Willingness to pay for QALY: perspectives and contexts in Japan

Abstract

Objectives: Theoretically, willingness-to-pay (WTP) for quality-adjusted life years (QALY) can vary depending on social and personal preferences and on ex-ante and ex-post settings. However, empirical investigations into the theoretical differences are lacking. In Japan, setting the threshold has been controversial since a pilot project to implement health technology assessments (HTA) launched in 2016. The study aim is to estimate WTP values for one additional QALY from different perspectives, health statuses, and contexts to confirm the difficulty in setting a uniform price threshold using WTP.

Methods: More than 1,000 respondents representing a cross-section of the Japanese population answered each of nine questionnaire decks in an online panel. WTP values were estimated on three different perspectives (personal, social, and socially inclusive) and on two contexts (ex-ante and ex-post). This study primarily used the non-parametric spike model based on double-bounded dichotomous choice (DBDC) settings to obtain the conditional WTP values.

Results: WTP per QALY was higher in the severe health status than in the moderate health status from all perspectives. Respondents from the socially inclusive perspective estimated the highest WTP value for a new drug. Respondents were also asked about life-threatening diseases in ex-post and ex-ante. The WTP value in ex-ante was higher than in ex-post, and demographic factors affecting the WTP were different in both situations. The various WTP values were obtained from these surveys.

Limitations: The analysis was based on data collected from an internet panel, which could be biased. There is also a risk that respondents answered the questionnaire differently if asked in everyday situations.

Conclusion: Use of a uniform price threshold may not be appropriate in policy settings, because it may not reflect diverse preferences based on different situations, such as disease type and severity. Setting a price threshold as Japan institutes an HTA system requires further research.

Keywords:

• Quality-adjusted life years
• willingness-to-pay
• health technology assessment
• incremental cost-effectiveness ratio
• Japan

JEL classification codes:

• I10
• I18

Introduction

Countries with universal healthcare systems allocate resources based on assessments to ensure rationality and fair decisions. Cost-utility analyses, using QALY as an outcome measure, can be useful for new health interventions having been one of the ways to allocate resources in some countries among various healthcare interventions^1–4.

Japan’s health insurance system, under which almost every drug is covered, has been in place since 1961⁵. Under that system, the cost of drugs is reimbursed within 60–90 days once a drug has been approved⁶. The efficacy and safety of the drugs were considered mainly during the decision-making process of approval and reimbursement. There was no mandatory clause to consider cost-effectiveness. However, the rapid growth of total healthcare expenditures has led to growing concern over the sustainability of the healthcare system. The increase of total healthcare expenditures, which accounted for JPY 42.2 trillion (USD 422 billion, USD 1 = JPY 110) in fiscal 2017⁷, is not solely attributed to an aging population, but also to the introduction of high-cost medications, such as direct acting antivirals for hepatitis and immuno-checkpoint blockade drugs.

Reimbursement prices of medications in Japan, both initial prices and revised prices, are set by the government⁸. After the drug goes on the market, market mechanisms are used and the price is revised once every 2 years in principal. The wholesale price is investigated by the government, and the new redemption price is determined by adding consumption tax and price adjustment rate (2% of market price for securing drug supply to the market) to the weighted average wholesale price for hospitals and pharmacies⁸. Moreover, there have already been various price revision systems put in place to reduce the reimbursement price of medications that exceed estimated sales such as urgent drug price revision⁹. However, apart from the existing drug pricing system, the government has tried to implement health technology assessment (HTA) into the public healthcare system. The pilot program was introduced in April 2016 and entire one was started from April 2019¹⁰. Unlike countries that had already introduced HTA systems, Japan intends to use HTA data for price revisions of existing drugs. The unique characteristic of the Japanese HTA provisional system would be that the incremental cost-effectiveness ratio (ICER) value of each product would be directly reflected in the price revision rate. However, concerns were raised that there was no room to allow for uncertainties. In terms of the function for converting ICERs to price revision rates, a “slope” like linear function which was used in the pilot phase (2016-2019) was replaced with a “step” like function, by which the revision rate are chosen from three distinct levels according to ICER values, for entire phase (2019-)^11,12. According to the government’s guideline, the QALY is to be the preferred method¹³.

Using HTA data, prices of candidate drugs would remain the same if their ICER values are less than JPY 5 million per QALY gained. The value of JPY 5 million can be regarded as the benchmark value of cost-effectiveness, or a quasi “threshold value”. At first glance, JPY 5 million seems to be consistent with values suggested in previous studies. Ohkusa stated that the social willingness-to-pay (WTP) value for one additional QALY was JPY 5.2–7.4 million in 2003¹⁴ and JPY 6.0–7.0 million in 2006¹⁵. Shiroiwa et al.¹⁶ conducted similar surveys using double-bind dichotomous choice (DBDC) methods and estimated the WTP values from social and personal perspectives to be JPY 5.0–5.8 million and JPY 4.7–5.4 million in 2010, respectively. In a further survey, Shiroiwa et al.¹⁷ concluded that various factors, including the initial health status and the magnitude of improvement or decrement in the quality-of-life (QoL) scores could affect WTP values, which ranged from JPY 2 million to JPY 8 million. Nimdet et al.¹⁸ conducted a systematic literature review of WTP studies and argued that WTP values for life-saving cases are substantially higher than those for improving QoLs. We could not justify the “benchmark” ICER value, or the ICER value of the starting point of price reduction, simply by referring to previous studies of the WTP value for one additional QALY. Theoretically, threshold values for ICERs would be fundamental tools for incorporating HTA data into the decision-making process. However, in an actual setting, few HTA agencies explicitly set threshold values. Schwarzer et al.¹⁹ reviewed availabilities of threshold values in 10 HTA agencies and found that only two institutions, NICE in the UK, and HITAP in Thailand, explicitly announced threshold values. Skoupa et al.²⁰ conducted a similar survey of 10 European countries and found that only two additional HTA agencies, those in Poland and Slovenia, introduced the threshold value. Defining the value is more ambiguous than was expected. Even NICE, whose “official” threshold value of GBP 20,000–30,000 is often referred to, did not have a clear-cut rationale for setting the threshold, which was decided on an empirical rather than theoretical basis. NICE also set up a system for life-extending end-of-life treatment, under which the threshold value is doubled to GBP 50,000^21,22.

Tsuchiya and Watson²³ proposed that personal, social, and socially inclusive perspectives and ex-ante and ex-post contexts be taken into account when questionnaires for WTPs are constructed. Dolan et al. (2003)²⁴ originally presented these variations of perspectives and contexts. The three perspectives concern whom the respondent is asked to think about. Personal perspective refers to the situation in which assessors themselves receive the treatment at their own expense. In the socially inclusive perspective, members of a group collectively pay for treatment of patients who are also members. Thus, assessing respondents can also be patients and pay into a collective budget for expenses. In the social perspective, respondents play roles as assessors and can be neither patients nor payers.

The two contexts refer to the relative point in time when the preference is elicited and the degree of certainty with the need for healthcare differs between them. In the ex-ante approach, participants have not yet suffered from an illness, but would with certain probability, and are asked how much they would pay to lower their risks or insurance payments. In the ex-post approach, respondents already suffering from an illness were asked how much they would pay for a specific treatment option.

No information is currently available on the extent that these factors could affect the WTP value, particularly in Japan. The objective of our study, using a DBDC survey, is to estimate the WTP value for one additional QALY, as well as how it would vary according to the following external factors: (i) disease severity (moderate and severe); (ii) perspective (personal, socially inclusive, and social); (iii) life-threatening status (yes or no); and (iv) conceptual (ex-ante or ex-post).

Methods

Survey structure

Tsuchiya and Watson’s²³ online supplement (2017) introduced stylized preference elicitation scenarios and questions which specified the timing of the illness and the nature of the risk, the patient (the user of the treatment), the payer of the treatment, and the assessor of the value of the treatment. Based on their scenarios and questions, we used a questionnaire survey to estimate WTPs for one additional QALY after asking demographic factors such as age, gender, marital status, employment status, education, and income. We qualified WTPs into two particular conditions of health—the moderate state = 22,233 (Japanese EQ-5D utility score of 0.536) and the severe state = 55,424 (Japanese EQ-5D utility score of 0.169)—to compare WTPs from personal, socially inclusive, and social perspectives, as shown in Table 1. In addition, WTP per QALY for situations of life-threatening disease was also studied. In this case, we asked WTPs for 0.5 QALY gains and 2.5 QALY gains under ex-ante, in which the illness has not occurred yet. In addition to the ex-ante situation, we asked WTPs for 0.5 QALY gains under ex-post, in which respondents know they have the illness (Table 1).

Table 1. Pattern of questionnaire.

No	Health state (EQ-5D-5L)	Perspective	QALY gained	Context
1	Severe	Personal	1.662	Ex-post
2	Moderate	Personal	0.928	Ex-post
3	Severe	Socially inclusive	1.662	Ex-post
4	Moderate	Socially inclusive	0.928	Ex-post
5	Severe	Social	1.662	Ex-post
6	Moderate	Social	0.928	Ex-post
7	Life-threatening	Personal	0.5	Ex-post
8	Life-threatening	Personal	0.5	ex-ante
9	Life-threatening	Personal	2.5	ex-ante

WTP questionnaires

In the personal scenario, respondents were asked to assume they were in one of the health statuses described above. If the respondent did not take an approved medicine, this health status would last for 2 years. The medicine used would lead to complete recovery, but the patient would have had to pay the whole cost of care because the drug would not have been covered by public health insurance. WTP for this treatment was estimated by the answers. In the socially inclusive scenario, respondents were asked their WTP to regain their own health, as well as that of their neighbors, in a system in which everyone in the community shares payments equally. In the social scenario, respondents were asked to assess WTP for a patient in a distant place. In a personal scenario, respondents are users, payers, and assessors at the same time. In a socially inclusive scenario, respondents are members of user-and-payer groups. In a social scenario, respondents are assessors and are neither users nor payers.

WTPs for ex-ante and ex-post were elicited only for life-threatening diseases. The ex-ante context assumed that respondents had a 20% chance of cancer with a life expectancy of 6 months. The respondents were asked their WTP premiums for insurance plans covering the full amount of treatment that would extend life expectancy by 6 or 18 months. In the ex-post scenario, personal WTP was asked when the cancer was first diagnosed. Similar to the previous ex-ante context, the medicine would extend the respondent’s life expectancy by 6 months. The recipient would pay the whole cost of care in both cases, because the treatment was not covered by public health insurance.

Double-bounded dichotomous choice and bid value

In this study, double-bounded dichotomous choice (DBDC) was used to estimate WTP per QALY. This technique is more efficient than a single-bound one as more information is elicited about each respondent’s WTP²⁵. First, five bid value in Table 2 was shown randomly and the respondents were asked whether they would pay the value for the treatment. Second, bid levels were presented to the respondents sequentially with another dichotomous choice. The first bid value was randomly selected from five levels from JPY 1.5 million to JPY 15 million. Depending on the first answer, the next bid values were changed and second-stage bid values were shown. If respondents answered “yes” to the first question, a higher bid value was shown in the second question and vice versa. The range of second-stage bid values was JPY 0.75–30 million. Bid values in DBDC are shown in Table 2.

Table 2. Bid value of questionnaire.

No	First bid-value Q1 (million yen)	Second bid-value (million yen)
		Q1 = No	Q1 = Yes
1	1.5	0.75	3.0
2	2.5	1.25	5.0
3	5	2.5	10
4	10	5.0	20
5	15	7.5	30

Questionnaire example

Examples of WTP questions in three scenarios were as follows.

1. Personal scenario: “Please imagine that you have disease A. Without treatment, this disease lasts for 2 years and then will be cured. The health status of disease A is shown below (health status based on EQ-5D-5L is shown). New treatment B completely cures disease A. Please note that treatment B is not covered by public health insurance and that you have to pay for all of this treatment. Would you purchase treatment B if the treatment cost is JPY XXX?”

2. Socially inclusive scenario: “Please imagine that you and some residents in your region have disease A (the same explanations of disease A, the health status, and treatment B). The whole cost of B is paid by the money collected from all residents in your region. You and other patients will pay some of the cost because the whole cost is shared among all inhabitants. How much should your region pay for treatment B for a patient?”

3. Social scenario: “Please imagine Prefecture X, which is distant from your region. Healthcare cost in X is paid by the collected money from residents in X. You do not have to pay that because you do not live in X. Disease A is only endemic in X. Those who live outside X, including you, have no chance of getting disease A (the same explanations of A, health status, and treatment B). The entire cost of treatment B is paid by the money collected from all residents in X. How much should Prefecture X pay for treatment B for a patient?”

The second dichotomous choice question is followed by the above statements in each scenario. In the scenario of a life-threatening disease, similar explanations to the personal scenario were presented with the exception that disease A is not curable and life expectancy is clearly stated with or without the treatment. In this scenario, life expectancy was set for 6 months and treatment prolonged it to 2 years (Supplementary Figure S1).

In total, each respondent answered nine WTP questions in which three perspectives were asked in two non-life-threatening health states, and two contexts (two ex-ante questions and an ex-post question) were asked in a life-threatening scenario. The patterns of the questionnaire are shown in Table 1.

Respondents

We used an online panel service (Intage Inc.) to recruit respondents from February–May 2018. This panel was drawn from a pool of ∼1.6 million people and consisted of respondents aged 21–69 years. Subjects were asked to provide consent to participate in the survey after reading the objectives and a document on the confidentiality of the survey. Participants could withdraw from the survey at any time. Recruitment for the survey ended when the targeted sample size of 1,000 was reached. Sampling was stratified according to age and sex to reflect the general distribution in Japan.

Statistical analysis

In the DBDC method, an acceptance curve showing the relationship between the probability of answering “yes” and bid values was estimated using the non-parametric Turnbull method²⁶. The mean WTP was acquired by calculating the area under the acceptance curve. The mean WTP for one QALY gained was calculated by dividing mean WTP by QALY gained shown in Table 1 for each WTP question. For example, if mean WTP was estimated to 7 million yen for question 1 in Table 1, the mean WTP for QALY was 4.21(7/1.662) million yen. This conditional WTP (cWTP) comes from respondents who had non-zero WTP. The sample mean WTP was calculated by multiplying cWTP and the percentage of people answering “yes” to the first question. If 80% of respondents answered “yes” to the first question, the sample mean WTP for question 1 was 3.37 (4.21*0.8) million yen. Logistic regression analysis was performed to evaluate the relationship between responses to the first question (willingness-to-pay more than JPY 1) and demographic factors. We used STATA15.1 for the analysis.

Results

More than 1,000 respondents representing a cross-section of the Japanese population were sampled from the online panel. Demographic characteristics are shown in Table 3. Sampling stratified by sex, age, and other factors was successful. Respondent demographics were similar to those of the Japanese general population²⁷.

Table 3. Demographics of respondents.

	Percentage
Average age	47
Gender
Male	52%
Female	48%
Education
High school	32%
Junior college	14%
Graduation from 4-year college oruniversity	34%
Graduate school	4.2%
Others	16%
Employment
Full-time worker	46%
Part-time worker	17%
Self-employment	6.0%
Homemaker	21%
Others	9.0%
Marriage status
Single	29%
Married	62%
Divorce, bereavement	8.0%
Income (JPY 10,000)
<200	40%
200–<600	35%
600–<1000	9.0%
1000+	2.0%
Non-respondent	15%

The mean WTP values per QALY from each questionnaire are shown in Table 4. First, the differences of WTP per QALY from the three perspectives and in the two health conditions were evaluated (Questionnaires 1–6). WTP per QALY in the severe health status category is higher than that in the moderate health status for all perspectives. Therefore, respondents were willing to pay more in severe health conditions overall. Comparing perspectives, the socially inclusive perspective estimated the highest WTP values for a new drug, followed by social and then personal. WTP values for life-threatening diseases in the ex-ante and ex-post contexts were evaluated (Questionnaire 7–9). The WTP for life-threatening diseases in ex-ante was significantly higher than in ex-post. To summarize the results, the values of WTP per QALY vary from 2.6 million yen to 14.9 million yen. It can be concluded that the values of WTP per QALY are shown to be significantly different according to the structures of the questionnaires. Logistic regression analysis was performed to evaluate the relationship between responses to the first question (willingness to pay more than JPY 1) and demographic factors, such as income, age, marital status, pattern of employment, and education level, as shown in Table 5 and Supplementary Table S1. The WTP values in ex-post were significantly correlated with income. In the case of ex-ante WTP, demographic factors did not affect the WTP. Comparing the survival gain in the life-threatening disease (Questionnaire 8 and 9), the WTP per QALY for a 6-month extension was significantly higher than for 18 months, indicating that the length of survival gain is not correlated with the value of WTP.

Table 4. Estimate of WTP values from the overall questionnaires.

No	Health state (EQ-5D-5L)	Perspective	QALY gained	Context	WTP per QALY*
1	Severe	Personal	1.662	Ex post	391
2	Moderate	Personal	0.928	Ex post	260
3	Severe	Socially inclusive	1.662	Ex post	638
4	Moderate	Socially inclusive	0.928	Ex post	364
5	Severe	Social	1.662	Ex post	504
6	Moderate	Social	0.928	Ex post	299
7	Life-threatening	Personal	0.5	Ex post	421
8	Life-threatening	Personal	0.5	Ex ante	1490
9	Life-threatening	Personal	2.5	Ex ante	470

* JPY 10,000.

Table 5. Relation between response to the first question and demographic factors.

	Q7 (ex-post)		Q8 (ex-ante)		Q9 (ex-ante)
	Estimate	p-value	Estimate	p-value	Estimate	p-value
Income	0.00061*	0.037	0.00045	0.11	0.00025	0.38
Age	0.00032	0.95	−0.0040	0.47	0.0073	0.19
Marriage status	0.093	0.53	0.046	0.76	−0.23	0.12
Pattern of employment	−0.18	0.32	−0.12	0.49	−0.085	0.63
Education level	0.099	0.47	0.12	0.39	−0.12	0.36

* p < 0.05.

Discussion

There are many controversial issues surrounding the “threshold value” in the healthcare decision-making process^{16,18,19,21,28,29}. Given that the Japanese government is introducing a rigid HTA system under which the ICER value will directly reflect the price revision rate, the role and impact of “threshold value”, interpreted as the starting point of price reduction, will become extremely important. Under this new system, the role and function of the appraisal process, in which factors other than the ICER itself are taken into account, is limited compared to that of foreign HTA bodies. There are some arguments to justify a narrower appraisal process that omits factors, such as social and ethical issues, since HTA data will affect the reimbursement price and not the coverage status. However, the need for an appraisal process should be decided based on more conceptual factors^30,31. The fundamental role of the appraisal process is not to overwhelm decisions derived from ICERs, but to ensure the flexibility of results. There are several factors that are not captured via the assessment process, or cost-effectiveness evaluations, in the HTA system that is being implemented. Therefore, since less flexibility has been allowed in the Japanese system, the appraisal process in Japan should be restructured in much the same way or even more carefully than those of foreign HTA bodies.

WTP values vary according to the severity of diseases (moderate, severe, and life-threatening), perspectives (personal, socially inclusive, social), and contexts (ex-ante, ex-post) as Tsuchiya and Watson²³ stated. In comparing perspectives, respondents in the socially inclusive category estimated the highest WTP values for a new drug. This result mirrors the actual situation in Japan in which citizens pay 30% or less for healthcare and the national health insurance service is supported by social health insurance. Consistent with previous studies by Shiroiwa et al.³², life-threatening cases showed the highest WTP value, and the WTP for severe cases was higher than that for moderate ones¹⁰. The WTP for ex-ante was 3-times as much as that for ex-post (JPY 14.9 million vs. JPY 4.7 million) in this study. In the ex-ante approach, the degree of uncertainty is higher than in the ex-post one. On the other hand, the ex-post respondents considered other factors, such as their income, when answering the question as shown. Therefore, ex-ante WTP can take demand-side uncertainty into account, and the questionnaire in ex-ante is generally appropriate for identifying the preferences in the case of a life-threatening disease^33–35. From the analysis of demographic factors, income significantly affects the WTP in ex-post in Table 5. On the other hand, none of the demographic factors affected the WTP in ex-ante. These results indicate that a variety of factors can affect the WTP value independently.

The WTP value for one additional QALY is used by healthcare decision-makers and is often referred to as the threshold value of the ICER per QALY. However, many underlying issues remain unresolved when we attempted to use a hypothetical WTP value for decision-making³⁶. Our results show there is no doubt that the WTP value itself must be varied from one situation to another, such as the difference of disease severities, disease areas, life expectancies, and contexts. These issues have been taken into account in several foreign HTA agencies; for example, NICE in the UK established a rule for life-extending end-of-life treatment under which threshold values are doubled to GBP 50,000 per QALY²². ZIN in the Netherlands considers the characteristics of the disease and intervention and chooses one value from three candidate values, EUR 20,000, EUR 50,000, and EUR 80,000/QALY^37,38. In the pilot project in Japan that began in 2016, the uniform threshold value of JPY 5.0 million per QALY was used, regardless of the characteristics of the interventions. However, multiple values have been proposed for the formal launch of the program in April 2019, including a higher threshold value of JPY 7.5 million for anti-cancer medications.

Other issues around the implementation of WTP value into the threshold value are still controversial, including the equity of the concept of the QALY itself; uncertainties around estimated values; and the possibility of more ad-hoc considerations^39–41.

The results from this study should be referred to carefully if it is to be used in the decision-making process. This study focused on the specific situation about access to new drugs. Moreover, the analysis was based on data collected from a panel, which could be biased. The sampling was not stratified based on sex and age. On the other hand, income, profession, and health status were not considered. There is also a risk that respondents would have answered the questionnaire differently if they had been asked in everyday situations.

Conclusion

Using data from a WTP survey covering several situations, this study determines that the use of a uniform price threshold may not be appropriate in policy settings, because it may not reflect the diverse preferences based on many factors, including illness type or severity.

Transparency

Declaration of funding

Office of Pharmaceutical Industry Research provided funding for the study.

Declaration of financial/other interests

MH is an employee of Takeda Pharmaceutical Company Limited. JME peer reviewers on this manuscript have no relevant financial or other relationships to disclose.

Acknowledgements

We thank Dr Sadao Nagaoka and Mr Masami Morita for their valuable suggestions.