Rintasyöpäpotilaiden tunnehallinta suurten kielimallien avulla (EMBRACE)

This intervention is a digital precision emotional management system based on a large language model, classified as a behavioral intervention. The system automatically analyzes text data from social media and in-app platforms to detect six basic emotions (joy, anger, sadness, fear, disgust, and surprise) in near real-time, and generates personalized patient labels by integrating patients' physical symptoms. Leveraging an evidence-based knowledge base, the system delivers tailored strategies for emotion regulation and symptom management, including emotion management techniques, symptom relief recommendations, and health education content. Patients in the intervention group autonomously use this system for 6 months in addition to receiving usual care, and complete biweekly skill translation assessments to facilitate the application of intervention skills in daily life.

This intervention is a digital precision emotional management system based on a large language model, classified as a behavioral intervention. The system automatically analyzes text data from social media and in-app platforms to detect six basic emotions (joy, anger, sadness, fear, disgust, and surprise) in near real-time, and generates personalized patient labels by integrating patients' physical symptoms. Leveraging an evidence-based knowledge base, the system delivers tailored strategies for emotion regulation and symptom management, including emotion management techniques, symptom relief recommendations, and health education content. Patients in the intervention group autonomously use this system for 6 months in addition to receiving usual care, and complete biweekly skill translation assessments to facilitate the application of intervention skills in daily life.

Participants allocated to the control group receive usual emotional and symptom management for breast cancer, with no exposure to the digital intervention system. Standard care includes routine follow-up examinations, symptom management, psychological support, health education, and guidance on wound care, drainage tube management, diet, rest, and physical activity. All care is delivered in accordance with established clinical nursing procedures at the treating hospital.

During follow-up, participants experiencing severe psychological distress (e.g. suicidal ideation) are referred to psychiatric services in line with standard clinical practice. After completion of the trial, control-group participants are offered the option to access the intervention system voluntarily.

Participants allocated to the control group receive usual emotional and symptom management for breast cancer, with no exposure to the digital intervention system. Standard care includes routine follow-up examinations, symptom management, psychological support, health education, and guidance on wound care, drainage tube management, diet, rest, and physical activity. All care is delivered in accordance with established clinical nursing procedures at the treating hospital.

During follow-up, participants experiencing severe psychological distress (e.g. suicidal ideation) are referred to psychiatric services in line with standard clinical practice. After completion of the trial, control-group participants are offered the option to access the intervention system voluntarily.

In this study, the EORTC QLQ-C30 Summary Score is used as the primary analysis endpoint, and assessments are conducted using the Chinese version of the EORTC QLQ-C30 questionnaire.

The EORTC QLQ-C30 Summary Score is calculated using the "13-domain mean method." Raw scores are first obtained from five functional scales and eight symptom scales, with all scores transformed to a 0-100 scale. All symptom scale scores are then reverse-scored, as higher original symptom scores indicate greater symptom burden; after transformation, higher scores uniformly indicate better outcomes, corresponding to milder symptoms. The arithmetic mean of the transformed scores across the 13 scales is subsequently calculated to generate the Summary Score, which remains on a 0-100 scale, with higher scores indicating better overall health-related quality of life.

The EORTC QLQ-C30 questionnaire comprises 15 dimensions, including five functional domains (physical, role, emotional, cognitive, and social functioning), three symptom domains (fatigue, pain, and nausea/vomiting), one Global Health Status/Quality of Life domain, and six single-item symptom scales (e.g., dyspnea, insomnia). Scoring is performed by first calculating the mean of the raw scores for each domain or item, which is then linearly transformed into a standardized 0-100 score using the formula (mean raw score-1)/range of response options×100. For the functional domains and the Global Health Status/Quality of Life domain, higher scores indicate better functioning or quality of life (positive scoring), whereas for the symptom domains, higher scores indicate greater symptom severity and poorer health status (negative scoring).

The Generalized Anxiety Disorder Scale (GAD-7) was used, consisting of 7 items, to assess the degree to which patients were troubled by anxiety-related symptoms such as tension, worry, and difficulty in relaxation over the past 2 weeks; the total score of this scale ranges from 0 to 21, with higher scores indicating more significant emotional symptoms.

The Patient Health Questionnaire-9 (PHQ-9) was used, which consists of 9 items and has a total score ranging from 0 to 27. The higher the score, the more significant the emotional symptoms are.

Each symptom score is calculated as the mean of frequency, severity, and distress ratings (range 0-4), with absent symptoms scored as zero. Higher MSAS scores indicate greater symptom burden.

The unplanned hospitalization rate is defined as the ratio of the total number of unplanned hospitalization events to the total number of person-months in the study cohort, and is expressed as the number of events per person-month. Unplanned hospitalization is defined as admission via emergency or urgent care. Planned hospitalizations for the purposes of chemotherapy or radiotherapy (ICD-9 codes: V58.0, V58.1, V58.11, V58.12) and rehabilitation (ICD-9 codes: V57.xx) are excluded. When the admission type is unknown, hospitalizations with emergency department charges greater than zero are classified as unplanned hospitalizations; otherwise, they are categorized as other hospitalizations.

Time from the date of randomization to the occurrence of the first unplanned hospitalization event.