Progress in measuring family spillover effects for economic evaluations

Abstract

Despite considerable progress in measuring family spillover effects for economic evaluations, much work remains. This editorial describes recent advances in the field and identifies areas for future research. In particular, we point out the need to clarify the potential for double-counting from including spillover effects related to caregiver time and spillover effects related to family quality of life outcomes in Reference Case analyses. We believe that research on family spillover effects has the potential to increase interest in the field of economic evaluation and its use as a tool for decision making.

Keywords:

• cost
• economic evaluation
• family spillover effects
• QALYs
• reference case analysis

The idea that certain health conditions and services affecting one person can have implications for the physical and mental health of other persons – especially family members – has been recognized for some time. However, few researchers attempted to include these ‘spillover’ effects in economic evaluations. For example, in their study on the cost of birth defects, Waitzman et al. noted that their estimates were conservative because they did not include the psychosocial and time costs of the birth defect on the family [1]. The US Panel on Cost–effectiveness Analysis in Health and Medicine recommended the inclusion of spillover effects on caregiver time in a Reference Case analysis as a cost in the numerator of the cost–effectiveness ratio. The panel also recognized potential spillover effects on family quality of life outcomes, and recommended their inclusion in sensitivity analysis as quality-adjusted life years (QALYs) in the denominator of the cost–effectiveness ratio, but argued against their inclusion in a Reference Case analysis [2].

The rationale for the latter recommendation is found in the panel’s cost–effectiveness analysis of food fortification to prevent neural tube birth defects. The authors recognized the potential implications on the quality of life for the family from having a child born with a neural tube defect, but claimed that including spillover effects on the family ‘is not recommended in a Reference Case analysis because the methods for capturing these impacts are in early stages of development. Also, data on these quality-of-life impacts are not available [3].’ It should also be noted that measurement of the Global Burden of Disease does not account for family spillover effects either [4].

In the past decade, however, research on incorporating family effects in economic evaluations has increased greatly to the point that one can argue they should be included in a Reference Case analysis, yet as we point out, more guidance on incorporating family spillover quality of life effects and time costs remains a priority. Guidelines for the Reference Case analysis within the National Institute for Health and Care Excellence recommend measuring QALYs for all affected parties, including other family members, where warranted [5]. The National Institute for Health and Care Excellence guidelines, however, do not include spillover time costs (loss of both work and leisure time) because the analysis is conducted from the perspective of the health system. In a societal analysis, both spillover time costs and quality of life effects should be considered.

Basu and Meltzer significantly advanced research on family spillover effects by developing a theoretical framework for their inclusion that showed how medical treatment can produce direct and indirect health effects on all family members, including the patient [6]. They also recognized potential ethical issues from including family effects, as illnesses affecting married persons or persons with children would have greater benefits from treatment and lower cost–effectiveness ratios relative to conditions that affect unmarried adults under this perspective. Basu et al. contributed to empirical methods for evaluating spillover effects on health by developing a modified time trade-off method to estimate spillover effects on the spouse (or partner) from a patient experiencing prostate cancer-related health states [7]. In this study, the patient’s partner was asked to tradeoff his or her own life based on the expected burden to their self only from the patient incurring health states such as incontinence, impotence or dying. This pilot study demonstrated the feasibility of direct elicitation approaches to measuring family spillover effects.

Prosser et al. evaluated spillover effects of chronic conditions using a novel standard gamble approach that included both family members who experienced the condition and a community sample without direct experience [8]. In their design, the chronic conditions (Alzheimer’s or dementia, arthritis, cancer or depression) could be studied across family relationships (child, spouse and parent) and within experienced and community samples. Their research comparing conditions is especially important, as some have proposed that spillover effects may be particularly important for some mental health conditions (autism, ADHD, intellectual disability, schizophrenia) because social supports for families affected by these conditions may be limited. Prosser et al. found that spillover effects yielded by both samples were similar for all the specified chronic conditions and across family relationships. The large sample and innovative design provides important insight into the existence of spillover effects and points to the need for additional research to identify the magnitude of these effects for specific conditions. In particular, further investigation is needed comparing specific mental health conditions relative to physical conditions.

A number of studies sought to estimate spillover effects using generic instruments to preference-weight health outcomes. Using generic instruments to estimate family spillover effects require specific attention to instrument choice and study design issues. Neumann et al. used the health utilities index mark 2 (HUI2) to investigate whether caregiver utility scores declined (spillover disutility) with advancing stages of Alzheimer’s disease [9]. The null finding may have been attributable to the choice of instrument, as a subsequent study using the same database found decreasing health-related quality of life among caregivers with more sensitive instruments [10]. Tilford et al. examined a similar question with caregivers of children with spina bifida using the Quality of Well-being scale and found a decrease in caregiver scores as the severity of the child’s lesion increased [11]. Surprisingly, few studies have examined the question of what generic preference-weighted instruments are best suited to studying family quality of life effects for economic evaluations [12], despite numerous studies devoted to instrument comparisons across patient conditions.

As with any observational study, confounding can bias study findings and careful attention to this issue is warranted. For example, data were collected on children with hearing loss and their caregivers [13] and used to test the hypothesis that cochlear implantation would produce a gain in QALYs over and above the gain to the child from a spillover effect on the caregiver. The initial analysis revealed a large difference in HUI3 scores between the two groups of caregivers, but subsequent analysis indicated that the difference was due to selection bias and use of the HUI3 instrument in this context. Deaf caregivers were less likely to obtain cochlear implantation for their children, and given the HUI3 has a hearing domain, these caregivers had much worse health state preference scores creating the appearance of a spillover effect. In contrast, SF6D scores did not differ between groups irrespective of whether deaf caregivers were included or excluded from the analysis.

One area of potential importance for measuring spillover effects is sleep. We recently conducted an observational study of caregiver spillover effects associated with sleep treatment for clinically diagnosed children with autism using a registry [14]. The sample contained a large proportion of children with sleep problems who could benefit from treatment. Both the SF6D and the EQ5D were used to estimate potential caregiver spillover effects and produced similar marginal effects despite the large difference in average scores. The findings suggested that caregiver health would improve as much from treatment of the child as would the health of the child. Given the cost of behavioral treatments, the additional health gain to the caregiver greatly increases the probability that such treatments will be cost-effective.

Remarkably, we are unaware of any studies that have sought to capture family spillover effects in QALY terms using prospective randomized trial designs, although studies have evaluated caregiver interventions using generic instruments within randomized trials [15]. A recent trial by Romeo et al. comparing mirtazapine and sertraline to prevent depression in patients with dementia produced intriguing findings on cost–effectiveness [16]. The study measured QALYs for patients and unpaid caregiving time in the economic evaluation. The authors found a small gain in QALYs for the mirtazapine group, but a large decrease in unpaid caregiving time that greatly improved the cost–effectiveness profile for mirtazapine to treat depression. The authors explained the findings as mirtazapine potentially ameliorating sleep disturbances in patients with dementia, which improved the quality of life of the patient and relieved an important source of distress for the caregiver. Unfortunately, the authors did not design the study to capture caregiver spillover effects in QALY terms.

Had the Romeo et al. study measured caregiver QALYs in addition to unpaid caregiver time, it would have raised an important unanswered question in the literature addressed at the beginning of this essay. Should both unpaid caregiver time and caregiver QALYs be included in economic evaluations? Some may argue such inclusion amounts to double counting along the lines of the US Panel’s decision not to include productivity effects in the numerator of the cost–effectiveness ratio by arguing they are captured in the QALY [17]. Because generic instruments are silent on this issue [18], new guidance can be expected as the field moves forward, but our own perspective is that both spillover time and health effects belong in the evaluation.

Further questions in the spillover literature remain as well. Al-Janabi contrasted the evaluative scope (family vs patient) with the evaluative space (QALYs vs Well-Being) [19], the space issue again needs to be addressed under the expanded scope. Measures to describe spillover effects should extend beyond QALYs to capture more aspects of life than health alone. Research is progressing on measuring well being of caregivers [20] and decomposing spillover effects into family and caregiver components [21]. Much more work can be expected in this area in the future.

A recent review of family spillover effects [22] demonstrates significant progress in methods and data for measuring disutility and that much more work remains. We believe this research agenda provides a shot in the arm to the economic evaluation literature that can increase both interest in the field and the use of economic evaluation by decision-makers. Much research is needed to identify the conditions and treatments that produce substantial spillover effects and those that can be safely ignored. Evidence from causal research designs, especially randomized trials, is lacking and necessary to improve understanding of findings from cohort studies. Finally, guidance is needed on a range of practical issues to permit widespread implementation of spillover findings in the field. The result of this research will be better estimates to guide decision-making. For example, a family perspective for cost–effectiveness of services for children with autism is likely to be more precise if the potential benefit to caregivers and siblings is included. A family perspective is better aligned with constituent understanding and interests in this condition, creating a more useful application and appreciation for cost–effectiveness analysis. We suspect the same could be true for other conditions as well.

Disclaimer

The content of this article is solely the responsibility of the authors and does not represent the official views of the National Institute of Mental Health or the National Institutes of Health.

Financial & competing interests disclosure

N Payakachat is funded by the National Institute of Mental Health (Grant R03MH102495). The authors have no other relevant affiliations or financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript apart from those disclosed.

No writing assistance was utilized in the production of this manuscript.