Optimising breathlessness triggered services for older people with advanced diseases: a multicentre economic study (OPTBreathe)

Background In advanced disease, breathlessness becomes severe, increasing health services use. Breathlessness triggered services demonstrate effectiveness in trials and meta-analyses but lack health economic assessment. Methods Our economic study included a discrete choice experiment (DCE), followed by a cost-effectiveness analysis modelling. The DCE comprised face-to-face interviews with older patients with chronic breathlessness and their carers across nine UK centres. Conditional logistic regression analysis of DCE data determined the preferences (or not, indicated by negative β coefficients) for service attributes. Economic modelling estimated the costs and quality-adjusted life years (QALYs) over 5 years. Findings The DCE recruited 190 patients and 68 carers. Offering breathlessness services in person from general practitioner (GP) surgeries was not preferred (β=−0.30, 95% CI −0.40 to −0.21); hospital outpatient clinics (0.16, 0.06 to 0.25) or via home visits (0.15, 0.06 to 0.24) were preferred. Inperson services with comprehensive treatment review (0.15, 0.07 to 0.21) and holistic support (0.19, 0.07 to 0.31) were preferred to those without. Cost-effectiveness analysis found the most and the least preferred models of breathlessness services were cost-effective compared with usual care. The most preferred service had £5719 lower costs (95% CI −6043 to 5395), with 0.004 (95% CI −0.003 to 0.011) QALY benefits per patient. Uptake was higher when attributes were tailored to individual preferences (86% vs 40%). Conclusion Breathlessness services are cost-effective compared with usual care for health and social care, giving cost savings and better quality of life. Uptake of breathlessness services is higher when service attributes are individually tailored.


Sample size of the discrete choice experiment
To calculate the sample size of the DCE, we have considered the following: 1) Regression analysis needs a sample size larger than the number of independent variables. We will be able to enter up to 20 independent variables and retain up to 14, if the rules of thumb suggested by Altman apply (n/10 variables and square root of sample size respectively). 1 For each predetermined subgroup of the main sample (i.e., diagnosis group), a sample size should be larger than 30 2) Due to collecting multiple observations per participant, relatively small sample sizes are required.
In the main phase, we plan to recruit 140 patients and expect approximately 25-30% of carers of them participate as well, this gives us an estimate of 190 participants. We will record multiple observations from each participant (18 data lines per participant-3 alternatives Χ 6 choice sets). This is within the middle range of samples size of DCEs reviewed, including six in palliative care 2-7 which had 80 to 350 participants.
Thus our study will be consistent with usual best practice.

Design of choice questions
Combination of all attributes produced 864 (3 Χ 2 Χ 4 Χ 3 Χ 3 Χ 4) profiles. Pairs of profiles were to be presented to respondents. Due to practicality and cognitive burden, we reduced the number of pairs with a D-optimal design strategy, 26 after removing implausible combinations, resulting in 18 sets of choices (36 profiles). To further reduce the number of choices presented, we used block design by randomisation. Each respondent faced six choices (see Supplementary Figure S1 for an example choice) and one warm-up at the beginning which was not used in the analysis. We included a third alternative (Neither, in which case patients do what they do to manage and treat their breathlessness) in each choice, which represents the current treatment or management of breathlessness and prevents us from overestimating preferences or acceptance.
Administration of discrete choice experiment questionnaire DCE questions were administered with background, contextual information and task instructions. We gave a written vignette to respondents and read it aloud to make respondents familiar with the settings and services that BSS was suggesting. We also explain the common and/or different aspects of the BSS from the pulmonary rehabilitation service provided in NHS. We described attributes again prior to starting. Then, BMJ Publishing Group Limited (BMJ) disclaims all liability and responsibility arising from any reliance Supplemental material placed on this supplemental material which has been supplied by the author(s) there was a warm-up exercise where we asked the participants to describe their current health care service to manage breathlessness in terms of attributes (this tested their understanding of the task, attributes and levels as well as how well they knew the characteristics of the service they were on) and a warm-up example choice set with annotations. We used icons for each level in attributes, unveiling and reading aloud attributes one by one to ensure participants consider each and every one of them when making choices. The vocabulary was user-friendly, checked and improved in pre-pilot and pilot stages, asking participants to explain why they chose the option for each task to detect heuristics and need for improvements in the choice tasks.  treatments, home visits by therapists and support from a social worker. Better mobility and independence at home and outside home, and more social activities are anticipated. Fewer hospital admissions are expected, and patients need to wait 2 weeks to get the first appointment. BSS II offers two consultations with GPs or nurses at GP surgeries, reviewing medicinal treatments. There is no additional support provided. Better mobility and independence at home and fewer visits to GP surgeries are anticipated. Waiting time for the first appointment is 8 weeks. Probabilities of taking part in the BSS is derived from the discrete choice experiments data analysis by gender.  Notes: 95% confidence intervals are in the parentheses. BSS I involves the consultations with specialist at outpatient clinic, reviewing both medicinal & nonmedicinal treatments, home visits by therapists and support from a social worker. Better mobility and independence at home and outside home, and more social activities are anticipated. Fewer hospital admissions are expected, and patients need to wait 2 weeks to get the first appointment. BSS II offers two consultations with GPs or nurses at GP surgeries, reviewing medicinal treatments. There is no additional support provided. Better mobility and independence at home and fewer visits to GP surgeries are anticipated. Waiting time for the first appointment is 8 weeks. Probabilities of taking part in the BSS is derived from the discrete choice experiments data analysis by gender. Costs are in 2014 UK sterling pounds.  Figure S2. Decision tree for Markov model analysis of breathless support service Notes: Probabilities of taking Breathlessness Services as well as age-and sex-specific, all cause and respiratory mortality (see Table S3) was used in defining the transitional chances. Simulations with 10,000 replications were estimated to generate costs and outcomes per person for 5 years (20 cycles).