Willingness-to-pay and time-trade-off as the measures of burden of psoriasis: A pooled analysis of community, hospital, and web-based samples

Background The study aimed to compare willingness-to-pay (WTP) and health utility measured by time-trade-off (TTO) in psoriasis.

based on the expected utility theory. In previous studies, the reported utility in skin diseases ranged from 0.64 in bullous disorders, to 0.907 in psoriasis, and 1.0 in alopecia, cosmetic concerns, and urticaria 7 .
According to previous data, TTO seems to fail to distinguish the burden of different minor skin diseases 7 .
Willingness-to-pay (WTP), asking how much the responder is willing to pay for a hypothetical treatment, is also considered as a preference-based measure of disease burden in monetary terms 7 . The WTP method shows its potential in dermatology since the burden of many skin conditions tends to be transient and relatively mild and respondents are not willing to accept a shorter life expectancy without the skin disease in TTO or even a small probability of death in SG 8 . In the last two decades, the WTP metrics have been introduced in the eld of dermatology and occasionally used in several skin diseases, such as psoriasis, atopic eczema, vitiligo, port wine stains, and rosacea 9−16 . Among which, both Schiffner R. et al. and Lundberg L. et al. reported TTO and WTP in psoriasis patients 9,13 ; however, their properties of detecting different emphasis on the burden of psoriasis regarding data distribution, concurrent validity, and discrimination have not been thoroughly compared previously. Furthermore, to our knowledge, there is no previous study in dermatology that used community-based patients to capture the patients' WTP, and there is no literature that elicited WTP of patients with psoriasis in China before. Our study aims to estimate WTP and TTO of in a pooled sample of psoriasis patients, and to compare WTP and TTO regarding the sensitivity in measuring the burden of psoriasis.

Study Design and Patients
This was a cross-sectional study with a pooled sample of patients with diagnosed psoriasis recruited from one of the following settings ( Figure S1).
1) The Dongfeng-Tongji cohort 17 : The cohort recruited retired workers initially, and it began to recruit workers in-service since 2017 in addition to the previously recruited subjects. We used a sample of participants enrolled between July 2017 and October 2018. Certi ed dermatologists participated in the eld survey and diagnosed skin diseases including psoriasis. Patients with psoriasis were included in the pooled analysis.
2) The dermatology clinic of a tertiary hospital in Changsha: Patients with psoriasis who gave their consent were consecutively recruited throughout 2017.
3) The Chinese Psoriasis Online Voluntary Registry 18 : The registry is a non-pro t and coordinated effort to collect information directly from patients who are at least 18 years of age or older and have been diagnosed with psoriasis. With over 140,000 registered participants, the registry helped patients with psoriasis in the eld of health education and social support. Unfortunately, this registry was not mentioned in a review of on psoriasis registry worldwide owing to a lack of publications 19 . We randomly invited a group of 250 patients to participate in this survey in September 2017. A subsidy was offered if a patient nished the survey to facilitate the response rate.

Measures of Clinical Severity and QoL
In this study, PASI was used as the gold standard to measure the clinical severity of psoriasis. Despite certain limitations such as the lack of consideration of symptoms, PASI is the most extensively used psoriasis clinical severity score and the most thoroughly validated instruments 20 . The following cut-off points were used: mild (0-10), moderate (11)(12)(13)(14)(15), and severe (> 15). For the web-based patients, we used the self-reported clinical severity (0-10). The self-reported clinical severity was elicited by asking "how severe do you rate your clinical symptoms today (0-10)"; among which, "0" represents free from any psoriasis symptom, and "10" represents the most severe symptoms. The self-reported clinical severity was also categorized into three groups: mild (0-3), moderate (4-7), and severe (8-10).
In this study, DLQI 21 was used to measure the QoL for all participants. DLQI is the most widely used QoL instruments in dermatology, and its psychometric properties in psoriasis have been widely validated.

Measurement of WTP
WTP was elicited using the contingent valuation method 22 , a widely used approach for the valuation of goods and services by potential consumers. Before the WTP amounts were elicited, the participants received education about psoriasis and that all payments were out-of-pocket with no insurance coverage.
WTP was elicited for two scenarios: WTP for controlling the disease and WTP for curing the disease. For WTP cure , an open-ended question was used to ask about the total amount of money that they would be willing to pay to cure psoriasis. In order to directly compare WTP and TTO in percentage manner, we used relative WTP in this study. Rather than asking absolute money as WTP cure , the WTP control was elicited by asking about the percentage of monthly income that the participant patients would like to pay to fully control the clinical symptoms of psoriasis. All hypothetical questions assumed no occurrence of side effects. The WTP cure in this study is expressed in USD with an exchange rate of 1 USD = 6.9523 CNY (2017).

Measurement of Utility by TTO
In this study, the TTO approach asks the subject to choose between a xed guaranteed life duration in his or her current health with psoriasis or a shorter duration with no psoriasis. In order to elicit accurately, simply, and understandably, we obtained the life expectancy for Chinese population 23 and informed participants their average life expectancies. Before elicitation, we thoroughly inquired the participants about their health status, including skin health and health status in general 24 . Participants were asked to choose from two options, option (A), remaining life-years (A years) with psoriasis in current status, or option (B), a shorter life duration (B years) without psoriasis or comorbid conditions. Ping-pong fashion was followed until arrival the maximum of time that patients are willing to trade-off for no-psoriasis status. The life duration in option B is varied until the respondent is indifferent between option A and option B. The TTO-derived utility of psoriasis is then calculated as B/(A + B).
Other Variables Age, gender, educational level, monthly income level, and duration of psoriasis were collected in the survey. Duration of psoriasis was dichotomized at the median (10 years) for the convenience of comparison.

Statistical Analysis
Medians and interquartile ranges (IQR) were used to describe the distribution of WTP and TTO, and Kruskal-Wallis H test was used to test the differences among groups. Concurrent validity was examined by correlations to PASI and DLQI. Spearman correlation coe cients were rst estimated between WTP and TTO with PASI, DLQI, and self-reported severity. The locally estimated scatterplot smoothing (LOESS) method was then used to examine the nonlinear correlations 25 . The receiver operating characteristic (ROC) curve and AUC were used to compare the capability of WTP and TTO to discriminate patients with moderate-to-severe psoriasis from mild cases. A bootstrap method was used for the multiple comparisons of the AUCs among the groups. Income and duration of disease were adjusted for in models. Statistical analysis was performed using the R software; the R packages "ggplot2" and "pROC" were used for plotting 26,27 . A P < 0.05 was considered statistically signi cant.

Data Distribution
The WTP cure showed a normal distribution after logarithmic transformation, while TTO showed a rightskewed distribution (Fig. 1). The distribution of WTP control largely overlapped across PASI and DLQI groups. The Kruskal-Wallis tests showed consistent results with the visual description: the medians of WTP cure were not signi cantly different across PASI or DLQI groups (Table S1). In contrast, the distributions of WTP control were well separated across PASI and DLQI groups. The distributions of TTO overlapped across PASI groups, but were well separated across DLQI groups. Statistical tests also showed consistent results (Table S1). The medians of WTP and TTO were not signi cantly different across the groups of self-reported severity. In addition, WTP control showed a oor effect (3% range in the lower 10% centile) while TTO showed a ceiling effect (0.05 range in the upper 10% centile, Table S2).

Concurrent validity
WTP control was signi cantly associated with PASI (r = 0.36, P < 0.001) but not with DLQI, while TTO was signi cantly associated with DLQI (r=-0.20, P = 0.001), but not with PASI (Table S3). Partial correlation coe cients adjusting for income and duration of psoriasis were consistent with the unadjusted results.
Nonlinear associations were further examined. WTP cure was associated with neither PASI nor DLQI. The association of WTP control with PASI was signi cant among the patients with mild psoriasis, while the con dence interval of the curve became very wide among those with moderate-to-severe psoriasis (Fig. 2). TTO was signi cantly inversely associated with DLQI, but was not correlated with PASI.  Figure S2). In contrast, the AUC of TTO (AUC = 0.64, 95% CI: 0.58-0.70) was signi cantly larger than that of WTP cure (AUC = 0.55, 95% CI: 0.58-0.74) and WTP control (AUC = 0.62, 95% CI: 0.56-0.67) in the discrimination for DLQI > 10, using the bootstrap method.

Discussion
For the rst time, the study reported WTP of Chinese psoriasis patients and, investigated WTP and utilities (TTO) in real-world patients with psoriasis. To our knowledge, the previous literature on WTP and TTO of psoriasis patients were based on hospital sample; and this may lead to a bias to capture the overview of impaired QoL in patients with psoriasis. This study included both community and hospital samples, providing a more comprehensive estimate of the burden of psoriasis in China. The median TTO estimate of our study population was 0.9 (IQR: 0.70-0.92), which was close to the upper limit of previous estimates 7 . Since our pooled sample is more representative for the patient population in China, the utilities elicited from the presenting study are more convincing and valuable.
This study was among the rst to systematically compare the properties of WTP and TTO as measures for the burden of psoriasis regarding data distribution, ceiling and oor effects, linear and nonlinear correlation, and capability of discrimination. In order to compare WTP and TTO in a more uni ed and direct approach, we used relative WTP in the present study since absolute WTP fails to generalised outside China. The results indicated that the WTP for monthly control (%) was highly correlated with the clinical severity of psoriasis (PASI) while TTO was highly correlated with non-preference-based QoL measure (DLQI). WTP cure in absolute monetary terms was relevant to neither the clinical severity nor the QoL, which was in accordance with the conclusions from previous studies that relative WTP might be a better indicator compared with absolute WTP 7 .
In our study, TTO showed a unique capability of representing QoL across all groups. TTO, a classic method to elicit health state utility, has been widely used across diseases. By asking people "how many years of current life with a certain illness they would like to give up in exchange for the rest living free from this condition," it allows direct comparisons of QoL across different disease areas. In this study, TTO demonstrated a consistent positive correlation with DLQI in patients with varying levels of scores. This result is interesting but not surprising. Several previous studies substantiated that DLQI is good at detecting impaired QoL 28, 29 . The more the patients felt embarrassed or self-conscious, or ashamed of their physical appearance, or experienced di culties in social interactions, the more they are willing to accept a shorter life expectancy without psoriasis. However, in our study, TTO showed ceiling effect and failed to show superiority in detecting clinical severity compared to WTP control , especially when psoriasis is in a mild status. Since TTO serves as a ruler of comparison across all diseases, the tool cannot be ne enough to detect nuances, especially in skin diseases with rare death or disability. We also identi ed a ceiling effect that the range of TTO in the upper 10% of patients with psoriasis was only 0.05. Indeed, even when their QoL was truly impaired, it was hard for patients with a skin disease to trade their life in return of free from cutaneous signs and symptoms.
Consistent with the suggestions from Chen S. et al. 8 , we continue to advocate the application of the WTP metrics in psoriasis. Indeed, WTP, as a monetary indicator, has inherent aws such as its correlation to income level. However, our results showed WTP control as a percentage of income could mitigate the effect of income originally existing in absolute WTP. In our study, the distribution of WTP for disease control (%) was well separated across PASI and DLQI groups, especially among those in a mild disease status(PASI < 10), indicating that WTP rather than TTO estimates can better re ect disease burden in mild psoriasis. Paralleled with the clinical severity of psoriasis, the dynamic conceptualization of the money paid out-ofpocket for a possible treatment may be more straightforward and sensitive to the patients when comparing with TTO. These ndings, if generalised, could provide further evidence to support the use of relative WTP to measure the burden of disease in dermatology, since most of the skin diseases are relatively mild and temporary compared to other major diseases.
There are limitations in this study. First, there is a lack of gold standard for measuring the burden of psoriasis. However, PASI and DLQI itself are generally acknowledged tools to capture the burden of patients with psoriasis, and they are complementary to each other in the evaluation of psoriasis 30 . As a result, PASI and DLQI were used as the references. Second, the sample size of patients with PASI > 10 was relatively small, and a robust conclusion could hardly be drawn among those with moderate-to-severe psoriasis. Further study is needed to verify our ndings in larger samples and different populations.
Despite the limitations, our study can serve as an essential contribution to the existing literature. By the use of a pooled sample from a population-based cohort, a tertiary hospital, and the biggest psoriasis online registry in China, the current study reports the rst data of WTP of psoriasis patients in China and can be more comprehensive than hospital-based studies. The characteristics of the patients varied substantially across the different sources of sample, which further demonstrates the heterogeneity of patients in different settings and underscores the importance of including different patient populations in the sample. More importantly, we found that WTP control is superior in detecting minor clinical severity status of psoriasis compared to TTO, while TTO is a more sensitive indicator for the impaired healthrelated QoL associated with psoriasis. In general, relative WTP and TTO are both recommended to depict the complete picture of the burden of psoriasis regarding multifaceted emphases.