outline
The EQ-VT study was conducted using a representative sample of the Dutch population in terms of age and gender. Respondents were recruited by a marketing research firm and interviewed by trained interviewers. Quality control measures were implemented to ensure data integrity, and regression methods used to analyze the data and generate a new set of EQ-5D-3L values were implemented. Finally, the new set of values was compared to the old set of values.
health status assessment
In our study, we used cTTO to assess the value of health status. For health states that respondents perceive to be better than death, cTTO presents a traditional TTO task in which participants make an adaptive choice between a period of X years in perfect health and 10 years in a valued health state. Through an iterative process, For health conditions that are perceived to be worse than death, a lead time TTO applies. In this case, the respondent chooses between X years in perfect health and a fixed period consisting of 10 years in perfect health and 10 years in a specified health state. Here X varies again until the point of indifference is reached and the health value is now determined by ( This approach allows cTTO to generate state values ranging from -1 to 1(5, 6).
experimental design
Each respondent completed 12 cTTO tasks consisting of an assessment of the worst possible health state, 33333, plus 11 additional health states selected from a subset of 196 of the 243 possible EQ-5D-3L health states. This set was derived by removing certain states from the full set of 243 EQ-5D-3L states: 11111 (all states), 33333 (worst state, already included), and 45 states deemed implausible. According to Viney et al., implausibility was defined as the presence of level 3 for mobility (“limited to bed”) and level 1 (“no problems”) for self-care or daily activities. (12). For the remaining 196 states, Lamers et al. Dutch EQ-5D-3L values are established and classified into 11 severity groups based on utility score (2). Respondents were randomly assigned one health condition from each severity block, stratified by health condition severity with condition 33333, for a total of 12 cTTO respondents per respondent. This experimental design ensured a broad representation of the EQ-5D-3L system while excluding logically inconsistent states.
Recruitment, sampling and survey administration
Dutch market research company Bureau Fris recruited participants from an existing panel to obtain a sample representative of the general Dutch population by age and gender. The target sample size was 400 participants, and participants received €30 as compensation for their time. Respondents were sampled according to age and gender, and efforts were made to collect a sample that was closely representative of the Dutch population in terms of geographical region and level of education. Interviews were conducted remotely by four trained interviewers using Zoom videoconferencing software, a validated approach shown to produce results similar to face-to-face interviews (13, 14).
Survey Structure
The survey began with respondents providing informed consent. This was followed by a series of questions about the respondent’s demographic characteristics, including age, gender, highest level of education, and region of residence. Next, respondents self-completed the EQ-5D-3L instrument.
Respondents then took the cTTO assessment task. They first completed two practice problems to familiarize themselves with the cTTO task and its procedures. In these practice tasks, respondents ranked ‘being in a wheelchair’ and ‘being in much better health than being in a wheelchair’ or ‘being in much worse health than being in a wheelchair’ as important depending on whether the first question ended with a lead-time TTO or a regular TTO.
Respondents then completed three additional practice EQ-5D-3L problems (21112, 32323, and 13311) to further learn about the cTTO process. After a practice round, each respondent ranked 12 EQ-5D-3L health states as important using the cTTO. Upon completing all cTTO tasks, respondents were presented with a ranking of their responses and asked to flag items they believed were out of order using the feedback module (15).
Finally, respondents completed a brief questionnaire about their preferences for different time periods, then concluded the interview and thanked them for their participation.
quality control
Several quality control measures were implemented to ensure data integrity. Interviewers received training from the research director and conducted five practice interviews with acquaintances, but they were not included in the final sample.
After each batch of 5 to 10 interviews collected by interviewers, data were taken and examined for protocol compliance and interviewer effectiveness. An interview was flagged as potentially problematic if: (1) The respondent did not enter a lead time TTO in either of the two wheelchair example tasks, suggesting that they did not receive a complete explanation of the method. (2) respondents spent less than 3 minutes total on the two wheelchair example questions, making it unlikely that the cTTO task was adequately described; (3) Respondents completed 12 actual cTTO questions within 6 minutes. or (4) state 33333 did not receive the lowest value and no other state received a value that was more than 0.5 lower (10).
If more than 40% of an interviewer’s interviews were flagged, the data was removed and the interviewers were retrained. Interviewer effects were also examined by comparing the distribution of interviewers’ responses. If the regression analysis revealed a positive slope for the level sum score, the respondent was excluded from the final sample. This means that serious health conditions are considered better than less serious health conditions.
statistical analysis
cTTO data were analyzed using a heteroscedastic Tobit model with the utility function described in Eq. (1):
$$\begin{sorted}&\\&\:{U}_{j}=&\\&{\beta\:}_{0}+{\beta\:}_{1}{MO2}_{j}+{\beta\ :}_{2}{MO3}_{j}+{\beta\:}_{3}{SC2}_{j}+&\\&{\beta\:}_{4}{SC3}_{j}+{\beta\:}_{5}{ UA2}_{j}+{\beta\:}_{6}{UA3}_{j}+{\beta\:}_{7}{PD2}_{j}+&\\&{\beta\:}_{8}{PD3}_{j }+{\beta\:}_{9}{AD2}_{j}+{\beta\:}_{10}{AD3}_{j}+{\epsilon\:}_{j}\end{aligned}$$
(1)
here, \(\:{U}_{j}\) It represents the utility assigned to a health state. \(\:j\)The variable is \(\:{MO2}_{j}\), \(\:{MO3}_{j},\)… .,\(\:{AD3}_{j}\) A dummy variable indicating the presence of the corresponding problem level for each dimension. For example, if you are in good health \(\:j\) If 12,313 \(\:{SC2}_{j}\), \(\:{UA3}_{j}\) and \(\:{AD3}_{j}\) 1 and all other variables in the model are 0. that \(\:\beta\:\) The parameters are estimated quality weights associated with level dimension dummies. \(\:{\beta\:}_{0}\) It’s an intercept. \(\:{\Epsilon\:}_{j}\) is the error term that is assumed to depend on the independent variables, as shown in Eq. (2):
$$\begin {sorted}&\&\:{\Sigma\:}_{JJ}=&\&\Text{e}\text{xt{p}\text{p}({\gamma\:}_{0}+{ \:}_{1}{mo2}_{JJ}+{\Gamma\:}_{2}{mo3}_{JJ}+{JJ}+{\Gamma\:}_{3}{JJ}_{JJ}_{JJ}_{J&{\Gamma\:}_{4 }{SC3}_{JJ}+{\Gamma\:}_{5_{JUA2}_{JJ}+{\Gamma\:}__{JUA3}_{JJ}_{JJ}_{JJ}_{7}{7}{PD2}_{JJ}+&\\\\ {\Gamma\:}_{8}{PD3}_{JJ}+{\Gamma\:}_{Ad2}_{JJ}_{JJ}_{JJ}_{10}{AD3}_{AD3}_{JJ})\END{Sort}$$$
(2)
It is also assumed that values of -1 are censored, which is the lowest possible value derived from the cTTO task. This is described in Eq. 3:
$$\:U=\left\{\begin{array}{c}U”\:if\:U”>-1\\\:-1\:if\:U”\le\:-1\end{array}\right.$$
(3)
here \(\:you\) Indicates cTTO response \(\:you”\) Fundamental potential value. After estimating the Tobit model, the resulting utility function can be used to predict values for all 243 EQ-5D-3L health states using Eq. 3 Connect the predicted potential utility to the cTTO response censored at the -1 lower bound.
To compare new sets of values with previous values, scatter plots, Bland-Altman plots, and summary statistics including mean difference, mean absolute error (MAE), root mean square error (RMSE), and Lin’s concordance correlation coefficient (CCC) ( 16 ) were used. – (17) This analysis provides a comprehensive assessment of the agreement between the two sets of values.
