Economic Valuation of Informal Care: Conjoint Analysis Applied in a Heterogeneous Population of Informal Caregivers

Introduction

Informal care, care provided by family and friends of the care recipient, gets increasing attention. This is not surprising because informal care is a substantial part of long-term care and substitutes formal home and nursing home care [1]. Also in the context of economic evaluations of health care, there is increasing attention for informal care. Adoption of a societal perspective, as commonly advocated, dictates that all relevant costs and health effects resulting from an intervention need to be incorporated in an economic evaluation [2]. Informal care should therefore not be ignored for several reasons. First of all, informal care can be a substantial part of total care provided and might thus contribute to the health-related quality of life of the care recipients under study. Second, provision of informal care may result in substantial costs (although perhaps not direct financial costs within the health-care sector), especially due to the time invested by the informal caregivers. It has been shown that substantial amounts of time are invested, for instance, in caring for patients with rheumatoid arthritis [3], suffering from the aftermath of a stroke [3], and a heterogeneous population of caregivers [4]. Third, the provision of informal care can result in health losses in informal caregivers [5,6], even leading to increased risks of mortality [7], and it has been argued that such health changes should be captured in an economic evaluation [8,9]. It is worth noting that even when an economic evaluation takes a less comprehensive perspective than the societal one, for instance the common health care perspective, the health effects in informal caregivers should not be ignored as they are the central outcome of interest [10]. After all, it is inconsistent to neglect the health effects in informal caregivers while attempting to maximize health with a given health-care budget.

Depending on the perspective chosen and evaluation type adopted (i.e., cost-benefit analysis, cost-effectiveness analysis, or cost-utility analysis), different methods can and need to be applied to incorporate informal care in economic evaluations (see for an overview, [9]). When a societal perspective is taken, both health effects and time costs need to be incorporated in the analysis. This can be done by using a single method that captures both elements (at least in theory), such as contingent valuation methods [11,12], or by trying to separate time costs and health effects, measuring the latter in terms of quality of life [6,8,13]. It is clear that the latter option is not compatible with a cost-benefit analysis, although the former is compatible with all three types of economic evaluation (even though it may provide less detailed information to the decision-maker, e.g., [9] or [14]).

In the literature, it has been suggested to incorporate the changes in the amount of informal caregiver time as direct non–health-care costs into the numerator of the cost-effectiveness ratio in economic evaluations of health-care interventions [2]. Two monetary valuation methods are often recommended in this context. First, the opportunity cost method, valuing hours spent on informal care at a wage rate (or a hypothetical wage rate caregivers would have earned on the market), and second, the proxy good method, valuing informal care hours at the wage rate of a professional caregiver [2,3,9,15]. Both methods, however, are rather insensitive to the heterogeneity (involving, e.g., different types of care and amounts of time) and dynamics (e.g., variation over time in informal care demands) of informal care. Moreover, they do not capture the full effects (e.g., positive and negative) of providing informal care for the informal caregivers, even when abstracting from possible health effects, which are also not captured. Therefore, both methods do not value the full impact of providing informal care (see [9,16] for elaborate discussions).

In theory, relatively new methods in the field of valuation of informal care, like contingent valuation (CV), conjoint analysis (CA) (also called choice experiments [17]), and the well-being valuation method are sensitive to the heterogeneity and dynamics of informal care. They are capable of capturing all relevant aspects of informal care, are sensitive to the different circumstances informal caregivers are faced with, and are capable of reflecting the true preferences of informal caregivers [11,12,18,19].

The method applied in this article is the stated preference method, CA [20]. In CA, respondents are normally asked to choose between different hypothetical scenarios or to rate them, highlighting different aspects, attributes of the commodity under valuation. CA stems from mathematical psychology [21] and is often applied in, for instance, the marketing literature [22,23], and in the transport economics literature [24], where it also has been applied to value time, in particular travel time. Van den Berg et al. [18] provide a detailed discussion of the advantages and disadvantages of CA to value informal care, also in comparison with other valuation methods. Here, we stress only that the application of CA to value informal care appears to be particularly useful, given the heterogeneous nature of the commodity informal care. Indeed, CA was especially developed to deal with the different components of a commodity. Not surprisingly, therefore, the popularity of CA in the field of health economics is increasing. Ryan and Gerard [25] provide an overview of its application in the field of healthcare, also in relation to economic evaluations (see for instance [26]). Different ways of performing a CA have been used. Although we asked our respondents to rate four different hypothetical situation and their own real-life situation, a majority of applications in health care use binary choices or paired comparisons instead of ratings [25]. We had two main reasons to elicit ratings. First, asking to rate a full set of scenarios generates more information in one step than using choice experiments. Second, having ratings for the hypothetical scenarios as well as for respondent's own real-life situation helped us to investigate the influence of individuals' own current situation on the ratings. CA has been used to value time also in the field of health care, for instance, to value waiting time, travel time, and to elicit preferences [25].

This article reports the use of a new CA to elicit the value of informal care in order to be able to incorporate this on the cost side of an economic evaluation of health care. Van den Berg et al. [18] report a first application of the method to value informal care provided to care recipients with rheumatoid arthritis using a relatively small sample (N = 135), showing promising results. Therefore, we set out to apply the CA to a relatively large sample of 865 informal caregivers providing care for care recipients with various diseases, including neurological, musculoskeletal, psychological, and circulatory diseases. The main focus of the article was to value the full impact of providing informal care on the informal caregivers by asking informal caregivers in a heterogeneous population to rate four different hypothetical informal care situations. We included a hypothetical monetary compensation in the four situations in order to derive a monetary compensation for providing different amounts of informal care and different care tasks. Moreover, we collected information about real-life care situations, including the amount and nature of provided care, health-related quality of life, and subjective burden due to providing care and analyzed whether they influenced the scenario ratings. We also asked informal caregivers to rate their own real-life caregiving situation compared with the four hypothetical situations. This enables us to test whether or not the rating of respondent's own real-life situation was associated with the scenario ratings. Asking informal caregivers to rate their own situation is a novel methodological attempt to deal with the informal caregivers' real-life circumstances.

The outline of this article is as follows. First, we introduce the application of CA in informal caregiving and describe the specific application of the CA in this study. Then we present the econometric model. Subsequently, we present the data and the results. We also derive a monetary value of informal care provided to a heterogeneous sample of care recipients. Finally, we discuss the results and conclude the article.

Development of CA for Informal Care

We define informal care as “a quasi-market composite commodity consisting of heterogeneous parts produced by one or more members of the social environment of the care recipient”[9]. The term quasi-market commodity is used to indicate that there exists no formal market for informal care and subsequently no market prices are available. Therefore, this study aims to derive a monetary value for informal caregiver's time. A heterogeneous commodity implies that informal care consists of different care tasks, like housework and personal care. The amount of informal care can also differ substantially between different caregiving situations, for example, according to the demand of the care recipient and the available amount of professional care or other informal care. Van den Berg et al. [9] give a detailed discussion of the heterogeneous nature and dynamics of the commodity informal care. We included some of these variations in our scenarios and tried to capture other parts in the independent variables.

Our scenario attributes were informal care hours, informal care tasks, and a monetary compensation. Table 1 gives an overview of the attributes chosen and the levels distinguished within each attribute.

We opted to leave the definition of light and heavy housework in the scenarios to the interpretation of respondents. When we developed our scenarios, there was no information about the nature and amount of care that informal caregivers in a heterogeneous population normally provide. Nevertheless, there was some information on a disease specific level, viz a population informal caregivers providing care for people with rheumatoid arthritis (RA) [27]. Therefore, the three care tasks were chosen because they included the most important informal care tasks [27]. We chose a weekly time investment that fell within the range of what most respondents would consider realistic. Riemsma et al. [27] found that informal caregivers provided care for an average of 33 hours per week for people with RA. We felt this amount would be an upper bound for our population, because of a different caregiver population, and also because Riemsma et al. [27] distinguished 28 care tasks as opposed to the 16 tasks we choose. (It is well known that time measurement is complicated, and that the measured amount of time depends on the questions posed—see, e.g., Juster and Stafford [28] for an overview.) We therefore expected our population to report somewhat lower time investments in informal care, and opted to include 7, 14, and 21 hours informal care per week, respectively, as relevant values of the time attributes. We selected our monetary compensation to encompass the Dutch market prices and health-care sector tariffs for unskilled housework of €8.53 per hour.

The three attributes and the selected levels result in 3² times 4 is 36 possible scenarios. Because we did not want to overburden the respondents by giving them too many scenarios, we reduced these to a manageable number of 16 scenarios by means of an orthogonal array. Such orthogonal array is still able to estimate main effects of attribute levels, but not interactions among attribute levels. In an orthogonal array, each level of one attribute occurs with each level of another attribute with equal or at least proportional frequencies. We used the Statistical Package for the Social Sciences (SPSS, Chicago, IL) orthoplan the procedure to arrange an orthogonal design. We then chose one reference scenario (a relatively extreme scenario) and distributed the remaining 15 among 5 groups of 3 scenarios based on reasonable variation in the scenarios. Thus, we ended up with 5 sets of 4 scenarios (each including the same reference scenario). The sets were randomly distributed over our respondents.

Because our objective was to derive a monetary valuation of informal care, we focused on the trade-offs between a monetary compensation and providing additional hours of care on the one hand and other care tasks on the other hand. Of course, the trade-off between informal care time and care tasks could also provide valuable information. Besides the information gathered in the CA exercise, we collected extra information on, for instance, the informal caregivers' objective and subjective burden and the rating of their own real-life situation compared with the hypothetical situations as described in the scenarios. This information will be used as additional independent variables to see how they are associated with the scenario ratings (and possibly indirectly the marginal rates of substitution).

Our central objective was to derive a monetary value of informal care consistent with the heterogeneous nature of this commodity. We therefore first asked the informal caregivers some questions about their current caregiving situation; including the number of years that the informal caregiver had already provided informal care, and how many hours they spent on informal care tasks during the last week according to a list of 16 care tasks. We distinguished between three types of care tasks: 1) household activities of daily living (HDL) like cleaning and cooking; 2) activities of daily living (ADL) like personal care; and 3) instrumental activities of daily living (IADL) like contacts with health care. Van den Berg and Spauwen [4] report that this way of time measurement overestimates informal care time (compared with the gold standard of the diary method).

Subsequently, we introduced a hypothetical caregiving situation with a set of four scenarios, and we asked respondents to rate them. See Figure 1 for an example of the exact question.

The hypothetical situations were presented at the back of the survey on a page that had a different color compared with the remainder of the survey. Rather than placing them on a separate page as in the previous study [18], respondents could fold this page in such a way that the scenarios could be placed next to the rating questions. Also, other parts of the survey, e.g., socio-demographics, were extended.

To get a better picture of the current informal care situation, we also measured health-related quality of life of informal caregivers using the EuroQoL (EQ-5D) [29]. We furthermore measured subjective caregiving burden. Subjective burden of informal care can be measured by a number of validated instruments [30–32]. We applied the Caregiver Strain Index (CSI) [33], because it contains a total sum score, unlike other instruments that focus on different subaspects of providing informal care, like financial problems or lack of family support. The CSI has a minimum score of 0 (indicating no subjective burden in terms of strain) and a maximum score of 13 (indicating much strain). A scoreof seven or higher means that the informal caregiver is substantially strained [34].

Econometric Model

In asking informal caregivers to rate four hypothetical caregiving situations, we assume these ratings to be a proxy of informal caregivers' (direct and indirect) utility and (direct and indirect) disutility (U^ic) derived from the four scenarios. This means that our dependent variable is latent. We only observe the respondents' ratings of 1 to 10 (which were transformed to the range of 0–9). Given the fact that respondents could only give a rating from 1 to 10, they were supposed to choose the rate that most closely represents their own feelings. These ratings are proxies for an informal caregiver's expected utility derived from the hypothetical scenario.

Formally:

(1)

where y* is an unobserved latent variable denoting respondent i's (i = 1, . . . , N) utility derived from scenario t (t = 1, . . . , 4), is a K-dimensional vector of scenario attributes presented to respondent i at scenario t. Furthermore, is a M-dimensional vector of respondents' and informal care characteristics, including the rating of an informal caregivers' own real-life situation, while ε_it is the error term. The ordered probit model is appropriate in this context; Greene [35] and Scott Long [36] discuss the model in more detail. We correct for clustering of respondent's scenario rating in the standard errors [37].

If we replace β in Equation 1 with the scenario attributes and with y_it, we get our empiric model:

(2)

where H is hours of informal care, LH is light housework, HH is heavy housework, and C is monetary compensation. In order to derive a monetary value of informal care, we kept the informal caregivers' utility constant while varying the level of the different components (attributes) of utility. These components consist in our application of care tasks, hours of care provided, and an hourly monetary compensation. By varying these attributes, one can derive the marginal rate of substitution between the attributes. Inclusion of a monetary compensation as one of the attributes in our application in the form of willingness to accept (WTA) enables us to derive the MRS between the other attributes and the monetary compensation. One has to be careful with the interpretation if one includes prices or costs as an attribute. (See for a critical discussion [38].) For instance, to derive informal caregiver's WTA for extra informal care provided (MRS_HC), we kept U^ic constant. Formally, this would be:

(3)

Data

Results

Discussion and Conclusion

In this article, we applied a CA to determine a monetary value of informal care in a heterogeneous population of care recipients. An important advantage of CA compared with more conventional methods like the opportunity cost method and the proxy good method, is CA's ability to capture more accurately the informal caregivers' preferences concerning this heterogeneous commodity.

First, it is worth noting that we chose to elicit real informal caregivers' preferences to provide care instead of the preferences of the general population. We did so because informal caregivers have experience in making choices about providing informal care. Therefore, they may be considered the best informed people, and, from a traditional welfare economic point of view, the appropriate individuals to state their preferences regarding informal care decisions. This selection may lead, however, to somewhat lower estimates of the monetary value of informal care as compared with asking the general public because of the selection of respondents. Indeed, we only include individuals who have already shown a willingness to provide informal care in the elicitation of preferences, while people who are unwilling to provide such care are expected to require a relatively higher compensation, ceteris paribus. (See Dolan et al. [40] for a more elaborate discussion of the different perspectives that could be used to elicit preferences.)

Another interesting question in relation to the selection of respondents is whether current informal caregivers can transcend their own caregiving situation in order to express their preferences about hypothetical caregiving situations described in the scenarios used in the CA. On the one hand, caregivers are expected to use their life experience when stating their preferences. Nevertheless, on the other hand, they need to abstract from their own specific situation to express their preferences for the different hypothetical situations. Current informal caregivers may be more capable of indicating their preferences for hypothetical caregiving situations, but the appraisal of different caregiving situations may also be influenced by their current caregiving situation, as was confirmed in our analysis. Respondents' ratings appear to be systematically associated with informal caregivers' characteristics. First of all, informal caregivers' ratings of their own real-life situation are positively correlated with the ratings of the hypothetical caregiving situations. Also, informal caregivers' subjective burden, caring for somebody on a waiting list for professional care and living together with the care recipient, have a statistically significant influence on the ratings. This seems to indicate that the informal caregivers' rating of the hypothetical care situations reflects, at least partly, their own experiences. This was not the case in [18], probably because CA was applied to a relatively homogeneous sample of caregivers in that article, viz caregivers for people with rheumatoid arthritis.

We find that on average, informal caregivers require an extra compensation of 81% per hour for providing 21 instead of 7 hours informal care per week. This implies a compensation of €12.36 per hour at a mean hypothetical compensation in the presented scenarios. The required compensation varies with the hypothetical initial number of hours of informal care provided. Such findings show that attaching one and the same value to every hour of informal care may be an oversimplistic approach to value informal care in economic evaluations. In that sense, more research is necessary in order to understand better what drives these valuations and how they could best be used in economic evaluations.

A limitation of our design is that we did not define light and heavy housework in the scenarios. This does not enable us to control for the heterogeneity of interpretations between light and heavy housework, which may result in measurement error. Moreover, unless the interactions among attribute levels are insignificant, a main effects design as used here can result in biased estimates of main effects. Another limitation of our presented scenarios is that the sample mean of provided informal care per week was far beyond the hypothetical amount of provided informal care, the scenario maximum: 105 versus 21 hours per week.

We feel that this application of CA shows that it is an interesting method to derive a monetary compensation of informal care. It would be useful as well as challenging to add additional attributes and levels to the scenarios in future research. This could, however, put greater (cognitive) burden on the respondents, probably at the costs of higher and perhaps selective nonresponse of respondents or item nonresponse. On the other hand, it would be interesting to deal with greater heterogeneity of informal care by adding, for example, more care tasks. For instance, instead of personal care, one could distinguish support with washing the care recipient and support with dressing to get more detailed information about informal caregivers' preferences. This would also solve the problem raised regarding the subjective interpretation of care tasks in our study. Regarding hours of care provided, one could create greater variation. Also, the equal spreading of hours provided over the days of the week in our design could be changed. It is worth testing whether or not it makes sense to provide 7 hours at one day or every day of the week for just 1 hour. More information on the nature of the illness or the relationship with the care recipient or even the health impact of caregiving could also be inserted. Moreover, an attribute dealing with the trade-off between professional and informal care would probably provide interesting information. Oral surveys with trained interviewers may help to overcome part of the item nonresponse problem.

In sum, we suggest that CA may be a promising alternative for the existing methods to value informal care, like the opportunity cost method and proxy good method.

Source of financial support: The Netherlands Organization for Health Research and Development (ZON-MW) (Grant-No. 945-10-044).