Feasibility and construct validity of the parent willingness-to-pay technique for children with juvenile idiopathic arthritis

Authors


Abstract

Objective

To assess the feasibility and construct validity of the willingness-to-pay (WTP) technique for measuring health care preferences in families of children with juvenile idiopathic arthritis (JIA).

Methods

Parents were asked to estimate the monthly US dollar amount they would be willing to pay to obtain for their child the following hypothetical drugs: ARTHRO, which guarantees complete clinical response; and NO-STOM-ACHE, a drug that eliminates gastrointestinal (GI) symptoms. A yes/no question was used with random assignment of the starting bids. Parents who agreed to pay the starting bid were then asked whether they would be willing to pay 200% and then 400% of this initial bid. Socioeconomic data and information on medications, disease activity, patient physical function, wellbeing, and health-related quality of life (HRQOL) were obtained.

Results

Sixty-two families of children with JIA were interviewed. GI symptoms were present in 54%, and 53% of the children had joints with active arthritis or limited range of motion. Four parents (7%) were unwilling to pay anything for any of the studied medications. The mean amount (median; mean percentage of available family income) families were willing to pay was $395 ($300; 15%) for ARTHRO and $109 ($80; 4%) for NO-STOM-ACHE. Correlation and regression analysis supported that, adjusted for the available family income, the WTP for ARTHRO was associated with disease activity, pain, and the HRQOL of the patients. After correction for the starting bids and the available family income, the WTP for NO-STOM-ACHE was associated with the patient's HRQOL, pain, and the amount of GI discomfort.

Conclusion

The WTP technique is feasible and has construct validity for measuring health care preferences for children with JIA. Relatively large WTP estimates support a possible important negative impact of the disease on families of children with JIA.

INTRODUCTION

The maximum amount that an individual is willing to pay for goods and services is a common measure of the value for these goods and services. A popular method of estimating the money values for health-care programs is the use of the willingness-to-pay (WTP) survey technique known as contingent valuation (1). The term was coined because the respondent is asked to consider the contingency of a market's existing for the goods or services being valued. The rationale for economic evaluations of health services arises from the economic concept of opportunity costs (2). Allocating scarce monetary and human resources to the development of better medications for arthritis, for example, means foregoing the opportunity to use those resources in other welfare-enhancing ways. In this context, economic evaluations can be defined as the process of ensuring that the value of what is gained from an activity outweighs the value of what is sacrificed (3). For example, the WTP approach can be used to measure an individual's valuation (e.g., preference) of how much he or she is willing to pay (e.g., sacrifice) to have access to new arthritis medications.

Alternative approaches of determining health care preferences include the estimations of changes in the health-related quality of life (HRQOL) of patients treated with these new medications. Preference-based HRQOL measures, such as the standard gamble or time trade-off techniques are suitable to capture individual differences of the HRQOL of patients with a certain disease phenotype. Conversely, non–preference-based HRQOL measures, such as the Child Health Questionnaire or the Pediatric Quality of Life Questionnaire (4–6), will detect no such differences (7), but rather measure the HRQOL of the average patient with a certain disease phenotype. Gains or losses in HRQOL are an alternative, more indirect approach for measuring health care preferences; changes of HRQOL of patients treated with new arthritis drugs serve as a surrogate marker of the patients' health care preferences. Conversely, WTP has the advantage to be a more direct approach for measuring health care preferences. Monetary values derived by the WTP method can easily be compared with WTP estimates for other health programs and different types of medical interventions across disease groups. Different from patient HRQOL, WTP also allows for the consideration of disadvantages and hardships encountered by the family members of the patients. Given these theoretical advantages for assessing patients' health care preferences, WTP has gained increasing popularity in recent years.

Applications of WTP in health care have previously been reviewed (8–10). However, the WTP method has not been assessed for its usefulness in children with juvenile idiopathic arthritis (JIA). The objectives of this study were to 1) assess the feasibility or acceptability of the WTP method when performed by parents of patients with JIA; 2) obtain preliminary estimates of the WTP of clinical remission and resolution of gastrointestinal (GI) symptoms for children with JIA; and 3) examine the construct validity of the method by studying the association of WTP with other nonmonetary outcome measures.

PATIENTS AND METHODS

Patients.

A convenience sample of families of children with JIA (11) was recruited from the pediatric rheumatology clinic of the Cincinnati Children's Hospital Medical Center (CCHMC) within a 3 month period. The rheumatology clinic is a tertiary referral center for children with rheumatic diseases, and ∼400 children with JIA are regularly seen at 1–4 month intervals. The primary caretaker of the patient was chosen to complete the study questionnaires and provide the WTP estimates, assuming that the children would lack a sufficient understanding of the family finances. The treating physician performed a standard clinic assessment and was asked to provide an assessment of JIA disease activity. Each family was interviewed twice at the time of clinic visits, at least 4 weeks apart, but generally on 2 consecutive visits.

Outcome measures.

WTP and, as validity checks, traditional outcomes (pain, number of involved joints—those with active arthritis or limited range of motion—and physician assessment of disease activity) as well as measures of physical function, HRQOL, and wellbeing were obtained. In addition, the parents were asked to provide socioeconomic information because these have been suggested to impact on the WTP (1). Therefore, data were collected on the monthly pretax family income, costs of health insurance and out-of pocket health care expenses, and the number of work hours lost due the care of the child with JIA.

Willingness-to-pay measurement.

WTP was elicited from the parents of children with JIA only, but not from the children themselves, because, depending on their age, children are prone to have an incomplete understanding of monetary issues. In addition, even older children are generally not in charge of the family finances and rarely have a regular monthly income. Prior to eliciting WTP of the parents, each family was assured that they would not be charged; that the WTP estimates were strictly used for research and not for clinical decisions, and the results would not be reported to the insurance company. Each participating parent was asked to complete the study questionnaire twice. This allowed us to assess whether parents were consistent in their estimates, e.g., whether the 2 consecutive WTP values provided by the parents “made sense” (12). For example, if at the time of the first interview a bid of $200 was rejected by the parent, then, at the time of the second interview, we would expect the parent to reject all bids of $200 and higher.

Hypothetical medications.

Each parent was introduced to 2 hypothetical drugs. The first hypothetical drug, ARTHRO, was a once-daily oral medication without any side effects that would provide clinical remission, e.g., reduce morning stiffness to <5 minutes and provide complete resolution of active arthritis and all joints with limited range of motion. The second hypothetical drug was NO-STOM-ACHE, a once-daily oral medication without any side effects that would prevent abdominal pains, nausea, vomiting, early satiety, and indigestion.

Starting bids.

The use of different starting bids is thought to avoid the risk of anchoring WTP values around a specific starting bid value (13). For each of the 2 study medications, local pediatric rheumatologists, nurses, and social workers were asked to provide lower estimates for costs that appeared sensible considering the patient population seen in the clinic and the proposed effects of the hypothetical drugs. Initial starting bids for each medication were then chosen to enable WTP estimates as low as 50% below the lowest suggested WTP estimate. For ARTHRO, possible initial starting bids were $50, $125, $200, and $300. For NO-STOM-ACHE, they were $5, $25, $30, and $40. At the time of the first study visit, families were randomized to one of the initial starting bids depending on the last digit of the patient medical record number. Starting bids for each of the studied medications at the time of the second study visit were based on the mean WTP per initial starting bid group.

Closed ended question approach.

In the absence of an accepted standard to determine WTP, we used the previously described closed-question approach that minimizes the cognitive task of the respondent and avoids large numbers of nonresponses (1, 14, 15). Followup questions to starting bids were given because they are thought to increase the precision of WTP estimates (16).

WTP was measured by presenting, on separate study forms for each study medication, the properties of the study medication together with the following sentence stem, “Would you, in order to obtain the treatment results, be willing to pay $ ___ per month in excess to your present medical expenses for anti-rheumatic therapies for your child?” The randomized initial starting bids were written in the blank. The parents were then asked to accept or reject the initial starting bid (dichotomous choice yes/no) for each medication separately. If they accepted the starting bid, then the dollar amount for the subsequently offered bid was doubled, and the parents were asked again to accept or reject the offer. This was repeated once so that parents could accept or reject up to 4 times the initial starting bid. The procedure was stopped if parents rejected the offered bid. At the time of the second study visit, this procedure was repeated using the appropriate starting bids.

Disease activity.

The treating physician rated global disease activity on a 100-mm visual analog scale (VAS) anchored at either end with “not active” or “very active.” This VAS was presented together with the sentence stem: “The arthritis of the patient is currently.”

Childhood Health Assessment Questionnaire (CHAQ).

The CHAQ consists of 2 components: disability and discomfort. Disability is assessed in 8 domains (dressing and grooming, arising, eating, walking, hygiene, reach, grip, and activities) (17). The disability index is calculated as the unweighted average of the 8 domain scores and yields a disability score between 0 (no disability) and 3 (most severe disability). Discomfort is determined by the presence of pain over the preceding 1-week period and is measured by a 10-cm VAS anchored at either end by “no pain” and “very severe pain” (VAS-pain).

Global rating of wellbeing.

Parents were asked to rate the patient's wellbeing during the preceding week using a 10-cm double-anchored linear analog scale presented with the sentence stem: “My/my child's overall well-being is” (18–20). The lower endpoint of the scale was marked with “extremely bad” and the upper endpoint of the scale was defined as “excellent” (VAS-well).

Pediatric Quality of Life Inventory (PedsQL).

Besides the PedsQL generic core scale (PedsQL-GC) (4), the PedsQL rheumatology module (PedsQL-RM) (5), a disease-specific HRQOL module for children with arthritis, is available. The PedsQL-GC has 4 domains and assesses physical, emotional, social, and school functioning. From the sum of the raw scores a total health summary score ranging from 0 to 100 can be calculated, with higher scores indicating higher HRQOL. The PedsQL-RM encompasses 5 domains: pain and hurt, daily activities, treatment, worry, and communication. As done for the PedsQL-GC, a summary score was calculated for the PedsQL-RM based on the unweighted averages of the domain scores.

Statistical analysis.

Data was analyzed using SAS 8.2 (SAS Inc., Cary, NC) and EXCEL 2000 (Microsoft Inc., Redmond, WA).

WTP estimates were adjusted for the available family income, which was defined as the monthly pretax family income minus assumed 25% taxes, minus monthly costs of health insurance, and minus monthly out-of-pocket health care expenses of the family for their child with JIA.

Because of the skewness of the data, nonparametric Kruskal-Wallis analysis of variance (ANOVA) was performed to test the effects of the randomized starting bids on the WTP estimate of the parents. Spearman correlation coefficients (rs) were calculated to evaluate the relationship between the measured outcomes. We hypothesized that outcome parameters (physical function, HRQOL, pain, well-being, disease activity) including the number of involved joints (ARTHRO) and GI symptoms (NO-STOM-ACHE) were weakly (0.2 ≤ rs < 0.40) correlated to the WTP and that the consideration of the available family income would strengthen the relationship between the outcome parameters and the WTP. Spearman correlation coefficients between 0.41 and 0.60 were considered moderate, whereas those between 0.61 and 0.80 constitute strong–moderate correlations (21).

Regression analysis and linear modeling were done to determine potentially important relationships between WTP and the other outcomes measures. Regression analysis was aimed at identifying the best fitting models using Mallow's CP as selection criterion. Standardized beta coefficients (STBs) were calculated to determine the relative importance of the predictors of WTP. STB values ≥0.2 are thought to be of clinical importance even if statistical significance is not reached in a well fitting model (22). Given the small sample size, results of the regression analyses can only be used for hypothesis generation rather than as a model to estimate parent WTP for their child.

RESULTS

Patient and parent demographics.

Families of children (n = 62) with JIA were studied. Approximately 15% of the JIA patients treated regularly at the CCHMC pediatric rheumatology clinic were included in the study. Less than 5% of the approached patients refused participation, mostly because of time conflicts. Families were recruited from the rheumatology clinic during routine visits after informed consent. Verbal assent was obtained from children ages ≥8 years. The mean age of the enrolled children was 10.5 years (SD 4.9 years, range 0.5–18 years). The mean time between study visits was 3.5 months (SD 0.6 months). Results of visit 1 were similar to those of visit 2 and are shown in Table 1.

Table 1. Demographics*
 nPercentage of totalRangeMeanStandard deviationMedian
  • *

    VAS = visual analog scale; NSAIDs = nonsteroidal antinflammatory drugs; TNF = tumor necrosis factor; MTX = methotrexate; CHAQ = Childhood Health Assessment Questionnaire; PedsQL = Pediatric Quality of Life Inventory.

Traditional outcomes      
 Age of assessed patient62 0.5–18.910.54.8910.8
 Active joint count62 0–19.00.72.570
 Joints with limited range of motion62 0–27.02.24.510
 Any involved joints at the time of the study visit33530–462.846.541
 10-cm VAS of disease activity by physician60 0–9.53.02.812.4
Medications      
 Any systemic daily or weekly medication5792    
 NSAIDs4979    
 Any gastointestinal-protective medication1727    
 Any second-line agent3150    
  Prednisone35    
  TNF-blocking medication1423    
  Methotrexate2540    
  MTX dose in mg/kg body weight  0.13–0.960.420.240.41
Economic information      
 Before tax family income per month, $55 500–7,0003,9001,8884,500
 Costs of health insurance per month, $55 0–1,200240310200
 Out-of-pocket health expenses per month, $57 10–200555535
 Hours of work missed due to arthritis per month, $58 0–455.282
 Health insurance      
  No insurance2134    
  Medicare/Medicaid58    
  Private insurance3048    
  Information refused610    
Patient-centered outcomes      
 10-cm VAS of well-being62 2.8–10.07.91.488.1
 10-cm VAS of pain62 0–9.92.62.891.3
 CHAQ62 0–2.130.490.580.25
 Any gastointestinal symptoms during the last week2954    
 10-cm VAS of gastointestinal symptoms60 0–9417.52.584.5
 PedsQL generic core scale59 26.5–10078.319.7684.8
  Physical function59 12.5–10074.725.2382.8
  Emotional function59 20.0–10078.822.4185.0
  Social function59 15.0–10081.321.0985.0
  School function (not for children ages 2–4 years)46 40.0–10081.718.3190.0
 PedsQL rheumatology module59 30.3–10075.916.8476.8
  Pain and hurt59 0–10064.831.2568.8
  Daily activity59 30.0–10088.418.50100.0
  Treatment59 21.4–10068.322.2665.0
  Worry (not for children ages 2–4 years)46 50.0–10082.016.8583.3
  Communication (not for children ages 2–4 years)46 16.7–10079.622.0283.3

The participating patients had, on average, 2.84 involved joints, and 53% of the patients had at least 1 joint involved with arthritis at the time of the study visits. The vast majority of children was treated with nonsteroidal antiinflammatory drugs and 50% also required a second-line drug to control the arthritis, with methotrexate being the most commonly used of these agents (Table 1). Five children (5 of 62; 8%) were treated with intermittent joint injections instead of using regular systemic medications. Fifty-four percent of the parents reported GI symptoms of their children, and 27% of the patients were treated with daily gastroprotective agents, mostly H2 blockers and proton pump inhibitors.

Seven families (7 of 62; 11%) refused to provide information on family income. The mean monthly family pretax income was $3,900 (median $4,500). The majority of the patients had either health insurance or health expenses were covered by state programs. On average, $240 was spent on health insurance, and an additional $55 was spent for copayments and medications not covered by the health plans. Based on the monthly income before taxes and after consideration of monthly health care expenses, estimated 25% taxes, and out-of pocket health expenses, the mean available monthly income was $2,638 (median $3,190). A mean of 5.2 working hours per month were lost by the family due to the patient's JIA. Lost working hours were mostly due to clinic appointments.

Descriptive analysis of the outcome measures.

Twenty-two patients (22 of 62; 36%) had a CHAQ score of 0 and therefore normal physical function, whereas 64% of the parents reported limited physical function of their child with arthritis (mean CHAQ score 0.48, median 0.25). Based on parents' reports, the HRQOL of the children was, on average, 78.3 and 75.6 as measured by the PedsQL-GC and the PedsQL-RM, respectively (Table 1).

Acceptability of WTP technique.

All parents were willing to engage in the WTP assessments on both visits. Four of the 62 parents (6%) refused to pay anything in addition to the current health expenses to achieve clinical remission (WTP for ARTHRO = 0). Similarly, 6 of the parents (10%) refused to pay anything for NO-STOM-ACHE, with 4 parents refusing to pay for both ARTHRO and NO-STOM-ACHE. Two of these children had no GI problems and well-controlled arthritis; 1 child had active arthritis but no GI complaints; and 1 child had active arthritis and severe GI problems but the available family income was low.

Consistency of WTP estimates.

Consistency was assessed by comparing the WTP values obtained at the time of the first interview to those measured at the time of the second interview. For ARTHRO, 54 of the 62 parents (87%) provided estimates that were consistent. At the time of the second interview, previously accepted bids during the first interview were rejected by 8 of 62 parents (13%). Inconsistent WTP estimates for ARTHRO were given for children who were rated as clinically unchanged (n = 4), improved (n = 3), or worse (n = 1) by the treating physicians. Similarly for NO-STOM-ACHE, 56 of the 62 parents (90%) provided consistent responses, whereas 10% of the responses were inconsistent. Children with inconsistent WTP estimates for NO-STOM-ACHE had either no GI symptoms (n = 4) or had less GI symptoms at the time of the first as compared with the second interview (n = 2) as rated by the parents.

WTP for ARTHRO.

On average, families were willing to pay an additional $395 (median $300) or 15% ($395/$2,638) of the monthly available family income to obtain a drug that guarantees clinical remission of arthritis (Table 2). The distribution of the WTP estimates for ARTHRO of the studied families is shown in Figure 1. The WTP of the families was unlikely influenced by the starting bids of ARTHRO because there were no important differences between the starting bids of the different WTP strata defined by the percentiles of the WTP for ARTHRO (Table 2). This was supported by the results from the Kruskal-Wallis ANOVA and Wilcoxon's rank sum test (P = 0.43; not significant).

Table 2. Willingness-to-pay estimates for ARTHRO*
 nWTP for ARTHRO, $Median starting bid for ARTHRO, $Mean starting bid for ARTHRO, $
  • *

    ARTHRO = hypothetical medicine that guarantees complete clinical response; WTP = willingness to pay; IQR = interquartile range.

  • Available family income computation: Monthly before tax income − monthly costs of health insurance − 25% taxes − monthly out-of-pocket health care expenses.

0–25th percentile15163125127
26th–50th percentile18300125172
75th–100th percentile29600125142
Mean ± SD62395 ± 329  
Median (IQR) 300 (450)  
WTP as % of available monthly  income55   
 Mean 15  
 Median 9  
Figure 1.

Distribution of willingness-to-pay (WTP) values for ARTHRO.

WTP for NO-STOM-ACHE.

WTP for NO-STOM-ACHE had a mean of $109 per month (median $80) and was lower than that for ARTHRO (Table 3). The distribution of the WTP estimates for NO-STOM-ACHE of the studied families is shown in Figure 2. On average, the WTP for NO-STOM-ACHE for children with GI symptoms was higher than for children without GI symptoms (mean/median $127/$80 versus $83/$50). Different from ARTHRO, the starting bids of NO-STOM-ACHE were moderately correlated to the WTP estimates for NO-STOM-ACHE (rs = 0.48, P < 0.0002). Kruskal-Wallis ANOVA and Wilcoxon's rank sum test supports that the WTP estimates for NO-STOM-ACHE were influenced by the initial starting bids (P < 0.02).

Table 3. Willingness-to-pay values for NO-STOM-ACHE*
 nWTP for NO-STOM-ACHE, $Median starting bid for NO-STOM-ACHE, $Mean starting bid for NO-STOM-ACHE, $
  • *

    NO-STOM-ACHE = hypothetical medicine that eliminates gastrointestinal symptoms; WTP = willingness to pay; IQR = interquartile range.

  • Available family income computation: Monthly before tax income − monthly costs of health insurance − 25% taxes − monthly out-of-pocket health care expenses.

0–25th percentile2020515
26th–50th percentile18802553
75th–100th percentile2313030200
Mean ± SD62109 ± 125  
Median (IQR) 80 (140)  
WTP as % of available monthly income55   
  Mean 4  
  Median 2  
Figure 2.

Distribution of willingness-to-pay (WTP) values for NO-STOM-ACHE.

Construct validity of the WTP estimates to express healthcare preferences.

Univariate analysis.

WTP for ARTHRO was weakly correlated to the VAS-well, VAS-pain, and HRQOL measured by the PedsQL-GC, especially to its physical function domain (Table 4). When correcting for the available family income, the correlations between the WTP for ARTHRO and VAS-well, pain (VAS-pain, PedsQL-RM pain and hurt domain), the physician assessment of disease activity, and physical function (CHAQ, PedsQL-GC physical function domain) were stronger, as we had hypothesized. The time missed from work was also significantly correlated with the WTP for ARTHRO corrected for the available family income. In the univariate analysis, the WTP for ARTHRO was unrelated to the number of active, limited, or involved joints, irrespective of whether the available family income was considered or not.

Table 4. Correlations of willingness-to-pay values with other measures outcomes*
Spearman correlation coefficients (p value if ≤ 0.05)ARTHRONO-STOM-ACHE
WTPWTP adjusted for available incomeWTPWTP adjusted for available income
  • *

    Data are Spearman correlation coefficients (P value if ≤ 0.05). ARTHRO = hypothetical medicine that guarantees complete clinical response; NO-STOM-ACHE = hypothetical medicine that eliminates gastrointestinal symptoms; WTP = willingness to pay; VAS = visual analog scale; CHAQ = Childhood Health Assessment Questionnaire; PedsQL = Pediatric Quality of Life Inventory.

  • Available family income computation: Monthly before tax income − monthly costs of health insurance − 25% taxes − monthly out-of-pocket health care expenses.

Age of assessed patient0.130.040.0010.04
VAS well-being−0.20 (< 0.05)−0.41−0.13−0.22
VAS pain0.33 (< 0.012)0.46 (< 0.004)0.36 (< 0.006)0.40 (< 0.01)
CHAQ0.070.30 (0.05)0.160.30 (0.05)
Number of joints with active arthritis0.03−0.020.1−0.05
VAS disease activity0.140.32 (< 0.04)0.170.34 (< 0.04)
VAS gastrointestinal symptoms0.150.31 (< 0.04)0.120.34 (< 0.03)
Income0.18 0.07 
Out of pocket health expenses0.17 0.12 
Cost of health insurance−0.02 −0.13 
Available family income0.23 0.15 
Hours missed at work0.160.28 (0.05)0.160.10
PedsQL generic core scale−0.22−0.23−0.23−0.29
 Physical function−0.30 (< 0.04)−0.35 (0.05)−0.25−0.32
 Emotional function−0.17−0.22−0.20−0.21
 Social function−0.16−0.24−0.19−0.33 (0.05)
 School function−0.22−0.10−0.14−0.21
PedsQL rheumatology module−0.12−0.30 (0.05)−0.09−0.20
 Pain and hurt−0.05−0.31 (0.05)−0.01−0.16
 Daily activities0.01−0.190.08−0.14
 Treatment0.06−0.080.80.03
 Worry−0.18−0.19−0.1−0.002
 Communication−0.14−0.17−0.15−0.19

WTP for NO-STOM-ACHE was weakly correlated to the VAS-pain and the starting bid only. Corrected for the available family income, WTP for NO-STOM-ACHE was moderately correlated to the VAS-pain but not to the pain and hurt domain of the PedsQL-RM. After correction for the available family income, the WTP for NO-STOM-ACHE was correlated higher to patient pain (VAS-pain), GI symptoms (by VAS), and HRQOL as measured by the PedsQL-GC (especially social and emotional functioning domains) and the PedsQL-RM (Table 4).

Multivariate analysis: predictors of WTP of parents of children with JIA.

Exploratory analysis was done to identify potential predictors for the WTP. Regression analysis and linear modeling was used to identify such predictors. Under the consideration of model fit and after correction for the available family income, WTP for ARTHRO was best predicted (R2 = 53%), by the number of involved joints (STB = 0.54, P =0.02), the PedsQL-RM (STB = −0.38, P = 0.06), and the physician assessment of disease activity (STB = 0.55, P = 0.03). After correction for the impact of the starting bids and the available family income, the WTP for NO-STOM-ACHE was best predicted (R2 = 26%) by the VAS for GI symptoms (STB = 0.45, P = 0.01). An alternative good model to predict the WTP for NO-STOM-ACHE was the HRQOL measured by the PEDSQL-RM (R2 = 30%, STB = −0.49, P = 0.01).

DISCUSSION

We studied the WTP of parents of children with JIA and confirmed that this is a feasible and potentially useful approach for measuring health care preferences. The WTP method has previously been used in patients with arthritis (15, 23–26). Similar to findings in adults with rheumatoid arthritis (23), parents were willing to pay a considerable proportion of their monthly income for complete relief of arthritis symptoms of their children with JIA, suggesting a quite substantial negative impact of the disease on the families of children with JIA. As in earlier WTP studies in arthritis, the WTP method was readily mastered by the parents (8, 15, 23, 24, 26). Similar to other studies in which parents were asked to provide WTP estimates, parents of children with JIA appeared to have few problems in providing monetary values for certain health interventions for their child's health (14, 27–30).

Besides the available family income and similar to previous studies, disease activity, pain, restriction in activities of daily living, and HRQOL were all correlated to the WTP of the parents (15, 24). Not surprisingly, the number of hours missed at work also appeared to be important for the WTP of the parents for ARTHRO, supporting the hypothesis that WTP is not only influenced by the health of the child under consideration but also by external factors influencing the entire family (14). Based on multivariate analysis and similar to other studies in arthritis, the number of involved joints appeared to impact on the WTP for ARTHRO (23, 31).

GI symptoms were quite prevalent in the cohort. WTP for guaranteed GI symptom relief was astonishingly high with a mean of $109. However, different from the WTP for clinical remission (ARTHRO), WTP for NO-STOM-ACHE was significantly influenced by the chosen starting bids, thus is likely higher than what would be expected if either more patients were interviewed or the initial starting bids were chosen differently. This stresses the inherent problems of using starting bids. Starting bid bias has been previously observed in other WTP studies and steps to avoid it were taken in our study (varying starting bids, followup questions) (16).

There is no commonly accepted best method for measuring WTP, but the closed-ended approach or dichotomous-choice approach is preferred by some because it more accurately resembles the market situation. Individuals are accustomed to deciding whether to buy a product when confronted with a certain price. Therefore, like others (13, 15, 32), we used a closed-ended, double-bounded approach with followup questions to elicit the WTP of the parents. This is opposed to the direct, open-ended approach of measuring WTP, where the respondents are asked to suggest a certain highest monetary amount themselves (16). Previous research suggests that open-ended measurements of WTP more often lead to refusal to engage in the WTP process and frequently result in unrealistic WTP estimates (33). Moreover, the closed-ended approach more realistically reflects the types of decisions individuals make every day (34), in which goods and services are offered at a certain price. Besides the starting bid bias discussed above, another disadvantage of the closed-ended method is that respondents tend to say yes to amounts above their maximum willingness to pay, resulting in an overestimation of the WTP (34). However, followup questions, as performed in the current study, increase the precision of the WTP estimates (25). WTP can also be measured by using so-called payment cards for different options of the WTP from which the subjects can choose (35) or discrete-choice questions, in which patients are presented a single price for a treatment that would yield a specified health change. Instead of the bidding game performed in this study, payment cards and the discrete-choice questions have all been used in the past and each of these approaches has its merits and disadvantages (13, 36, 37). As shown in the case of NO-STOM-ACHE, a starting bid bias can prevail, even when randomized starting bids and followup questions are used.

To our knowledge, consistency of WTP estimates of parents has not been examined in the past. In the current study, consistency of parent WTP was assessed by assuming that previously rejected bids would be unacceptable to parents later on. Under this assumption and similar to other studies (12), ∼10% of the WTP estimates provided by the parents were inconsistent. However, one could argue that changes in JIA disease activity and GI symptoms between visits accounted, at least in part, for inconsistent WTP estimates. For example, parents of children whose arthritis has improved between the 2 study visits may want to pay less for ARTHRO because their children's disease is already well controlled by the available medications. If we used a less stringent definition of inconsistency of WTP values and assumed that only parents whose children's arthritis or GI symptoms remained the same or improved between visits should not accept bids that are higher than those previously rejected, then still 5 parents (8%) would have provided inconsistent WTP for ARTHRO and 4 parents (6%) for NO-STOM-ACHE.

Like others, we used parent proxies to measure the WTP for health preferences of the children with JIA. The advantages, disadvantages and the rationale of using proxy ratings in general and for pediatrics in particular, have been reviewed (38–43). All previous studies of WTP for pediatric health preferences used adult proxy reporters. Consistent with the findings of our study, Donaldson et al (30) found that WTP is a feasible approach to elicit pediatric health preferences from parents. A study by Liu et at (14) suggests that the WTP of mothers to avoid influenza is approximately twice as high for their children as for themselves. A study by Sach et al (27) assessing the WTP for pediatric cochlear implants confirms findings of our study that parent WTP is relatively high and depends on the family income. High WTP of parents to avoid pain of children is supported by previous studies (29, 44). This is consistent with the findings of our study where pain of the children appeared to be an important decision criterion for the WTP of parents.

It has been proposed that WTP cannot be used in countries with health care systems provided by the state, such as Norway, Great Britain, or Canada, or for patients whose drug expenses are entirely covered by state programs. Although the absolute amounts for the WTP estimates will depend on the health system and the financial situation of the society, the WTP technique has been successfully performed in the setting of socialized health systems in the past (15, 25). Irrespective of the health care system, WTP data has the potential to provide health care purchasers and providers with information on intensity as well as direction of health preferences of certain populations (2, 30).

In our experience, the performance of the WTP methods was more easily understood and more willingly performed than the standard gamble technique, which we previously used to assess a similar group of families. Nonetheless, the approach to measure health care preferences is strongly influenced by the research question posed or clinical problem to be assessed. HRQOL estimates based on the different methods are not directly comparable, and health preferences derived by different methods are not interchangeable.

Based on the clinical features of JIA patients in our clinic, the studied patients appear to be representative for our clinic (45, 46). No data are available to evaluate the external validity of the WTP estimates for clinical remission (ARTHRO) and complete resolution of GI symptoms (NO-STOM-ACHE) in JIA.

Previous research suggests that WTP estimates may be affected by the age of the patient under consideration and that WTP for children is lower than for adults and infants (47). Age of the studied patients was not significantly associated with the WTP estimates in either the univariate or multivariate analysis. It has previously been shown that WTP values for guaranteed treatment effects, like those promised by ARTHRO and NO-STOM-ACHE, are often higher than those that would be present in a real-life setting in which only a certain percentage of the patients benefit from a medical intervention (48). WTP was likely affected by the knowledge of the parents that they are not accountable for the accepted WTP amount stated in the study. Previous research in patients with asthma who were offered an asthma management program suggests that the actual WTP may be as low as 30% of the hypothetical WTP (13). Nonetheless, WTP may provide a more tangible measure of health care preferences than, for example, change scores of HRQOL measures. It remains to be determined whether there is a best approach to measure health care preferences. Future research in the measurement properties of WTP estimates is necessary to determine whether WTP has real advantages over other methods (2, 36).

There are several limitations to our study. We did not followup refused bids by decreasing the subsequent bidding amount offered, but discontinued the exercise instantly. This might have led to increased WTP for the studied medications. However, the primary objective of the study was to establish the feasibility and construct validity of the WTP method, rather than determining exact WTP estimates for clinical remission or resolution of GI symptoms. Thus the objectives of the studies were unlikely influenced by the lack of followup on refused bids. In addition, the analysis is based on data from a small group of patients with often well-controlled arthritis, adding to potential problems with the generalizability of the concrete WTP estimates.

In summary, the WTP appears to be a useful tool for assessing health care preferences of families of children with JIA. Like other approaches to measure health preference, the WTP technique has disadvantages and advantages (1, 2, 49). The method is feasible and has construct validity when performed by parents of children with JIA. More insight is required regarding test–retest validity, responsiveness to change, and the relationship between hypothetical and actual WTP values in JIA. Of special interest would be to compare the WTP of antiinflammatory medications with differences in efficacy or side effect spectra.

Ancillary