Options for possible changes to the blood donation service: health economics modelling

Article history

The research reported in this issue of the journal was funded by the HS&DR programme or one of its preceding programmes as project number 13/54/62. The contractual start date was in May 2015. The final report began editorial review in May 2017 and was accepted for publication in November 2017. The authors have been wholly responsible for all data collection, analysis and interpretation, and for writing up their work. The HS&DR editors and production house have tried to ensure the accuracy of the authors’ report and would like to thank the reviewers for their constructive comments on the final report document. However, they do not accept liability for damages or losses arising from material published in this report.

Declared competing interests of authors

Crispin Wickenden, Catharina Koppitz, Gavin Cho, David J Roberts and Gail Miflin are all employees of NHS Blood and Transplant. Richard Grieve is currently a member of the National Institute for Health Research Health Technology Assessment Commissioning Board.

Copyright statement

© Queen’s Printer and Controller of HMSO 2018. This work was produced by Grieve et al. under the terms of a commissioning contract issued by the Secretary of State for Health and Social Care. This issue may be freely reproduced for the purposes of private research and study and extracts (or indeed, the full report) may be included in professional journals provided that suitable acknowledgement is made and the reproduction is not associated with any form of advertising. Applications for commercial reproduction should be addressed to: NIHR Journals Library, National Institute for Health Research, Evaluation, Trials and Studies Coordinating Centre, Alpha House, University of Southampton Science Park, Southampton SO16 7NS, UK.

2018 Queen’s Printer and Controller of HMSO

Aims and objectives

The study’s aim is to identify cost-effective strategies for maintaining the blood supply. The study estimates the relative cost-effectiveness of alternative minimum interdonation intervals (12 vs. 10 vs. 8 weeks for men; 16 vs. 12 vs. 10 weeks for women). The study adopts formal methods to elicit preferences from donors using stated preference (SP) surveys to estimate the frequency at which they are willing to donate whole blood according to alternative potential changes to the blood donation service. We use these estimates from the surveys along with observed data on deferral rates from the INTERVAL trial to report the relative cost-effectiveness of alternative strategies overall, and particularly for subgroups of prime policy relevance, for example donors whose blood type is in high demand and BAME donors.

Report overview

This report details the three interlinked components of the study. Chapter 2 describes the use of the INTERVAL trial data to report the relative cost-effectiveness of alternative minimum donation intervals over 2 years. Chapter 3 reports the design, and results, of SP surveys that provide estimates of the frequency at which donors are willing to donate whole blood according to alternative future changes to the blood donation service. Chapter 4 estimates the cost-effectiveness of alternative strategies for maintaining the blood supply, drawing on findings from the surveys and the analyses of the INTERVAL trial. The design, analysis and interpretation of each component of the research have been informed by the key service provider (NHSBT), a public representative and current whole-blood donors.

At the design stage, we identified strategies to improve opportunities for existing blood donors to donate. The strategies were identified through a review of NHSBT documents describing future strategies and policies, the results of market research, an informal review of relevant published literature, consultation with NHSBT colleagues and insights from preliminary qualitative research undertaken with INTERVAL donors (see Appendix 1).

The following strategies were chosen for the evaluation presented in this report:

1. provision of a health report, including cholesterol and blood pressure tests, for all donors at each whole-blood donation visit9,43–45

2. extended opening hours (weekend or evening), for both static donor centres and mobile sessions for collecting whole blood

3. increase in the maximum number of whole-blood donations per year for current donors attending static donor centres, as per the INTERVAL trial.

Changes since the proposal

There were three main changes made to the research originally proposed. First, the cost-effectiveness analysis focused on alternative strategies for current whole-blood donors and did not consider alternative strategies for the recruitment of new donors. The NHSBT suggested that an evaluation of the effect of alternative strategies for pre-existing whole-blood donors would provide more relevant evidence to inform future strategy. Second, the approach taken to elicit donor preferences for alternative changes to the blood donation service was to undertake SP surveys rather than a discrete choice experiment (DCE). 46 The SP survey design was chosen as, unlike a DCE, it could capture respondents’ stated intentions regarding frequency of donation. Third, the time horizon chosen for the cost-effectiveness analysis was 1 year, rather than the 10 years originally proposed. This choice of time horizon was in accordance with NHSBT’s requirements, which, given the uncertainties about the future demand for red blood cells in the medium term, were to consider the shorter-term effects of the alternative strategies. Each of these changes was discussed and agreed with the project advisory group.

Introduction

This chapter will address the question, ‘what is the cost-effectiveness of reducing the minimum interval between whole-blood donations?’. This economic evaluation compares the costs and consequences of the alternative minimum donation intervals considered in the INTERVAL trial. Full details of the INTERVAL trial are provided elsewhere;42,47,48 here we focus on the essential elements of the cost-effectiveness analysis (CEA). The study followed the main principles set out in the INTERVAL trial protocol42 and statistical analysis plan. In particular, the study contrasted the costs and consequences of the alternative randomised arms in INTERVAL according to the intention-to-treat principle. 49 The main time horizon was 2 years, as per the follow-up period of the INTERVAL trial. We estimated the effect of randomisation to alternative minimum recall donation intervals on the mean number of successful whole-blood donations, overall donation deferrals, donation deferrals caused by low haemoglobin (Hb), quality of life (QoL) and cost. The study measured costs from the NHS and Personal Social Services perspectives as recommended by the National Institute for Health and Care Excellence. 50 The costs included were those costs of the blood donation visit that were anticipated to differ over the trial follow-up period and according to strategy, and included the relevant costs of blood collection but excluded processing costs or fixed costs. The cost analysis also included the costs of deferrals and any subsequent health-care costs of those deferrals that were caused by low levels of Hb.

Methods outlines the main features of the INTERVAL trial and the methods used in the CEA. Results summarises the main results, and Discussion discusses the main findings, and outlines how they will inform the subsequent CEA of alternative strategies for maintaining the blood supply beyond reducing the minimum interval between blood donations.

Methods

Selection and recruitment of participants, and exclusions from the cost-effectiveness analysis

Donors were eligible for inclusion in the trial if they were aged ≥ 18 years, met the routine criteria for whole-blood donation, were willing to be randomised, had an e-mail address and access to the internet (required to provide baseline and follow-up information), were willing to donate whole blood at a static centre, and returned the baseline questionnaire. Donors already registered at static donor centres, specific subgroups of donors attending mobile sessions who were willing to donate at a static centre for the duration of the trial, and new donors were all considered for inclusion. Those donors who were eligible to take part in the trial, and who consented, were randomised to the three sex-specific intervention groups in a 1 : 1 : 1 ratio. The CEA excluded those donors who withdrew consent for use of their data (n = 221), who died during or after the trial follow-up period (n = 142) until December 2016, when linked NHSBT national blood supply database (PULSE) data were extracted, or who did not have requisite PULSE data available (n = 37), leaving an overall sample for the CEA of 44,863. This is outlined in Figure 1.

FIGURE 1.

The CONSORT flow chart: participation, exclusions and completeness of main CEA (adapted from Di Angelantonio et al. 47).

Results

For both sexes, the baseline characteristics were similar across the randomised arms, as shown in Table 2. Overall, the mean [standard deviation (SD)] age for men and women was 44.7 years (14.2 years) and 40.9 years (14.0 years), respectively; 13% of men and 14% of women were categorised as having high-demand blood types, and 8% of men and 11% of women were classified as new donors. Over 90% of participants were self-defined as of white ethnic origin, and about 65% of the participants were recruited from static donor centres. The mean number of blood donations in the 2 years preceding the trial was 4.2 for men and 3.4 for women, with mean deferral for a low Hb level of 0.04 for men and 0.12 for women, and mean deferral for other reasons of 0.32 for men and 0.34–0.36 for women. The mean baseline QoL was 0.86 for men and 0.85 for women.

The resource use over 2 years is presented in Table 3. For men, the mean number of blood donation visits was 7.76, 6.60 and 5.68 in the 8-, 10- and 12-week arms, respectively, and for women the corresponding average number of visits was 5.10, 4.60 and 4.01 in the 12-, 14- and 16-week arms, respectively. The average rate of deferral for low levels of Hb, per session attended, was higher in the arms that had shorter minimum donation intervals. For men, this deferral rate was 5.71% in the 8-week arm compared with 3.73% in the 10-week arm and 2.55% in the 12-week arm. For women, this deferral rate increased from 5.05% (16-week arm) to 6.63% (14-week arm) and 7.92% (12-week arm). The corresponding mean numbers of Hb-related deferrals per donor over 2 years were also higher in the randomised arms with reduced donation intervals. The proportion of deferrals caused by other reasons, the average number of fainting episodes (see Table 3) and other donor-reported health-care events (reported in Table 4) were similar across the randomised arms.

The main cost-effectiveness results are presented in Table 5. For both sexes, the average SF-6D utility scores were similar across the randomised arms at the 2-year follow-up and at each of the intervening time points (see Appendix 5). For men, the average number of whole-blood donations over the 2-year follow-up period increased by 1.71 [95% confidence interval (CI) 1.60 to 1.80] for the 8- versus 12-week interval arm, and by 0.79 (95% CI 0.70 to 0.88) for the 10- versus 12-week interval arm. For women, the corresponding increase in the average number of donations was 0.85 (95% CI 0.78 to 0.92) for 12 versus 16 weeks, and 0.46 (95% CI 0.40 to 0.53) for 14 versus 16 weeks. The average costs per donor over 2 years increased for each of the reduced interval strategies. The ICERs were £9.51 (95% CI £9.33 to £9.69) for the 8- versus 12-week interval arm for men, and £10.17 (95% CI £9.80 to £10.54) for the 12- versus 16-week interval arm for women. Figure 2 shows that when the cost-effectiveness results are plotted on the cost-effectiveness plane, the distributions of the mean costs and mean number of donations are centred tightly around the means.

FIGURE 2.

Uncertainty in the incremental costs (£) and number of whole-blood donations, and their joint distribution, for reduced interval strategies vs. standard practice (control arm) over 2 years’ follow-up. (a) Male and (b) female.

The likelihood ratio test shows that the inclusion of interaction effect for subgroups by randomised group improved model fit (male: χ² = 79.28, p = 0.0002; female: χ² = 46.55, p = 0.0153). The subgroup results reported in Table 6 show that the ICERs were similar across almost all subgroups. The main exception was for the 14- versus 16-week contrast for women whose ethnicity was defined as black/mixed black. For this subgroup, the incremental effect of the reduced interval on the number of whole-blood donations was small, and so the accompanying ICER was large (> £200). However, the sample size for this subgroup is low, and the estimates are somewhat unstable (n = 330 across all three arms).

The sensitivity analysis found that the estimates of incremental cost-effectiveness were generally similar when alternative assumptions were taken to those in the base-case analysis, shown in Table 7. The base-case results were most sensitive to the inclusion of the additional staff costs required if the donor centres had no capacity to collect the additional units of blood. Under this scenario, the ICERs increased to £26.59 (95% CI £26.41 to £26.77) for the 8- versus 12-week interval arm for men, and to £27.25 (95% CI £26.88 to £27.62) for the 12- versus 16-week interval arm for women, compared with the base-case ICERs of £9.51 (95% CI £9.33 to £9.69) and £10.17 (95% CI £9.80 to £10.54), respectively.

Discussion

The main finding from this trial-based CEA undertaken in 25 static donor centres was that, compared with the control arm strategy, which was the current minimum interval specified by the NHSBT, the reduced minimum donation interval strategies increased the average number of donations, at a small additional average variable cost over 2 years. The rate of deferral because of low levels of Hb and the average number of deferrals per donor was higher for the reduced minimum interval strategies. The main finding was that reducing the minimum recall interval yields additional units of blood donated at an additional average variable cost of around £10. This finding was generally similar across the subgroups considered, including those with high-demand blood types, among whom it is particularly important to increase the volume of blood supplied. The time horizon was limited to 2 years in accordance with the follow-up period in the INTERVAL trial. The CEA did not consider the effects that the increased rates of deferrals may have on the rate at which donors leave the donation register, and any additional costs from, for example, recruiting new donors.

The main trial analysis reported that reducing donation intervals resulted in a higher proportion of donors (especially male) reporting symptoms such as fatigue, potentially as a result of iron deficiency or blood donation, but did not find any evidence of an effect on randomised arm according to the physical or mental summary score of the SF-36. 47 The CEA also found that donors’ QoL, measured according to the SF-6D utility score, was similar across arms for all time points concerned. There were no differences in the self-reported fainting episodes or adverse events across the arms. The CEA found that the reduced minimum donation intervals did not lead to an increase in health-care resource use, or in morbidity among donors. The SF-6D is a recommended measurement of health utility, an appropriate approach was taken to handling missing data, and the average utility scores were similar to those for the age- and sex-matched general population (0.81 for men and 0.79 for women). 55 Unlike other generic instruments, such as the EuroQol 5-dimension questionnaire, the SF-6D does include dimensions for vitality and fatigue, but there is still no guarantee that it is sensitive to the small differences in minor symptoms, such as dizziness or restless leg syndrome, reported across randomised arms in the main INTERVAL trial analysis. 47

The INTERVAL trial protocol specified a comprehensive system of reminders to donors to make and keep appointments to donate blood. The trial protocol therefore implied some additional resource use versus routine NHSBT practice, but as this reminder system may have been important in encouraging donors to attend donation sessions according to the frequency observed during the RCT, the costs of reminders for sessions attended were according to the trial protocol. Indeed, a non-randomised comparison of the donation frequency pre versus post randomisation suggests that, for both sexes, the donation frequency for the control arm increased by around 40%, which could reflect a ‘trial effect’ or other temporal differences beyond the INTERVAL trial. As part of an extension to the INTERVAL trial (Phase II), approximately 50% of INTERVAL donors agreed to continue on their previously allocated study donation interval for a further period of 6–24 months, and also to be randomised to receive the enhanced trial protocol reminders or the reminders undertaken as part of routine NHSBT practice. It would be useful to extend the economic analysis undertaken here to examine whether or not the more intensive reminder system undertaken as part of the original INTERVAL trial was in itself cost-effective. 56,57

A strength of the CEA is that it used data from a large, well-conducted RCT with complete follow-up data for the main end points of interest, and included as a control arm the current minimum donation interval in England. The large sample size meant that it was possible to report both the overall effect of alternative minimum donation intervals (on frequency of attendance to donate whole blood and deferrals) and the effect according to subgroups of key policy relevance. The trial provided an excellent vehicle for estimating the rate of deferral per donation on attendance according to alternative donor characteristics. In particular, the INTERVAL trial analysis found that, for 7% of Hb-related deferrals, the level of Hb was sufficiently low to imply additional health-care consultations and costs. These estimates of deferral rates will be used in Chapter 3 in adjusting the estimates from the stated preference survey to predict actual rates of successful donation. The estimated deferral rates will also be used in the subsequent economic evaluation of a wider range of strategies to increase donation frequency (see Chapters 3 and 4).

The economic evaluation of the INTERVAL strategies has the following limitations. First, although the INTERVAL trial followed donors for 2 years, the higher deferral rates from low levels of Hb reported in the reduced interval arms could lead to a higher rate of donors leaving the blood donation registry in the long run. This is plausible if the levels of Hb, which were on average lower in the reduced interval arms after 2 years, continue to diverge. 47 The Phase II INTERVAL extension study will provide some additional evidence about the relative rate at which donors leave the register, and the potential higher costs of replacing them with new donors. Second, the RCT was undertaken at 25 static donor centres; it is unclear whether or not it would be cost-effective to roll out the reduced interval strategy to mobile sessions. Third, the CEA did not include the full range of costs that may differ according to the minimum recall interval. In particular, data were not available on the number of non-attendances for each individual. In the sensitivity analysis, when we approximated these costs, we found that the ICERs of the reduced interval strategies increased somewhat, but generally remained below an additional variable cost of £30 for an additional unit of blood donated. The results were most sensitive to the assumption that there would be sufficient capacity within the static donor centres to collect the additional units of blood donated. This assumption may not be realistic if this strategy is rolled out to all donors attending static centres. However, if the reduced interval strategies are applied only to those groups whose blood type is in high demand, then current capacity (on average, 75%) may be sufficient to collect the additional units of blood; in which case the base-case ICER is more relevant (around £10 per additional unit of blood collected).

The INTERVAL trial considered only a single set of strategies for maintaining the future blood supply, and yet NHSBT may consider these strategies in conjunction with other changes to the blood service (e.g. extending opening hours at donor centres). Hence, studies that complement the INTERVAL trial in investigating the costs and consequences of other strategies for maintaining the supply of whole blood are required. Chapter 3 reports on surveys of non-INTERVAL but also ex-INTERVAL donors that were undertaken to elicit their preferences for alternative blood donation strategies (including reduced donation intervals). The subsequent economic evaluation then uses estimates from the survey, combined with estimates from INTERVAL for deferral rates, to estimate the cost-effectiveness of a range of strategies for maintaining the blood supply.

Introduction

This chapter reports the design and analysis of the SP surveys. Thus, it is concerned with our second objective: to investigate the frequency with which donors are willing to donate whole blood according to alternative future changes to the blood collection service.

This chapter has the following structure. Development of the stated preference survey describes the development of the SP survey, in particular emphasising the key considerations that drove our design choices. Selection of attributes and levels outlines the process by which the attributes and their levels were selected. Sample size and efficient design details the design of the survey, including the number and allocation of SP questions to participants. Administration of the surveys describes the administration of the surveys, which comprised a pilot survey, a survey of blood donors who had not participated in the INTERVAL trial (non-INTERVAL) and a survey of INTERVAL trial participants (ex-INTERVAL). Analysis of annual frequency of donation from the stated preference surveys describes the analysis of the intended annual frequency of donation. Survey respondents reports response rates to the non-INTERVAL and ex-INTERVAL surveys and describes the respondents in terms of a number of characteristics. Results reports the results of the analysis of the SP surveys and the chapter concludes with Discussion.

Development of the stated preference survey

Five main considerations influenced the design of the SP survey. First, the purpose of the survey was to help inform future changes to the blood collection service. Hence, the attributes chosen were those judged to reflect aspects of the blood collection service that the NHSBT could change in the short term. This has important implications for the selection of attributes, specifically attributes that are under the control of the NHSBT. Second, there were separate male and female questionnaires because of well-established differences between men and women in permitted donation frequency, and our explicit interest in the strategies contrasted in the INTERVAL trial. Third, the donation scenarios presented to donors distinguished between the last place (LP) the respondent donated and a different place (DP). This is necessary because some changes in the opportunity to donate would involve the donor donating in a DP, whereas for others the venue would not change. Fourth, we chose an unlabelled design, which in this context implies that an opportunity to donate was not indicated as being at a static donor centre or a mobile session. Fifth, our interest lay in making predictions regarding frequency of donation by existing blood donors. We were not concerned with the broader question of what factors influence decisions to become a donor or to stop donating.

Donors donate at different annual rates depending on a combination of their personal characteristics and on the opportunities that they have to donate. In effect, we are assuming that donors consider the marginal costs and benefits of donating blood. Changes that increase the cost to the donor of donating (e.g. increased travel time) will tend, other things being equal, to reduce the frequency with which they donate. However, changes that increase the benefits to donors of donating (e.g. provision of a health report) will increase their preferred frequency of donation.

The maximum number of donations allowed annually depends on the donor’s sex. Some donors may be unconstrained in that they are already donating at their preferred frequency, whereas others may be constrained by this maximum in the sense that they would like to donate more frequently. Changes to the maximum frequency with which donors can donate will not change the costs and benefits of donating to an individual donor. However, decreases in the interval required between donations will result in previously constrained donors being able to donate more often. More speculatively, an increased permitted frequency of donation might alter donors’ perceptions regarding a donation norm, which may encourage some donors to donate more frequently.

Selection of attributes and levels

The selection of attributes and levels always involves compromises given limits on the feasible number of scenarios about which to ask respondents. Our choice of attributes to describe the opportunities to donate was informed by a rapid literature review, input from policy-makers at the NHSBT and preliminary findings from qualitative research with blood donors participating in the INTERVAL trial [R Lynch and S Cohn, London School of Hygiene & Tropical Medicine (LSHTM), personal communication, October 2015]. The pilot study explored five attributes identified as pertinent to determining the impact of a number of policy-relevant strategies, namely travel time, opening hours, total donation time, provision of a health report and the maximum number of donations per year. The first three attributes influence the cost to the donor of the blood donation, whereas the provision of a health report potentially increases the benefits to the donor. The INTERVAL trial is investigating the safety of increasing the maximum frequency of donation (three or four times per year for women and 4–6 times for men). This attribute is included to understand how donors might respond if the limits were altered alongside the other possible future changes to the blood service. As noted earlier, this does not affect the costs or benefits of donation but rather the scope for donors to achieve their preferred donation frequency.

Donation venue was not included as an attribute, but the survey had two sections: the first asks the donor to think about donation opportunities at the ‘last place you gave blood’ (LP), and the second asks the donor to ‘imagine you were asked to donate at a different place’ (DP). This was in order to capture more of the context not included in the attributes (e.g. the community aspect of blood donation and familiarity of staff) that had been identified as important to some donors in the qualitative research conducted as part of the INTERVAL trial.

The appropriate levels for each attribute were defined according to summary estimates from the PULSE database, NHSBT market research and consultation with blood donors. Table 8 shows the attributes and levels used in the non-INTERVAL and ex-INTERVAL surveys (see Appendix 6 for those used in the pilot survey).

The analysis of the pilot survey results revealed a systematic over-prediction of the frequency of blood donation compared with the donation frequency observed in practice. One potential explanation, and a view expressed at our donor workshop, was that some donors might wish to donate more frequently but face additional constraints in practice, which cause observed behaviour to diverge from that predicted. Our analysis of these discrepancies (between the predicted donation frequencies and those observed) showed that these were consistent across various subgroups. 58 For the main survey of non-INTERVAL donors, we replaced the time taken to make the donation with an appointment availability attribute. The appointment availability attribute is a means of ensuring that the SP scenarios explicitly address differences in appointment availability. Two further changes from the pilot survey were made: we reduced the number of levels for the opening times attribute from eight to four and adopted a full factorial design in place of a main effects fractional design. The reduction in opening hours levels was made in order to increase the feasibility of a full factorial design and to facilitate analysis of preference differences between subgroups of donors.

For each set of attribute levels, we asked donors to state the frequency with which they would be willing to donate blood. The survey included the option ‘I would probably not donate’. Figure 3 gives an example of a typical question.

FIGURE 3.

Example of a question from the stated preference survey.

Sample size and efficient design

The purpose of the pilot survey was threefold: (1) to ensure that our overall systems for selecting and inviting donors to participate, and for recording their responses, worked as intended; (2) to obtain reliable information on the likely response rate; and (3) to identify any issues with respect to individual questions and the questionnaire as a whole.

For the pilot study, the NHSBT issued 5016 e-mail invitations to eligible donors. We had anticipated a response rate between 10% and 20% (based on previous NHSBT surveys); however, a response rate of 25% was achieved. The mean time to complete the whole survey was just under 6 minutes (5 minutes and 47 seconds). No calls relating to our pilot survey were logged at NHSBT’s national call centre.

The pilot had a main effects design because to do otherwise was infeasible given the sample size. A full factorial design for men would have involved 96 possible LP (1¹ × 2² × 3¹ × 8¹) and 384 possible DP scenarios (2² × 3¹ × 4¹ × 8¹). For women, 64 possible LP (1² × 2³ × 8¹) and 256 possible DP scenarios (2³ × 4¹ × 8¹) would have been required. An efficient design was adopted based on the need to estimate the marginal rate of substitution between attributes with reasonable precision (main effects only). Ngene^TM 1.1.2 (Choice Metrics Pty Ltd, Sydney, Australia) was used to establish an efficient design by considering each section of the survey as one choice set compared with an ‘opt-out’, resulting in eight LP and 12 DP scenarios for women (12 versions of the survey) and 24 LP and 24 DP scenarios for men (72 versions of the survey). Reflecting the greater number of male scenarios, two men were invited for every women in the pilot survey.

Sample size calculations for SP surveys are not straightforward, and although the response rate in the pilot (given its size) was informative, the likely response rate to the main survey remained uncertain. A sample size of 100,000 was chosen for the main non-INTERVAL survey. This sample size took account of the successful experience from the pilot and the practical challenges associated with inviting larger numbers. The main survey, being much larger, provided an opportunity to adopt a full factorial design. It appeared likely that there would be significant interaction effects; for example, a donor’s willingness to travel might be influenced by the opening hours and appointment availability. Taking into account the changes following the pilot survey with respect to attributes and levels, there are 96 potential LP scenarios for men (4² × 3 × 2) and 384 potential DP scenarios (4³ × 3 × 2). The LP scenarios can be divided into 48 blocks of two, and the DP scenarios into 96 blocks of four. For women, the 64 LP scenarios (4² × 2²) and 256 DP scenarios (4³ × 2²) can be divided into 32 blocks of two and 64 blocks of four, respectively. As in the pilot survey, two men were invited for every women in the non-INTERVAL survey, because there were many more possible scenarios in the male survey.

Administration of the surveys

The non-INTERVAL survey received ethics approval from the NHS (reference number 16/YH/0023) and LSHTM (reference number 10384) Research Ethics Committees in November 2015 and January 2016, respectively, and non-substantial amendments were made following the pilot survey in May 2016. For the final protocol for the survey of non-INTERVAL donors, see Report Supplementary Material 1. A total of 100,000 donors were randomly selected from the PULSE database to be invited to participate in the survey of non-INTERVAL donors, according to the following criteria: 17–70 years old, donation of at least one unit of whole blood in the past 12 months, e-mail address held by NHSBT and residence in mainland England. Donors were excluded from the surveys if they were temporarily suspended from giving blood (e.g. donors who had recently had a tattoo), had previously stated that they did not want to participate in surveys or had received a request to participate in a survey or research from the NHSBT (including the INTERVAL trial) in the preceding 6 months. Owing to the recent NHSBT communication policy, women with AB-positive (AB+) blood were also excluded. Selected donors were sent an e-mail invitation from the NHSBT with a link to the online survey built and hosted on FluidSurveys^TM (1 January 2015 version; SurveyMonkey^®, Palo Alto, CA, USA). For the e-mail invitation sent to non-INTERVAL donors, see Report Supplementary Material 2, and for the consent form and donor information sheet, see Report Supplementary Material 3. Donors who did not complete the survey (excluding those who refused consent) within the first 3 days were sent a reminder e-mail, and the survey closed 3 days after the reminder e-mail. Figure 4 provides a CONSORT-style diagram for the survey.

FIGURE 4.

A CONSORT-style diagram stated preference survey of non-INTERVAL participants. a, Complete PULSE data are not available for 24 of these donors; b, complete PULSE data are not available for 11 of these donors.

To assess the generalisability of the survey results, the characteristics of the donors who completed the survey were compared with a larger PULSE sample of donors who have donated at least once in the 12 months prior to March 2016.

We repeated the same SP survey with the ex-INTERVAL participants using similar exclusions to the previous surveys, but this time did not exclude donors if they had not given blood in the 12 months prior to the survey (in December 2016). For the final protocol of the survey of ex-INTERVAL donors see Report Supplementary Material 4. The ex-INTERVAL participants are of particular interest because many of them have experienced the higher rates of donation featured in some of the SP scenarios. Moreover, all INTERVAL trial donations were made at donor centres and the non-INTERVAL survey had relatively small numbers whose last donation took place in a donor centre (n = 2970), reflecting that most blood in England is collected at mobile sessions. All eligible ex-INTERVAL participants were invited, by e-mail, to participate in the survey. For the ex-INTERVAL donors the invitation e-mail and consent and donor information sheets are available in Report Supplementary Material 5 and 6, respectively. Figure 5 provides a CONSORT-style diagram for the ex-INTERVAL survey.

FIGURE 5.

A CONSORT-style diagram survey of ex-INTERVAL participants. a, Complete PULSE data are not available for 32 of these donors; b, complete PULSE data are not available for one of these donors.

Examples of the non-INTERVAL (which are the same as those administered to the ex-INTERVAL sample) and pilot surveys are reproduced in Appendix 7. In addition to the SP questions, all respondents to the pilot, non-INTERVAL and ex-INTERVAL surveys were asked to complete a series of background questions (see Appendix 8). These included questions about how often the donor recalled giving blood in the past 12 months and how often they wanted to give blood. Regarding the last donation visit, questions related to from where they travelled, how they travelled, how far they travelled, how long it took, how they made the appointment and how easy it was to get a suitable appointment. Finally, they were asked whether their total visit duration was < 1 hour or > 1 hour.

Analysis of annual frequency of donation from the stated preference surveys

The primary purpose of the SP surveys was to predict the average number of whole-blood donations per year given different opportunities to donate, for example with respect to opening hours and travel time. These predictions were central to the assessment of the cost-effectiveness analysis of alternative NHSBT strategies (see Chapter 4).

We chose to use an ordered logit model for the base-case analysis to recognise that the response variable was categorical, but also had a natural ordering (three times per year, two times per year, etc.). The sensitivity analyses examined whether or not the findings were sensitive to the choice of model by considering two alternative models: the two-part model, which recognises that the response ‘I would probably not donate’ is a different kind of response from once, twice or three times per year; and a gamma model, which treats the responses as continuous, but with a lower bound at zero.

The SP surveys for male and female donors had different levels for the minimum donation interval attribute, and so we analysed the responses for each sex separately. We investigated whether or not there was effect modification according to donor subgroup. We estimated two-way interactions of each attribute with the following subgroups: age, blood type, ethnicity and venue for last whole-blood donation (see definitions in Survey respondents). These estimated two-way interactions were used to report results according to donor subgroup.

We used the estimated coefficients from the ordered logit model to predict each individual’s donation frequency according to the alternative levels for each attribute. This was done in two stages. First, a predicted probability for each donor for each category of response variable was calculated from the coefficients of the ordered logit regression model. Second, for each level of each attribute, the expected annual donation frequency for each donor was calculated using the predicted probabilities estimated in the first stage for each response category. The average of these predicted annual donation frequencies according to each attribute level was reported, together with the 95% uncertainty intervals, to reflect the variation in these predictions across the survey sample. We also used the estimated coefficients from the ordered logit model to report the proportion of donors who said that they ‘probably would not donate’ according to each attribute level.

To assess the concern that donors may overstate their donation frequency, we then compared each individual’s predicted annual frequency with their actual donation frequency (recorded in the PULSE database) over the year preceding the survey. Annual donation frequency was predicted by setting attribute levels to represent the donor’s most recent experience, as recorded in PULSE in March 2016.

The algorithm used to determine the current service-level characteristics is shown in Figure 6. We then adjusted these predictions using the predicted probability of deferral from the INTERVAL trial arms assigned to the current minimum donation intervals (see Chapter 2). The average discrepancy between the predicted donation frequency and observed donation frequency was reported overall, for men and women, and for each subgroup, with 95% CIs to reflect the uncertainty in the predictions across individuals. However, these CIs do not reflect uncertainty in the estimation of the models used to predict intended donation frequency or deferral rates.

FIGURE 6.

Attributing current service-level characteristics.

Although the surveys were designed to facilitate the prediction of donation frequencies for a range of donation opportunities rather than in order to understand the factors which lead donors to continue or discontinue donating, one of the potential responses was ‘I would probably not donate’. This response can give some insight into, for example, whether increasing travel time might lead more donors to discontinue donation. We report the use of this response category by subgroup.

The non-INTERVAL sample and the eligible ex-INTERVAL participants were invited to complete exactly the same questionnaire. The preferences of the ex-INTERVAL respondents who had been in the intervention arms of INTERVAL may well differ from those of the non-INTERVAL sample, partly because of the former group having experienced shorter donation intervals. However, we would anticipate the preferences of the non-INTERVAL sample whose last donation was at a static donor centre to be more similar to those of ex-INTERVAL control arm participants. They might still differ by virtue of the ex-INTERVAL participants all having expressed a willingness to be randomised to arms with more frequent invitations to donate. If the preferences of the two groups are sufficiently similar, more precise predictions regarding future donation behaviour can be obtained by pooling the two samples.

Survey respondents

The overall response rates were 25.2% for the non-INTERVAL survey and 32.4% for the ex-INTERVAL survey (see Figures 4 and 5). Table 9 compares four groups of male whole-blood donors: respondents to the non-INTERVAL survey, invitees to the non-INTERVAL survey, eligible donors and all donors. These comparisons are made with respect to five donor characteristics: age (17–30, 31–45, 46–60 and ≥ 60 years), blood type (distinguishing high and standard demand), ethnicity [distinguishing white, black/mixed black, Asian/mixed Asian and not stated (for detail on how these are defined, see Appendix 3)], whether or not a nursery donor, and the venue where the donor last gave blood (static donor centre or mobile session). High-demand blood types are O–, A– and B–, and standard-demand blood types are O positive (O+), A positive (A+), B positive (B+), AB+ and AB negative (AB–). A nursery donor is defined as a donor who has given blood 1–4 times in the past 5 years, whereas those donating five times or more in the last 5 years are not nursery donors.

As would be expected, the breakdown of eligible male donors differs from all male donors because of the exclusions applied. The distribution of those invited by age, etc., as should be the case, closely follows that of the eligible donors. Comparing the characteristics of the respondents with those of the invited donors, older donors and those indicating white ethnicity are over-represented, and nursery donors, donors making a single donation in the last 12 months and donors who last donated at a static centre are under-represented. The patterns observed for male donors are repeated for female donors, shown in Table 10.

Tables 11 and 12, men and women, respectively, provide a comparison of ex-INTERVAL respondents, eligible ex-INTERVAL participants (all of whom were invited) and all those assessed for eligibility, using the same donor characteristics. For both men and women, nursery donors and donors with zero donations in the last 12 months are under-represented in those eligible for the SP survey. As can be seen for both men and women, respondents differ from those invited in that younger (older) donors are under-represented (over-represented) and nursery donors and those who made no or only one donation in the last 12 months are under-represented.

Results

The full results for the ordered logit models are presented in additional material available on the project website. The incremental effects of changes in the levels of the different attributes were estimated using the separate ordered logit models estimated for non-INTERVAL men and women and ex-INTERVAL men and womem. The predicted changes in the annual frequency of donation are shown in Figures 7–18. Figure 7 reports the results for the full male and full female samples of non-INTERVAL donors. Figures 8–11 report the results for four non-INTERVAL subgroups, namely 17- to 30-year-olds, high-demand blood types, black/mixed black donors and nursery donors. Figure 12 reports the results for the full male and full female samples of ex-INTERVAL donors. Figures 13–16 report the results for four ex-INTERVAL subgroups, namely 17- to 30-year-olds, high-demand blood types, black/mixed black donors and nursery donors.

FIGURE 7.

Incremental effect of alternative types of blood service on average number of donations per year for all non-INTERVAL respondents. For men (• n = 15,652) and women (▴ n = 8329). The uncertainty intervals reflect the 5th and 9th percentiles in the predicted donation frequency across the relevant target population.

FIGURE 8.

Incremental effect of alternative types of blood service on average number of donations per year for non-INTERVAL respondents aged 17–30 years. For men (• n = 1646) and women (▴ n = 1663). The uncertainty intervals reflect the 5th and 95th percentiles in the predicted donation frequency across the relevant target population.

FIGURE 9.

Incremental effect of alternative types of blood service on average number of donations per year for non-INTERVAL high-demand blood type respondent. For men (• n = 1551) and women (▴ n = 921). The uncertainty intervals reflect the 5th and 95th percentiles in the predicted donation frequency across the relevant target population.

FIGURE 10.

Incremental effect of alternative types of blood service on average number of donations per year for non-INTERVAL respondents of black/mixed black ethnicity. For men (• n = 98) and women (▴ n = 103). The uncertainty intervals reflect the 5th and 95th percentiles in the predicted donation frequency across the relevant target population.

FIGURE 11.

Incremental effect of alternative types of blood service on average number of donations per year for non-INTERVAL nursery respondents. For men (• n = 3542) and women (▴ n = 3024). The uncertainty intervals reflect the 5th and 95th percentiles in the predicted donation frequency across the relevant target population.

FIGURE 12.

Incremental effect of alternative types of blood service on average number of donations per year for all ex-INTERVAL respondents. For men (• n = 4754) and women (▴ n = 4179). The uncertainty intervals reflect the 5th and 95th percentiles in the predicted donation frequency across the relevant target population.

FIGURE 13.

Incremental effect of alternative types of blood service on average number of donations per year for ex-INTERVAL respondents aged 17–30 years. For men (• n = 253) and women (▴ n = 431). The uncertainty intervals reflect the 5th and 95th percentiles in the predicted donation frequency across the relevant target population.

FIGURE 14.

Incremental effect of alternative types of blood service on average number of donations per year for ex-INTERVAL high-demand blood type respondents. For men (• n = 643) and women (▴ n = 580). The uncertainty intervals reflect the 5th and 95th percentiles in the predicted donation frequency across the relevant target population.

FIGURE 15.

Incremental effect of alternative types of blood service on average number of donations per year for ex-INTERVAL respondents of black/mixed black ethnicity. For men (• n = 39) and women (▴ n = 52). The uncertainty intervals reflect the 5th and 95th percentiles in the predicted donation frequency across the relevant target population.

FIGURE 16.

Incremental effect of alternative types of blood service on average number of donations per year for ex-INTERVAL nursery respondents. For men (• n = 18) and women (▴ n = 65). The uncertainty intervals reflect the 5th and 95th percentiles in the predicted donation frequency across the relevant target population.

These figures have a common structure. Throughout, • refers to men and ▴ refers to women. Consider, for example, Figure 7. The first row describes the mean (95% uncertainty intervals) under the current service configuration (status quo). All subsequent rows report the average predicted annual donation frequency according to the specific attribute and level in question, and in the absence of any other changes in the opportunities to donate. So in the figures the next three rows indicate the predicted effect of changes in travel time. Thus, for example, if travel time were to increase by 30 minutes, it is predicted that men would, on average, make 1.4 fewer donations per year and women would make 1.1 fewer donations per year. Similarly, if opening hours were to change from the status quo to 2–8 p.m., the model predicts that men would provide 0.5 more donations per year and women 0.3 more donations per year.

There are four important general findings. First, the estimated incremental effects are very much in line with prior expectations. Thus, increases in travel time are predicted to reduce the annual number of donations. Extending evening opening hours is predicted to increase the annual number of donations. Improved availability of appointments is predicted to increase donations, whereas reduced availability of appointments is predicted to decrease the number of donations. Introduction of a health report is predicted to increase the number of donations. Reducing the minimum interval between donations (increasing the maximum number of donations permitted annually) is predicted to increase the number of donations. Second, although there are some subgroup differences, the similarities in the results across subgroups are more striking than the differences. Thus, for example, introduction of a health report is associated with a relatively small increase in frequency of donation for all subgroups. For all subgroups the greatest impact is with respect to increases in travel time. Reducing the interval between donations such that annual opportunities to donate increase by 1 is predicted to increase annual donations on average by about 0.5. Third, although there are differences in incremental effects between men and women, they are generally not large. The greatest differences are with respect to increased travel time and reduced appointment availability. Fourth, these general characterisations of the results apply to both the non-INTERVAL and ex-INTERVAL donors, although, interestingly, the predicted effect of reducing minimum interdonation intervals is about two-thirds greater for ex-INTERVAL donors than for non-INTERVAL donors.

Tables 13–16 report, for the different attribute levels, the predicted percentage of respondents from the ordered logit model who indicate ‘I would probably not donate’. This is done for all attribute levels and by the levels of donor characteristics. The baseline represents the individual-level service configuration the donor experienced at their last donation. Tables 13–16 present this information for male non-INTERVAL, female non-INTERVAL, male ex-INTERVAL and female ex-INTERVAL respondents, respectively. Increased travel time is associated with markedly higher proportions of ‘stop donating’ responses. Opportunities to donate in the evening and greater availability of appointments are associated with lower proportions of ‘stop donating’ responses. Provision of a health report is associated with a slightly lower proportion of ‘stop donating’ responses. Finally, increases in the maximum number of donations permitted annually are associated with lower proportions of ‘stop donating’ responses.

Several patterns by subgroup can be observed in these data. The older the age group, the less frequently they give the ‘I would probably not donate’ response. Generally, men are less likely than women to say that they would stop donating. Although the difference is not marked, high-demand blood type respondents are less likely than standard-demand blood type respondents to say that they will stop donating. For all attribute levels, the proportion of ‘stop donating’ responses by nursery donors exceeds that of non-nursery donors. Among non-INTERVAL donors, those whose last donation was made at a donor centre were less likely to indicate that they would stop donating than those who last donated at a mobile session. Among non-INTERVAL donors, there is a tendency for black/mixed black donors to give a ‘stop donating’ response more often than other ethnic groups. Among ex-INTERVAL donors, there is a tendency for Asian/mixed Asian donors to give a ‘stop donating’ response more often than other ethnic groups. More frequent donation in the past 12 months is associated with lower percentages of ‘stop donating’ responses for men and women in both the non-INTERVAL and ex-INTERVAL surveys.

There is a fairly mixed picture when comparing the percentages of ‘stop donating’ responses across ex-INTERVAL and non-INTERVAL donors, except that ex-INTERVAL nursery donors consistently make fewer ‘stop donating’ responses than non-INTERVAL nursery donors.

Figures 17 (men) and 18 (women) compare the incremental effects on annual donations for the non-INTERVAL donors whose last donation was at a static donor centre, and for ex-INTERVAL donors who were in the control arm (thus not exposed to an increase in the maximum number of donations permitted annually). Figure 17 shows similar patterns for the two groups; however, non-INTERVAL respondents are more responsive to increased travel time, and ex-INTERVAL respondents are more responsive to increases in the maximum permitted number of donations. The results for women follow a similar pattern to the male results but, generally, any differences between the two groups are not statistically significant, possibly reflecting the smaller numbers of respondents.

FIGURE 17.

Incremental effect of alternative types of blood service on average number of donations per year for male, high-demand blood type non-INTERVAL respondents who donated at a static centre and male ex-INTERVAL control arm respondents. For men, high-demand blood type non-INTERVAL respondents who donated at a static centre (• n = 137) and male ex-INTERVAL control arm respondents (• n = 211).

FIGURE 18.

Incremental effect of alternative types of blood service on average number of donations per year for female, high-demand blood type non-INTERVAL respondents who last donated at a static centre and female ex-INTERVAL control arm respondents. For women, high-demand blood type non-INTERVAL respondents who last donated at a static centre (▴ n = 78) and female ex-INTERVAL control arm respondents (▴ n = 78).

The discrepancy between mean ‘stated’ and mean ‘observed’ donations is shown for non-INTERVAL respondents in Figures 19 (men) and 20 (women). ‘Stated’ consistently exceeds ‘observed’ for men by nearly 20%, whereas the discrepancy for women was negligible. The comparable discrepancies for ex-INTERVAL respondents, shown in Figures 21 and 22, indicate stated donations as less than observed donations for both men (< 10%) and women (about 15%).

FIGURE 19.

Discrepancy between the mean number of ‘observed’ and ‘stated’ donations per year for male non-INTERVAL respondents. The error bars reflect the 95% CI around the average discrepancy in the predicted vs. observed number of donations per year.

FIGURE 20.

Discrepancy between the mean number of ‘observed’ and ‘stated’ donations per year for female non-INTERVAL respondents. The error bars reflect the 95% CI around the average discrepancy in the predicted vs. observed number of donations per year.

FIGURE 21.

Discrepancy between the mean number of ‘observed’ and ‘stated’ donations per year for male ex-INTERVAL respondents. The error bars reflect the 95% CI around the average discrepancy in the predicted vs. observed number of donations per year.

FIGURE 22.

Discrepancy between the mean number of ‘observed’ and ‘stated’ donations per year for female ex-INTERVAL respondents. The error bars reflect the 95% CI around the average discrepancy in the predicted vs. observed number of donations per year.

The incremental effects of changing attribute levels were estimated for all donors, and for high-demand blood type donors, using a gamma model and a two-part model (for both the non-INTERVAL and ex-INTERVAL respondents). The results are shown in Appendices 9–16. The overall pattern of incremental effects is very similar to that reported in the Figures 7, 9, 12 and 14 (based on the ordered logit model), respectively. The only systematic difference is that the estimated response to reductions in the minimum donation interval is slightly reduced when estimated by the gamma and two-part models as compared with the ordered logit model.

Discussion

The SP surveys provide plausible estimates of the effects of alternative future changes to the blood collection service on the stated frequency of donation. We find that reducing the minimum recall period between whole-blood donations, or changing appointment availability to include weekends or evenings, leads to moderately large increases in the stated frequency of blood donation. By contrast, the introduction of a donor health report is predicted to have a relatively small effect on the stated frequency. These findings are in line with our a priori hypotheses, and with feedback from whole-blood donors attending the translation workshops (see Appendix 17). A general concern with the use of SP surveys for service evaluation is that there may be ‘hypothetical bias’; responders tend to overstate their willingness to provide a publicly funded service with altruistic features. 59 However, we found that for men who completed the non-INTERVAL survey, the average annual frequency of donation from the survey responses was only moderately higher than the actual donation frequency, for women the discrepancy was minimal, and for both sexes the discrepancy was similar across subgroups. These findings applied to the current blood service characteristics experienced by the survey responders. This provides a reasonable basis for the subsequent CEA to assume that the relative estimates of service changes predicted from the survey responses predict the effect of future changes to the blood collection service on donation frequency.

A potentially important finding from both surveys is that donors are willing to donate more often if there are more frequent opportunities to donate, in particular if the minimum interval between invitations is reduced. This implies that the overall findings from the INTERVAL trial could apply to donors attending static centres outwith the trial setting. The survey results also extend the trial findings in showing that the reduced interval strategies are predicted to increase donation frequency when considered against, or alongside, alternative changes to the blood service. Of potential policy relevance is the finding that appointment availability at weekends is predicted to increase donation frequency for those subgroups where additional blood donations may be of relatively high value, for example donors with blood types that are of ‘high demand’, donors whose ethnicity was defined as BAME, and nursery donors, who tend to be more difficult to retain. The finding that increasing travel time will reduce donation frequency and increase the rate with which donors leave the register is not surprising, but does have potential policy implications. The NHSBT policy has been to close smaller mobile sessions (i.e. three and six beds), and invest more resources in new static donor centres. Further moves that imply that donors will have to travel longer distances would be predicted to reduce the pool of donors, including those donors whose blood type is in high demand. Future evaluations should recognise the predicted reduction in the donor pool when evaluating the cost-effectiveness of further ‘switching’ of blood collection from mobile sessions to static donor centres.

The SP surveys have several strengths. First, the sample sizes (around 24,000 non-INTERVAL donors; 9000 ex-INTERVAL donors) are large compared with previous surveys, including DCEs in the evaluation of health services,60 and enable the surveys to provide precise estimates, not just overall, but for subgroups of prime interest. Specifically, we provided precise estimates of the relative effects of alternative service changes for subgroups for which it is particularly important to identify strategies that increase the blood supply, namely blood donors with high-demand blood types, BAME donors and nursery donors (all by sex). This will allow the analysis to estimate interaction effects of the inter-related attributes required for the subsequent CEA of alternative strategies. Second, the study administered three surveys in total. The pilot survey was important to check that the survey design provided plausible estimates, and to refine the design by modifying the choice of attributes and levels to try to reflect ‘real-life’ donations. In particular, an additional attribute concerning the availability of blood donation appointments was included in both main surveys. Feedback from the first donor translation workshop was that the definition of the attribute was clear and that the levels specified were relevant both to donors attending donor centres and to those attending mobile sessions (see Appendix 17). The second (non-INTERVAL) and third (ex-INTERVAL) surveys elicit preferences (donation frequencies) from donors attending predominantly mobile sessions and static donor centres, respectively. It was also useful to administer the same SP survey to two groups of donors who had, and had not, experienced the INTERVAL trial protocols. Third, the analytical model (ordered logit) respected the particular form of response data. We also undertook a thorough sensitivity analysis that considered other modelling choices and found that the direction and magnitude of estimated effects were robust to these choices (see Report Supplementary Material 7). Fourth, this study used appropriate methods for eliciting preferences and adopted recommended design principles. 46 The study therefore provides a more robust basis for predicting future donation behaviour than previous surveys of blood donors’ preferences. 7–11

The limitations of the surveys are that, first, it was infeasible to include all the attributes and levels of future policy relevance. Inevitably, choices had to be made according to those aspects of the service most likely to be of future importance for service provision and in line with best practice for survey design; we have reported the rationale for these choices. Second, although the response rates to the surveys were higher than anticipated, around 75% of invited donors did not complete questionnaires. Furthermore, as in the INTERVAL trial, donors who did not have an e-mail address were excluded. Rather than assuming that the requisite survey data were ‘missing completely at random’, we assumed that they were ‘missing at random’, that is, conditional on the variables included in the model. 53 Nonetheless, the estimates could be biased because of unobserved differences between the responders and the target population of prime interest. Third, although the discrepancy analysis found that the donation frequency predicted was similar to that observed, this did not consider the relative effect of service changes. These limitations have to be recognised in interpreting the subsequent findings of the CEAs that use the SP survey results.

The estimates from the SP survey presented in this chapter are not in themselves sufficient to inform the choice of future strategies for the following reasons. First, some of the strategies of interest require that several of the attributes are combined. For example, the strategy that specifies that a static donor centre is open at weekends requires estimates from the interaction of appointment availability and opening times attributes. This can then provide the requisite estimates of the effects of a weekend or evening opening strategy in static donor centres (see Chapter 4). Second, the responders to the survey do not represent the target population of interest, which may differ according to the strategy. However, the estimates presented in this chapter can be used to predict the relative effects of the alternative changes to the blood service for the relevant populations for each of the strategies considered in the CEA of the alternative strategies. Third, the survey results suggest that the introduction of a health report would lead to a relatively small increase in annual donation frequency, whereas strategies that would extend appointment availability would lead to greater increases in annual donation frequency. However, it will also be important to consider the relative costs of the alternative strategies. Chapter 4 will address these issues, using the estimates from the SP surveys to evaluate the cost-effectiveness of alternative strategies for maintaining the supply of whole blood.

Introduction

This chapter draws on results of the INTERVAL trial analysis (see Chapter 2), and the findings of two large surveys of donors’ preferences (see Chapter 3), to answer the broader question of which strategies available to the NHSBT are likely to represent value for money for the NHS in England. The NHSBT’s strategic objective is to reduce costs while maintaining the blood supply to the NHS, particularly the supply of blood types in high demand (O–, A– and B–). The research presented in this chapter addresses the third project objective pertaining to the cost-effectiveness of alternative strategies for maintaining the supply of whole blood. We focused on six strategies, which are all designed to improve opportunities for donors to give blood or make blood donation more attractive to donors. These were (1) offering a health report, (2) extending opening hours at static donor centres to weekends, (3) extending opening hours at static donor centres to evenings, (4) shifting opening hours of mobile sessions to weekends, (5) shifting opening hours of mobile sessions to evenings and (6) reducing the minimum interval between donations for donors attending static donor centres. These strategies are not mutually exclusive and may be implemented in combination and in conjunction with other initiatives. All six strategies are likely to increase the total volume of blood collected, which will come at additional cost. Given that the overall demand for whole blood is falling, the NHSBT would need to combine these strategies with other initiatives to release resources, several of which are already in progress, such as the closure of three- and six-bed mobile sessions in favour of more efficient nine-bed sessions. This analysis provides evidence to inform future decisions about the allocation of resources. The NHSBT may be prepared to invest to encourage more frequent donations by particular types of donors, such as those with high-demand blood types. Alternatively, these strategies could be adopted in such a way that the collection of higher demand blood types could be substituted for the collection of other blood types, by offering the service change described in each strategy exclusively to these groups of donors, for example by offering preferential appointments at certain times of the day or week.

This chapter sets out the methods used to identify the relevant target populations for each strategy (see Methods, Target population), defines the relevant comparator according to service provision recently experienced by donors (the status quo; see Methods, Comparator) and outlines those subgroups relevant to the analysis (see Methods, Subgroups). We describe how the preferences of non-trial participants from the SP survey were modelled to predict the donation frequency of all recent blood donors and how these were combined with deferral rates estimated from the INTERVAL trial (see Chapter 2) to predict the likely effect of each strategy on the volume of blood collected (see Methods, Predicted effect on the volume of blood donated). Relevant costs measured from a NHS and Personal Social Services perspectives over a 1-year time horizon are described (see Costs), with further details provided in Appendices 19 and 20. The results are presented for each target population for all donors, and donors with high-demand blood types (see Results, Base-case analysis), with results for other subgroups provided in additional material available on the project website. Key assumptions made in the analysis are tested with sensitivity analysis (see Methods, Sensitivity analysis and Results, Sensitivity analysis) and are discussed in relation to the main findings (see Discussion).

Methods

Results

Discussion

This analysis has estimated the relative cost-effectiveness of offering whole-blood donors a health report at each donation (strategy 1), extending static donor centre opening hours to weekends (strategy 2) and weekday evenings (strategy 3), shifting mobile sessions to weekends (strategy 4) and weekday evenings (strategy 5), and reducing the minimum interval between donations (strategy 6) – all compared with the status quo (current practice). The results showed that the most cost-effective strategies are those that would be implemented solely in donor centres, rather than those that would be implemented at mobile sessions or at both types of venue (strategy 1). The findings were similar across all the subgroups of donors considered, including those donors whose blood types are in relatively high demand. For BAME subgroups, the incremental cost per additional unit of blood for the health report was £69 (compared with £135 for the overall sample), which implies that the health report was still relatively cost-ineffective, although it should be recognised that the sample size of BAME donors was relatively small for both surveys.

Strategy 6, reducing the minimum interval between donations to the shortest minimum interval for donors of both sexes at donor centres, is predicted to have a relatively low additional variable cost of £10 per additional unit of whole blood collected. The strategies to extend opening times in static donor centres to weekday evenings and weekends can also provide additional units of whole blood at moderate additional costs, an extra £23 and £29 per additional unit of blood collected, respectively. In other words, if the NHSBT were to adopt these strategies, it would cost an extra £10 to £29 in variable costs for each unit of blood collected beyond the current volume.

The NHSBT currently collects almost 1.6 million units of whole blood at a cost to the NHS of £120 per unit. This cost covers the whole supply chain (collections, processing, transport, testing, donor marketing, hospital services, storage and donor records), so is not directly comparable to the variable costs applied in this analysis. The proportion of the £120 charge per unit that is attributable to blood collection is around 40%, and, hence, the comparable amount paid for a unit of blood, which includes both variable and fixed costs, is around £48. The NHSBT has recently closed small mobile sessions as the costs per unit of blood collected were relatively high (> £50 per unit). The incremental cost for an additional unit of blood for all strategies except the health report, therefore, seems to be within an acceptable range, although the cost-effectiveness threshold remains unknown. This is particularly the case for approaches to providing more units of high-demand blood types: the NHSBT might be prepared to release more resources to pay for an additional unit of these blood types or substitute the collection of other blood types. In contrast, we find that the introduction of the health report is likely to be cost-ineffective.

The results were generally robust to the base-case assumptions concerning the source of data, as well as the choice of model used to predict donation frequencies from the survey data. The results were somewhat sensitive to the assumption that there is sufficient capacity at static donor centres to collect extra blood with existing levels of staff. When the costs of additional static donor centre staff were included, the strategy of reducing the minimum interval between donations became slightly less cost-effective compared with the status quo. The results suggest that up to 73,130 units of blood could be collected by reducing the minimum interval between donations for donors who give blood at static donor centres; this would be a 29% increase in the volume of blood collected at donor centres in 2015/16. Current staffing constraints would not be sufficient to support the collection of additional blood in this quantity. However, if the extra collection of blood was limited to high-demand blood types (up to 9900 units), it is feasible that this could be accommodated at times and locations sufficiently convenient for donors to give blood at the frequency predicted by their survey responses. Alternative approaches to providing more whole blood from the high-demand subgroup of donors would be to extend the opening times for static donor centres into weekday evenings or weekends. These strategies would provide additional capacity for whole-blood collection in static centres at a reasonable incremental cost per additional unit.

This analysis adds to the limited literature on the cost-effectiveness of alternative changes to the blood collection service. 18–28 Strengths of this analysis include the consideration of strategies of live policy relevance, the results of which provide timely evidence to inform decision-making. The analysis was based on the results of two large SP surveys, which had a combined sample size of > 30,000 whole-blood donors in England. Responders had given blood within the past 12 months and donors attending mobile sessions and static donor centres were well represented across the two surveys. The SP surveys were developed to offer flexibility in assessing the potential impact of changing different aspects of the blood service for two very different types of blood donation experience (static donor centres and mobile sessions), including aspects that do not currently exist (provision of a health report). Predictions from the SP survey model were similar to the observed donation frequency of female donors, differing moderately from the observed donation frequencies of male donors, and were similar across donor subgroups. The analysis draws on the deferral rates observed in the INTERVAL trial over 2 years and provides predictions about the population of prime relevance – the donors who live in mainland England and who gave blood in the past 12 months.

An important limitation of the analysis was that the 1-year time horizon did not allow for the costs and consequences of the strategies to be fully considered. In particular, although the increased rate of deferrals was captured in the analysis, the long-term effects of deferrals were not. The INTERVAL trial analysis found that, after 2 years, those in the reduced interval arms had lower haemoglobin and more haemoglobin-related deferrals. Beyond the health-care costs associated with treating donors with very low haemoglobin levels, there may well be longer-term costs and consequences, such as a higher proportion of donors leaving the donor register in response to repeated deferrals. Replacing donors is costly in terms of recruiting new donors and the opportunity costs of replacing a regular donor with another who may prefer to donate less frequently. The ensuing reduction in blood volume collected and the associated increase in costs could render this strategy less cost-effective than our analysis suggests. Further research is required on deferral rates over time. The INTERVAL extension study may provide opportunity to undertake such research, given that it will collect data on participants for up to 3 years.

The additional unit costs included in the model may not correspond to the full variable cost of collecting blood. Although the costs of deferrals to the service were included, we did not account for costs associated with donors who did not attend (DNA) their appointment, although there is no evidence to suggest that the rate of DNAs increases with higher donation frequency. Processing costs were excluded from the analysis, but in the likely event that the NHSBT does not collect more blood, but different blood types, these costs will not change. There may also be cost savings to the service not considered in the analysis. We did not consider efficiency savings of relying on a smaller pool of donors who give blood more frequently, which may be considerable, given the high cost of recruiting new donors and keeping them engaged in the process of regular blood donation. In addition, we did not incorporate the effects of patients switching from donation at mobile sessions to static donor centres (and vice versa). If opening hours increase at static donor centres, it is likely that some donors will opt to give blood at static donor centres instead of at their usual mobile session, an approach that may become more common if mobile sessions are closed. Savings made from the closure of less efficient mobile sessions might therefore improve the cost-effectiveness of strategies to improve opportunities to donate at static donor centres. Any cost savings would be specific to the particular session, but could potentially offset the additional costs required to implement these strategies.

Direct costs to donors were not considered. Opportunity costs faced by donors in donating whole blood more frequently may have been captured in their responses to the SP surveys, in terms of how often they would donate. Costs such as travel expenses may affect donation behaviour, although, because they are accrued in the pursuit of an altruistic act, including these costs in an analysis should be accompanied by the incorporation of benefits to donors in terms of increased utility from blood donation itself.

Uncertainty in the model predictions on donation frequency and associated resource use was incorporated in the probabilistic sensitivity analysis. Unit costs were assumed to be fixed; the largest were set by standard NHS sources and contracts, so are unlikely to vary. Other sources of uncertainty, most importantly structural uncertainty in the costing model and uncertainty in the extent to which SP surveys predict actual donor behaviour in response to service changes, have not so far been explored.

In conclusion, this analysis found that, for whole-blood donors attending static donor centres, reducing the minimum intervals between donations or extending the opening times for blood collection to include weekday evenings or weekends would provide additional units of whole blood at moderate additional variable costs compared with current practice. In contrast, although the introduction of a donor health report or extending opening times for mobile sessions are also predicted to increase the volume of whole blood donated, the additional costs are somewhat higher. The results of the analysis were consistent across a large number of prespecified subgroups, including those of prime policy relevance: donors with high-demand blood types, BAME donors, young donors and nursery donors. Chapter 5 will consider the overall conclusions from all aspects of this research and the general strengths, limitations and overall implications for the blood service and for further research.

Patient and public involvement

The views of blood donors and a public representative informed the design of the research and the interpretation and implications of the findings. The choice of strategies for evaluation was informed by discussions with blood donors and suggestions from a public representative. Blood donors guided the design of the SP surveys by suggesting aspects of the service that might be important to change, including extending donation opportunities to weekday evenings and weekends. Blood donors also identified attributes that should not be included, for example financial payment for donating blood, and provided insights on what the questionnaire recipient might be thinking when attempting to answer the questions. The questionnaire design was also informed by findings from qualitative research on blood donors’ views, which suggested that it was important to recognise that donors may prefer to continue to donate at the same venue rather than a different setting, irrespective of other features of an opportunity to donate (R Lynch and S Cohn, LSHTM, personal communication, October 2015). The design of the SP survey was informed by a large pilot study of blood donors; donors and the public representative reviewed the donor information and consent forms prior to submission to the NHS Research Ethics Committee. The eventual CEA was driven by the responses to the surveys administered to large numbers of blood donors, both those who were previously participants in the INTERVAL trial (n = 9318) and those who were not (n = 25,187).

The translation workshops with donors were an important element of the research. The first translation workshop was undertaken following the preliminary analysis of the first SP survey, and it offered insight into the early interpretation of the preference results. It also endorsed the changes to the survey design, made in response to the results of the pilot study. The second translation workshop allowed the research team to gather views on the overall cost-effectiveness findings from the project and provided insights as to how they should best be communicated to donors and what, from the donors’ perspective, should be the priorities for future research.

Strengths

The study provided estimates of the relative cost-effectiveness of alternative strategies of direct policy relevance for donor subgroups of prime interest. The results presented extend the limited extant literature on the cost-effectiveness of alternative strategies for maintaining the blood supply. In particular, the direct input of the NHSBT during the design stage ensured that the study considered those strategies of prime relevance for key subgroups of interest. This required the use of evidence from three complementary sources: a large well-conducted RCT, SP surveys and the NHSBT donor registry. The INTERVAL trial provided estimates of the relative effect of reduced donation intervals on donor health and on a key concern – relative rates of deferral for donors attending static donor centres. The surveys investigated a key concern set out in the NHSBT blood 2020 strategy document (pillar 1),4 regarding improving the donors’ experience, informed by the understanding of their relative preferences for alternative changes to the blood service, including those such as the introduction of a donor health report that are not currently available.

The study also extended previous research on blood donation by following choice-based principles. A particular strength of the study was that it was replicated in two contrasting settings. Among ex-INTERVAL participants, all of whom had experience of donating at a static donor centre, and among non-INTERVAL participants, the majority of whom had donated at mobile settings. The main findings concerning donor preferences were similar across the surveys of non-INTERVAL and ex-INTERVAL donors. The large sample sizes in the INTERVAL trial and in both stated preference surveys enabled results to be reported across many different subgroups, including those of prime interest to the NHSBT. The PULSE data set provided a sampling frame for the study and allowed the representativeness of the study populations to be defined. It was also used to define the target population for the CEA and the current practice experienced by blood donors. This was important to ensure that the CEA results were provided for the population of prime interest.

The study provided a framework that could be followed in future evaluations, not just of alternative changes to the blood service but also of health services more widely. In particular, the research found that it was feasible to quickly elicit preferences from large samples of the public. The research also found that the discrepancy between stated and actual behaviour was moderate (men) or small (women); this finding from large survey and corresponding registry data extends the limited methodological literature in health care on stated versus revealed preferences. Finally, the CEA was accompanied by extensive sensitivity analyses, which found that the findings were robust to alternative standpoints, notably the source of data for the estimates of relative donation frequency, and the assumptions underlying the unit cost estimates.

Limitations

The main time horizon for the cost-effectiveness analysis was 1 year, and so the long-term effects of the alternative strategies on, for example, the rate and costs of donors leaving the register, or any effects of the choice of strategies on fixed costs, were not considered. In particular, the INTERVAL trial found that the reduced donation intervals did lead to increased rates of Hb-related deferrals and, if, in the long term, this does encourage donors to leave the donation register, this will make this strategy less cost-effective. Costs were measured from the perspective of the NHS, so any additional costs to donors, for example from time off work or increased travel time, were not considered. This was consistent with the measurement of consequences, which excluded any change in donor’s well-being from donating blood more often beyond that which would be detected by the generic measure of health-related QoL (SF-6D utility score). The research did not evaluate all the strategies of potential interest to the NHSBT, such as expanding the number of static donor collection centres or closing and/or merging of mobile sessions. The cost-effectiveness results relied on estimates from the 25.2% (non-INTERVAL) and 32.4% (ex-INTERVAL) responders to the SP surveys. The analysis adjusted for differences between the survey responders and the target population (e.g. according to donation history), but does not allow for those differences that remain unobserved (e.g. intrinsic motivation).

Conclusions and implications for the blood collection service

• Donors whose last donation was at a static donor centre would strongly prefer more opportunities to donate during weekday evenings or at weekends. For donors with blood types that are in high demand, providing improved opportunities to donate by extending opening hours to the evenings or weekends in all donor centres would be relatively cost-effective (£23 and £29 per additional unit donated, respectively).

• A strategy of reducing minimum intervals between donations from 12 to 10 or 8 weeks (men), or from 16 to 14 or 12 weeks (women), is preferred by donors, and can provide additional whole blood at £10 per additional unit. In the short run, this strategy might help maintain the required levels of red cells for blood types that are in high demand, such as O–, but will increase rates of Hb-related deferrals and so may lead to donors leaving the register and increase costs in the longer term.

• Moving mobile sessions to the weekends or providing a donor health report at each donation visit would not, on average, lead to sufficient increases in the frequency of whole-blood donations to justify the additional costs.

• Requiring donors to travel further to donate whole blood will encourage donors to leave the donation register; subgroups of younger or less experienced donors and some BAME (ex-INTERVAL) donors are among those most likely to stop donating. As it is important to retain these donor subgroups, if the NHSBT continues to close mobile sessions and increase travel time for donors, it may be important to adopt strategies targeted at retaining these particular subgroups of donors.

Implications for further research

• If any of these strategies are implemented, it will be important to monitor the preferences and donation frequency for different donor subgroups, and in alternative settings. Such an evaluation could build on the research framework presented here and consider the costs and consequences of scaling up the strategy, in real time, by using large-scale surveys of preferences calibrated to actual donation behaviour as recorded in the PULSE registry.

• Further research is required to investigate the effectiveness and cost-effectiveness of tailoring opportunities to donate blood according to the preferences of particular donor subgroups. There is a particular requirement to understand the preferences of BAME donors, as their blood is in high demand, and only a relatively small sample of these donors were surveyed in the HEMO study.

• Improved understanding of how to predict which donors are likely to have Hb-related deferrals would allow stratification of the donor population by likelihood of deferral. This could allow increased blood collection using shorter interdonation intervals for a defined subpopulation of donors whose blood type is in high demand.

• Further research is warranted on evaluating novel marketing strategies to encourage the recruitment of particular subgroups of new donors (e.g. those with O– blood type, BAME donors, younger donors), to ensure that there is the required mix of blood donors to sustain the future blood supply.

Health Economics Modelling study management group

Professor Richard Grieve (chief investigator), Professor John Cairns, Kaat De Corte, Professor Neil Hawkins, Dr Mark Pennington, Dr Silvia Perra, Dr Zia Sadique and Sarah Willis.

Health Economics Modelling study advisory group

Professor Janet T Powell (independent chairperson), Dr Gavin Cho, Laura Hontoria Del Hoyo, Dr Gail Miflin, Dr Carmel Moore, Professor David J Roberts, Crispin Wickenden and Peter Zollman.

The INTERVAL study

Participants in the INTERVAL RCT were recruited with the active collaboration of NHSBT, which has supported field work and other elements of the trial. DNA extraction and genotyping was funded by the NIHR, the NIHR BioResource (http://bioresource.nihr.ac.uk/) and the NIHR Cambridge Biomedical Research Centre (www.cambridge-brc.org.uk). The academic co-ordinating centre for INTERVAL was supported by core funding from the NIHR Blood and Transplant Research Unit in Donor Health and Genomics, the UK Medical Research Council (G0800270), the British Heart Foundation (SP/09/002) and the NIHR Cambridge Biomedical Research Centre. A complete list of the investigators and contributors to the INTERVAL trial is provided in Moore et al. 42

Contributions of authors

Richard Grieve (Professor, Health Economics Methodology) conceived and designed the overall study, directed the acquisition, analysis and interpretation of the results, and drafted and critically revised the manuscript.

Sarah Willis (Research Fellow, Health Economics) contributed to the design of the study, the delivery of the preference surveys and the acquisition of survey, PULSE and cost data; conducted the CEA and interpreted the results, and drafted and critically revised the manuscript.

Kaat De Corte (Research Degree Student, Health Economics) contributed to the design of the study, contributed to the acquisition, analysis and interpretation of the survey, PULSE and INTERVAL data; conducted the discrepancy analysis, and drafted and critically revised the manuscript.

M Zia Sadique (Assistant Professor, Health Economics) conducted the analysis and interpretation of the INTERVAL trial data, the SP survey data and predictions used in the CEA, and drafted and critically revised the manuscript.

Neil Hawkins (Professor, Health Economics and Health Technology Assessment) contributed to the design of the CEA, the analysis and interpretation of the data, and the drafting and critical review of the manuscript.

Silvia Perra (Research Fellow, Statistics) contributed to the design of the study, conducted the analysis of the pilot SP survey, contributed to the acquisition of the PULSE data, the analysis of the preference survey data, and the drafting and critical review of the manuscript.

Mark Pennington (Senior Lecturer, Health Economics) contributed to the design of the study, the acquisition, analysis and interpretation of the cost data, and the drafting and critical review of the manuscript.

Jenny Turner (Research Manager) co-ordinated the HEMO study, organised the workshops with blood donors, designed and created the study website, and contributed to the drafting and critical review of the manuscript.

Carmel Moore (Senior Scientific Research Co-ordinator, Trials) conducted the scientific co-ordination of the INTERVAL trial, contributed to the interpretation of the results pertaining to the INTERVAL trial and the drafting and critical review of the manuscript.

Crispin Wickenden (Head of Donor Management, NHSBT) contributed to the design and delivery of the preference surveys and use of PULSE data, the interpretation of the results, and the drafting and critical review of the manuscript.

Catharina Koppitz (Market Research Analyst, NHSBT) contributed to the design and administration of the preference surveys and use of PULSE data, the interpretation of the results, and the drafting and critical review of the manuscript.

Gavin Cho (Consultant Haematologist, NHSBT) contributed to the interpretation of the results and the drafting and critical review of the manuscript.

David J Roberts (Professor and Consultant Physician, Haematology) was joint chief investigator for the INTERVAL trial, and contributed to the interpretation of results and the drafting and critical review of the manuscript.

Gail Miflin (Medical and Research Director, NHSBT and Consultant Haematologist) contributed to the design of the study, interpretation of the results, and the drafting and critical review of the manuscript.

John A Cairns (Professor, Health Economics) helped design the overall study, directed the design and analysis of the preference surveys and the overall results, and drafted and critically revised the manuscript.

Publication

Willis S, De Corte K, Cairns JA, Sadique MZ, Hawkins N, Pennington M, et al. Cost-effectiveness of alternative changes to a national blood collection service [published online ahead of print 16 May 2018]. Transfus Med 2018.

Data-sharing statement

Owing to the nature of this study and the conditions attached to original data agreements, there are no data available for wider use. All queries should be submitted to the corresponding author in the first instance.

Patient data

This work uses data provided by patients and collected by the NHS as part of their care and support. Using patient data is vital to improve health and care for everyone. There is huge potential to make better use of information from people’s patient records, to understand more about disease, develop new treatments, monitor safety, and plan NHS services. Patient data should be kept safe and secure, to protect everyone’s privacy, and it’s important that there are safeguards to make sure that it is stored and used responsibly. Everyone should be able to find out about how patient data are used. #datasaveslives You can find out more about the background to this citation here: https://understandingpatientdata.org.uk/data-citation.

Disclaimers

This report presents independent research funded by the National Institute for Health Research (NIHR). The views and opinions expressed by authors in this publication are those of the authors and do not necessarily reflect those of the NHS, the NIHR, NETSCC, the HS&DR programme or the Department of Health and Social Care. If there are verbatim quotations included in this publication the views and opinions expressed by the interviewees are those of the interviewees and do not necessarily reflect those of the authors, those of the NHS, the NIHR, NETSCC, the HS&DR programme or the Department of Health and Social Care.

Non-INTERVAL donors

(PDF download)

Question 1: how many times did you give blood in the last 12 months?

If you cannot remember, please give your best guess.

Question 2: how many times did you want to give blood in the last 12 months?

Question 3: when you last gave blood, where did you travel from?

If you cannot remember, please give your best guess.

Question 4: roughly how far did you travel to the place where you last gave blood?

Question 5: how did you travel to the place where you last gave blood?

Please choose the answer that applies to the longest part of your journey.

Question 6: how long did it take you to travel to the place where you last gave blood?

If you cannot remember, please give your best guess.

Question 7: roughly how long did your last blood donation visit take?

From your scheduled appointment time to the time you arrived at the tea table and were free to leave (including waiting time).

Question 8: what prompted you to make your last appointment to give blood?

Please choose the answer that best applies to you.

Question 9: are you able to make appointments to give blood as often as you would like, at a day and time that suits you?

Translation workshop I: key themes from the discussion

Monday 21 November 2016, 11.15–15.15.

At LSHTM, London.

Attended by seven donors (all of whom were non-INTERVAL donors) and a public representative. Attendees were split into three groups for the key discussions, joined by colleagues from NHSBT, and each group was facilitated by a member of the research team (RG, SW and KDC).

This is a summary of the general feedback and discussion that followed each break out session at the workshop, focusing on the main themes.

Translation workshop II: key themes from the discussion

Monday 24 April 2017, 5.30 p.m.–8.30 p.m.

At LSHTM, London.

Attended by 13 donors (a mix of ex-INTERVAL participants and non-INTERVAL donors, nursery and more experienced donors). Attendees were split into three groups for the three discussions, joined by colleagues from NHSBT and each group facilitated by a member of the research team (RG, SW and KDC).

This is a summary of the general feedback and discussion that followed each break out session at the workshop, focusing on the main themes.

Base-case results

Sensitivity analysis: assuming no additional capacity to collect more blood at donor centres, for all donors

Sensitivity analysis: assuming no additional capacity to collect more blood at donor centres, for donors with high-demand blood types

Sensitivity analysis: using a two-part model to predict effect of strategies on the volume of blood collected, for all donors

Sensitivity analysis: using a two-part model to predict effect of strategies on the volume of blood collected, for donors with high-demand blood types

Sensitivity analysis: using a gamma model to predict effect of strategies on the volume of blood collected, for all donors

Sensitivity analysis: using a gamma model to predict effect of strategies on the volume of blood collected, for donors with high-demand blood types