clinically important difference (> 1 and > 1.5, respectively). 143,177 Although few measurable effects were identified for physical function, we acknowledge the qualitative data from patients and carers that captured feelings of expanding horizons, for example being able to independently complete activities of daily living. The diverse nature of improvements in physical function may be more readily captured using individualised measures, such as goal attainment scaling. 178

The interface between holistic services for breathlessness and pulmonary rehabilitation is an important consideration for people with chronic respiratory disease. Based on current evidence, we do not view holistic breathlessness services as a replacement for pulmonary rehabilitation, which is a highly effective and underutilised intervention. 81 However, these services may have a role for people who remain highly symptomatic despite completing pulmonary rehabilitation, or as a bridge for people who decline it, including those recently hospitalised with an acute exacerbation of disease. Holistic breathlessness services may also provide an additional opportunity for health gains in those unable to complete programmes with a major exercise component (e.g. in cases for which breathlessness limits exercise to an intensity associated with a training response). 28,84,179 With high levels of health service use in breathless patients and limited resource, evidence-based factors to prioritise access based on likelihood of benefit may be useful. When services are already in place or being established, our responder analysis suggests that there is merit in focusing efforts on patients with poor psychological health relating to their breathlessness.

Stakeholders’ priorities for clinical practice elicited through our TEC with strongest consensus called for flexible and accessible person-centred care, drawing on multiple expertise, and with the widest possible coverage. Stakeholders advocated for acknowledgement of the role of informal carers, valuing and being able to respond to breathlessness as a symptom in its own right, and sharing breathlessness management skills between other professionals and with informal carers. Free-text survey comments from participants underscored the need for person-centred, holistic care that acknowledged both psychological and emotional concerns. This fits with findings from the systematic review in terms of therapeutic components most valued by patients (e.g. tailored education, psychological support) and clinical benefit relating to anxiety and depression. The promotion of joint working around breathlessness was also suggested in a previous consultation focused on combined rehabilitation services for people living with COPD and heart failure. 180 This and other elements of stakeholders’ statements relating to clinical practice are in line with a palliative care approach of person-centred care, multidisciplinary input and inclusion of informal carers in the unit of care. 54 Working to build links with, or learn from, palliative care may be an efficient way to facilitate working in line with their priorities.

For policy

Recent international guidelines advocate for early integration of palliative care in people experiencing chronic disease. 96,97 Chronic or refractory breathlessness may serve as an appropriate referral indicator, especially in non-cancer conditions for which the unpredictable course of disease and difficulty predicting survival are barriers to timely palliative care referral and receipt. 54 Indeed, a symptom-triggered approach should reach more people than approaches based on prognostication.

Stakeholders’ priorities relating to policy elicited through our TEC particularly supported the recognition of informal carers in terms of their role, importance and support needs. When appropriate, this extended to supporting and encouraging their participation in health care. This is supported by recent evidence demonstrating the substantial contribution of informal carers to people with advanced disease and chronic breathlessness,35 including that their contribution saves approximately two-thirds of what would otherwise be formal care costs,51 and the impact that this has on their own health and well-being. 51,181

Our participants also stressed the importance of education, both to increase awareness among the general public and health and social care professionals, and to upskill professionals within existing services. Multiple stakeholders suggested that breathlessness services should be developed by adapting existing services rather than introducing something new. They noted that educating professionals in social care and communities about supporting people with breathlessness should also have value because, despite the high utilisation, people with breathlessness spend most of their time outside health services.

For research

The need for economic modelling of breathlessness services was the only research recommendation that received high consensus in our stakeholder consultation. This gap in evidence was reflected in the limited economic data available from studies included in the systematic review. The heterogeneity of services with respect to staffing, structure and content highlights a concurrent need to determine the most effective service delivery models. Although the patient and carer experience data highlight valued components (e.g. simple, portable interventions; involvement of carers), additional work is needed to understand which components of holistic services are most effective and how these might be integrated into, or alongside, the existing models of care. Consistency in the use of outcome measures in future studies would permit more detailed evaluations of comparative clinical effectiveness and cost-effectiveness. Considering our findings around the impact of these services on NRS distress and CRQ mastery, but the shorter administration format of the former, consistently using NRS distress due to breathlessness to measure service outcomes could be a good starting point. Alongside this, it is crucial to acknowledge that service users should value services and be willing to access them. Understanding what makes services attractive and acceptable to service users could help better serve patients and provide evidence to inform more efficient resource allocation. Discrete choice experiments could be used to identify which ‘attributes’ would be prioritised and preferred by patients and carers.

The remaining research recommendations generated by participants received strong or moderate agreement, but low consensus. Although this, in part, may relate to inadequate definitions or justifications for the populations and/or settings, and differing knowledge bases of contributing stakeholders around existing evidence, it may also reflect the diversity of issues that are in need of better evidence and the differing opinions over which issues to prioritise. Overall, participants emphasised that research should focus on developing effective solutions rather than revisiting the problem. Additional proposed topics for research included the role of psychosocial factors in breathlessness management, enhancing community support and testing the optimal ways to support informal carers of patients with breathlessness.

Summary of outputs and dissemination methods

Through the systematic review, we have disseminated evidence on symptom-triggered services as a means to access timely palliative care, especially in non-cancer conditions, and have provided evidence on the effectiveness of holistic breathlessness services. The responder analysis allowed for shared knowledge on patient characteristics that moderated treatment response from services. Furthermore, our participants were able to share consensus- and evidence-based priorities relating to clinical practice, policy and research.

We have shared findings via open-access publications in scientific journals. 92,171,172 Plain English summaries have been completed for each study component. 182,183 This project has been presented at several meetings to varied audiences (see List 1: presentations) and additional abstracts have been submitted [e.g. to the ERS International Congress (https://erscongress.org/)]. Throughout this project we have engaged with PPI representatives to provide feedback and inform the project from an external perspective. Furthermore, the Cicely Saunders Institute ‘Open House’ and ‘Conversation Starter’ events allowed for information-sharing with members of the public. To extend our reach further, information on our stakeholder event has been published in online news pieces in various local and international platforms (see List 2: online news articles). To enable engagement with a wider audience, we will share our publications, lay summaries and the full report via several social media platforms, including an online journal club on Twitter (Twitter, Inc., San Francisco, CA, USA; www.twitter.com).

Our stakeholder workshop allowed for education of professional bodies and lead service managers through presentation of relevant findings. We have shared the findings of this project directly with public bodies and policy-makers through submission to the Taskforce for Lung Health call for evidence. 184 We may have further opportunities to present findings at the Knowledge Exchange Seminar held in the Cicely Saunders Institute to educate colleagues, and at PPI workshop events to provide information to patient and public representatives (both events will have a rehabilitation theme).

Ensuring acceptable and appropriate research processes

For the stakeholder consultation, involvement of our PPI members in reviewing the workshop groupwork processes and refining the final list of statements from stakeholders ensured that these materials were clear and readable by participants from all backgrounds. Their help with piloting and redesigning the online survey to be more engaging and user-friendly was also important; for example, adding a background image and adding an option to be contacted by the researcher if patient/carer representatives wanted assistance completing the survey. This seemed to make these processes more acceptable to participants, including patient/carer representatives (who formed > 10% of our online survey respondents).

Including patient/carer voices throughout the consultation with stakeholders

The PPI members have been an important part of including patient/carer voices throughout our consultation, not only by improving the acceptability of our research methods, but also by suggesting additional patient/carer stakeholders to invite, and, through their own participation, by completing the workshop booklets and online consultation survey.

Ensuring interpretation of findings were grounded in patient/carer experiences

Having PPI members included across all project meetings ensured that there were opportunities to relate emerging findings to their real-life experiences, whether this was by highlighting inconsistencies in clinical practice or highlighting important issues related to breathlessness (e.g. relationships, including intimacy). Including a PPI member in the thematic analysis of the systematic review data ensured that the findings were not limited to researcher interpretations. This included raising the importance of reassurance for patients with breathlessness in how they are already self-managing, the need for psychological support that goes beyond just breathlessness, and the complexities of who is a ‘patient’ and who is a ‘carer’ when both may have substantial support needs.

Improving clarity and reach of dissemination

Our PPI group’s help with dissemination so far, commenting on the systematic review paper92 and co-authoring the stakeholder paper,172 has ensured increased reporting clarity. In our project meetings, they have also suggested additional ideas for disseminations to increase reach (e.g. through the NIHR INVOLVE network), and have agreed to help write plain English summaries to share alongside publications as they are accepted.

Patient and public involvement member reflections

Box 4 provides direct quotations taken from our PPI members reflecting on their work on the project and with the research team. Our PPI members said that they appreciated the opportunity to get involved in a variety of tasks and being given specific training (including one-to-one sessions) when required. They acknowledged that some of the more technical research methods could be difficult to understand at times but appreciated that the research team took time to explain and respond to questions about this during and outside meetings. Overall, they commented that they felt valued as part of the project team, leading to more rewarding and satisfying involvement.

Search strategy

1. exp Palliative Care/

2. exp Terminal Care/

3. exp Terminally Ill/

4. exp Hospices/

5. ((advanc* or progressiv* or agressiv* or end) adj2 (diagnos* or diseas* or illnes* or cancer* or malignan* or stage* or dementia* or failure* or heart*)).tw.

6. (last adj3 life).tw.

7. (Advanced disease or Cancer or Intrathoracic malignancy or Lung cancer or Chronic obstructive pulmonary disease or Non-malignant disease).tw.

8. 1 or 2 or 3 or 4 or 5 or 6 or 7

9. exp nursing/

10. exp physical therapy modalities/

11. exp occupational therapy/

12. (Multiprofessional or Multidisciplinary or Holistic or Complex intervention or Non-pharmacological intervention or non-pharmacological management or non-pharmacological or physiotherap* or Nurs* or Occupational therap*).tw.

13. dyspn?ea.tw.

14. (short* adj2 breath).tw.

15. (urge* adj2 breath*).tw.

16. breathless*.tw.

17. ((labo?red or difficult* or small) adj3 breath*).tw.

18. ((respirat* or breath*) adj3 (distress* or comfort* or discomfort*)).tw.

19. (air adj3 (hunger or starve* or need* or gasp* or pant*)).tw.

20. suffocat*.tw.

21. unsatisf* inspiration.tw.

22. exp Dyspnea/

23. (breathlessness intervention service or breathlessness support service).tw.

24. 9 or 10 or 11 or 12

25. 13 or 14 or 15 or 16 or 17 or 18 or 19 or 20 or 21 or 22

26. 24 and 25

27. 23 or 26

28. 8 and 27