To obtain data on the preferences of patients with depression for medication management by conducting a DCE, so as to help clinicians and policy makers in developing countries to formulate effective medication management strategies and improve medication adherence of patients with depression.

This study followed the design principles of the DCE study, and data collection was conducted from June 2024 to February 2025. Our study adhered to the checklist for reporting guidelines on good practices in health care regarding DCE developed by the International Society for Pharmacoeconomics and Outcomes Research (ISPOR) (35). Our study protocol has been published (36).

We determined attributes and levels according to the recommended methods (25, 35). The electronic databases, Wanfang database, Embase, Web of Science, China National Knowledge Infrastructure (CNKI) and PubMed, were used to review the relevant literature on adherence in people with depression. Snowballing method was used to obtain more comprehensive information. Based on this, potential attributes and levels were identified and an interview outline was determined. Subsequently, people with depression were interviewed one-on-one, and data saturation was reached after 15 interviews. Afterwards, focus group discussions were conducted with depressed patients who did not participate in the one-on-one interviews, and a total of 4 focus group discussions (FGDs) were conducted with 3 depressed patients in each group. One-on-one consultations were conducted with 1 health economist, 1 psychiatric nurse, 1 psychiatrist, and 1 pharmacist (all of whom worked for more than ten years) to obtain more attributes and to determine the applicability of potential attributes in current medical health care. Ten possible attributes have been obtained. Some researchers believe that the optimal number of attributes for DCE is 6 (37). Therefore, people with depression were required to vote on the importance of these attributes, ultimately including the six attributes that they consider the most important, namely “adverse reactions” “provider” “follow-up frequency” “cost” “follow-up method” and “ convenience of purchase”. See Attachment A for attributes and levels details (36).

A unlabeled design was used to construct the choice sets to avoid the attention of respondents to the target attributes (38). A fractional factorial design was performed using Ngene software to generate 36 choice sets, and they were randomly divided into 3 blocks to reduce the cognitive burden on responders. Each choice set included an “opt-out” option to avoid overestimation of participation rates (39). To examine the consistency of respondent selection, the second choice set for each block was repeatedly included as the thirteenth-choice set.

The survey questionnaire consists of three parts. The first part introduces the purpose, content, questionnaire filling requirements, and precautions of the research. The second part was a general information questionnaire used to collect the socio-demographic characteristics of the respondents. The third part is the DCE selection task. In this part, each attribute and its levels are described in detail, and an example of choice set is given for “warm-up”. Attachment B shows an example of choice set (36).

Twenty patients with depression were invited to participate in the pilot test, and the statements of the questionnaire were modestly adjusted to improve clarity and intelligibility based on the results and feedback from the pilot test.

The subjects of this study were depressed patients without other psychiatric disorders. Patients who could understand and fill out the questionnaire, has the experience of taking antidepressants currently or in the past six months, and are willing to participate in the survey were included.

Convenience sampling was adopted to determine the respondents. Data collection includes both online and on-site methods. Emails or We-chat accounts of patients who met the inclusion criteria were obtained from the admission registries of four hospitals in Sichuan Province, and questionnaires were provided to them via We-chat or email. At the same time, patients who meet the inclusion criteria are invited from the outpatient and inpatient departments of these hospitals to participate in the survey, and questionnaires are collected on-site. The data collection period is from June 2024 to February 2025.

Stata16.0 software was used to analyze the data. Descriptive statistics were performed on the socio-demographic characteristics of the respondents. The preferences of people with depression were estimated using a mixed logit model. All coefficients in the model are assumed to be normally distributed. The cost attribute was modeled as a continuous variable to estimate the willingness to pay (WTP) of patients with depression, and the remaining attribute levels were coded as dummy variables. The value of the coefficient β indicates the relative importance of the attribute level to the reference level in the medication management scheme, while the statistical significance of the coefficient β indicates whether the attribute level influences the selection of depressed patients. Grouping based on sociodemographic variables to explore the preference differences among people with depression with different demographic characteristics. The nlcom command is used to simulate the uptake rate, which is the change in the probability of depressed patients receiving medication management programs when one or more attribute levels change compared to the baseline medication management program. For all calculations, statistical significance was indicated at a two-sided p < 0.05 level.

A total of 500 patients with depression from four hospitals in Sichuan province were invited, of which 420 patients agreed to participate in the survey, and 411 patients completed the questionnaire. Among the 411 completed questionnaires, 38 participants (9.2%) were excluded due to failing the consistency test, resulting in 373 participants included in the final analysis. Among them, there were more females (n=218) than males (n=155) (accounting for 58.45% and 41.55% respectively). Most patients were aged 18 – 29 years (124 respondents, accounting for 33.24%). The number of patients with recurrent depression is 197, accounting for 52.82%. Patients with high school education account for 37.73%, and the highest number of respondents with monthly income between 2000 - 6000, accounting for 47.99%. Most of the respondents (69.97%) took multiple antidepressants. Details are provided in Table 1 .

The results of the mixed logit estimates for the total sample are reported in Table 2 . The WTP measures how much respondents are willing to pay to improve the characteristics of the medication management program, and the results are reported in Table 3 . We found that the coefficients at all attribute levels were significant, that is, p < 0.05, indicating that the six attributes included in this study are important for adherence to antidepressant medication in depressed patients. The results showed that the coefficient of the cost attribute was negative, suggesting that depressed patients preferred the less costly medication management option. The most salient finding was that adverse reaction profiles overwhelmingly determined patients’ preferences. The most important attribute for patients with depression is the adverse reactions of antidepressants. Their preference coefficients for slight and mild adverse reactions are 0.905 (95% CI 0.6629873 - 1.146418) and 0.832 (95% CI 0.591335 - 1.073632), respectively. Patients demonstrated the highest WTP to avoid severe side effects, valuing slight adverse reactions at 1,517.77 CNY (95% CI: 712.37 - 2,323.18) and mild reactions at 1,396.62 CNY (95% CI: 786.14 - 2,007.09) annually. This substantial WTP differential - exceeding all other attributes by 3 – 4 fold - underscores that tolerability should be the primary consideration when prescribing antidepressants. Follow-up modality emerged as the second most critical factor influencing adherence. Patients strongly preferred blended care models combining face-to-face and remote follow-ups (coefficient=0.777, WTP = 1,304.25 CNY) over telemedicine-only approaches. Psychiatrists were the most preferred follow-up providers (coefficient=0.513, 95%CI 0.3193968 - 0.7071995). It is worth noting that psychiatric nurses were also preferred by people with depression as follow-up providers (coefficient=0.232, 95%CI 0.3193968 - 0.7071995). 96%CI 0.0220205 - 0.4414176). Although medication accessibility showed statistical significance (coefficient=0.196), its relatively modest WTP (328.51 CNY) and wider confidence intervals (-6.56 to 663.59 CNY) position it as a secondary consideration compared to side effects and follow-up quality. This indicates that while patients demonstrated a positive preference for this attribute, they were not willing to pay a statistically significant additional amount for it. At the same time, the statistics of the SD coefficients for all attribute levels are significant, suggesting that there is heterogeneity in the preference of people with depression for attribute levels.

Subgroup analysis stratified by depression recurrence status revealed clinically distinct preference patterns, and the results are shown in Table 4 . For people with depression who have not experienced recurrence, the three most important attribute levels are, in turn, adverse reactions–slight (coefficient=1.386), follow-up methods-alternate between face-to-face & telephone/We-chat (coefficient=1.363), and adverse reactions-mild (coefficient=1.046). In addition, their preference for convenience of purchase-general convenience was negative (coefficient=-0.267), indicating that they prefer to purchase antidepressants on-site at pharmacies. For patients with recurrent depression, the three most important attribute levels are, in turn, adverse reactions-slight (coefficient=0.748), follow-up frequency-once every four weeks (coefficient=0.699), and adverse reactions-mild (coefficient=0.620). It is worth noting that compared with telephone/WeChat follow-up method, patients with recurrent depression have no significant preference for face-to-face and telephone/We-chat alternate, while their preference for attribute level face-to-face is negative (coefficient=-0.462), which indicates that they prefer remote follow-up, that is, follow-up through telephone/We-chat.

The probability of people with depression receiving a medication management program after a change in one or more attribute levels is simulated, and the main findings are reported in Figure 1 . When a single attribute level changes, adverse reactions have the greatest impact on preferences of people with depression. When adverse reactions change from severe to slight and from severe to mild, the probability of people with depression receiving medication management program increases by 0.362 and 0.333, respectively. When the follow-up method changed from telephone/we-chat to face-to-face and telephone/we-chat alternately, the acceptance probability of medication management program of patients with depression increased by 0.311. When multiple attribute levels change simultaneously, the optimal medication management program can increase the probability of people with depression accepting the program by 0.517. The optimal program should meet the following requirements: (1) psychiatrists as follow-up providers; (2)follow-up once every two weeks; (3) slight adverse reactions; (4)follow-up methods were face-to-face and telephone/we-chat alternately; (5) it is very convenient to purchase drugs; and (6) it costs 400 yuan per month. When changing fewer attribute levels, such as face-to-face and telephone/we-chat follow-up provided by psychiatrists and slight adverse reactions, the probability of accepting the medication management program increased by 0.478, which was close to the optimal medication management program.

This study has some limitations. Firstly, although we recruited patients from four hospitals in Sichuan Province to reduce sample bias, this may not fully represent the views of people with depression in other regions of China. Secondly, our use of convenience sampling, while logistically efficient, may have resulted in a sample that does not fully capture the diversity of adherence behaviors, particularly among high-risk or hard-to-reach populations. Caution is therefore warranted when extrapolating these findings to broader clinical settings. Thirdly, although we combined online and on-site survey responses to ensure sample diversity, we did not perform a formal sensitivity analysis to compare preferences between these subgroups. Differences in unobserved characteristics (e.g., digital literacy, socioeconomic status) may introduce bias. Future studies could stratify recruitment to assess mode effects. Finally, in order to reduce the cognitive burden of the respondents, only the six most important attributes and their levels were included in our study, and the excluded attribute levels may also be important, which limits our discussion of the results. It may be the research direction of future scholars to include more comprehensive attribute levels and carry out national large sample surveys.

Our findings provide actionable guidance for clinical practice to enhance medication adherence in depression. The strong patient preference for minimal adverse effects (with a 36.2% increase in program uptake for slight side effects) underscores the need for clinicians to prioritize well-tolerated antidepressants and proactively address side effect concerns through patient education and dose adjustments. The differential follow-up preferences based on relapse history—remote options (telephone/WeChat) for patients with prior relapse versus in-person pharmacy visits for non-relapsed patients—highlight the importance of personalized care pathways.

The near-optimal strategy combining psychiatrist-led follow-up (face-to-face or telephone) with low-side-effect regimens (47.8% uptake increase) offers a practical, evidence-based model for implementation. These results are particularly relevant for healthcare systems aiming to optimize resource allocation, as they demonstrate patients’ willingness to pay for better treatment experiences. By adopting these tailored approaches—including telehealth integration for high-risk patients and collaborative care models—clinicians may improve adherence rates and, consequently, treatment outcomes in real-world settings. This is especially valuable in resource-limited areas where improving adherence may reduce the economic burden of ineffective depression treatment.