Table 1 presents the agenda of the task force, developed through preliminary meetings and initial e-mail correspondence of its 17 members. The agenda encapsulates the key issues and common methodological challenges likely to be encountered by analysts seeking to incorporate health spillovers into CEA.

Table 1 Agenda of the SHEER task force3.1 Reference Perspective

A ‘reference perspective’ is necessary in order to inform the basis for the definition of health spillovers to be employed by the group, and to outline the context in which it is envisioned that these recommendations will primarily be applied. With a view to producing a practicable recommendation set, aligned with the primary aim of the task force, it is proposed that this guidance focus on incorporation of family and caregiver health spillovers (as defined subsequently) into a CEA conducted from a healthcare/health payer perspective. Such a perspective includes only those costs that fall under the healthcare budget and only health effects as outcomes/benefits [6]. Our interpretation of the healthcare/health payer perspective is analogous to that observed most commonly in a European context, where a single healthcare provider is more likely to predominate in a given jurisdiction. It is important to note, however, that in some contexts (e.g., the USA), where there are multiple healthcare providers, health costs and outcomes will most likely fall under the remit of multiple providers.

3.2 Definition of ‘Family’, ‘Caregiver’ and ‘Family and Caregiver Health Spillovers’

Working definitions of the terms ‘family’, ‘caregiver’ and ‘family and caregiver health spillovers’ were considered essential given the multiplicity of meanings that may be attached to each. All definitions contained within this paper are colour coded in green.

The definition of ‘family’ below is deliberately far-reaching and, moreover, specific to health economic evaluations, acknowledging that network composition and size will differ within and across health, cultural and other contexts. Yet, some measure of consistency is required in order to make valid comparisons between evaluations, as the wider the circle, the more ‘spillover effects’ are likely to be captured:

Linked with these definitions, and focusing on the impact of an individual’s illness on others’ HRQoL, the following working definition of ‘family and caregiver health spillovers’ has been endorsed by the task force:

This definition, purposely restricted to encompass only health spillovers onto family and caregiver HRQoL, does not negate the potential for impacts among persons outside this group, but rather focuses, for practical purposes, on those for whom the effects are anticipated to be largest. Deciding who is included or not inevitably requires some discretion on the part of the analyst, and likely depends, at least in part, on the health condition and age of the patient. Where there is ambiguity, sensitivity analyses with a wider/narrower set of included individuals may be undertaken (see recommendation 7 in relation to emerging good practice). As an economic evaluation compares different interventions, it also important to note that the effect of the intervention on the caregiver/family member may arise through a change in the patient’s health status and/or via another pathway [54]. For example, an ‘organisational intervention’ to discharge patients from hospital earlier is likely to affect family caregivers by shifting the location of care to the home and increasing the amount of informal care provided, rather than by changing the patient’s health status.

3.3 Emerging Good Practice and Related Future Research Recommendations

The succeeding sections detail 11 consensus recommendations for emerging good practice and 12 in relation to future research. The emerging good practice recommendations are summarised in Table 2, with their full text being presented thereafter. Recommendations for emerging good practice are colour coded in blue. As mentioned, a supporting text, expanding upon the discussions of the task force and the evidence review, accompanies the recommendation set. Where relevant, applicable recommendations for future research are highlighted in red. Table 3 outlines these and other avenues for further study.

Table 2 Summarised consensus recommendations for emerging good practiceTable 3 Recommendations for future research

The full results of the second-round vote are included in Appendix B of the ESM. Consensus was not reached as to the appropriateness of two proposed, emerging good practice recommendations; none of the proposed recommendations was deemed inappropriate.

3.3.1 When to Incorporate Health Spillovers into a CEA

There are two main schools of thought in this context, which relate to commonly espoused viewpoints in economic evaluations, often phrased as ‘perspective’ [55]. Under the healthcare/health payer perspective, the goal of the decision maker (to be informed by the economic evaluation) is typically assumed to be to maximise health from a fixed budget [56]. The decision maker, thus, aims to produce as much health (i.e., QALYs) as possible for the population from that fixed budget. If this is the decision rule, two things become important: (i) the measurement and inclusion of all relevant health changes in order to ensure health maximisation [8] and (ii) the capture of health effects related to displaced/foregone activity as well as those resulting from the new intervention [7]. Using a societal perspective, in contrast, the goal of the informed decision maker is assumed to be the maximisation of social welfare from a more flexible budget [57]. Here, all costs and benefits are relevant, regardless of where or on whom they fall [24].

Acknowledging that a payer perspective is the least inclusive of those most frequently employed [58], albeit adopted by NICE [30], the task force advocates for the consideration of health spillovers in all economic evaluations conducted from this and those more inclusive perspectives, such as the societal perspective, which is the default in the USA and the Netherlands, for example. In other words, family and caregiver health spillovers should be included in economic evaluations from all of the more frequently employed perspectives (i.e., healthcare/health payer, societal or public sector perspective). Consensus among members has been reached as to the need for an express statement to that effect, embodied in the recommendation below:

3.3.2 Opportunity Costs

If one assumes interventions are funded from a fixed budget, the opportunity costs of healthcare spending become relevant, including the health spillovers related to healthcare interventions displaced or foregone when a new intervention is adopted. However, the wider impacts of that care that is displaced/foregone will likely not have been considered when establishing benchmarks/thresholds of cost effectiveness. This may lead to a greater observed quantity of health displacement when funding new interventions where these effects have been considered [59]. To combat this and establish the value of the net gain in health, one should seek to account for all relevant gains of a healthcare intervention, including health spillovers, minus the lost health (also including health spillovers) due to care displaced or foregone [6, 57].

Given the focus of this report on health spillovers, the task force has refrained from further discussion of the ‘broader costs’ mentioned in the recommendation. It should, however, be recognised that caregivers and the wider family will also incur time and other costs, both within and outside healthcare sector.

In the absence of any guidance on disinvestment, healthcare interventions that are displaced or foregone are typically unknown, and as such, little is known about lost spillover effects related to this activity [6]. Much research is therefore needed to develop methods to quantify the average displaced/foregone spillover and this motivates the following future research recommendation:

3.3.3 Distributional Consequences of Inclusion

The distributional consequences of spillover inclusion (as well as omission) provoke much debate in the literature [5, 21, 29]. While systematic exploration of the trade-off between the maximisation of aggregate health and preservation of health equity [28] is not practicable at the level of an individual economic evaluation, the decision maker should at least be forewarned of this challenging issue to allow for informed decisions [6].

The implication of this recommendation is that consideration of spillovers could introduce bias in resource allocation (as accumulation of QALY gains among more isolated patient groups may be reduced and interventions targeting these groups would appear less cost effective). The limited evidence base in this regard compels further investigation of the extent and acceptability of these potential equity impacts, via, for example, the application of analytic approaches [60,61,62], as proposed by Dixon and Round [28].

3.3.4 Spillover Time Horizon

In addition to how wide a net to cast (see recommendation 7, below), there is also a question of how long, in terms of duration, to measure effects for (i.e., how to define the analytic time horizon). Typically, health costs and benefits cease at the time beyond which no incremental benefits of an intervention can be attributed to a cure or successful treatment, or at the end of life—all related to the patient. When considering health spillovers, however, family members may survive the patient and bereavement may be an important consideration in their HRQoL. Moreover, long-term health consequences of caregiving may surpass the time horizon of a patient’s illness, even when the patient survives. With respect to the time horizons applied to health spillovers, one must principally consider whether to match the patient’s timeline, both in treatment benefits/costs and survival, or to impose different timelines for family members/caregivers, so as to acknowledge that effects may not entirely coincide. Balance and consistency in accounting for these impacts are also crucial, across trial arms for example [27]. Last, it must be noted that a lack of established methods for including the impacts of bereavement, and the unpredictably of those effects (especially as the analytic horizon extends), make the consideration of bereavement particularly challenging [59]. Clearly, much research is needed in this regard.

3.3.5 Outcome Measurement

Both direct [10, 63] and indirect measurement methods [64,65,66,67,68,69] have been used in estimating spillover health (dis)utility. There is, however, no clear guidance on which methods or instruments are optimal. Relatedly, while expressing the need for the robust capture of caregiver health spillovers, a recent review by NICE described a dearth of good-quality evidence to inform the relative impact of new interventions on caregiver HRQoL, compared with current practice [37]. A lack of empirical evidence, therefore, makes the task of method/instrument selection challenging.

The group also discussed the potential for employing estimates of the relative value of care-related quality-of-life and HRQoL outcomes (e.g., Carer Experience Scale [70] or the CarerQol instrument [71] relative to EQ-5D outcomes). These empirical weightings (or “exchange rates”) could be used to convert caregiver outcomes captured in the form of care-related quality of life to HRQoL outcomes. This would allow for the aggregation of caregiver outcomes with patient outcomes measured via HRQoL [72, 73]. The threshold of consensus was, however, not reached for this method to be included as a recommendation. With regard to avenues for future research in measuring caregiver and family outcomes, the development of novel preference-based measures aimed at capturing health spillovers was proposed (which inherently necessitates determination of the measures’ underlying conceptual domains and attribute/item selection prior to preference-based valuation [69, 74]).

The task force has refrained from explicit recommendations on whether to include family and caregiver spillover outcomes as utilities (in which case the caregiver/family member is treated as a separate entity in the analysis) or disutilities/utility increments (where the extent to which caregiver/family member impacts are relieved or exacerbated as a result of patient health changes, implying, for example, that caregiver/family member disutility would improve when the patient dies) in decision models [27]. Implicitly, recommendation 4, in advocating for spillover measurement over a time horizon sufficient to capture all relevant impacts (including bereavement), suggests consideration of family members and caregivers as separate entities is favoured. Appropriate means of addressing this issue will require further exploration.

3.3.6 Sources of Data: Use of Primary Versus Secondary Data

Given our context, the SHEER membership have endorsed the following definitions of ‘primary’ and ‘secondary’ data for the purposes of this recommendation.

As one might expect, the evidence review identified trial-based evaluations as tending to include primary data more frequently, whereas models used secondary data. Described barriers to the inclusion of primary data related to missing data on family/caregiver outcomes [20], insufficient time horizons in trials to detect change in utilities resulting from spillovers [19, 75], and differing outcome measurement methods being employed among patients and family/caregivers [76]. Challenges to the use of secondary data pertained to the context specificity of spillovers [13], and differing interpretation of estimates from the same source [20, 27]. Taking this exacting data environment into consideration, the task force recognises that secondary data sources will often need to be employed, especially where impacts are to be modelled over the lifetime of the specified family/caregiver cohorts [77]. Nevertheless, the collection of primary data is currently favoured.

3.3.7 Aggregation Methods

In view of the utilitarian foundations of health economic evaluations, the objective, and conventional practice, is the identification and additive summing of all relevant effects (regardless of the source). Effects may be quite broad [11, 21] and, accordant with our working definition of family and caregiving health spillovers, there is a need to pragmatically capture these within an economic evaluation, given resources and other constraints. The task force has developed a stepwise aggregation procedure, as delineated below:

Although we recommend additive summation of spillover health effects with patient health effects, this summation may obscure the cost per QALY for patients, hampering comparisons of efficiency with evaluations where spillovers are omitted. A ‘multiplier approach’ (where spillover health effects are stated in a ratio to patient health effects, with this ratio being applied to the incremental cost-effectiveness ratio) has been proposed to counter this issue (see Al-Janabi et al. [7] for further details on this approach). Consensus was, however, not reached on recommending that spillovers and patient effects be presented in the form of a multiplier.

3.3.8 Outcome Value Weighting

Criteria for resource allocation are usually context specific, depending on social values and judgement. In particular, a small but growing literature shows that the social value of QALYs may vary across settings and individuals [72]. In the context of health spillovers, debate has developed as to the appropriateness of differentially weighting family/caregiver and patient outcomes. Some argue for a lower weighting to be applied to family/caregiver outcomes in order to focus on patients [76] or combat perceived inequity [28], others for higher weighting of caregiver outcomes, owing to a duty of care to family caregivers, given their role in supporting society [78]. Considering the limited peer-reviewed empirical basis for differential weighting of family/caregiver outcomes [72], and the abovementioned utilitarian foundations of health economic evaluations, a general agreement among members has emerged for the application of equal weighting in the reference case.

It is important to note that this recommendation pertains to equity weighting (reflecting social value and societal preference) and not to weights applied to caregiver outcomes in instances where different metrics are used among patients and caregivers (i.e., the earlier mentioned “exchange rate” between care-related quality of life and HRQoL [72, 73]). Future research is needed to investigate methods for the elicitation of spillover equity weights.

3.3.9 Outcome Double Counting

When deriving utilities using indirect techniques (i.e., generic preference-based HRQoL instruments), health spillovers may inadvertently be included, either at the measurement stage or the valuation stage, posing a risk of double counting. In relation to measurement, patients may, for example, take into account the impact of their condition on family members when reporting their own HRQoL [79], such that spillover effects are reflected in utilities. Similarly, the double counting of spillover effects may occur if the caregiver/family member fails to compartmentalise changes in their own health resulting from the health status of the patient [80]. There is also evidence that when valuing health states, respondents consider being a burden to others [81, 82]. Problems with double counting can also occur when the effects of informal caregiving are reflected in the monetary value of informal care time [4, 83]. The nature and extent of double counting requires further exploration and, given the state of current practice, the analyst should seek to minimise, rather than eliminate, the risk of double counting.

3.3.10 Consistency in Inclusion and Reporting of Spillovers

There is varied practice in including spillovers in economic evaluations [20, 22,