Textbox 1 describes 2 disease briefs developed and reviewed by 3 clinical oncologists. Disease briefs included a mock cancer diagnosis and treatment options (delivery, side effects, and efficacy).

Five students from the University of North Carolina School of Medicine were recruited and trained to portray oncologists in the simulated patient-clinician encounters. A member of the research team served as a standardized patient during training. The training process included 2 hours of simulation script review with our research team, 1 hour of observation of an oncologist conducting both VB and NVB simulated encounters, and 6 hours of in-person practice, during which students practiced both types of simulations. To ensure consistency throughout the study, students will be debriefed after each simulated encounter to reinforce standardized engagement with participants.

We developed 2 simulation scripts (values-based and nonvalues-based), with guidance from a simulation expert and 3 clinical oncologists.

A review of literature was performed to assess the best practices in communication and SDM during clinical encounters [21,32-37]. The resulting VB script was designed to model a “high-quality” encounter in which clinicians focus on discussing and incorporating patients’ values into treatment decisions. The quality of the encounter was structured using three levels of situational awareness [38], modified for clinician and patient engagement.

This script was developed to ensure the oncologist addressed all 3 levels during the encounter, supporting a values-aligned decision-making process.

The NVB simulation script was designed to model a “lower quality” encounter in which the oncologist focuses on treatment decisions without explicitly addressing the patient’s values. In these encounters, any shift toward values-based discussion must be initiated by the patient, referred to as “patient agency.”

For example, if a participant mentions caring for their grandkids and expresses concern about how treatment may affect that role, the clinician is prompted to transition to level 2 of the VB framework (processing information about how the patient’s values are impacted). This may also prompt the clinician to go to level 1 (gathering information about what the patient values most). After addressing the patient’s concern, the clinician returns to the NVB script to inquire if the patient is then ready to make a treatment decision. During NVB encounters, clinicians are instructed not to advance through the values-based levels unless explicitly prompted by the patient. This structure ensures that values-based engagement in NVB encounters occurs only in response to patient-initiated cues.

See Figure 2 for the simulation script schema. This schema was used to support both the development of simulation scripts and medical student training for conducting VB and NVB encounters.

We developed a novel rating scale called the Values Elicitation and Clarification of Treatment Options Rating Scale (VECTORS). VECTORS is a 7-item rating scale that includes 3 domains: rapport building (1 item), clinician engagement (3 items), and patient engagement (3 items). VECTORS uses a Likert-type scale with 3 response options representing a 10-point scale (0=no effort observed, 5=moderate effort was observed, and 10=exemplary effort was observed). Higher scores reflect higher quality of values-clarification processes.

Table 1 displays the rating score criteria for each of the 7 items in VECTORS. VECTORS was established following the same situational awareness framework and review of literature as described under the simulation script development to rate the quality of values-clarification processes that occurred during patient-clinician encounters.

Characteristic information will include participants’ age, sex, race, ethnicity, cancer diagnosis, education level, household income, employment status, comfort level with technology, self-reflection and insight, and preferred role in problem-solving and decision-making (PSDM).

Self-reflection and insight will be assessed subjectively using the validated Self-Reflection and Insight Scale (SRIS) [39]. The SRIS is a 12-item assessment that measures participants' (1) engagement in self-reflection, (2) need for self-reflection, and (3) insight into self-reflection using a 5-point Likert-type scale. Higher scores indicate higher self-reflection and insight. The SRIS is an appropriate measure to assess participants' perceptions on whether the information in either the VOICE report or the ACS guide was useful for treatment decision-making.

To determine participants’ preferred role in PSDM, mean scores will be calculated for both problem-solving (PS) and decision-making measures for each participant, then placed into 1 of 3 classifications: hand over (mean score <3), share (mean score between 3 and 3.99), or keep (mean score ≥ 4) [40]. Participants who “hand over” both PS and decision-making are categorized as “passive.” Patients categorized as “shared,” either “hand over” or “share” PS, but “share” or “keep” decision-making. Four subcategories exist for patients categorized as “shared”: (1) leaning passive, (2) shared equally, (3) leaning autonomous, and (4) divide and share. Patients categorized as “autonomous” want to retain control of both PS and decision-making (ie, PS equals “keep,” and decision-making equals “share” or “keep”). Two subcategories exist for patients categorized as “autonomous”: (1) leaning shared and (2) consumerist. It is considered theoretically implausible for an individual to assume control for PS, but not for decision-making. Categorization of PSMD roles is displayed in Table 2.

The primary outcome is the quality of values-clarification processes during simulated treatment decision encounters for advanced cancer, assessed using VECTORS to compare intervention (VOICE) and control (ACS) groups, as well as VB and NVB encounters.

Two independent raters will review the recorded simulations and use VECTORS to estimate the quality of values-clarification observed during each encounter. Rater 1 will assist with recordings; therefore, it will not be blinded. Rater 2 will be blinded, as the recorded simulations will not indicate whether simulations were VB or NVB, or whether participants were in the intervention or control groups. We will calculate the Cohen κ coefficient to evaluate interrater reliability. We will conduct a sensitivity analysis by performing data analysis on the dataset that includes ratings from rater 1, then perform the same data analysis using the dataset that includes ratings from rater 2. We will compare the data analysis results to look for differences.

Data from secondary measures will either be collected through observed ratings by study team members (VECTORS and OPTION-5), participant surveys (CollaboRATE and PrepDM), or interviews, as described below and depicted in Figure 1.

SDM communication behaviors will be subjectively assessed using the OPTION-5 scale [36,37]. OPTION-5 is a validated scale used to assess the extent to which SDM communication behaviors occur during an initial diagnosis encounter. OPTION-5 is a 5-item validated survey that has been used to measure the extent to which clinicians involve patients in treatment decisions, including oncology patients, and is therefore an appropriate assessment for this study [37,41-44]. Five aspects of the clinicians’ behavior during the simulated encounters will be evaluated and rated using a 5-point Likert scale, ranging from “0=no effort observed” to “4=exemplary effort is observed” [37]. The five aspects include the extent to which the clinician (1) draws attention to or confirms alternate treatment options exist, (2) supports the patient to become informed or deliberate about treatment options, (3) gives information or checks understanding about the options available, (4) elicits the patients’ preferences about the options available, and (5) makes an effort to integrate the patients' preferences into the treatment decision [36,37]. Two independent raters will use the OPTION-5 scale to assess the level of SDM observed during each simulated patient-clinician encounter, as described for VECTORS. Interrater reliability and sensitivity analyses will be performed, as described for VECTORS.

The level of SDM that occurs in the simulated encounters from the patient’s perspective will be subjectively assessed immediately following simulations using the 3-item validated CollaboRATE scale [35]. CollaboRATE is a patient-reported assessment that measures patients’ perceived effort of clinicians to (1) understand patients’ health issues, (2) listen to what matters most about their health issues, and (3) include what matters most when choosing a treatment [35].

Usefulness in preparing participants to engage in SDM will be measured immediately following the final simulation using the PrepDM scale. PrepDM scale [45], a validated 10-item scale, will be used to assess participants’ perceived usefulness of the tool in preparing them to communicate with oncologists during the simulated encounter.

Performance actions of participants will be assessed after encounters, including (1) whether participants brought either the tailored digital summary report from VOICE or the ACS guide to the encounter and (2) whether participants informed the clinician they completed and reviewed VOICE or the ACS guide.

All participants will engage in 2 semistructured interviews immediately following each simulation to assess (1) situational awareness (ie, ability to gather, process, and project information) [38] during the simulated encounter and (2) perceptions of whether VOICE or the ACS guide prepared them for the encounter. Qualitative data from postsimulation interviews will be summarized and grouped into themes, which will be used to understand the impacts of using VOICE or the ACS guide prior to an encounter.

This pilot study was not powered, as its primary objective was to gather preliminary data necessary to inform the design and power calculations of a future, larger-scale study. We aimed to estimate the quality of values-clarification processes during simulated encounters (as measured by VECTORS, ranging from 0 to 100) for intervention and control groups in 2 different simulated encounters (VB and NVB). We will attempt to enroll 11 participants in each pathway to gather preliminary data, for a total of 44 participants.

The preliminary data gathered from this pilot study, as well as other study outcome measures, will inform future work using VECTORS in real-world and simulated encounters.

Descriptive statistics will be used to summarize participants’ demographic characteristics, comfort level with technology, personality (BFI-2-XS [Big Five Inventory-2 Extra-Short Form]) [46], self-reflection and insight (SRIS) [39,47], and preferred role in PSDM [40].

Simulated encounters will be quantified using VECTORS to assess the estimated quality of values-clarification processes that occur between intervention and control groups during both VB and NVB encounters. Interrater reliability for VECTORS will be assessed by calculating the Cohen κ weighted coefficient and discrepancies in ratings. Kappa statistics between 0.61 and 0.80 indicate substantial agreement between raters, and 0.81 to 1.00 indicate almost perfect agreement between raters [48]. Analyses of the 3 domains in VECTORS include the raw scores (rapport building: 0-10; clinician engagement: 0-30; patient engagement: 0-30). Analyses of VECTORS overall scores involve taking the sum of all 3 domains, producing a raw score of 0-70, then rescaling to a score of 0-100.

Descriptive statistics will be used to report the mean and SD for VECTORS overall and by each domain. Mean and SD will be reported as cross-tabulations for all encounters, and by intervention and control for VB and NVB encounters, and for whether clinicians transitioned from NVB to VB encounters due to patient agency (ie, patients initiated a discussion pertaining to their values).

Linear regression models using generalized estimating equation methods will be fit using all patient data from both encounter types (VB vs NVB) to account for repeated measures per patient. These models will objectively estimate differences by group and by encounter type and test for potential interactions to show if the intervention works better in one scenario compared to the other.

Simulated encounters will be subjectively analyzed using the OPTION-5 instrument to assess the extent to which SDM communication behaviors occurred during the simulation. The OPTION-5 scores from each simulation will be summed, producing a score from 0 to 20, then rescaled to a scale of 0-100. Interrater reliability for OPTION-5 will be assessed as described in the primary outcome for VECTORS.

Patient self-reported scores from the CollaboRATE assessment will be summed, then rescaled to a scale of 0-100. CollaboRATE scores will also be dichotomized (perfect score vs nonperfect score) to allow the use of logistic regression. CollaboRATE often shows ceiling effects; therefore, dichotomizing responses improves the variation among participant scores [49,50].

The PrepDM total mean scores will be quantified by averaging 10 subscales. Items will be summed, then divided by 10 to provide a mean score. Scores will be rescaled to a 0-100 scale by subtracting 1 from the summed score and multiplying by 25, with higher scores representing higher perceived levels of preparation in decision-making [45].

Descriptive statistics will be used to report the mean and SD by intervention and control groups for secondary outcome measures: OPTION-5, CollaboRATE, and PrepDM. Mean and SD will also be reported by VB and NVB encounters for OPTION-5 and CollaboRATE, and by whether participants brought the tool (VOICE or ACS guide) for PrepDM scores.

Descriptive statistics will also be used to summarize the number of participants who (1) brought either the tailored digital summary report from VOICE or the ACS guide to the encounter or (2) informed the clinician they completed and reviewed VOICE or the ACS guide.

Kruskal-Wallis tests will be performed to objectively evaluate differences in distributions in OPTION-5, CollaboRATE, and PrepDM scores.

Linear regression models will be performed for OPTION-5, as described in the primary measure for VECTORS. Log-binomial regression models will be performed with all binary patient data from both types (VB vs NVB) to account for repeated measures per patient. These models will objectively estimate differences by group and by encounter type and test for potential interactions to show if the intervention works better in one scenario compared to the other.

Qualitative data from the simulated patient-clinician encounters and postsimulation interviews will be analyzed by 3 members of our research team. The blinded rater responsible for scoring VECTORS and OPTION-5 will not be involved in coding interviews to maintain independence between assessments. The qualitative analysis will follow a hybrid approach with 3 phases [51]. In phase 1, data from interviews will be categorized by a priori themes from the interview guide. In phase 2, a posteriori codes will be created and summaries from participant responses developed, looking for meanings and patterns in the interviews. Discrepancies will be collectively addressed to ensure consensus on codes. In phase 3, the a priori and a posteriori codes will be combined into family codes to structure the findings from the interviews [51]. All codes will undergo a comprehensive review and will be grouped into themes through discussion of the research team.