Taken together, results across three independent data sets for a wide range of health states provide strong support for the hypothesis that being a caregiver influences the way one responds in a TTO health state valuation task. Compared to respondents who were not caregivers, caregivers tended to be less willing to trade any time in the TTO tasks. Furthermore, when caregivers did trade, they tended to trade less time, resulting in consistently higher utility scores than non-caregivers. The difference between caregivers and non-caregivers in health state utility was often quite large, with between-group differences up to 0.22 on the standard utility scale anchored to 0 (dead) and 1 (full health). Utility differences of this magnitude are large enough to have a substantial impact on the outcome of a cost-utility model. Thus, caregiver status may be an important demographic variable to consider when collecting, interpreting, or using utility data. The current results also serve as a reminder that health state utilities are shaped not only by the health states themselves, but also by characteristics of the respondent valuing the health states.
SG data were not available in the data sets for Studies 2 or 3, but in Study 1, caregiver status appeared to have less influence on SG responses than on TTO responses. For example, caregiver status had a less pronounced impact on willingness to gamble in SG than on willingness to trade in TTO (Table 3). In addition, mean differences in utility scores between caregivers and non-caregivers were smaller for SG than TTO for three of the four health states in Study 1 (Table 5). Perhaps the SG task is less susceptible than TTO to the respondent’s current life situation because of the abstract nature of the task. In the TTO, participants may consider the direct impact of shorter lifespan on their dependents, whereas the probabilities involved in the SG task may not lead respondents to think of their own responsibilities as directly. Still, SG utilities had a smaller utility range for caregivers than for non-caregivers, the caregivers did have statistically significantly higher SG utilities for health state C, and between-group differences for health states B and D may be considered clinically important (Table 5). In sum, while results of the TTO task appear to be more strongly influenced by caregiver status, SG responses may also be affected. Future research is needed to clarify the impact of caregiver status on SG responses.
Current findings add to results reported by van der Pol & Shiell suggesting that people are less willing to trade time in the hypothetical TTO task if they know that others are directly dependent on them. Whereas the van der Pol & Shiell study focused only on a sample of 30 parents of young children, the current study included both parents of dependent children and individuals who considered themselves to be caregivers of an adult. This broader group was used because respondents in previous utility assessment studies (conducted by authors of the current studies) have mentioned the impact of caring for an elderly or disabled relative. However, the great majority of caregivers in the current study cared for children rather than adults (i.e., 83 of 106 caregivers; 78.3%). As a sensitivity analysis, current analyses were re-run in two other subgroups: including parents of dependent children and parents of any child regardless of whether the child is still dependent. Results of these sensitivity analyses followed the same general pattern as results of the primary analysis. The parent subgroups were generally less willing to trade time (i.e., resulting in higher utility scores) than comparison groups. Future research, perhaps involving qualitative interviews asking participants to explain their TTO responses, is needed to explore the separate influence, if any, of parenting and other types of caregiving on utility scores. The current analysis may also be replicated in larger samples with various types of caregivers, including larger numbers of caregivers of adults.
Parenting and other types of caregiving may reduce willingness to trade due to increased altruism and a sense of responsibility for another individual’s well-being. In addition, even if a child is grown, being a parent may reduce willingness to trade because parents may want to live longer to experience specific milestones in their child’s life. For example, parents have reported that they wanted to live long enough to see their child’s wedding or graduation when responding in previous TTO studies conducted by the authors of the current study. Such extrinsic goals, which occur at a specific point in the future, have been proposed as one factor that could limit willingness to trade time in TTO or gamble in SG.
Current results also have implications for the conceptual foundations of utility assessments. The standard preference elicitation and quality-adjusted life year (QALY) models ask respondents to evaluate health states based on the effects on their own welfare. However, results reported in this paper and elsewhere in the literature suggest that respondents also consider the effects of their health on the welfare of others when valuing health states. Several authors have developed conceptual frameworks that broaden the utility model to take into account not only the preferences of respondents for their own health outcomes, but also their preferences for effects on other people such as spouses, children, and adult dependents. These frameworks are often referred to as a family or household utility function[25–27]. Current findings suggest that future research to develop and implement these family or household utility approaches could be useful, although this methodology could be complex, and the scores may not be entirely comparable to conventional utilities.
Study limitations related to sample selection should be acknowledged. During screening, efforts were made to avoid over-representing respondents from any particular age, ethnic/racial background, gender, or employment group. However, recruitment was not stratified or designed to be nationally representative in any of the three studies used for this secondary analysis. Furthermore, the recruitment strategy, requiring that all study participants respond to newspaper or online advertisements, could have introduced selection bias. The extent to which current results may differ from utilities derived from a nationally representative sample in the US, UK, or any other country is not known.
Another potential limitation is that a respondent’s personal experience with the condition described in health states could influence utility scores. In Studies 2 and 3, none of the participants were ever diagnosed with the condition described in the health states (ADHD and schizophrenia, respectively), although it is possible that they knew people who had received these diagnoses. It is not known whether participants in Study 1 had ever been diagnosed with the condition described in the health states (osteoarthritis). Personal knowledge of the conditions being rated could have an impact on utility scores, and it is not known whether this was a confounding factor in the current analysis.
It is also possible that aspects of the interview methodology could have introduced bias. For example, although health states were initially presented to participants in random order during ranking or VAS introductory tasks, health states were then presented in the TTO tasks in the order in which participants ranked them. At multiple points during the interviews and at the end of the interviews, participants were asked to review and compare all of their utility scores in order to minimize ordering effects and learning effects that could affect differences among health states. Furthermore, if there were any ordering or learning effects, any resulting biases would likely be the same for caregivers and non-caregivers, and consequently, this issue is unlikely to have an impact on the issues examined in the current secondary analysis.