The “Stock of Time” Method: A New Approach to Calculate Indirect Costs and Benefits in Economic Evaluations

Leisure Time

Our model assumes that changes in working time and in the quantity of leisure time are not included in the health state, while the quality of leisure time is included. Nyman ³ stated that income effects are not included in the HRQoL measures but effects on leisure time are, because 3 of the 4 most widely used generic health status questionnaires include a question related to role functioning (EQ-5D, SF-6D, and QWB). For example, the EQ-5D questionnaire contains 5 dimensions: mobility, self-care, usual activities, pain/discomfort, and anxiety/depression. Respondents are asked to self-assess their functioning on each of these dimensions, on a 3-level scale. The ability to participate in paid or unpaid work is subsumed in the third dimension of the EQ-5D questionnaire, labeled “usual activities.” The 3 levels from which respondents choose on this dimension are:

According to Nyman (p. 1394): ³

“By explicitly including these questions, the EQ-5D, SF-6D, and QWB capture the intrinsic value of the role functioning aspect of the health state, the aspect that relates most closely to the amount of leisure that the respondent would have in that health state.”

However, the question on role functioning has no references to the amount of leisure time nor to time for work. To the extent that respondents do indeed take into account the effects of their health condition on work and leisure, the EQ-5D questionnaire would seem to pick up the quality of leisure time and working time, not the quantity. Moreover, it would be inconsistent to assume that people do incorporate the quantity of leisure time in their HRQoL scores but not the quantity of working time, as the question is the same for working hours and leisure time. So, if the effects of a change in working hours are not included in the HRQoL score and have to be added separately (in terms of production loss), the same would hold for the effects of a change in leisure time.

There is a small body of literature that tries to establish if effects on income and leisure are included when assessing health states, based on questionnaires among health professionals ¹⁹ and the general public.^20–22 Respondents were asked to value 1 or more health states. These studies find that most respondents (ranging from 61% to 84%) said they spontaneously include effects on leisure time. The questionnaires do not distinguish between effects on the quality of leisure time and the quantity of leisure time. The studies find hardly any differences in HRQoL scores between respondents who did and did not consider leisure while valuing the health state. This seems logical: it is hard to imagine respondents would not spontaneously take the quality of leisure time into account as all 5 dimensions on the EQ-5D are relevant for the quality of leisure time (as impaired mobility, pain, and anxiety decrease the quality of leisure time). However, they found significantly lower HRQoL scores for respondents who thought the health state would affect their ability to enjoy leisure. This is consistent with our hypothesis that the quality of leisure is spontaneously included while rating HRQoL. The results of the scarce literature are thus consistent with the model in that the HRQoL scores of respondents take into account the ability to enjoy leisure (the quality of leisure time). The studies provide no evidence that the amount of leisure time is spontaneously included in the HRQoL score.

If the assumptions of the model are valid, the amount of leisure time has to be added separately to the direct health effects reflected in the QALY gain. Moreover, we propose that the change in leisure time is not measured as the inverse of the change in working hours. Surveys that measure the change in working hours due to a change in health thus should also measure the change in leisure time and recovery time. Empirical research shows that poor health is related to more recovery time and less time for paid work.^11,12

To incorporate the total stock of time in health evaluations, 2 practical issues have to be tackled: what to include in recovery time and what is the value of leisure time? As to the definition of recovery time, Grossman ¹⁷ distinguished time invested in health (such as preventive activities) and lost time due to sickness or injury. The definition in that empirical literature is in accordance with lost time due to sickness or injury, as this requires more sleep and personal care. Preventive activities, such as physical exercise or screening, are not distinguished in the empirical literature. Sports and fitness, for example, are defined as leisure time.^10,11 Not only empirically, but also conceptually, it will be difficult to distinguish preventive activities from leisure time, as people may enjoy these activities. We therefore suggest considering these activities as leisure time. There is also discussion about including “eating and drinking at home.” Podor and Halliday ¹² saw this as leisure time, whereas Gimenez-Nadal and Molina ¹¹ viewed it as an activity to stay alive. Although eating and drinking are necessary for staying alive, the time spent on eating and drinking is a choice and might as well be counted as leisure time.

As to the valuation of leisure time, this is at the margin equal to the net wage rate (see Appendix 2). For nonmarginal changes, additional assumptions have to be made as the marginal value of leisure time diminishes as the amount of leisure time increases. It might be zero for someone who does not work. For the economic evaluation of nonmarginal changes of leisure time, the average value of leisure time gained or lost has to be taken into account, not the marginal value. This value will be higher than 0% and lower than 100% of the net wage. A study of Verbooy et al. ²³ finds a willingness to pay for leisure time of €12.27 per hour for people with paid employment (leaving €0 responses out) based on a contingent valuation study in the Netherlands in 2018. As the average net wage rate in 2018 was about €18 per hour (a gross wage rate of €22 per hour), the value of leisure time would have been roughly 70% of the average net wage rate. We recommend performing sensitivity analysis with different percentages of the net wage.

Valuation of Changed Production

Most guidelines for health evaluations do not take into account productivity changes due to health changes. ² Guidelines that take a societal perspective do incorporate productivity changes. ² However, there is an ongoing debate about the method to value such changes in production due to health changes. Both the HCA and the FCA are proposed, with relatively few studies using the latter method. ²⁴

The HCA assumes that the value of the production change is the present value of all lost or gained future production of the individual. ⁶ Guidelines recommend using the average gross wage to calculate the value of production changes. ² In case of a change in longevity, there has been discussion about double counting, as the consumption in lost or gained life-years is already captured by the value of a QALY,^25,26 which is consistent with our model. The “net human capital approach,” as Johannesson ²⁶ called it, has not become standard practice in health evaluations, although the Danish guideline proposes to perform a sensitivity analysis with this approach. ²⁷

The FCA takes a short-run perspective. In the short run, there might be unemployment, and workers can be replaced by unemployed persons. However, business cycles with high and low unemployment alternate. The FCA would curiously imply that there is no productivity gain of getting sick people—who have already been replaced—back to work, even if recovery occurs in a tight labor market. Moreover, if a sick worker can be replaced by an unemployed person, it would not be costless, because the unemployed person taking over the work of the sick person would lose leisure time. ²⁸ Friction cost proponents indicated that leisure is not treated as “costless” but rather that gains and losses of leisure at a societal level even out between the sick worker and the replacing worker.^29,30 They thus implicitly assume that health does not influence recovery time.

Numeric Examples

The following numeric examples indicate that the results of our model —the stock of time approach (STA)—will often yield results that are in between the HCA and the FCA. The example is inspired by an economic evaluation of the costs of head and neck cancer, which calculates the production losses of the cancer using the HCA and the FCA. ³¹ We simplified the underlying data to give a clearer insight in the consequences of the methods and added possible effects of a hypothetical intervention to cure this disease.

Individuals are assumed to be diagnosed with the cancer at an age of 50 y. Because of the illness, 20% of the patients die at an age of 51 y. The amount of time available for work and leisure before the illness was 16 h per day: 24 h minus 8 h of recovery time (5,840 h per year). Because of the illness, survivors decrease working hours by, on average, 2 h per day (730 h per year). To illustrate the effect of recovery time, we consider 2 situations. In situation 1, the average amount of recovery time stays the same, which means leisure time increases with 730 h a year, as working time decreases with 730 h. In situation 2, recovery time increases with 4 h per day—1,460 h per year. This means the amount of leisure time decreases with 730 h per year.

Table 1 presents the total welfare loss of the illness due to a change in working hours and leisure time, calculated using the 3 methods: the HCA, FCA, and STA. The calculations assume that premature death leads to a loss of 14 productive years (from age 51 y to retirement age at 65 y). We assume a gross wage rate of €20 per hour and an average tax rate of 20%. Furthermore, for the FCA, a friction period of 16 wk is assumed. For simplicity, a discount factor of 0% is used. The average willingness to pay for leisure time is assumed to be 70% of the net wage. Table 2 also contains the direct welfare losses in terms of lost QALYs, which are valued at €300,000 per patient.

OPEN IN VIEWER

Table 1

Average Welfare Loss per Patient Calculated with the HCA, FCA, and STA

Welfare losses	HCA	FCA	STA, Situation 1	STA, Situation 2
1. Value of lost production due to premature death (20% of patients)	−€102,200	−€2.246	−€20.440	−€20.440
2. Value of lost production per survivor	−€163,520	−€3.594	−€163.520	−€163.520
3. Value of lost/gained leisure time survivors	—	€0	€91.571	−€91.571
Total welfare loss production and leisure time	−€265.720	−€7.600	−€92.389	−€275.531
Value lost QALYs	−€300,000	−€300.000	−€300.000	−€300.000
Total welfare loss	−€565,720	−€307,600	−€392.389	−€555.091

HCA, human capital approach; FCA, friction cost approach; QALY, quality-adjusted life-year; STA, stock of time approach.

Change in production in HCA and STA cumulated over a period of 14 y. Change in production in FCA cumulated over a period of 16 wk.

OPEN IN VIEWER

Table 2

Average Costs and Benefits per Patient Calculated with the HCA, FCA, and STA

Costs and benefits	HCA	FCA	STA, Situation 1	STA, Situation 2
Benefits (10% of welfare loss)	€56,572	€30,760	€39,239	€57,553
Cost of the intervention	−€50,000	−€50,000	−€50,000	−€50,000
Net benefits	€6,572	−€19,240	−€10,761	€7,553

HCA, human capital approach; FCA, friction cost approach; STA, stock of time approach.

As Table 1 indicates, the largest welfare loss due to premature death occurs in the HCA. This is because all lost production is valued. The FCA leads to far lower costs because only 16 wk of production loss are counted, instead of 14 y. According to the STA, the production loss due to premature death is only the external effect on the rest of society, which is 20% (the tax rate) of the production loss in the HCA calculation. The HCA thus leads to a large overestimation of the welfare loss due to premature death, while the FCA leads to an underestimation given a friction period of 16 wk. The value of lost production of survivors amounts to €163.520 in the HCA and the STA. According to the FCA, this welfare loss is only €3.594, as it is calculated over the friction of 16 wk, instead of over 14 y. The gain or loss in leisure time of survivors is not taken into account in the HCA, while the FCA assumes the value of gained leisure time is zero, because it evens out with the lost leisure time of the unemployed who take over the paid work after the friction cost period.

In situation 1, the leisure time for the patient increases, which decreases the total welfare loss due to the lost working hours. The STA calculations value the increased leisure time of the patient at 70% of the net wage rate, so, in total, at 56% of the gross wage (70% × 80%). Compared with the HCA, the STA leads to a lower total welfare loss by taking into account the gain in the leisure time of the patient. As compared with the FCA, the STA leads to a higher total welfare loss because the value of lost production according to the FCA is measured over a period of 16 wk, instead of 14 y.

In situation 2, the patient loses leisure time, leading to a larger welfare loss than in situation 1 according to the STA. The STA result now is almost equal to the HCA. The reasoning behind this result is, however, quite different. On one hand, the value of lost production due to premature death in the STA is smaller than in the HCA; on the other hand, the HCA does not take the value of lost leisure time of survivors into account.

A health intervention to cure or prevent the illness will have higher benefits in case of larger welfare losses due to the illness. Suppose the intervention would decrease the welfare losses by 10% and the cost of the intervention is €50,000. Table 2 shows that this health intervention would lead to estimated welfare gains if the indirect welfare effects are calculated with the HCA and with the STA, but only in case illness leads to a decrease in leisure time (situation 2). Using the FCA will lead to estimated negative welfare gains, as will the STA if the illness does lead to a gain in leisure time.

The overestimation and underestimation of the HCA and the FCA depend heavily on the nature of the illness or intervention. The HCA’s overestimation will be especially high if there is a significant change in longevity. The underestimation of the FCA will be particularly large in cases of longer periods of sickness and a loss of leisure time due to sickness.