Trauma narratives and healing of offspring caregivers of breast cancer patients: a mixed qualitative methods study Authors: Yifan Zhao, Xueli Qiu, Ziang Xiao, Michelle R. Kaufman Published: 2025 ABSTRACT BackgroundCaregivers’ health is a crucial public health concern. Although breast cancer’s psychosocial burden extends to offspring caregivers, their trauma narratives and seeking of social support remain understudied, particularly in China. This study aimed to explore how adult offspring caregivers share trauma stories in online support groups (OSGs) and identify how narrative practices may facilitate self-healing.MethodsWe employed a mixed qualitative study approach – using large language models and thematic analysis to examine offspring caregivers’ sentiment patterns and narrative characteristics at different treatment stages for the parent for whom they were caring. The sample comprised 705 posts and comments from offspring of breast cancer patients, collected from the Douban OSG “Having a Breast Cancer Patient in My Family.”ResultsAt the early stage of treatment, most offspring caregivers of breast cancer patients exhibited negative sentiments. A combination of positive, negative, and neutral sentiments was observed in the middle stage. In the late stage, positive sentiments predominated. Furthermore, we identified three narrative themes aligning with sentiment trajectories among offspring caregivers: collapse, interactive, and constructive narratives. These findings suggest the self-healing function of media-based narratives. We argue that the deeper exploration of online trauma narratives related to breast cancer is necessary for more effectively conceptualizing the role of OSGs in providing health-related support.ConclusionsTrauma narration of offspring caregivers, particularly in online spaces, is an underexplored, but potentially crucial area. Our study revealed the role of the OSG impact on individual emotional and behavioral experiences that could be integrated into interventions to enhance psychological treatments.Supplementary InformationThe online version contains supplementary material available at 10.1186/s40359-025-03854-y. Background Breast cancer is the most common malignant tumor among women worldwide, with the highest incidence and mortality rates, particularly in women living in transitioning countries [1]. In China, breast cancer ranks first among malignant tumors in women, and it is expected to affect 1.02 million women in the next five years [2]. A breast cancer diagnosis is a potentially traumatic event [3]. Clinically, trauma is defined by the Diagnostic and Statistical Manual of Mental Disorders, fifth edition, text revision (DSM-5-TR), as exposure to actual or threatened death, serious injury, or sexual violence through direct experience, witnessing, learning it occurred to a close other, or repeated/extreme exposure to aversive details [4]. A life-threatening diagnosis such as breast cancer can be considered a traumatic stressor, as it directly exposes individuals to the threat of serious illness and potential mortality. This experience can impair or obstruct resilience and healing capacities [3]. The adjustment following the diagnosis can lead to difficulties such as anxiety, depression, guilt, isolation, worthlessness, or even post-traumatic stress disorder (PTSD) [5–7]. PTSD, which sometimes arises from such trauma, is characterized by a re-experience of traumatic events (invasive memories, flashbacks, or nightmares); avoidance of thinking about and recalling traumatic events, or avoidance of activities, situations, or characters that remind one of traumatic events; and continuing to feel that they are currently facing a serious threat [8]. The traumatic psychological impact can persist for years, even after remission [9, 10]. This significant psychological distress and burden can extend to caregivers, particularly offspring caregivers of breast cancer patients, as they witness a loved one’s suffering and endure substantial lifestyle and psychosocial disruptions [11, 12]. Sometimes the burden caregivers bear even exceeds that experienced by patients themselves [13–16], leading to a decline in their physical health [17, 18]. Salient factors include early unknowns surrounding critical illness, the complexity of healthcare team processes, care decision making, financial concerns, disruptions to social life, and an overwhelming care role [19–21]. In collectivist cultures, societal pressures to suppress personal grief while performing filial duties further compound these challenges [22]. However, studies of trauma narratives and healing have focused mostly on patients, with little attention being given to offspring caregivers particularly in the Chinese context. The nature of the kinship relationship between the caregiver and care recipient leads to a unique perspective on the trauma of caring for a breast cancer patient. First, “Xiao” (filial piety) is a traditional Chinese family moral code emphasizing the children’s responsibilities to care for their parents. It has long been promoted as an ethical component of elder care [23–25]. Unlike spouse caregivers and other family members, offspring caregivers often assume a more prominent role and must balance caregiving with other social responsibilities. Second, between 1980 and 2015, China implemented the “one-child policy,” which means that many offspring caregivers may be the only child and lack support from siblings to take care of their parents. Third, in China, there are currently insufficient community care resources, and related services are still not in place [26]. Although studies have showed that face-to-face family visits, guidance, and group support for family caregivers can help to reduce their care burden and improve their negative emotions and care skills [27–29], due to a lack of respite services, caregivers hardly have the chance to go out to participate in education, training, and other in-person support activities [30]. Therefore, it is difficult for family caregivers to obtain traditional carer support, and this generation may face greater challenges when providing care for their elderly mothers with breast cancer, often requiring additional social support [31]. Online social support groups (OSGs) provide a distance-delivered and easily accessible opportunity to relieve the caregiving burden [32–34]. By 2024, China’s internet penetration rate had reached 78.0%, amounting to 1.1 billion users [35]. This extensive connectivity has facilitated the proliferation of OSGs, particularly on platforms like Douban groups and dedicated healthcare forums. Caregivers are increasingly using OSGs to transmit and obtain information, provide and receive emotional support, socialize and form interpersonal relationships, and experience comradeship with others sharing similar stressors, thereby helping to reduce their perceived anomaly [36, 37]. Previous studies have indicated that OSGs may even offer opportunities for deeper connections and more substantial support compared to face-to-face interactions [38]. This is primarily because anonymity allows for more intimate sharing, asynchronicity enables thoughtful emotional processing without real-time pressure, and minimized social status promotes sharing ideas and resources among peers [39, 40]. Such platforms are vital for sharing trauma narratives. Sharing trauma stories online facilitates understanding and integration of thoughts and feelings, providing meaning to distressing events. This process allows such experiences to fade from conscious thought and enables individuals to move forward gradually [41]. Furthermore, connections established between the storyteller and the listener appear to provide self-therapy [42] and help individuals reaffirm their identities amidst the uncertainties and crises induced by the disease. These connections also foster peer support and social connections, potentially aiding others in similar situations [43, 44]. Notably, this process aligns closely with narrative therapy, a concept proposed by Michael White and David Epston: externalizing problems, fostering self-awareness, and reconstructing narratives to achieve self-healing [45]. OSGs’ decentralized nature—allowing anyone to act as a storyteller or listener—aligns with narrative therapy’s de-expertization and listening-based ethos, while features such as likes and comments, alongside emotional contagion, enhance the externalization of suffering [46]. Consequently, narrative therapy remains a valuable framework for analyzing trauma narratives within online spaces. Despite their potential, the role of OSGs in mediating trauma narratives among offspring caregivers in Chinese contexts remains underexplored [47]. This study aims to investigate the trauma narratives and social support needs of offspring caregivers of breast cancer patients during different treatment stages in China. Sentiment analysis (SA) provides an opportunity to understand caregivers’ needs and the emotional tone within a text [48]. With the advent of large language models (LLMs), there is great potential for their employment on sentiment analysis problems [49]. The considerable scale of data used for LLM training can substantially enhance performance across sentiment analysis tasks, particularly in sentiment classification accuracy [50]. To our knowledge, no prior studies have used LLMs to examine the sentiments of breast cancer offspring caregivers in OSGs. A better understanding of OSGs’ use among offspring caregivers may facilitate future development of OSGs-based intervention programs. Such programs are especially needed in developing countries like China where community mental health resources are limited [51]. Therefore, to address these objectives, the study, grounded in narrative therapy theory, draws on existing literature as well as first-hand materials. It proposes two research questions:RQ1: What are the sentiments expressed in OSGs of the offspring caregivers of breast cancer patients at different stages of the patient’s treatment?RQ2: Is there a dominant narrative surrounding trauma among offspring caregivers within the OSG? What is the trajectory of traumatic narrative changes? RQ1: What are the sentiments expressed in OSGs of the offspring caregivers of breast cancer patients at different stages of the patient’s treatment? RQ1: What are the sentiments expressed in OSGs of the offspring caregivers of breast cancer patients at different stages of the patient’s treatment? RQ2: Is there a dominant narrative surrounding trauma among offspring caregivers within the OSG? What is the trajectory of traumatic narrative changes? RQ2: Is there a dominant narrative surrounding trauma among offspring caregivers within the OSG? What is the trajectory of traumatic narrative changes? Methods Study design To gain an in-depth understanding of the topic, we used a mixed qualitative methods approach, combining large language models (LLMs) and thematic analysis, to analyze the same dataset for cross-validation. First, we used the LLM to identify the sentiment patterns among offspring caregivers of breast cancer patients at different treatment stages. Second, we explored narrative themes related to the different treatment stages of these caregivers, to provide a deeper understanding and interpretation of the patterns that LLMs identified. After independently analyzing the two components, we integrated the results for interpretation. This analysis aimed to uncover the changing trajectory of dominant narrative types and sentiments during different stages of treatment. Study site: Douban.com In identifying a web source for analyzing breast cancer trauma narratives among offspring, the Chinese website Douban (www.douban.com) emerged as an ideal venue due to its abundance of user-generated cancer-related stories. Founded in 2005, Douban is the largest social and knowledge-sharing community website in China. As a user-generated content (UGC) platform, it offers a multifaceted community network that integrates services such as blogs, dating, groups, and collections [52]. Among these features, Douban groups serve as gathering places and hubs for topical discussions, allowing users to create communities, engage with peers, and connect with others who share similar interests. The group “Having a Breast Cancer Patient in My Family” is not the only OSG we found specifically dedicated to offspring caregivers sharing their own trauma experiences, but it is one of the most active communities of its kind. Created on October 4, 2010, this community has grown substantially and remains highly active, with 3,867 registered members and 984 posts as of August 2024. This high level of activity, combined with its unique focus, makes it an ideal source for examining authentic trauma narratives within this demographic. Data collection The data collection process was conducted in two sequential stages. In the first stage, data extraction (April 1–15, 2024), we used Python-based web crawlers to extract all publicly available posts, comments, publication dates, and user names from the Douban group “Having a Breast Cancer Patient in My Family.” This extraction comprehensively covered content published from the group’s establishment on December 9, 2010 through March 31, 2024. The result was an initial dataset comprising 984 posts and 5,310 comments from 314 distinct users. To protect user privacy, all collected data were anonymized immediately upon acquisition. In the second stage, data screening (April 16–30, 2024), we removed duplicate entries and those with insufficient word count, with each post and comment treated as an independent data unit. The inclusion criteria included: (1) the post or comment was published or responded to by offspring caregivers of breast cancer patients in either text or picture format; (2) the content was a trauma narrative, including monologues or interactive narratives, such as updates on the mother’s illness, requests for support, emotional expressions, and related experiences; (3) the narrative included information about the offspring caregiver, including their user ID and records of their mother’s treatment stage. Advertisements, commercial promotions, respondent recruitment, and posts and comments from deregistered users were excluded from data analysis. After this screening process, 705 posts and comments were retained for final analysis. Data analysis First, we combined Mattingly’s [53] framework of the illness trajectory and categorized three stages of breast cancer treatment based on disease progression (Fig. 1): the early stage refers to the period when a patient is newly diagnosed and begins planning their treatment; the middle stage refers to when the patient is undergoing treatments; and the late stage refers to when the patient experiences one of these outcomes: recovery, death, or recurrence. The first author classified posts and comments into the three treatment stages, with the second author cross-checking the classification. When there was disagreement, the two authors discussed which stage the data should be classified to reach a consensus. To analyze narrative and sentiment evolution at the individual level, we identified a subsample of users whose contributions spanned the entire care continuum. From the total sample of 314 users, we selected those who had authored at least one post classified into the early stage, one into the middle stage, and one into the late stage. This objective criterion identified 19 users (6.05% of the active user sample) who formed our longitudinal cohort for deeper qualitative analysis of individual trajectories. After the classification of treatment stages was finalized, the data were used for both sentiment analysis through LLMs and thematic analysis. Fig. 1Process of breast cancer treatments as reported by offspring caregivers posting on Douban Process of breast cancer treatments as reported by offspring caregivers posting on Douban Large language models-based sentiment analysis To assess the sentiment patterns of different treatment stages, we utilized a novel approach that leverages LLMs to assist in qualitative coding [54]. Unlike traditional machine learning (ML) based classifiers, the LLM-based approach does not require large-size training data for specialized tasks. By training on a vast corpus of natural language, LLMs can capture nuances and subtleties in text, which makes them suitable for our analysis. For this study, we integrated gpt-4o-2024-05-13 from OpenAI in API format, employing a structured codebook approach with few-shot examples (see Additional File 1). This codebook defined three broad sentiment categories: positive, neutral, and negative. The model was guided by a system prompt instructing it to (1) review the text and its context, (2) consult the structured codebook definitions, and (3) assign a single sentiment code (positive, neutral, or negative) from a predefined code set, ensuring consistency and systematicity. We set the temperature parameter to 0 for deterministic outputs, with all other parameters at their default values. The coding framework followed the [Code/Description/Examples] format. For instance, Code: POSITIVE; Description was defined with the description that positive sentiments encompass feelings such as happiness, joy, and love, as well as optimism, hope, and energy (e.g., certainty, pride, victory) [55]. Examples: “We must have confidence. It’s been two years since my mother was diagnosed with breast cancer. She is now leading a normal life. She can eat, sleep, work, and play mahjong.” To validate the consistency and accuracy of the LLM’s sentiment classification, we assessed agreement. First, to ensure a shared understanding of the coding framework between the human coders and to calibrate the LLM, a preliminary calibration exercise was conducted. The first author and a second author independently practiced coding on 10 sample narratives. They then discussed their coding decisions to resolve any discrepancies and establish a consensus on the application of the codebook. Meanwhile, the LLM was configured for the task using the same codebook and few-shot examples that would be employed in the formal analysis. This dual process ensured that both the human coders and the LLM were aligned with the coding criteria before proceeding. Next, for formal validation, we selected a stratified random sample of 71 posts for human validation, which included 24 from the early stage, 35 from the middle stage, and 12 from the late stage. The first author performed blind coding without knowledge of the LLM’s results. Afterward, we calculated the inter-rater reliability (IRR). Cohen’s κ based on a cross-tabulation (Table1) of the classifications from the LLM and the human coder. The observed agreement between the LLM and human coder was high, κ ≈ 0.891 [56]. The sentiment base rates within the validation sample were positive 35.2% (25/71), neutral 18.3% (13/71), and negative 39.4% (28/71). This distribution closely mirrored the sentiment distribution observed in the overall sample. Table 1Cross-tabulation of sentiment codes assigned by the GPT-4o model and human codersHuman: positiveHuman: neutralHuman: negativeIn totalLLM: positive252027LLM: neutral013114LLM: negative022830In total25172971 Table 1Cross-tabulation of sentiment codes assigned by the GPT-4o model and human codersHuman: positiveHuman: neutralHuman: negativeIn totalLLM: positive252027LLM: neutral013114LLM: negative022830In total25172971 Cross-tabulation of sentiment codes assigned by the GPT-4o model and human coders Thematic analysis We employed inductive and deductive thematic analysis to explore trauma narratives among offspring whose mothers were undergoing various cancer treatments. The first and second authors imported 705 Douban posts and comments into ATLAS.ti and independently reviewed all entries, making notes to identify initial patterns and potential trends [57]. The two authors continuously refined the codes during the open coding process, and any disagreements were resolved through discussion and consensus. For each member of the offspring data group, we recorded information such as user IDs, patients’ treatment stages, and their motivations for posting to the Douban group. Douban prohibits minors from registering accounts, implying that the offspring caregivers were adults. To protect the confidentiality of group members, all usernames were replaced with pseudonyms, and direct quotations have been paraphrased to prevent the identification of or matching specific users with their original online posts. To ensure paraphrasing preserved the original intent, emotional tone, and meaning, we used a rigorous two-step process. First, the first author rephrased quotations by substituting identifying details (e.g., specific locations/timelines/personal references) with generic equivalents (e.g., “a local hospital”), maintaining precise semantic content, sentiment, intensity, and contextual intent. Second, the second author independently verified alignment of meaning and tone between original and paraphrased texts. Any discrepancies were resolved by consensus to retain substantive nuances. Results Among 705 posts, 226 posts (32.1%) were identified as posted during the early stage, 344 posts (49.2%) were during the middle stage, and 132 posts (18.7%) were during the late stage. During the early stage, most offspring caregivers exhibited negative sentiments. In the middle stage, a combination of positive, negative, and neutral emotions was observed. In the late stage, positive emotions predominated. LLM-based sentiment analysis result Analysis of all posts revealed the following stage distribution and sentiment patterns (Fig. 2). During the early stage of treatment, negative sentiment predominated, and positive and neutral sentiments were less common. Conversely, in the middle stage, positive sentiments slightly surpassed neutral emotions. In the late stage, positive sentiments were the most prevalent. Fig. 2The proportion of positive, neutral, and negative sentiments expressed in posts at different treatment stages The proportion of positive, neutral, and negative sentiments expressed in posts at different treatment stages In addition, the sentiment curve of the 19 active offspring caregivers showed that most of the offspring sentiment patterns were not single at various treatment stages (Fig. 3) and were consistent with the results of the previous LLM sentiment analysis. In the early stage of treatment, offspring sentiments were predominantly negative, accounting for 73.7% (n = 14). During the middle stage, the narratives of all 19 offspring began to diversify: positive sentiments represented 52.6% (n = 10), neutral emotions accounted for 21.1% (n = 4), and negative emotions comprised 26.3% (n = 5). By the late stage of treatment, positive emotions dominated at 73.7% (n = 14), while negative emotions accounted for 26.3% (n = 5). Fig. 3Types of sentiments and trauma narratives among offspring caregivers of breast cancer patients Types of sentiments and trauma narratives among offspring caregivers of breast cancer patients Thematic analysis results Collapse narrative at the early treatment stage During the early stage of treatment, offspring caregivers often expressed profound pessimism about their mothers’ conditions. A total of 64.2% (n = 145) used the term “mental collapse” to describe their inability to cope with the physical and emotional toll of their mother’s illness, which we called the collapse narrative. Phrases like “my heart was in a mess,” “the sky fell,” “a bolt from the blue,” and “the world was dark” vividly depicted their turmoil. When their mothers refused to cooperate with treatment, some caregivers described falling deeper into despair and frustration. One offspring caregiver of a patient who had just been diagnosed complained:“Since my mom got sick, she’s become more sensitive and stubborn. She can’t communicate well anymore, refuses to go along with medical tests, often has wild thoughts, and is unwilling to accept anything new. I feel so tired now. Is my mother the only one like this?” (@Kira). “Since my mom got sick, she’s become more sensitive and stubborn. She can’t communicate well anymore, refuses to go along with medical tests, often has wild thoughts, and is unwilling to accept anything new. I feel so tired now. Is my mother the only one like this?” (@Kira). Furthermore, some caregivers expressed distress over family decisions to conceal the diagnosis, further heightening feelings of betrayal and helplessness. For example, a caregiver of one-child family shared: “My mother hid her diagnosis from me to avoid impacting my studies. I never imagined such a crisis would hit our already financially strained family.” Additionally, due to a strong desire for their mothers’ recovery and limited knowledge of breast cancer, offspring caregivers often developed suspicions and distrust toward medical professionals. For instance, one caregiver questioned the rationale behind certain medical decisions, asking, “Why did the doctor inform me about the surgery before completing the pathology report?” and “Why is there only a brief pathology diagnosis without detailed information? I don’t understand.” Some caregivers even shared images of their mothers’ diagnostic reports in the OSG to seek professional opinions from others. It is noteworthy that metaphors were prevalent in trauma narratives shared by offspring caregivers. They frequently used war-related imagery to convey the urgency, difficulty, and complexity of battling cancer, portraying it as an aggressive external invader. One caregiver described the illness using metaphors of “obstacle” and “oppression”: “This invisible, bomb-like anxiety resembles a layer of hardened stone, with all other pleasures in life merely resting on its surface.” Interactive narrative at the middle treatment stage Interactive narratives refer to reciprocal disclosures that seek both informational and emotional support. During the middle treatment stage, offspring caregivers gradually accepted their mothers’ breast cancer diagnoses and actively sought support in the OSG. Caregivers’ posts focused on exchanging information, seeking advice, and sharing experiences. The themes of their posts included treatment plans, postoperative diets, medical insurance, clinical trials, and prosthetic breast options. When caregivers articulate their confusion, they often receive supportive responses from other members. For instance, a caregiver expressed concern about her mother taking Herceptin1 before receiving targeted therapy. One member responded: “My mother did not experience any symptoms when using Herceptin alone. She was generally in good health, although she did have some facial spots that faded after treatment. Try not to worry too much.” The mentality of “try not to worry too much” was commonly shared within the group. Other caregivers expressed similar sentiments. “Our mother will be fine, " “We must give the patient the greatest confidence and support, " and “persistence in completing treatment is victory” repeatedly appeared in their posts. Additionally, some offspring caregivers journaled their mothers’ battles with cancer in the OSG in what they termed anti-cancer diaries. Such diaries, which chronicled the experience of accompanying and supporting their mothers, evoked strong psychological resonance within the community, illustrating the profound impact of shared narratives. An offspring caregiver journaled her mother’s chemotherapy experience, writing:“She completed chemotherapy on January 19 and has not gotten out of bed for three days. Today, I looked through her old photos. She was so beautiful. How can someone who cared for her hair every two weeks accept hair loss and the need for wigs? My heart aches when I think about this.” (@Lucas). “She completed chemotherapy on January 19 and has not gotten out of bed for three days. Today, I looked through her old photos. She was so beautiful. How can someone who cared for her hair every two weeks accept hair loss and the need for wigs? My heart aches when I think about this.” (@Lucas). This diary elicited considerable empathy from group members. One caregiver responded, “All the sadness comes from love… Children who love their mothers all feel the same.” Another caregiver shared his experience of caring for her mother during surgery: “If I could offer advice based on my current perspective and experience, it would be to take better care of her, to help her process the emotional challenges following her illness, to show her your love, and to help her rebuild the confidence to fight the disease.” Through these sentiment interactions, caregivers in the virtual community seemed to gradually cultivate a sense of belonging, identity, attachment, and support. It is important to note that mutual support lies at the heart of the OSG. Writing about traumatic experiences provides personal meaning and benefits those who engage with these narratives. By sharing their stories, individuals seemed to seek emotional connection and social support, which strengthens communal bonds and fosters mutual recognition and understanding. This was the original intention of the group creator, who remarked:“This place feels like a family. Everyone needs to stay strong. I discovered my mother had breast cancer a few years ago, but since it was a benign tumor, I didn’t worry much after surgery. However, recent exams revealed the cancer has returned, necessitating chemotherapy. I began gathering information and realized that many others face similar challenges. Douban is known for its supportive atmosphere, but I hadn’t found a relevant group, so I created one. I hope this group will facilitate information sharing and mutual encouragement for our families. Please recommend it to others so that we can offer support to more people.” (@Francis). “This place feels like a family. Everyone needs to stay strong. I discovered my mother had breast cancer a few years ago, but since it was a benign tumor, I didn’t worry much after surgery. However, recent exams revealed the cancer has returned, necessitating chemotherapy. I began gathering information and realized that many others face similar challenges. Douban is known for its supportive atmosphere, but I hadn’t found a relevant group, so I created one. I hope this group will facilitate information sharing and mutual encouragement for our families. Please recommend it to others so that we can offer support to more people.” (@Francis). Constructive narrative at the late treatment stage Although breast cancer has the risk of spreading, metastasis, and recurrence in the late stage of treatment, most offspring caregivers posting in the OSG seemed to adopt a positive attitude and focus on their mothers restoring physical health. Many offspring whose mothers have not yet passed the five-year recurrence period actively engaged in prayers within the support group. Some of them reported going to temples to pray to gods and Buddhas for psychological comfort, hoping for their mother’s recovery and the well-being of other breast cancer patients. Some offspring caregivers did not leave the “Having breast cancer in my family” group after their mothers passed away and continued to leave messages under their posts to express their longing for their mothers. Other community members contributed by leaving supportive messages and offering encouragement, expressing their condolences and solidarity. “I wish that we would all get better and better, as we still have a long way to go in life!” As Lagerkvist puts it, “Death has made a mediated return to the public sphere” (p. 9) [58]. Online support groups, in particular, have permeated caregivers’ daily lives and appear to have become an important new space for mourning. Death was a topic exclusively discussed at the late stage. There were two contrasting perspectives on death: some offspring caregivers expressed a deep fear of the death that cancer may bring, while more caregivers faced the death that cancer may bring calmly. They gradually realized that “fate is sometimes unavoidable and cannot be changed. What is more important is to cherish the present, love yourself and your family, and love life.” One offspring reflected:“Life is inherently uncertain. We pursue many things, often forgetting what is most important to us along the way. This realization is an understanding of life and perhaps a necessary lesson. Let’s focus on improving ourselves and our relationships with the world and those around us in the years ahead.“(@Gina). “Life is inherently uncertain. We pursue many things, often forgetting what is most important to us along the way. This realization is an understanding of life and perhaps a necessary lesson. Let’s focus on improving ourselves and our relationships with the world and those around us in the years ahead.“(@Gina). Some caregivers did not perceive that death itself is not inherently tragic; rather they felt a sense of regret for not fulfilling their filial duties before their mothers passed away: “The regret is that I was ignorant and did not accompany her adequately at the beginning.” “I am so unfilial. The lives of the elderly are so lonely, I really should spend more time with them.” Even though some offspring caregivers have returned to normal life, they remain willing to stay in the virtual community to provide support to patients and their families, seeing themselves as valuable resources and important helpers for new members. Discussion We explored the evolving trajectory of trauma narratives among offspring caregivers of breast cancer patients through the use LLMs to conduct a sentiment analysis and thematic analysis, building on Shi et al.‘s work [31]. The results revealed that the trajectory of trauma narratives generally aligns with sentiment changes. At the early stage, caregivers’ sentiments are predominantly negative, transitioning to mixed sentiments in the middle stage, and ultimately becoming positive in the late stage. We correspond these stages to the collapse narrative (early stage), interactive narrative (middle stage), and constructive narrative (late stage). Notably, this correspondence is a broad trend rather than a strict mapping. For instance, while constructive narratives often co-occur with positive emotions, they may also incorporate negative emotional expressions—such as admitting enduring caregiving challenges or remembering the deceased. In addition, our findings support previous research indicating that for caregivers who actively dealt with their emotions and thoughts, active OSG participation was positively related to psychological well-being [59]. This may be due to the caregivers’ gradual acceptance of their mother’s cancer diagnosis, which leads to emotional stability and a greater willingness to seek support from OSGs. Cabling suggested that the higher the stage of a person’s cancer, the less likely the person is to post in OSG [60]. If they do post, the post tends to focus on personal side effects and the anxiety that goes with it. However, our study did not observe a direct relationship between treatment stage and OSG activity or post volume. As the cancer treatment stage progressed, offspring caregivers focused on seeking informational and emotional support on the OSG. Especially in one-child families, the issue of caring for mothers with breast cancer has become more urgent. Notably, even after the mother’s passing, some caregivers continued to digitally mourn and support others in the OSG. Consistent with previous research results, social media has a profound impact and potential utility for cancer survivors and caregivers [61–64]. Platforms such as OSGs, Douban, Weibo, and Xiaohongshu (Little Red Book) in China and Facebook and X (Twitter) in the U.S. and other Western societies may provide social support for individuals who are experiencing complex and challenging cancer experiences. Furthermore, our findings suggest that narratives and interactions within OSGs serve a therapy function and might be considered a possible supplement to more traditional professional treatment. This aligns with a systematic review [65], which concluded that blogging and vlogging facilitate recovery for people with mental illnesses by enabling self-expression and social connection. Specifically, such platforms help build support networks, reduce isolation, and foster empowerment through the sharing of personal narratives. While this survey suggests this is the good news, its scope remains quite broad. Building on these prior studies, our research specifically explores OSG trauma writing and its therapeutic potential. Our study found similar results to Tan [42], who conceptualized blogs as a form of self-therapy and self-help. We regarded that Douban has a similar healing function. By sharing traumatic experiences on social media platforms, offspring caregivers transform their internal emotions and psychological states into external, perceptible symbolic texts. These texts invite the support of community members, who act as invisible paraprofessional therapists, offering support through interactions such as likes, follows, and comments. Such engagement helps guide those posting to shift their perspectives on cancer, enabling them to become either active controllers or calm disengagers to achieve reconciliation with traumatic experiences. Our study of OSGs reveals similarities with Tan’s [42] findings on blogs regarding self-reflective writing interventions, including the online disinhibition effect and Bruns’ [66] described interactions consisting of comments, links, and subscriptions between readers. The primary difference, however, may lie in the authenticity of traumatic experiences and social needs, which are often more pronounced in OSGs. OSGs tend to attract individuals who have a strong desire to share, are inclined to write, and are deeply engaged in building and maintaining relationships with others, partly due to the platform’s immediacy and the rapid pace of content creation. Additionally, our study found that war metaphors in OSG narratives function as a potential self-healing tool for offspring caregivers. Although war metaphors in illness narratives (e.g., “cancer as a battle”) are often critically examined for their potential to increase stress, our findings reveal a crucial distinction. For offspring caregivers, adopting the “war” frame allows them to assume the active, protective role of a guardian building a “fortress” for their mother. This shift in agency—from passive victim to active defender—is key. It helps offspring caregivers who contend with traumatic reactions feel seen and heard, enabling them to reconstruct their traumatic experiences into a coherent narrative of resilience and purpose, thereby facilitating meaning-making and integration of the self [67]. Thus, even within binary oppositions similar to Ramšak’s “murder metaphors” [68], the function of the metaphor is transformed by the caregiver’s standpoint into an empowering one. Theoretical implications This study unravels a series of theoretical insights delineating the adaptation and inspiration of narrative therapy within the media landscape of China. First, our research reveals the pivotal role of the OSG in narrative therapy. We propose that the self-healing function of media-based trauma narratives not only bridges the gap between communication and medicine but also provides a fresh perspective on health communication studies. Second, our research highlights the potential digital technology offers for individuals to create social connections with others going through similar traumatic experiences to rebuild themselves. This approach encourages a deeper reflection on the current dynamics between digital technology, human psychology, and the relationship between media and public health. Limitations and directions for future research Our study contributes to current literature and understanding of offspring caregivers’ trauma narratives with some limitations. First, we cannot determine whether it was the OSG or the time factor that enabled offspring caregivers to achieve inner healing. Future research could employ in-depth interviews or cyber ethnography to further investigate the role of OSG in the healing process of offspring caregivers. Second, the analysis of users who posted across all treatment stages (n = 19) focused on a self-selected, highly engaged subgroup. While their sentiment trajectories aligned with the broader cohort’s patterns, their experiences may reflect a higher degree of involvement or motivation. Consequently, the generalizability of the longitudinal narrative themes to less active caregivers requires further investigation. Even so, the generalizability of qualitative data beyond those producing the data (in this case textual posts) cannot be guaranteed. Third, we did not thoroughly investigate how caregiver gender or the differences between recurrence and palliative care contexts might intersect with the narrative patterns we identified. This is because these factors were not consistently present in our dataset. Addressing these aspects is crucial for future research, especially considering that many patients are from the one-child policy generation, making the gender of the sole offspring caregiver an important factor warranting targeted exploration. Fourth, our focus on breast cancer caregivers’ online narratives may overlook offspring caregivers of other cancer types. Finally, we acknowledge that self-healing is not universally achieved by all cancer patients and offspring caregivers. It is also influenced by personal factors (e.g., resilience, emotional stability, and coping mechanisms) and external support (e.g., support from family, friends, or healthcare professionals, access to counseling or therapy). Conclusion This study broadens the empirical scope of narrative therapy within health communication research by focusing on offspring caregivers of breast cancer patients in the OSG. It addresses the cancer experiences of this often-overlooked group, highlighting their sentiment changes, trauma narratives, and ultimately achieving inner healing during their mothers’ treatment stages. These findings contribute to a more comprehensive understanding of the role of the OSG in health communication and its impact on individual emotional and behavioral experiences. It could guide the development of tailored emotion management interventions, psychological treatments, and programs for offspring caregivers with the potential to positively impact public health outcomes. Thus, health communication scholars should continue to explore the character of and continuities between content published within online platforms by cancer and other disease caregivers. Supplementary Information Supplementary Material 1. Supplementary Material 1.