Advertisement

Determination of Cost-Effectiveness Threshold for Health Care Interventions in Malaysia

  • Yen Wei Lim
    Correspondence
    Address correspondence to: Asrul Akmal Shafie, Discipline of Social and Administrative Pharmacy, School of Pharmaceutical Sciences, Universiti Sains Malaysia, Minden, Penang 11800, Malaysia.
    Affiliations
    Discipline of Social and Administrative Pharmacy, School of Pharmaceutical Sciences, Universiti Sains Malaysia, Penang, Malaysia
    Search for articles by this author
  • Asrul Akmal Shafie
    Affiliations
    Discipline of Social and Administrative Pharmacy, School of Pharmaceutical Sciences, Universiti Sains Malaysia, Penang, Malaysia
    Search for articles by this author
  • Gin Nie Chua
    Affiliations
    Discipline of Social and Administrative Pharmacy, School of Pharmaceutical Sciences, Universiti Sains Malaysia, Penang, Malaysia

    Health Economics Research Unit/Academic Primary Care, Institute of Applied Health Sciences, University of Aberdeen, Aberdeen, Scotland
    Search for articles by this author
  • Mohammed Azmi Ahmad Hassali
    Affiliations
    Discipline of Social and Administrative Pharmacy, School of Pharmaceutical Sciences, Universiti Sains Malaysia, Penang, Malaysia
    Search for articles by this author
Open ArchivePublished:June 02, 2017DOI:https://doi.org/10.1016/j.jval.2017.04.002

      Abstract

      Background

      One major challenge in prioritizing health care using cost-effectiveness (CE) information is when alternatives are more expensive but more effective than existing technology. In such a situation, an external criterion in the form of a CE threshold that reflects the willingness to pay (WTP) per quality-adjusted life-year is necessary.

      Objectives

      To determine a CE threshold for health care interventions in Malaysia.

      Methods

      A cross-sectional, contingent valuation study was conducted using a stratified multistage cluster random sampling technique in four states in Malaysia. One thousand thirteen respondents were interviewed in person for their socioeconomic background, quality of life, and WTP for a hypothetical scenario.

      Results

      The CE thresholds established using the nonparametric Turnbull method ranged from MYR12,810 to MYR22,840 (~US $4,000–US $7,000), whereas those estimated with the parametric interval regression model were between MYR19,929 and MYR28,470 (~US $6,200–US $8,900). Key factors that affected the CE thresholds were education level, estimated monthly household income, and the description of health state scenarios.

      Conclusions

      These findings suggest that there is no single WTP value for a quality-adjusted life-year. The CE threshold estimated for Malaysia was found to be lower than the threshold value recommended by the World Health Organization.

      Keywords

      Introduction

      For many years, clinical evidence has been the only evidence required in deciding how to fund health care interventions or drug reimbursement. Nevertheless, many countries are now considering the cost of drugs as part of the important criteria in decision making because of the finite resources available in the health care sector [
      • Barnieh L.
      • Manns B.
      • Harris A.
      • et al.
      A synthesis of drug reimbursement decision-making processes in organisation for economic co-operation and development countries.
      ]. Consequently, economic evaluations have become increasingly important tools to support efficient resource allocation in the health care sector, especially in resource-constrained settings [
      • Eichler H.
      • Kong S.X.
      • Gerth W.C.
      • et al.
      Use of cost-effectiveness analysis in health-care resource allocation decision-making: How are cost-effectiveness thresholds expected to emerge?.
      ].
      The results of economic evaluations, especially cost-effectiveness analysis and cost-utility analysis, are usually summarized as an incremental cost-effectiveness ratio (ICER). The ICER represents the incremental cost per incremental gain in outcomes of one intervention compared with another. To draw conclusions on the cost-effectiveness (CE) of health care interventions, the ICER is usually compared with a reference value, the CE threshold, sometimes referred to as the ICER threshold or the ceiling threshold [
      • Cleemput I.
      • Neyt M.
      • Thiry N.
      • et al.
      Threshold Values for Cost-Effectiveness in Health Care.
      ,
      • Thavorncharoensap M.
      • Teerawattananon Y.
      • Natanant S.
      • et al.
      Estimating the willingness to pay for a quality-adjusted life year in Thailand: Dose the context of health gain matter?.
      ].
      The CE threshold represents the willingness to pay per quality-adjusted life-year (WTP/QALY) gained and is a vital component of decision making involving economic evaluation [
      • Drummond M.F.
      • Sculpher M.J.
      • Torrance G.W.
      • et al.
      ]. In previous arguments on the importance of an explicit threshold value, Johannesson and Meltzer [
      • Johannesson M.
      • Meltzer D.
      Editorial: some refections on cost-effectiveness analysis.
      ] claimed that without a CE threshold, cost-effectiveness analysis cannot be considered a proper decision-making tool because it would lack a systematic and universally recognizable decision criterion.
      A number of countries such as the United Kingdom, Ireland, and the Slovak Republic have explicitly stated their own threshold values [
      • Barnieh L.
      • Manns B.
      • Harris A.
      • et al.
      A synthesis of drug reimbursement decision-making processes in organisation for economic co-operation and development countries.
      ]. For instance, the National Institute for Health and Care Excellence (NICE) in the United Kingdom has set a threshold value of £20,000 to £30,000 per QALY gained [
      • Barnieh L.
      • Manns B.
      • Harris A.
      • et al.
      A synthesis of drug reimbursement decision-making processes in organisation for economic co-operation and development countries.
      ,
      • Appleby J.
      • Devlin N.
      • Parkin D.
      NICE’s cost-effectiveness threshold. How high should it be?.
      ,
      • Shiroiwa T.
      • Sung Y.K.
      • Fukuda T.
      • et al.
      International survey on willingness-to-pay (WTP) for one additional QALY gained: What is the threshold of cost effectiveness?.
      ]. In Malaysia, however, there is no such explicit threshold value. In current practice, decisions regarding new health care technologies are made without a transparent decision criterion. This situation leaves more room for arbitrariness and ad hoc considerations in the decision-making process. It also prevents the determination of the true opportunity cost of a new medical intervention, which, in turn, imposes inefficiencies and inconsistencies in decision making, and threatens the sustainability of the health care funding system [
      • Eichler H.
      • Kong S.X.
      • Gerth W.C.
      • et al.
      Use of cost-effectiveness analysis in health-care resource allocation decision-making: How are cost-effectiveness thresholds expected to emerge?.
      ,
      • Donaldson C.
      • Baker R.
      • Bell S.
      • et al.
      ]. Although the World Health Organization (WHO) has made a generic recommendation for CE threshold in developing countries to take the value of 1 to 3 times the gross domestic product (GDP) per capita per disability-adjusted life-years, such an approach does not accurately reflect the specific needs and the economic and disease burden of the general population in each country. Therefore, establishing a Malaysian CE threshold expressed in terms of cost per QALY is vital, because it will provide a solid criterion for decision making. This study was conducted primarily to determine a CE threshold value for health care interventions in Malaysia. A secondary goal was to identify the factors that affect WTP per QALY.

      Methods

       Study Design and Samples

      A cross-sectional, contingent valuation survey was conducted between December 1, 2012, and December 31, 2014. A sample size of 608 was required to detect a minimum difference of 0.05 between health states at a 0.05 significance level and with 0.80 statistical power. To account for the 40% of nonresponse expected in population survey, the sample size was increased to 1000. Respondents were interviewed in person for 10 to 20 minutes. All respondents chosen were Malaysian adults aged between 20 and 60 years and able to understand either English or Malay. The questionnaire was available in both languages. Both the English and Malay versions of questionnaire were tested and validated in pilot studies to ensure that there is no translational bias. The questionnaire was first designed in English and translated into Malay using the standardized patient-reported outcomes translation procedure [
      • Wild D.
      • Grove A.
      • Martin M.
      • et al.
      Principles of good practice for the translation and cultural adaptation process for patient-reported outcomes (PRO) measures: report of the ISPOR Task Force for Translation and Cultural Adaptation.
      ]. Forward translation was done by native English speakers who resided in Malaysia and came from a medical background with experience in translating/managing the translation of patient-reported outcome measures. Both languages were used in questionnaire development because most Malaysians younger than 60 years are literate in at least one of the two languages. Malay is the official language of Malaysia, whereas English is offered as a compulsory second language subject as part of the national education syllabus. The country has an adult literacy rate of up to 93.1% [
      UNICEF
      ]. During the interview process, respondents were given the flexibility of choosing their language of preference during the survey. As such, the risk of selection bias because of language was expected to be low.
      Stratified multistage cluster sampling was used on the basis of the sampling frame provided by the Population and Housing Census of Malaysia [
      Department of Statistics Malaysia
      Population and Housing Census of Malaysia.
      ]. Three states and a federal territory in Peninsular Malaysia, namely, Penang, Kedah, Selangor, and Kuala Lumpur, were clustered into four regions. The samples were then allocated to each region on the basis of the total population of each region. After this step, 20 enumeration blocks were selected from each region. In the third stage of stratification, 120 and 66 living quarters were selected in each city and each rural area, respectively, in proportion to the 65% urban dwellers nationally. Full-time students were excluded because their financial dependency might bias the valuation of WTP.

       Study Instrument

      The study questionnaire was developed by a group of practitioners and academics in HTAsiaLink, a network of health technology assessment organizations in Asia. This collaborative study was conducted simultaneously in three other member countries: Korea, Japan, and Thailand.
      Each questionnaire was divided into four parts (Appendix 1 in Supplemental Materials). Part 1 consisted of 11 items on respondents’ socioeconomic background (sex, age, ethnicity, educational level, occupation, marital status, number of household members, monthly household income, status in the household, presence of health problem, and private health insurance). Part 2 consisted of an assessment of the respondents’ current health state using the three-level EuroQol five-dimensional questionnaire (EQ-5D) and valuation of their current and hypothetical health states. During this part of the study, each respondent was asked to imagine being in one of the seven hypothetical health states (Appendix A in Supplemental Materials). The health state descriptions were derived from the EQ-5D definitions. They were chosen to represent “mild,” “moderate,” or “severe” health conditions as well as “extended life for terminal illness” and “life-saving intervention for immediate death” [
      EuroQol Group
      EuroQol—a new facility for the measurement of health-related quality of life.
      ]. Health states with a utility value of more than 0.70 were categorized as mild conditions, whereas those with utility values of 0.35 to 0.70 and less than 0.35 were classified as moderate and severe health conditions, respectively [
      • Szende A.
      • Oppe S.
      • Devlin N.
      EQ-5D Value Sets: Inventory, Comparative Review and User Guide.
      ,
      • Shafie A.A.
      EuroQol 5-dimension measure in Malaysia.
      ]. This gave rise to two mild health states (11121 and 11212), two moderate health states (11323 and 22222), and one severe health state (22232) (Fig. 1).
      Fig. 1
      Fig. 1Example of an information sheet of a given health state described to respondents.
      The terminal health state group was represented by two scenarios: one involving extended life for terminal illness and the other involving life-saving intervention for immediate death situations. Both versions shared the same hypothetical severe health state (22232) but the valuation scenarios being described were different. Each respondent valued only one hypothetical health state during the interview. Descriptions of the hypothetical health states were illustrated on a separate color-printed card that was used by the interviewers. Utilities of the current and hypothetical health states were measured using the three-level EQ-5D and the visual analogue scale (VAS) [
      • Shafie A.A.
      • Hassali M.A.
      • Liau S.Y.
      A cross-sectional validation study of EQ-5D among the Malaysian adult population.
      ,
      • Yusof F.A.
      • Goh A.
      • Azmi S.
      Estimating an EQ-5D value set for Malaysia using time trade off and visual analogue scale methods.
      ].
      Part 3 consisted of a contingent valuation exercise in which each respondent was asked for the amount he or she was willing to pay for a scenario involving the hypothetical health state selected in part 2 (Fig. 2) with two QALY gained levels, 0.2 QALY and 0.4 QALY. Keeping small QALY gains would enable “health losses” to be considered by the respondents such that WTP values are subjected to “budget constraints” to avoid extreme WTP values [
      • Donaldson C.
      • Baker R.
      • Bell S.
      • et al.
      ].
      Fig. 2
      Fig. 2Example of a question asked for treatment scenarios.
      A bidding game technique and a double-bounded dichotomous choice approach were applied in eliciting the maximum WTP value for each respondent. In each version of the questionnaire, respondents were asked to place a value on a hypothetical health state scenario on the basis of a certain starting bidding amount. The starting bidding amounts were calculated in proportion to the Malaysian GDP per capita in 2010. To test and control for anchoring effect commonly associated with contingent valuation, the starting bidding values were varied at 5%, 10%, 20%, 40%, 80%, or 120% of the GDP per capita. This yielded six different starting bidding amounts (MYR1,300, MYR2,600, MYR5,200, MYR10,400, MYR20,800, and MYR31,000). Respondents were randomly assigned to one of the six starting bidding amounts by rolling a dice.
      The bidding amount was adjusted on the basis of the answers provided by respondents. If the answer was “yes” for the first bid, the bidding amount was increased by one level in the subsequent bid. If respondents, however, answered “no” for the first bid, the second bidding amount was decreased by one level. Respondents who accepted both bids were asked for their maximum WTP in an open-ended question. Nevertheless, respondents who rejected both bids were asked whether they were willing to pay a small amount. If the answer was “yes,” they were asked an open-ended question about the maximum amount they were willing to pay, which should be lower than the minimum offered bid. Respondents who were not willing to pay even a small amount were directed to indicate their reasons. In part 4, respondents were asked to provide their feedback on the survey’s difficulty regarding the VAS used and on the contingent valuation exercise, each rated on a five-point Likert scale.

       Data Management and Statistical Analysis

      In this study, the mean WTP amount was estimated using both the nonparametric Turnbull method and the parametric interval regression analysis. To obtain the mean value of WTP per QALY, the WTP values derived from the bidding amounts were divided by the amount of utility gained. Protest zero cases were determined and were excluded from the analysis. All the statistical analyses were performed using STATA 9.0 (StataCorp, College Station, TX) and SAS 9.3 (SAS Institute Inc., Cary, NC).

       Nonparametric Turnbull analysis

      The nonparametric Turnbull method was used to estimate the survivor distribution characterizing WTP responses. The Turnbull method makes minimal assumptions about the distribution of WTP and estimates the mean and median WTP as lower bound estimates. The estimated lower bound of WTP is distributed asymptotically normal. Because of the independence of the true underlying distribution, the lower bound estimate of WTP is appealing. In addition, this estimate offers a conservative lower bound for all non-negative distributions of WTP. In practice, it represents the minimum expected WTP for all distributions of WTP defined from 0 to infinity [
      • Haab T.C.
      • McConnell K.E.
      Referendum models and negative willingness to pay: alternative solutions.
      ,
      • Haab T.C.
      • McConnell K.E.
      Valuing Environmental and Natural Resources: The Econometrics of Non-Market Valuation.
      ]. Given that tj is the bidding amount to price j, M* represents the pooled price ranges, and Fj* is the pooled probability of a “no” response to price j, the lower bound estimate of WTP can be calculated by the following equation:
      ELB(WTP)=j=0Mtj(Fj+1Fj).


       Parametric interval regression analysis with Weibull distribution

      Because of its distribution-free property, the drawback of the nonparametric Turnbull method is that it provides estimates of only a fraction of the distribution that falls into particular intervals defined by the lower and upper dollar thresholds used. Moreover, this nonparametric approach does not provide any point estimates, such as a median, and it cannot handle covariate analysis. To address these problems, a parametric interval regression model was used. We assumed that WTP distribution follows a Weibull distribution because of its non-negative distribution and goodness of fit to our sample data [
      • Carson R.T.
      • Wilks L.
      • Imber D.
      Valuing the preservation of Australia’s Kakadu Conservation Zone.
      ]. The cumulative density function of the Weibull distribution is defined as follows:
      F(z;θ;σ)=1exp([z/σ]θ),


      where z is the WTP, θ is the shape parameter, and σ is the scale parameter.
      To assess the factors that affected respondents’ WTP per QALY, all the sociodemographic characteristics (such as sex, age, ethnicity, educational level, marital status, and monthly household income), utility measures, and valuation scenarios (such as treatment, life-extending, and immediate death scenario) were included in the interval regression model as the potential explanatory variables. They were then removed using backward-elimination method, for items that were not significant at 25% level [
      • Carson R.T.
      • Wilks L.
      • Imber D.
      Valuing the preservation of Australia’s Kakadu Conservation Zone.
      ,
      • Shiroiwa T.
      • Igarashi A.
      • Fukuda T.
      • et al.
      WTP for a QALY and health states: More money for severer health states?.
      ].

      Results

      One thousand one hundred respondents were approached and the overall response rate was recorded at 92.1%, with a total of 1013 respondents aged between 20 and 60 years being interviewed face-to-face. Their sociodemographic characteristics are presented in Table 1.
      Table 1Sociodemographic characteristics of the Malaysian population
      CharacteristicMean ± SD or N (%)Malaysian population
      Data for Malaysian populations were obtained from the Department of Statistics, Malaysia, in the Population Distribution and Basic Demographic Characteristics Report 2010. Data shown are from Malaysian citizens and were updated in 2010.
      (%)
      Age (y)39 ± 11.5
      Sex
       Male363 (35.8)50.7
       Female650 (64.2)49.3
      Ethnicity
       Malay778 (76.8)67.4
       Chinese136 (13.4)24.6
       Indian97 (9.6)7.3
       Others2 (0.2)0.7
      Education
       Primary school or less117 (11.5)
       Secondary school644 (63.6)
       Bachelor’s degree224 (22.1)
       Higher than bachelor’s degree28 (2.8)
      Occupational sector
       Part-time student20 (2.0)
       Pensioner55 (5.4)
       Housewife274 (27.1)
       Unemployed64 (6.3)
       In full-time employment600 (59.2)
      Marital status
       Single213 (21.0)35.1
       Married758 (74.8)59.6
       Divorced/separated10 (1.0)0.8
       Widowed32 (3.2)4.5
      Number of household members4.8 (1.9)
      Monthly household income
      Data for monthly household income were obtained from the Department of Statistics, Malaysia, in the Malaysia Household Income and Basic Amenities Survey Report [22].
       <MYR50034 (3.4)1.2
       MYR500–999125 (12.3)6.2
       MYR1,000–2,999432 (42.6)45.4
       MYR3,000–4,999223 (22.0)23.1
       MYR5,000–10,000163 (16.1)18.2
       >MYR10,00036 (3.6)5.9
      Status in household
       Head of the household306 (30.2)
       Spouse of the head of the household459 (45.3)
       Son/daughter of the head of the household215 (21.2)
       Parent of the head of the household6 (0.6)
       Relative of the head of the household17 (1.7)
       Others10 (1.0)
      Health problem
       Yes230 (22.7)
       No783 (77.3)
      Private health insurance
       Yes398 (39.3)
       No615 (60.7)
      Hypothetical health state scenario
       Treatment scenario with mild health state 11121151 (14.9)
       Treatment scenario with mild health state 11212154 (15.2)
       Treatment scenario with moderate health state 11323152 (15.0)
       Treatment scenario with moderate health state 22222149 (14.7)
       Treatment scenario with severe health state 22332154 (15.2)
       Extended life for terminal illness128 (12.6)
       Life saving for immediate death125 (12.3)
      Questionnaire language answered
       English344 (34.0)
       Malay669 (66.0)
      Duration of interview (min)
       English8.7 ± 2.8
       Malay8.6 ± 2.8
      low asterisk Data for Malaysian populations were obtained from the Department of Statistics, Malaysia, in the Population Distribution and Basic Demographic Characteristics Report 2010. Data shown are from Malaysian citizens and were updated in 2010.
      Data for monthly household income were obtained from the Department of Statistics, Malaysia, in the Malaysia Household Income and Basic Amenities Survey Report
      Department of Statistics Malaysia
      Malaysia Household Income and Basic Amenities Survey Report.
      .
      In the measurement of health status and utility, most of the respondents reported no problem to all the dimensions in the EQ-5D. For moderate problem scenarios, pain/discomfort was reported more by the respondents (n = 186 [18.4%]) when compared with other dimensions. None of the respondents experienced extreme problem in mobility, self-care, and usual activities except for pain/discomfort and anxiety/depression. Figure 3 shows a summary of the responses reported by respondents to the current health status described in the EQ-5D description.
      Fig. 3
      Fig. 3Summary of the responses to current health status in the EQ-5D description. EQ-5D, EuroQol five-dimensional questionnaire.
      In the VAS valuation, the mean score of current health state was estimated at 81.6, with an SD of 14.5. The responses to five different hypothetical health states in VAS valuation are summarized in Appendix B in Supplemental Materials. The mean score of VAS shows a strong inverse correlation to the severity of health state descriptions. This indicates a considerable agreement that most of the people will place a lower utility score when facing a severe health state.

       WTP Measures

      The results for the Turnbull estimates and the probability of respondents responding “no” to each of the bid values for 0.2 QALY and 0.4 QALY gained are presented in Table 2. The responses to the bidding amounts of MYR20,800, MYR31,000, and MYR39,000 violate the monotonicity assumption for a standard distribution function for scenarios involving both 0.2 QALY and 0.4 QALY gained. The responses to these bidding amounts were pooled with the responses for MYR10,400 bids to maintain monotonicity. The probability of mass point estimates is presented in the last two columns. Besides the responses to the aforementioned bidding amounts, responses to all other bidding amounts satisfy the monotonicity assumption.
      Table 2Turnbull nonparametric estimator for 0.2 QALY and 0.4 QALY gained
      0.2 QALY0.4 QALY
      Unrestricted estimateTurnbull estimateUnrestricted estimateTurnbull estimate
      tj (MYR)Tj
      Data were derived from long format.
      Nj
      Data were derived from long format.
      Fj (= Nj/Tj)Fj
      Data were derived from long format.
      fj
      Data were derived from long format.
      tj (MYR)Tj
      Data were derived from long format.
      Nj
      Data were derived from long format.
      Fj (= Nj/Tj)Fj
      Data were derived from long format.
      fj
      Data were derived from long format.
      700155660.4260.4260.426700147500.3400.3400.340
      1,300196940.4800.4800.0541,300183680.3720.3720.031
      2,6003181560.4910.4910.0112,6003251310.4030.4030.031
      5,2002811540.5480.5480.0575,2003001500.5000.5000.097
      10,4003132080.6650.6120.06410,4003261980.6070.5750.075
      20,8002201460.664Pooled backPooled back20,8002461580.642Pooled backPooled back
      31,00042130.310Pooled backPooled back31,00054190.352Pooled backPooled back
      39,0003350.152Pooled backPooled back39,0003340.121Pooled backPooled back
      39,000+1.0001.0000.38839,000+1.0001.0000.425
      MYR, Malaysian Ringgit; QALY, quality-adjusted life-year; Fj, probability of a negative response to price j; Fj*, pooled probability of a negative response to price j; fj*, difference within Fj* (fj* = Fj*Fj*˗1); Nj, number of respondents rejecting the choice; tj, bidding amount for price j; Tj, total number of respondents.
      low asterisk Data were derived from long format.
      After excluding all 0 WTP values, the mean lower bound estimates of WTP for 0.2 QALY and 0.4 QALY gained were estimated at MYR4568.30 ± 138.13 and MYR5124.20 ± 134.61, respectively (Appendix C in Supplemental Materials). The WTP per QALY was estimated to be approximately MYR22,840 (~US $7,000) for 0.2 QALY gained and MYR12,810 (~US $4,000) for 0.4 QALY gained (October 2014: MYR3.2 = US $1) [
      International Monetary Fund
      ].
      The mean, median, and SD for scenario with both 0.2 QALY and 0.4 QALY gained were estimated using an interval regression model on the basis of the Weibull distribution, which provided the best fit for the distribution pattern over the range of interval-censored data; these results are presented in Table 3. The WTP value was calculated by integrating the survival function from 0 to infinity, resulting in a mean WTP value of MYR5,693.95 with a standard error of the mean (SEM) of 211.78 for 0.2 QALY gained, and a WTP/QALY value of approximately MYR28,470 (~US $8,900). For scenarios with 0.4 QALY gained, the mean WTP value was estimated at MYR7,971.62 (SEM 287.06) and the mean WTP/QALY value was approximately MYR19,929 (~US $6,200).
      Table 3Summary of interval regression model results based on a log-exponentiated Weibull distribution for the determination of the CE threshold
      Variable0.2 QALY0.4 QALY
      Estimate (SE)P valueEstimate (SE)P value
      Intercept4.93 (3.32)0.1374.81 (3.54)0.174
      Age (y)–0.02 (0.01)<0.001–0.02 (0.01)<0.001
      Male (female = ref.)–0.23 (0.14)0.099–0.19 (0.14)0.170
      Education (primary school or less = ref.)
       Secondary school0.44 (0.23)0.0580.48 (0.22)0.033
       Bachelor’s degree or higher0.65 (0.28)0.0211.00 (0.28)<0.001
      Marital status (single = ref.)
       Married–0.28 (0.19)0.144–0.18 (0.19)0.341
       Divorced/widowed–1.19 (0.38)0.002–1.16 (0.37)0.002
      Monthly household income (<MYR500 = ref.)
       MYR500–9990.37 (0.43)0.3870.50 (0.42)0.232
       MYR1,000–2,9990.93 (0.41)0.0231.05 (0.40)0.009
       MYR3,000–4,9991.66 (0.43)<0.0011.70 (0.42)<0.001
       MYR5,000–10,0002.31 (0.45)<0.0012.17 (0.44)<0.001
       >MYR10,0002.18 (0.56)<0.0012.04 (0.56)<0.001
      Scenario (treatment = ref.)
       Extended life for terminal illness0.88 (0.21)<0.0010.93 (0.21)<0.001
       Immediate death0.64 (0.21)0.0020.38 (0.20)0.057
      Self-utility score (EQ-VAS current health)0.01 (0.01)0.2570.00 (0.00)0.682
      Ethnicity (other = ref.)
       Malay3.07 (3.25)0.3443.55 (3.48)0.308
       Chinese3.38 (3.25)0.2993.73 (3.48)0.284
       Indian2.77 (3.25)0.3953.26 (3.48)0.349
      CE, cost-effectiveness; EQ-VAS, EuroQol visual analogue scale; MYR, Malaysian Ringgit; QALY, quality-adjusted life-year; ref., reference; SE, standard error.

       Factors Affecting WTP

      Taking covariates into account for the determination of CE thresholds in the interval regression model (Table 3) revealed that factors such as age, sex, level of education, income, and valuation of scenario affected respondents’ WTP. Respondents with a higher education bid a higher WTP amount in both the 0.2 QALY and the 0.4 QALY gained scenarios. Compared with respondents with a primary school education, those with a bachelor’s degree or higher were more willing to pay for the scenarios described in the questionnaires.
      In addition, WTP was found to be positively associated with respondents’ income. This is true and significant across all income groups (except the MYR500–999 group). By comparing the results with the base-case scenario of monthly household income less than MYR500, it was found that respondents with monthly household income ranging from MYR5,000 to MYR10,000 had the highest WTP across income groups for both the 0.2 QALY and the 0.4 QALY gained scenarios. WTP was also found to be affected by the valuation scenario. Respondents were willing to pay more to extend life for terminal illness than for treatment in both QALY scenarios. Higher WTP was also noted for the avoidance of immediate death when compared with treatment in the 0.2 QALY scenario.
      Respondents’ age, however, was found to be negatively associated with their WTP: as respondents’ age increased, their WTP decreased. Other factors such as respondents’ marital status were also found to affect WTP. Respondents who were divorced or widowed had a lower WTP compared with respondents who were single.

      Discussion

      On the basis of the estimates obtained with the nonparametric Turnbull method, the CE threshold value was found to range from MYR12,810 to MYR22,840 (~US $4,000–US $7,000). In contrast, the WTP/QALY value obtained using the parametric interval regression model was MYR19,929 to MYR28,470 (~US $6,200–US $8,900). In interval-censored data, the results of parametric and nonparametric models cannot be directly compared because of the overlapping time intervals of censoring and the fact that nonparametric analysis does not provide a probability for right-censored observations [
      • Lindsey J.K.
      A study of interval censoring in parametric regression models.
      ]. Nevertheless, both models have their different robustness in presenting the results.
      There are certain insights that might be gained from comparing the estimation of CE threshold with these two methods. The performance of the parametric approach seems to be highly satisfactory, especially when the Weibull distribution is chosen, because it allows a reasonably wide range of distributional shapes [
      • Lindsey J.K.
      A study of interval censoring in parametric regression models.
      ]. The CE threshold estimated by the parametric method would be more informative because it allows for covariate analysis to assess factors that might affect the threshold value. Thus, the result from parametric regression analysis was primarily reported. Nevertheless, this is by no means implying that the nonparametric method is inferior to the parametric method. The result explored by the nonparametric Turnbull method could serve as a reference.
      The estimated CE threshold value found in this study ranged between 0.90 and 1.28 times the GDP per capita, which was lower than the WHO-recommended threshold value of 1 to 3 times the GDP per capita [
      World Health Organization
      Cost-effectiveness thresholds.
      ]. Comparing the results in this study with the currently used WHO-recommended threshold value, it is likely that many health care technologies that might not represent a good value are reimbursed. It is premature to conclude that the WHO-recommended threshold is too high for reimbursement. Nevertheless, the findings in this study are consistent with findings from other contingent valuation studies, in which the estimated WTP/QALY value is lower than the proposed threshold value [
      • Thavorncharoensap M.
      • Teerawattananon Y.
      • Natanant S.
      • et al.
      Estimating the willingness to pay for a quality-adjusted life year in Thailand: Dose the context of health gain matter?.
      ,
      • King J.T.
      • Tsevat J.
      • Lave J.R.
      • et al.
      Willingness to pay for a quality-adjusted life year: implications for societal health care resources allocation.
      ,
      • Shafie A.A.
      • Lim Y.W.
      • Chua G.N.
      • et al.
      Exploring the willingness to pay for a quality-adjusted life-year in the state of Penang, Malaysia.
      ].
      The CE threshold was presented here as a range of a lower and an upper threshold instead of a single value. The probability of rejection approaches 1 when the cost per QALY is at the upper limit of the threshold range. On the contrary, the probability of rejection is near to 0 when the cost per QALY is at the lower limit of the threshold range [
      • Shiroiwa T.
      • Sung Y.K.
      • Fukuda T.
      • et al.
      International survey on willingness-to-pay (WTP) for one additional QALY gained: What is the threshold of cost effectiveness?.
      ]. This could be useful in practice in which a certain amount of flexibility is required for decision making. In fact, Shiroiwa et al. [
      • Shiroiwa T.
      • Sung Y.K.
      • Fukuda T.
      • et al.
      International survey on willingness-to-pay (WTP) for one additional QALY gained: What is the threshold of cost effectiveness?.
      ], Ahlert et al. [
      • Ahlert M.
      • Breyer F.
      • Schwettmann L.
      What you ask is what you get: willingness-to-pay for a QALY in Germany.
      ], and Devlin and Parkin [
      • Devlin N.
      • Parkin D.
      Does NICE have a cost-effectiveness threshold and what other factors influence its decisions? A binary choice analysis.
      ] pointed out that there is no single value for a QALY and that the threshold should be treated as a range of cost.
      The lack of an empirical reimbursement threshold in Malaysia might negatively affect the transparency for decision making in the health care sector. Ahlert et al. [
      • Ahlert M.
      • Breyer F.
      • Schwettmann L.
      What you ask is what you get: willingness-to-pay for a QALY in Germany.
      ] mentioned that no matter how explicitly and openly the decisions are taken, they need to somehow reflect the preferences of the population and thus the estimation of a monetary value that a member of society places on additional QALYs is desirable. In addition, many economists argued that individuals are the best judges of their own well-being and that social welfare depends only on the welfare of individuals in society [
      • Dolan P.
      The measurement of individual utility and social welfare.
      ]. Hence, the present study’s attempt to derive a societal standard for WTP/QALY from data describing human behavior or preferences using contingent valuation method fits this argument.
      There are a few identifiable limitations in this study. First, the hypothetical scenario was designed in generic dimension of health, and not for specific diseases. Although the findings in this study are not in favor of the assumption that “a QALY is a QALY is a QALY,” in which QALYs cannot fully reflect the preferences of people in terms of their health states [
      • Shiroiwa T.
      • Igarashi A.
      • Fukuda T.
      • et al.
      WTP for a QALY and health states: More money for severer health states?.
      ,
      • Dolan P.
      • Shaw R.
      • Tsuchiya A.
      • et al.
      QALY maximisation and people’s preferences: a methodological review of the literature.
      ], the CE threshold estimated in this study could serve as a universal threshold for Malaysia because it represents the opportunity cost for the health interventions, which is ranged from the treatment to life-saving scenarios that cover from mild to severe conditions in general.
      As with all interviewer-administered surveys, there exists the possibility of interview effect. To control for this and to ensure data comparability across interviewers, all interviewers were trained by study investigators using a standardized protocol before fieldwork. In addition, because of the hypothetical nature of the scenarios used in the survey, there are concerns about respondents’ ability to understand and interpret the contingent valuation task reliably. In light of this concern, the instrument was piloted extensively before the actual data collection and information obtained from each stage of piloting was used to improve the framing of questions, descriptions of scenarios, and overall realism of the task.
      As with any contingent valuation study using a dichotomous choice approach, WTP estimates may be subjected to anchoring effect in which respondents’ valuations are influenced by the starting bidding values. By varying the starting bidding values used in this study, no significant difference was found between respondents’ WTP across different starting values. To also test for scope sensitivity, the mean WTP elicited for different sizes of QALY improvement was compared. Using a one-tailed t test, the mean WTP for 0.2 QALY gained (MYR12,106) was found to be statistically significantly lower than the mean WTP for 0.4 QALY gained (MYR14,008) (P < 0.001).
      Last but not least, this study was carried out in Peninsular Malaysia, excluding East Malaysia (consisting of the states of Sabah and Sarawak), because of the limited budget and time. This may limit the generalizability of the study and its representativeness of the entire Malaysian population. Nevertheless, multistage stratified cluster sampling with data derived from computerized household lists from the Malaysia’s Department of Statistics census [
      Department of Statistics Malaysia
      Population and Housing Census of Malaysia.
      ] was used in this study to reduce systematic differences. Moreover, the starting bidding and QALY values used in this study were weighted to the general population distribution.
      In the United Kingdom, the high price of oncology drugs is posing problems for NICE in making decisions on reimbursement because of the high ICER per QALY. In this context, some have argued that NICE should review its CE threshold for end-of-life drugs [
      • Raftery J.
      NICE and the challenge of cancer drugs.
      ]. In January 2009, NICE introduced supplementary advice to accommodate a higher ceiling threshold for oncology drugs that may be life-extending for patients with short life expectancy. The additional advice applies when the CE ratio exceeds the NICE upper end threshold of £30,000 per QALY (£35,000–£45,000), provided that the treatment extends life by at least 3 months compared with available alternatives [
      • Raftery J.
      NICE and the challenge of cancer drugs.
      ,
      • Chabot I.
      • Rocchi A.
      Oncology drug health technology assessment recommendations: Canadian versus UK experiences.
      ].
      The rationale for this advice was the assumption that society valued life obtained from patients at end of life more than life obtained from other patients. NICE accommodated a higher threshold for end-of-life drugs, but this higher threshold must still be met for a positive recommendation [
      • Chabot I.
      • Rocchi A.
      Oncology drug health technology assessment recommendations: Canadian versus UK experiences.
      ]. It, however, remains unclear how the higher limit for end of life is determined by NICE. Hence, this study could provide an empirical insight in discussion on the need of setting a specific threshold value in this case. Given the increasing demand for economic evidence in decision making for the reimbursement of health care technologies, it would be desirable to conduct high-quality economic analyses alongside clinical studies. To contribute to this goal, an empirical study to identify the monetary value of a QALY is essential. This study represents the first effort to provide an empirical CE threshold for Malaysia.
      The findings of this study indicate that there is no single value of a QALY. The CE threshold for Malaysia was estimated to range between MYR19,929 and MYR28,470 (~US $6,200–US $8,900), which is lower than the currently used WHO-recommended threshold value. Educational level, estimated monthly household income, the description of health state scenarios, and age of the respondents are some of the identifiable factors that may have affected the value of the determined CE threshold. Although the findings in this study still deserved further investigation, we believe that CE thresholds based on people’s preferences as elicited by contingent valuation would contribute to more rigorous scientific decision making in the health care sector for the future.

      Acknowledgment

      We thank the HTAsiaLink members who provided their expert opinions and technical support in this study: Dr Yot Teerawatthaon, Dr Montarat Thavorncharoensap, Dr Jeonghoon Ahn, Professor Takashi Fukuda, Dr Takeru Shiroiwa, and Dr Ataru Igarashi.
      Source of financial support: This study was funded by Universiti Sains Malaysia via the Accelerated Programme for Excellence Delivering Excellence Grant (grant no. 1002/PFARMASI/910330). The funding agreement ensured the authors’ independence in designing the study, interpreting the data, and writing and publishing the report.

      Supplementary material

      References

        • Barnieh L.
        • Manns B.
        • Harris A.
        • et al.
        A synthesis of drug reimbursement decision-making processes in organisation for economic co-operation and development countries.
        Value Health. 2014; 17: 98-108
        • Eichler H.
        • Kong S.X.
        • Gerth W.C.
        • et al.
        Use of cost-effectiveness analysis in health-care resource allocation decision-making: How are cost-effectiveness thresholds expected to emerge?.
        Value Health. 2004; 7: 518-528
        • Cleemput I.
        • Neyt M.
        • Thiry N.
        • et al.
        Threshold Values for Cost-Effectiveness in Health Care.
        Belgium Health Care Knowledge Centre, Brussels, Belgium2008
        • Thavorncharoensap M.
        • Teerawattananon Y.
        • Natanant S.
        • et al.
        Estimating the willingness to pay for a quality-adjusted life year in Thailand: Dose the context of health gain matter?.
        Clinicoecon Outcomes Res. 2013; 5: 29-36
        • Drummond M.F.
        • Sculpher M.J.
        • Torrance G.W.
        • et al.
        Methods for the Economic Evaluation of Health Care Programmes. 3rd ed. Oxford University Press, New York, NY2005
        • Johannesson M.
        • Meltzer D.
        Editorial: some refections on cost-effectiveness analysis.
        Health Econ. 1998; 7: 1-7
        • Appleby J.
        • Devlin N.
        • Parkin D.
        NICE’s cost-effectiveness threshold. How high should it be?.
        BMJ. 2007; 335: 358-359
        • Shiroiwa T.
        • Sung Y.K.
        • Fukuda T.
        • et al.
        International survey on willingness-to-pay (WTP) for one additional QALY gained: What is the threshold of cost effectiveness?.
        Health Econ. 2010; 19: 422-437
        • Donaldson C.
        • Baker R.
        • Bell S.
        • et al.
        European value of a quality adjusted life year—final publishable report. 2010; (Available from:) ([Accessed November 18, 2012])
        • Wild D.
        • Grove A.
        • Martin M.
        • et al.
        Principles of good practice for the translation and cultural adaptation process for patient-reported outcomes (PRO) measures: report of the ISPOR Task Force for Translation and Cultural Adaptation.
        Value Health. 2005; 8: 94-104
        • UNICEF
        Malaysia: statistics. 2013; (Available from:) ([Accessed December 12, 2014])
        • Department of Statistics Malaysia
        Population and Housing Census of Malaysia.
        Department of Statistics Malaysia, Malaysia2010
        • EuroQol Group
        EuroQol—a new facility for the measurement of health-related quality of life.
        Health Policy. 1990; 16: 199-208
        • Szende A.
        • Oppe S.
        • Devlin N.
        EQ-5D Value Sets: Inventory, Comparative Review and User Guide.
        Springer, Dordrecht, The Netherlands2007
        • Shafie A.A.
        EuroQol 5-dimension measure in Malaysia.
        in: Michalos A.C. Encyclopedia of Quality of Life and Well-Being Research. Springer, Dordrecht, The Netherlands2014: 2041-2044
        • Shafie A.A.
        • Hassali M.A.
        • Liau S.Y.
        A cross-sectional validation study of EQ-5D among the Malaysian adult population.
        Qual Life Res. 2011; 20: 593-600
        • Yusof F.A.
        • Goh A.
        • Azmi S.
        Estimating an EQ-5D value set for Malaysia using time trade off and visual analogue scale methods.
        Value Health. 2012; 15: S85-S90
        • Haab T.C.
        • McConnell K.E.
        Referendum models and negative willingness to pay: alternative solutions.
        J Environ Econ Manage. 1997; 32: 251-270
        • Haab T.C.
        • McConnell K.E.
        Valuing Environmental and Natural Resources: The Econometrics of Non-Market Valuation.
        Edward Elgar Publishing, Cheltenham, UK2002
        • Carson R.T.
        • Wilks L.
        • Imber D.
        Valuing the preservation of Australia’s Kakadu Conservation Zone.
        Oxf Econ Pap. 1994; 46: 727-749
        • Shiroiwa T.
        • Igarashi A.
        • Fukuda T.
        • et al.
        WTP for a QALY and health states: More money for severer health states?.
        Cost Eff Resour Alloc. 2013; 11: 22
        • Department of Statistics Malaysia
        Malaysia Household Income and Basic Amenities Survey Report.
        Department of Statistics Malaysia, Malaysia2009
        • International Monetary Fund
        Representative exchange rates for selected currencies for October 2014. 2014; (Available from:) ([Accessed November 1, 2014])
        • Lindsey J.K.
        A study of interval censoring in parametric regression models.
        Lifetime Data Anal. 1998; 4: 329-354
        • World Health Organization
        Cost-effectiveness thresholds.
        2005 (Available from:) ([Accessed February 7, 2013])
        • King J.T.
        • Tsevat J.
        • Lave J.R.
        • et al.
        Willingness to pay for a quality-adjusted life year: implications for societal health care resources allocation.
        Med Decis Making. 2005; 25: 667-677
        • Shafie A.A.
        • Lim Y.W.
        • Chua G.N.
        • et al.
        Exploring the willingness to pay for a quality-adjusted life-year in the state of Penang, Malaysia.
        Clinicoecon Outcomes Res. 2014; 6: 473-481
        • Ahlert M.
        • Breyer F.
        • Schwettmann L.
        What you ask is what you get: willingness-to-pay for a QALY in Germany.
        CESifo Working Paper No. 4239, CESifo Group, Munich, Germany2013
        • Devlin N.
        • Parkin D.
        Does NICE have a cost-effectiveness threshold and what other factors influence its decisions? A binary choice analysis.
        Health Econ. 2004; 13: 437-452
        • Dolan P.
        The measurement of individual utility and social welfare.
        J Health Econ. 1998; 17: 39-52
        • Dolan P.
        • Shaw R.
        • Tsuchiya A.
        • et al.
        QALY maximisation and people’s preferences: a methodological review of the literature.
        Health Econ. 2005; 14: 197-208
        • Raftery J.
        NICE and the challenge of cancer drugs.
        BMJ. 2009; 338: 271-272
        • Chabot I.
        • Rocchi A.
        Oncology drug health technology assessment recommendations: Canadian versus UK experiences.
        Clinicoecon Outcomes Res. 2014; 6: 357-367