The impact of shared decision-making on patient-reported outcomes in Traditional Chinese Medicine in Shanghai, China: a cross-sectional study using structural equation modeling.

1
Introduction
When patients are diagnosed with cancer, they are challenged to manage information about their condition, treatment options, and potential adverse effects, as well as their options involve trade-offs between aggressive management and quality of life, often making high-risk decisions with uncertainty about the benefits of their treatment.1,
2 To deliver high-quality cancer care, it is essential for healthcare providers to remain attentive to patients’ evolving and unmet needs.3 Toward this end, the concept of shared decision-making (SDM) has come to play an increasingly important role in supporting patient-centered cancer care. It emphasizes the importance of clinicians clearly describing the potential risks and benefits of different treatment options, while encouraging patients to explicitly express their preferences and values.4,
5
As an integral component of quality healthcare, growing scholarly attention has focused on examining how SDM influences patient-reported health outcomes, healthcare quality indicators, and healthcare utilization.6 Previous studies have pointed to the effectiveness of SDM on metrics of interest. Specifically, greater patient-perceived SDM is consistently associated with higher decision satisfaction.7, 8, 9 Patients who actively participate in treatment decisions often report greater satisfaction, feel better informed, and able to know what matters most to them.10 Furthermore, when clinicians use SDM to encourage active patient participation in treatment choices, the level of patient-perceived effective decision-making is improved,11 thereby promoting improved adherence to selected options and reducing post-decision regret.12,
13 Ultimately, patients are more likely to achieve improved health outcomes when treatment decisions are based on the best clinical evidence, aligned with their personal values, mutually endorsed by both parties, and are feasible to implement.14 These, in broad terms, represent typical outcomes that can be measured – described by Shay as cognitive-affective, behavioral, and health outcomes.4 While SDM may influence health outcomes directly, its impact is more commonly exerted through indirect or mediated pathways, particularly via patients’ emotional-cognitive responses and subsequent behaviors.14
Traditional Chinese Medicine (TCM) technology, represented by herbal supplements, has emerged as an important complementary and integrative medicine modality in cancer care when conventional therapies fail to relieve symptoms or cause additional adverse effects.15,
16 TCM emphasizes dialectical identification and personalized treatment, yet in clinical practice such individualized approaches may yield uncertain outcomes regarding efficacy, potential toxicity and psychological risk, administration routes and medical costs.15,
17 Thus, patients and clinicians must jointly make decisions about TCM treatment that aligns with the patient's medical condition and individual inclinations best. While the single or multivariate effects of SDM on decision-making satisfaction, treatment decision usefulness, and health-related quality of life (HRQoL) have been previously investigated, empirical support remains inconclusive regarding the synergistic effects and potential mechanisms through which SDM influences these patient-reported outcomes in TCM. To elucidate the potential decision-making benefits and areas of improvement for SDM implementation in the context of TCM, this study examined a theory-based structural path model of SDM and patient-reported outcomes by using structural equation modeling (SEM) among patients with lung cancer.

2
Methods
2.1
Conceptual framework and hypothetical model
The present study was built on the above theories and literature review by proposing a refined conceptual framework based on the patient-centered communication theoretical model and its adapted version (Fig. 1).4,
14 This study hypothesized that SDM would positively influence patients' satisfaction with the decision-making process, facilitated acceptance of TCM treatments and made treatment regimens more responsive to their needs, thereby improving the likelihood of their HRQoL. This study was designed to address this hypothetical assumption by providing theoretical, empirical and practical considerations.

2.2
Study design, setting and participants
Participants were recruited from TCM and/or Oncology departments of nine tertiary hospitals in Shanghai, China through a cross-sectional multicenter survey with a structured questionnaire from September 2020 to January 2021. As a national demonstration area for the development of TCM, Shanghai featured a relatively mature TCM service system alongside high utilization rates of such services in cancer care.18 The city’s extensive service coverage also lent its representativeness for conducting research in the field of TCM. Eligible participants included 1) adult patients (aged ≥ 18 years) with a confirmed diagnosis of lung cancer and have or had received anti-cancer TCM injections, regardless of the specific injection type, and 2) were willing to provide written informed consent. Participants with illnesses or symptoms considered likely to impede effective communication were excluded from the study. In our study, the SDM processes were not standardized or centrally regulated across hospitals, allowing us to capture patients’ perceptions of SDM as they naturally occurred in routine TCM clinical settings.

2.3
Sample size and data collection procedure
The minimum sample size was calculated using the algorithm developed by Westland for determining the lower bound of simple size in SEM.19 We used the given latent variables (n = 3) and the number of observed variables (n = 14) in this study to achieve an effect size of 0.3, a power of 0.8, and a probability level of 0.05. The minimum sample size was 110 individuals for each model. A structured self-reported questionnaire was used to collect data, which was revised in conjunction with expert interviews following a pre-test of 20 patients. In the formal survey, uniform instructions were provided to participants regarding the purpose of the survey and how to complete the questionnaire, and questionnaires were administered only after obtaining informed consent from each participant. Participants were informed that their responses would remain anonymous and confidential, and that all data would be used solely for research purposes. They completed the questionnaires independently, with researchers available on site to address any questions or provide clarification as needed during the process. When the questionnaires were completed, the researchers reviewed each questionnaire for completeness, and any missing information related to sample characteristics was subsequently collected through follow-up visits.

2.4
Measurements
2.4.1
The demographic questionnaire
Demographic characteristics of participants included gender, age, education level, employment status, annual household income per capita, and clinical profile, including the type and stage of lung cancer, period since diagnosis, and route of administration.

2.4.2
Perceived shared-decision making
The 9-Item Shared Decision-Making Questionnaire (SDM-Q-9) was utilized to assess the degree to which patients felt involved in the decision-making process. The questionnaire contains nine items, each corresponding to a distinct step in the SDM process, and it has demonstrated high reliability among cancer patients in numerous medical settings in China.20, 21, 22 The original SDM-Q-9 employs a six-point Likert scale; however, in this study it was adapted to a five-point Likert scale ranging from 1 (completely disagree) to 5 (completely agree) to improve clarity for respondents and ensure consistency with other patient-reported outcome measures used in the survey. The total score of the scale ranged from 9 to 45, with higher SDM-Q-9 scores reflecting higher levels of perceived SDM processes. The Cronbach's α value for the SDM-Q-9 was 0.932.

2.4.3
Decision-making satisfaction
We developed two items to assess decision-making satisfaction: (1) Overall, I am very satisfied with my medical counseling experience; and (2) the consistency of the reported role preference and the actual role experienced in the physician encounter. The second item was designed based on the modified Control Preference Scale (CPS), which presents five distinct decision-making roles and requires participants to identify the one that most accurately reflects their preferred position.23, 24, 25 Participants were asked two questions about their preferred and perceived roles in the decision-making process, and responses were then assessed for consistency by categorizing their responses into “(completely) active”, “(completely) passive”, or “collaborative” roles. The items were rated on a five-point scale from 1 (completely disagree/inconsistent) to 5 (completely agree/consistent), and item 2 was reverse coded. The final score was the sum, and higher scores indicated a higher level of decision-making satisfaction. The Spearman-Brown coefficient for the two items was 0.649.26

2.4.4
Treatment decision usefulness
We also formulated two items to measure usefulness of treatment decision: (1) I am very supportive of receiving anti-cancer TCM injections; and (2) the treatment decision fully addresses my concerns and priorities. The response range was from 1 (completely disagree) to 5 (completely agree). Therefore, the score was calculated using the sum of the two questions with higher scores indicating greater perceived usefulness of treatment decision. The Spearman-Brown coefficient for the two items was 0.665.26

2.4.5
Health-related quality of life
HRQoL measures assess an individual's functioning and well-being employing the EuroQoL-5 dimensions-5 levels (EQ-5D-5 L) scale. The scale is an instrument with a descriptive system and a visual analog scale (VAS). The descriptive system asked participants to indicate their health status across five dimensions: mobility, self-care, usual activities, pain/discomfort, and anxiety/depression, with each dimension having five levels: no problems, slight problems, moderate problems, severe problems, and unable to/extreme problems.27 The psychometric property of the descriptive system has been validated in Chinese populations.28,
29 Patients' health utility values were calculated using the Chinese-specific EQ-5D-5 L value set developed by Luo et al.,29 with scores ranging from −0.391 (worse heath) to 1.000 (full health). The EQ-VAS enabled participants to rate their current overall health status on a vertical visual analogue scale ranging from 0 (worst imaginable health) to 100 (best imaginable health). To examine whether the choice of HRQoL indicator influenced the results, both the EQ-5D index value and EQ-VAS score were used separately to represent the HRQoL for patients.

2.5
Statistical analyses
SPSS Statistics version 20.0 and Amos version 29.0 (IMB Corp) were used for statistical computing. Descriptive analyses used frequencies and percentages to report categorical variables. Continuous variables were summarized as mean ± standard deviation (SD), or median with interquartile range, as appropriate for the data distributions. The distribution differences of perceived SDM-Q-9 score across individual characteristics was examined by t-test or ANOVA. Nonparametric tests were used, if requirements for parametric test procedures were not met.
SEM is used to specify and test linear relationships between directly measurable variables (observed variables) and variables that cannot be directly measured and represented by multiple observed variables (latent variables).30 To examine the associations among SDM, decision-making satisfaction, treatment decision usefulness, and HRQoL, Pearson correlation analysis was firstly used to examine the direction and degree of relationships among variances. Then, confirmatory factor analysis (CFA) was conducted to evaluate whether the measurement for the latent variable was executed correctly. Composite reliability (CR) was used to demonstrate composite reliability, and average variance extracted (AVE) was used to evaluate the convergent validity of the constituent factors. The structural model was then using SEM evaluated the direct and indirect effects between latent factors in the hypothesized theoretical model. The robust maximum likelihood estimator was used as not all variables were normally distributed. Subsequently, bias-corrected bootstrapping method (5000 samples) was used to employed the statistical significance of all effects within each pathway by computing a 95 % confidence interval (CI). After that, multi-group SEM was used to examine the structural variance of latent variables across groups with statistically significant demographic characteristics in SDM-Q-9 scores. Path-by-path comparisons were evaluated using critical ratios of differences (CRD) to determine whether significant differences existed in the structural paths across groups. Model fit was evaluated using multiple indices: absolute fit measurement (standard root mean square residual (SRMR), root mean square error of approximation (RMSEA)), incremental fit measurement (comparative fit index (CFI), Tucker-Lewis index (TLI)) and parsimony-adjusted measurement (χ2/DF).31 All P-values were two-sided, with statistical significance set at α ≤ 0.05.

3
Results
3.1
Characteristics of the participants
A total of 350 lung cancer patients were recruited for this study, out of which 347 (response rate: 99.1 %) were included in the final analysis, and the male-to-female ratio of participants was 1.2:1. The participants' mean age was 64.8 ± 8.7 years (range 40 to 87 years), mostly being retired (63.7 %). The majority of participants had been diagnosed with non-small cell lung cancer (82.4 %), and most were at an advanced stage (stage Ⅳ: 40.1 %), and their mean duration of disease was 27.3 ± 26.1 months (range 0.1 to 156 months). The proportion of those who utilized TCM as monotherapy or complementary TCM was quite comparable. Univariate analysis showed that there were differences in SDM-Q-9 scores among patients with different gender, residence, tumor stage, period since diagnosis, and route of administration (P < 0.05). The full sample characteristics and the comparison of SDM-Q-9 scores in different sample characteristics are presented in Table 1.

3.2
Mean, standard deviations and correlations of the study variables
Table 2 shows the means, standard deviations and bivariate correlations among the variables. The Pearson's correlation analysis showed that SDM-Q-9 score, decision-making satisfaction, treatment decision usefulness, EQ-5D-5 L index, and EQ-VAS score had significant positive correlations with the other variables, respectively, with correlation coefficients ranging from 0.197 to 0.582.

3.3
Measurement model of the hypothesized model for latent variables
Supplements 1–2 present the CFA procedure to evaluate the validity of the constituent factors. The modified model yielded a good fit, with SRMR = 0.028 (< 0.05), RMSEA = 0.047 (< 0.08), CFI = 0.985 (> 0.90), TLI = 0.980 (> 0.90), χ2/DF = 1.773 (< 3). The standardized factor loadings of the constituent factors ranged from 0.570 to 0.856 and were all statistically significant (P < 0.001), which indicated that the constituent factors responded well to the latent constructs. The CR and AVE of the latent variables were within the acceptable range. As presented in Supplement 3, the model demonstrated satisfactory reliability and validity based on the fit indices, supporting its suitability for SEM analysis.

3.4
Structural model of the hypothesized model
Separate structural models were constructed for each HRQoL indicator (i.e., EQ-5D-5 L index value and EQ-VAS score). For the model using the EQ-5D-5 L index value, the final model fit indices were as follows: SRMR = 0.027 (< 0.05), RMSEA = 0.042 (< 0.08), CFI = 0.986 (> 0.90), TLI = 0.982 (> 0.90), χ2/DF = 1.601 (< 3). Similarly, the model based on the EQ-VAS score also showed acceptable fit: SRMR = 0.028 (< 0.05), RMSEA = 0.044 (< 0.08), CFI = 0.985 (> 0.90), TLI = 0.979 (> 0.90), χ2/DF = 1.678 (< 3). Fig. 2 illustrates both models along with their standardized path coefficients, with factors of latent variables omitted for clarity.
Table 3 presents a comprehensive overview of the standardized effects in the model. In model 1 (using the EQ-5D-5 L index value to represent HRQoL), SDM not only directly positively affected both decision-making satisfaction (β = 0.438, 95 %CI: 0.206 – 0.639) and treatment decision usefulness (β = 0.380, 95 %CI: 0.172 – 0.577), but also indirectly positively affected treatment decision usefulness through decision-making satisfaction (β = 0.104, 95 %CI: 0.012 – 0.304). Treatment decision usefulness had a direct positive effect on the EQ-5D-5 L index value (β = 0.308, 95 %CI: 0.088 – 0.577). Therefore, SDM exerted an indirect positive effect on the EQ-5D-5 L index value either through treatment decision usefulness alone (β = 0.117, 95 %CI: 0.028 – 0.290) or sequentially through decision-making satisfaction and treatment decision usefulness (β = 0.032, 95 %CI: 0.004 – 0.136).
Similarly, in model 2 (using the EQ-VAS score to represent HRQoL), the direct effect of SDM on decision-making satisfaction (β = 0.438, 95 %CI: 0.206 – 0.639) and treatment decision usefulness (β = 0.380, 95 %CI: 0.172 – 0.577) were positively significant. The indirect effect of SDM on treatment decision usefulness through decision-making satisfaction also positively significant (β = 0.099, 95 %CI: 0.009 – 0.300). Treatment decision usefulness also displayed a significant direct effect on the EQ-VAS score (β = 0.216, 95 %CI: 0.018 – 0.450). Therefore, the mediating effect of treatment decision usefulness was significant, amplifying the positive impact of SDM on the EQ-VAS score, either independently (β = 0.083, 95 %CI: 0.009 – 0.224) or through decision-making satisfaction (β = 0.021, 95 %CI: 0.002 – 0.112).

3.5
Multi-group analyses by sample characteristics
The details of the multi-group analyses are presented in supplements 4–8. In brief, both models indicated that SDM positively affected decision-making satisfaction in male patients (model 1: β = 0.649, 95 %CI: 0.427 – 0.820; model 2: β = 0.644, 95 %CI: 0.411 – 0.816), but no significant effect among females (model 1: β = 0.000, 95 %CI: −0.208 – 0.178; model 2: β = 0.027, 95 %CI: −0.141 – 0.304). Conversely, SDM directly exerted a significant positive effect on treatment decision usefulness in female patients (model 1: β = 0.544, 95 %CI: 0.232 – 0.799; model 2: β = 0.533, 95 %CI: 0.191 – 0.798), whereas no such effect was found in males (model 1: β = 0.140, 95 %CI: −0.112 – 0.384; model 2: β = 0.150, 95 %CI: −0.098 – 0.398). Additionally, model 2 indicated a significant direct effect of treatment decision usefulness on the EQ-VAS score in patients diagnosed for more than 12 months (β = 0.498, 95 %CI: 0.117 – 1.076), while this effect was not significant in those diagnosed within 12 months (β = 0.033, 95 %CI: −0.185 – 0.359). No significant differences in structural paths were observed according to residence, tumor stage and route of administration.

4
Discussion
4.1
Summary of key findings
This study conducted a multicenter survey among patients receiving TCM for lung cancer to examine the multifaceted associations of SDM on patient-reported outcomes. Our findings indicated that SDM had a positive direct effect on both decision-making satisfaction and treatment decision usefulness. Moreover, treatment decision usefulness mediated the association between SDM and HRQoL. In contrast, the data failed to demonstrate a direct significant effect of SDM on HRQoL, whether HRQoL was measured using EQ-5D-5 L index value or EQ-VAS score. Multi-group analyses further revealed potential heterogeneity by gender and period since diagnosis, whereas no significant differences according to residence, tumor stage or route of administration, supporting the robustness of the main findings. By confirming the potential positive effects of SDM on affective-cognitive, behavioral, and health outcomes, our findings provide partial support for the hypothesized relationships among these constructs.

4.2
Comparison with previous studies
As the most frequently used outcome measure for SDM, decision-making satisfaction is a key component of affective-cognitive outcomes in healthcare decision processes.32 We found that patients with higher SDM-Q-9 score produces a direct positive effect on their TCM decision-making satisfaction, which is consistent with previous studies showing that patients who perceive greater involvement in SDM tend to report more favorable affective-cognitive outcomes.11,
33,
34 Additionally, there has been a shift towards understanding how patient-clinician communication, including SDM, may be associated with more distal behavioral and health outcomes.4,
14,
35,
36 On behavioral outcomes, previous studies suggest that SDM increases patients' treatment adherence and new technology uptake by their fostering trust, understanding, and confidence in the chosen course of action.37 Our study found that SDM exerted both direct and indirect positive effects on treatment decision usefulness with decision-making satisfaction acting as a mediator. When patients believe that the chosen treatment meaningfully addresses their disease-related concerns, they are more likely to view the decision as useful and to demonstrate greater engagement in subsequent treatment behaviors.38,
39 The implementation of SDM is associated with satisfactory patient-reported health outcomes has been documented.40,
41 In terms of HRQoL, some studies have reported positive results between SDM and health outcomes, while others indicate a negative or no relationship.4,
42 In our analysis, whether QoL was measured by the EQ-5D index or EQ-VAS score, the significant indirect effect in the absence of a significant direct effect was interpreted as a full mediation (or indirect-only mediation) of the effect of SDM on HRQoL by treatment decision usefulness. In other words, only when patients perceive that treatment decisions through SDM to be effective will they experience better quality of life outcomes, which correlate with their actual health conditions.

4.3
Clinical and scientific implications
Collectively, the findings from this study and in prior studies suggest that the role of SDM between patients and clinicians has intrinsic value to patient-reported outcomes, provides further support for healthcare institutions and policymakers to continue promoting patient-centered care models in routine TCM clinical practice. SDM emphasizes the effective two-way exchange of information and perspectives between clinicians and patients, a process that is further influenced by clinicians' explanations and emotional support.11,
43 Better SDM enables clinicians to more effectively communicate with patients about the uncertainty regarding the safety, effectiveness, cost, and ethical concerns of the potential treatments, ultimately arriving a mutually acceptable and ideally optimal decision.44 TCM practitioners should value the shaping of relationships and the maintenance of interactions with patients, thereby fostering patients’ understanding and internalization of their disease and improving their cancer care experiences.
Highlighting the importance of understanding the patients’ perspectives is critical to advancing the science of measuring SDM. In this study, we integrated the patients' preferred roles in decision control into satisfaction metrics, as most advocates of SDM take the view that patients should be engaged in decisions to the extent that they desire. Respecting these preferences, rather than compelling patients to engage in all decisions, is recognized as a key component of decision satisfaction.12,
45 The results suggest that patients are more satisfied with their treatment choices when their perceived level of involvement aligns with their preferred roles in the decision-making process.46,
47 Thus, simply assessing whether patients perceive themselves as involved in decision-making may not fully capture their actual experiences and preferences. It is essential to also consider whether patients genuinely desire to participate in decision-making to gain deeper insights into their expectations and experiences regarding the decision-making processes.
It was found that increasing control or involvement in healthcare can make patients more certain about their choices to support possibly adherence, ultimately leading to better HRQoL.48,
49 For cancer patients, SDM interventions not only facilitate a better understanding of their condition and treatment options but also alleviate anxiety stemming from uncertainty and information asymmetry, thereby enhancing psychological adaptation. These positive affective‐cognitive effects may, in turn, translate into improved treatment adherence and self-management behaviors, ultimately exerting a lasting impact on long-term cancer outcomes. Thus, while ensuring the effectiveness of the treatment interventions themselves remains a prerequisite for achieving improved quality of life, fostering patient-centered decision-making processes and enhancing clinician–patient communication and collaboration are also critical components for promoting long-term health outcomes, particularly among patients with a longer period since diagnosis.

4.4
Strengths and limitations
The main strength of this study is that it captured the perspectives of a broad sample of lung cancer patients across multiple centers. This allowed us to explore the relationships between SDM and patient-reported outcomes within the context of TCM clinical practice, thereby contributing to the growing body of research on SDM. Furthermore, the statistical analysis was conducted using SEM based on a conceptual framework, which enabled simultaneous addressing the multiple relationships of dependence among variables.
Several limitations must be acknowledged when interpreting the findings. First, the cross-sectional design does not inherently ensure the causality of the study variables, but the predictive structure of the model appears reasonable. Second, standardizing SDM processes across hospitals is challenging, and the identified relationships may have been influenced by such heterogeneity. That said, the sample diversity and representation likely reflect the general reality of TCM clinical practice for lung cancer in Shanghai. The findings should be generalized with caution to other regions with differing patient populations, healthcare resources, or patterns of TCM utilization. Third, decision-making satisfaction and treatment decision usefulness were each measured using only two items, which may limit internal consistency and content validity. Although reliability within our sample was acceptable, future studies should consider adopting more comprehensive instruments to better capture these constructs. Finally, this study was based on the participants' subjective interpretation of each questionnaire item, and social desirability bias may have led to certain answers being under- or overestimated.

4.5
Conclusions and suggestions for future research
In conclusion, our findings indicate that patients' decision-making satisfaction and perceived treatment decision usefulness can be directly improved through better SDM. Moreover, treatment decision usefulness played a mediating role in the relationship between SDM and HRQoL, which could exert a positive effect of SDM on HRQoL in cancer survivors. By offering initial evidence of the potential positive impacts of SDM on affective-cognitive, behavioral, and health outcomes, our study contributes to understanding the link between SDM and patient-reported outcomes in TCM consultations. These results should be interpreted with caution given the limitations related to the study setting and population, but may nonetheless inform future intervention studies in developing appropriate and effective programs of clinician–patient decision-making interactions. Future studies may need to employ experimental or longitudinal designs to ascertain whether patient-reported outcomes alter before and after SDM implementation, thereby determining causation between variables and producing accurate results.

Author contributions
Conceptualization: Fuming Li, Yue Teng, Yan Wei. Methodology: Fuming Li, Shimeng Liu, Liu Liu, Yan Wei. Software: Fuming Li, Liu Liu, Yan Wei. Validation: Yue Teng, Liu Liu. Formal analysis: Fuming Li. Investigation: Yue Teng, Liu Liu, Juntao Yan. Resources: Yue Teng, Yan Wei, Yingyao Chen. Data curation: Fuming Li, Liu Liu, Juntao Yan. Writing – Original Draft: Fuming Li. Writing – Review & Editing: Fuming Li, Shimeng Liu, Liu Liu, Yan Wei, Yingyao Chen. Visualization: Fuming Li. Supervision: Shimeng Liu, Yan Wei, Yingyao Chen. Project administration: Shimeng Liu, Yan Wei, Yingyao Chen. Funding acquisition: Yan Wei.

Declaration of competing interest
Shimeng Liu is an editorial board member of this journal but editorial board member status had no bearing on editorial consideration. The authors have no other conflict of interest to declare.

Funding
This research was supported by the Humanities and Social Sciences Youth Foundation, Ministry of Education of the People's Republic of China (NO. 18YJCZH187).

Ethical statement
This research was reviewed and approved by the institutional review board of School of Public Health, Fudan University (registration number IRB00002408&FWA00002399). Informed consent was obtained from all participants.

Data availability
The data that support the findings of this study are available within the article and its supplementary material.