Derivation of the mapping functionsPopulation

Children and adolescents with tic disorders who visited in Fujian Medical University Union Hospital from 2018 to 2021 were invited to participate in our survey. Participants could be diagnosed as tic disorders. The study was approved by the institutional review boards.

Inclusion criteria: (1) Male or female child or adolescent, ≤ 17 years of age at the time of signing the informed consent/assent. (2) Diagnosis of TS that met the Chinese Classification of Mental Disorders-third edition (CCMD-3). (3) Patient and designated guardian(s) able to comprehend and comply with protocol requirements. (4) Patient and designated guardian(s) volunteered to participate in this study and written informed consent was obtained.

Measurements

All patients received an informed consent informing them about the study and asking for their voluntary participation. Those who agreed to participate were provided with the Yale Global Tic Severity Scale (YGTSS) [18, 19] and the EuroQol five-dimension questionnaire youth version (EQ-5D-Y) [20], along with some questions regarding sociodemographic characteristics. The questionnaires were evaluated by the caregiver. In addition, the sociodemographic characteristics data and questionnaires was distributed and collected by trained personnel.

The YGTSS is widely used to assess the severity of tic disorders, and its internal consistency and test-retest raliability have been verified [18]. The YGTSS includes three subscales: motor tics, vocal tics and a separate impairment scale. Motor and vocal were included in our survey. The motor and vocal tics are rated separately on a 0–5 scale across five dimensions: number, frequency, intensity, complexity and interference [18]. The scores can be summed to produce the Total Motor Tic score (range 0–25) and the Total Vocal Tic score (range 0–25), and the combined Total Tic score (range 0–50) [18].

The EQ-5D is the most widely used preference-based scale for calculating health-related quality of life [21, 22]. EQ-5D-Y is a version specifically for children and adolescents. The EQ-5D-Y measures the health state (such as 11,111, 21,231 etc.) of patients from five dimensions including mobility, self-care, performance of usual activities, pain or discomfort, and anxiety or depression. Then, the utility index can be calculated by the value set and we adopted the EQ-5D-Y-3 L Value Set for Chinese population [23].

Statistical analysis

To describe the sample, we used frequencies and percentages for categorical variables and means and standard deviations (SDs) for quantitative variables. Unless critical information was missing, all data was used for developing the mapping functions and validating these functions.

Mapping algorithms

We used two different utility mapping algorithms to convert YGTSS onto the EQ-5D-Y-3 L.

1.

General linear models (GLMs). A GLM requires the dependent variable is continuous and the residuals must be normally distributed. However, the EQ-5D-Y utility are defined in an interval, so GLM is not always appropriate [24].

2.

Beta regression models. To avoid above issue, we built beta regression models, which use the logit function as a link and allow modeling outcomes with skewed distributions [25]. The EQ-5D-Y-3 L is a three-level scale, which is prone to complete health states (11,111), resulting in a utility value of 1, which is called the “ceiling effect”. The “ceiling effect” makes the data tend to be skewed and with truncated tails (censored data). The beta regression models require the response variable has to be restricted to the open interval (0, 1); therefore, we transformed the boundary points of the EQ-5D-Y utility to slightly lower or higher values by applying the formula [Y(N-1) + 0.5]/N, where Y is the observed EQ-5D-Y utility, N is the number of participant in our survey.

Four models strategy were developed used in the 2 statistical approaches. The following predictor variables were considered: Model 1: the YGTSS Total tic scores; Model 2: the YGTSS subscales; Model 3: the YGTSS Total tic scores plus covariates; Model 4: the YGTSS subscales plus covariates. Sex, age, height and weight were included as covariates. As a dichotomous variable, sex was assumed to be 0 for females and 1 for males.

The Akaike (AIC) and the Bayesian (BIC) information criteria and adjusted R-squared for GLM and beta models were calculated to compare the goodness-of-fit. We also compare the predictive performance of the 4 models by calculated the mean absolute error (MAE) and the root mean squared error (RMSE). All statistical analyses were conducted with R softwre version 4.1.2.

Cost-utility analysisModel structure

A decision tree model was constructed to simulate the clinical management of children and adolescents with tic disorders for 1 year (52 weeks) under two different doses of aripiprazole and placebo treatment, using TreeAge Pro Version 2022 (Fig. 1). The clinical events included in the model consist of adverse reactions leading to treatment discontinuation, as well as the efficacy of the drug after administration: post hoc response, partial response, and non-response. Post hoc response, partial response, and non-response were defined as a reduction in the YGTSS total score of > 50%, 25–50%, and < 25% from baseline, respectively [26].

Fig. 1

Schematic representation of model structure. A decision-tree analysis assessed aripiprazole as monotherapy management of Tourette disorders over a time horizon of 52 weeks

Model input

From the perspective of China’s healthcare system, this study only examined direct medical costs, including the cost of drugs, outpatient care, inpatient stay, rehabilitation, office-based physicians, ancillary therapy and auxiliary material/sundries, as shown in Table 1. There were no more research reporting the direct medical costs from the perspective of China’s healthcare system. Consequently, all cost data origined from the research of Dodel et al., and they were converted by the equivalent value of China using purchasing power parities (PPPs) [14]. Then, the costs were converted based on the average exchange rate of the US dollar in December 2022 (USD 1 = 6.98 RMB), with a discount rate of 5%. The unit price of aripiprazole in China was obtained from the Yaoyuan network (XXX) [27]. The treatment doses of aripiprazole were chosen based on the research by Floyd Sallee et al. Therefore, the doses of children and adolescents with tic disorders were 5 mg/day (Low dose) or 10 mg/day (High dose).

The utility of children and adolescent with tic disorders before and after treatment were converted based on the research of Floyd Sallee et al. using our mapping model [26]. The probability of post hoc response, the probability of partial response and the discontinuation rate due to AEs were also originated from the research of Floyd Sallee et al. [26]. We assumed that the disutility of adverse reactions leading to treatment discontinuation was 10% of the initial utility.

Sensitive analysis

The uncertainties of key parameters were analyzed using one-way sensitivity analysis (OWSA) and probabilistic sensitivity analysis (PSA). For drug costs and direct medical costs, one-way sensitivity analysis were performed across a wide range (± 20%) to capture all possible scenarios. The range of probability was ± 10%. For all key parameters, PSA was applied to reflect the impact of their stochastic characteristics on the results. In the PSAs, we performed 1000 Monte Carlo iterations on the uncertainty of all key parameters within 95% confidence intervals. Reasonable values were used in the absence of these confidence interval values (e.g., 20%). The ranges and distributions of the parameters used in the sensitivity analyses are given in Table 1. According to the WHO’s recommendation, three times China’s per capita GDP in 2022 was used as the threshold value (36832.95 US dollars).