Yan Zuo

Background: Generative AI (GenAI) is increasingly entering health care through documentation support, communication tools, educational content generation, and other knowledge-intensive functions. However, organizational adoption remains uneven, and concerns related to privacy, security, output reliability, governance, workflow fit, and infrastructure continue to limit broader implementation. Although the literature increasingly discusses the skills health care workers may need in the GenAI era, less is known about how hospital managers view the organizational readiness required before such skills can be expected across the workforce. Hospital human resource (HR) managers are especially important in this regard because they are involved in training, competency development, workforce planning, and organizational change. Objective: This study aimed to explore how hospital HR managers perceived the relevance of GenAI-related skills, the organizational barriers to broader implementation, and the forms of preparation they considered realistic in the current stage of adoption. Methods: We conducted a descriptive qualitative study in 2 tertiary hospitals in China—1 in Beijing and 1 in Sichuan Province. Purposive sampling was used to recruit HR managers involved in staffing, training, competency development, or workforce planning. Semistructured telephone interviews were conducted between January 15 and February 9, 2025. Interviews were audio-recorded, transcribed verbatim, and analyzed using inductive thematic analysis. The researchers familiarized themselves with the transcripts, generated initial codes, grouped related codes into categories, and developed themes through iterative comparison, team discussion, and refinement. Rigor was supported through reflexive memoing, audit trail documentation, team debriefing, and participant validation. Results: Fifteen HR managers participated. Participants did not describe mature or formalized GenAI competency systems within their institutions. Instead, they described an early-stage organizational environment characterized by strategic recognition of GenAI, informal and uneven experimentation, uncertain governance, and substantial variation in readiness across roles and settings. Three themes were identified: (1) GenAI was viewed as increasingly inevitable and relevant to hospital work, but not yet as a stable universal competency requirement; (2) organizational unreadiness, including infrastructure limitations, lack of approved systems, unclear policy boundaries, and workforce heterogeneity, constrained any move toward universal mandates; and (3) participants preferred phased, role-specific preparation over immediate compulsory standards. They emphasized that readiness depended on secure access, governance, workflow alignment, and differentiated training pathways rather than broad top-down requirements. Conclusions: Hospital HR managers perceive GenAI as increasingly important but not yet governable through universal competency expectations. Their accounts suggest that hospitals are not simply deciding whether to adopt GenAI, but are negotiating when, for whom, and under what safeguards GenAI-related skills become legitimate workforce expectations. Readiness-based, phased, and role-specific approaches may therefore be more appropriate than premature competency mandates in the early stages of hospital GenAI implementation.

To investigate changes in care-seeking behaviors among Chinese cancer survivors following the lifting of COVID-19 control measures, identify demographic, clinical, and socioeconomic predictors of professional care use, and to assess perceived convenience and severity of care needs post-pandemic. We conducted a cross-sectional survey in Southwest China between May 2023 and January 2024. A total of 477 valid responses were collected using convenience sampling from outpatient clinics, rehabilitation units, and online survivor networks. Group differences in care-seeking before and after COVID-19 control were tested with Chi-square and Kruskal-Wallis tests. Paired Chi-square tests were applied to compare pre- and post-COVID-19 behaviors. Multivariable Poisson regression with robust error variance was used to identify independent predictors of professional care-seeking. Hospital visits declined from 54.1% before control to 40.0% after, while online consultations increased from 8.2% to 22.2% and self-care rose from 27.0% to 33.1%. Perceived convenience of care worsened, with inconvenience reports rising from 32.7% to 43.4%, and the proportion of survivors reporting severe or very severe illness increased. Older age, higher income, higher education, cohabitation, multiple comorbidities, and better functional status were significant predictors of professional care-seeking. Post-COVID-19, cancer survivors in China shifted toward online and self-care modalities, while hospital use declined. Access to professional care is shaped by socioeconomic and functional factors, which leads to disparities in survivorship care. Expanding telemedicine, strengthening community oncology services, and addressing socioeconomic and functional barriers are essential to ensure equitable access to care. Building resilient, integrated models that combine hospital, community, and digital modalities may better support survivors during future public health emergencies.

Background: Urban–rural disparities may hinder HPV vaccination in Tibetan regions. This study quantified disparities in HPV vaccine knowledge, willingness, and uptake among women in Tibetan settings. Methods: A multi-center cross-sectional survey of women aged 18 to 45 years was conducted. HPV knowledge was measured using a 10-item score (range 0–10). Vaccination uptake was defined as self-reported receipt of at least 1 dose. Among unvaccinated participants, willingness was dichotomized as willing versus unwilling or undecided. Multivariable linear and logistic regression models accounted for within-site correlation using cluster-robust standard errors by survey site. Results: Among 812 participants (568 urban, 244 rural), overall uptake was 29.8% (242/812) and was higher among urban than rural residents (35.2% vs. 17.2%, p < 0.001). The median knowledge score was 6 (IQR 2 to 7) overall; urban residents had higher knowledge than rural residents (median 6 [IQR 3 to 8] vs. 3 [IQR 0 to 6]; p < 0.001). Among unvaccinated participants ( n = 570), willingness was associated with urban residence (adjusted OR 1.32, 95% CI 1.01 to 1.73), knowledge (per 2-point increase: 1.46, 1.28 to 1.67), Bachelor’s-or-higher education (1.83, 1.28 to 2.62), and prior HPV testing (1.62, 1.08 to 2.44). In the base uptake model, urban residence (1.67, 1.21 to 2.30) and knowledge (per 2-point increase: 1.34, 1.23 to 1.46) were associated with uptake. In the staged model adding prevention-service contact, Pap smear history (1.88, 1.39 to 2.56) and HPV testing history (2.41, 1.55 to 3.75) were strong predictors and the urban association attenuated (1.26, 0.93 to 1.71). Among unvaccinated participants, cost (46.7%) and safety concerns (33.7%) were common; rural residents reported substantially higher access or travel burden (47.5% vs. 7.1%) and supply or availability constraints (45.0% vs. 2.7%). The most common maximum willingness-to-pay category was 300 to 999 CNY (37.5%). Conclusion: Large urban–rural gaps in knowledge and vaccination uptake persist despite high willingness among unvaccinated women, indicating a substantial intention-to-behavior gap driven by affordability and delivery constraints. Interventions should pair education with financing and service delivery strategies that reduce travel burden and improve reliable vaccine availability in rural Tibetan settings.

Background: Survival after head and neck cancer treatment has improved, shifting attention toward survivorship quality and post-treatment service gaps. Evidence from China that quantifies unmet supportive care needs using standardized instruments across multiple regions remains limited. Methods: We conducted a multicenter cross-sectional study in three tertiary hospitals in East, Middle, and Southwest China. Adults with histologically confirmed head and neck cancer were recruited during routine follow-up 3 to 24 months after curative-intent treatment. Supportive care needs were assessed with the Supportive Care Needs Survey Short Form (SCNS-SF34) and the head and neck module (SCNS-HNC), and head and neck-related quality of life was measured with University of Washington Quality of Life version 4 (UW-QOL v4). The primary outcome was any moderate-to-high unmet need on SCNS-SF34. We reported overall and hospital-specific prevalence, domain-level burden, and correlates from mixed-effects logistic regression with hospital clustering interpreted cautiously because only three centers were included. Results: Among 1,407 surveyed survivors, 1,239 questionnaires were valid for analysis (88.1%). Overall, 78.2% reported at least one moderate-to-high unmet need. Health system and information (34.2%) and psychological needs (31.2%) were the most prevalent SCNS-SF34 domains. Dry mouth or sticky saliva (54.6%) and swallowing difficulties (36.8%) were the leading head and neck-specific unmet needs. The median number of moderate-to-high SCNS-SF34 needs per patient was 3 (IQR 2 to 5). Lower UW-QOL composite score (per 10-point decrease: aOR 1.61), rural residence (aOR 1.25), monthly household income below 5,000 CNY (aOR 1.45), stage III to IV disease (aOR 1.25), and multimorbidity of at least two comorbidities (aOR 1.83) were associated with higher odds of unmet needs. In a hospital fixed-effect sensitivity model, the direction and magnitude of the main patient-level associations were materially unchanged. Conclusion: Unmet supportive care needs were common among Chinese head and neck cancer survivors in early survivorship, particularly in information, psychological, salivary, and swallowing domains. The observed associations suggest that routine needs screening and multidisciplinary follow-up pathways may help identify survivors who are more likely to report substantial unmet needs, especially in socioeconomically vulnerable groups. Any service-level interpretation should remain cautious because the evidence is observational and based on three tertiary centers.

Background: Financial toxicity is increasingly recognized as a consequential dimension of cancer care, yet multicenter evidence in China remains limited for patients with gynecologic cancers undergoing radiotherapy, a treatment pathway that often entails repeated visits and substantial non-medical costs. This study estimated the prevalence and severity of financial toxicity and examined its association with quality of life and psychological distress. Methods: A multicenter cross-sectional survey was conducted in three tertiary hospitals in China, led by West China Second University Hospital, Sichuan University. Adult patients with gynecologic cancers receiving external beam radiotherapy and or brachytherapy with definitive, adjuvant, or curative-intent salvage or consolidation intent were consecutively recruited. Financial toxicity was assessed using the COmprehensive Score for financial Toxicity (COST, 0 to 44; lower scores indicate worse toxicity). Quality of life was measured using the EORTC QLQ-C30, and psychological distress using the Distress Thermometer (DT, 0 to 10; clinically significant distress defined as DT ≥ 4). Multivariable regression models included hospital fixed effects to account for measured differences across centers and adjusted for sociodemographic, access-burden, and clinical covariates. Results: Among 1,533 returned questionnaires, 1,303 were valid and analyzed (85.0%). Mean COST score was 21.6 (SD 7.4); 17.5% had severe financial toxicity (COST ≤ 14), 48.0% moderate (15 to 24), and 34.5% mild (≥25). Mean QLQ-C30 global health status was 61.3 (SD 14.2). Mean DT score was 4.3 (SD 2.1), and 65.2% met criteria for clinically significant distress. In adjusted analyses, each 5-point decrease in COST was associated with lower global health status (β −4.17, 95% CI −4.62 to −3.72), lower emotional functioning (β −4.68, 95% CI −5.20 to −4.16), higher fatigue (β 4.47, 95% CI 3.93 to 5.01), and higher DT score (β 0.32, 95% CI 0.23 to 0.40), all p < 0.001. Each 5-point decrease in COST was associated with higher odds of clinically significant distress (OR 1.34, 95% CI 1.22 to 1.47; p < 0.001). Conclusion: Financial toxicity was common among gynecologic radiotherapy patients in China and was independently associated with poorer quality of life and higher psychological distress. Integrating financial toxicity screening with supportive care pathways during radiotherapy may help identify high-risk patients and guide targeted assistance.

Aim: To examine determinants of nurses' adoption of generative artificial intelligence outputs in clinical practice using a technology acceptance model and an integrated structural equation modelling framework. Design: Cross‐sectional online survey. Methods: Registered nurses in mainland China completed an anonymous questionnaire assessing perceived performance benefits, perceived ease of use, perceived information quality, perceived source credibility, social influence, facilitating conditions, adoption intention and adoption behaviour. Structural equation modelling was used to evaluate the measurement model and estimate a primary mediation model in which perceived performance benefits and perceived ease of use predicted adoption intention, and adoption intention predicted adoption behaviour. An integrated model added information quality, source credibility, social influence and facilitating conditions as additional determinants. Sensitivity analyses were conducted using an ordinal estimator to assess robustness. Results: The analytic sample comprised 330 nurses. In the primary model, higher perceived performance benefits and greater perceived ease of use were associated with stronger adoption intention, and stronger adoption intention was associated with higher self‐reported adoption behaviour. The integrated model showed that perceived information quality contributed to adoption intention beyond core expectancy beliefs, while perceived source credibility showed a small direct association with adoption behaviour. Social influence demonstrated a modest association with adoption intention, whereas facilitating conditions showed weaker associations after accounting for other determinants. Model conclusions were consistent across estimation approaches. Conclusion: Nurses' adoption of generative artificial intelligence outputs is shaped by perceived performance benefits, ease of use and perceived information quality, with adoption intention functioning as the proximal determinant of self‐reported use. Implementation strategies should focus on demonstrable workflow gains, reducing interaction burden and strengthening governance and verification to support safe adoption.

Background: Chatbots driven by large language model artificial intelligence (AI) have emerged as potential tools to enhance health information access for patients with cancer. However, their integration into patient education raises concerns among oncologists. Limited literature has examined the perceptions and attitudes of oncologists in terms of endorsing AI-driven chatbots for health information. Objective: This study aims to explore the perceptions and attitudes of Chinese oncologists toward endorsing AI-driven chatbots to patients with cancer. Methods: In this phenomenological qualitative study, we purposively sampled oncologists from 4 hospitals in Southwest and East China and conducted semistructured interviews with 24 participants between November 19, 2024, and December 21, 2024. The data saturation principle was observed to determine the end point of data collection. Data were analyzed using the Colaizzi method. Results: The participants were aged 42.0 (range 29‐53) years on average, including 9 (37%) female and 15 (62%) male participants. The participants had an average of 8.8 (range 1‐25) years in oncology. Of the participants, 7 (29%) had recommended AI chatbots to patients. Three key themes were revealed from analysis of interview transcriptions, including perceived benefits, significant concerns, and impacts on doctor-patient dynamics. Benefits included enhanced accessibility and potential support for chronic condition management. Concerns centered on liability, misinformation, lack of personalization, privacy and data security risks, and patient readiness and education. Oncologists stressed a dual impact of AI chatbots on doctor-patient dynamics, recognizing the potential for improved communication and risks of trust erosion due to overreliance on AI. Conclusions: While recognizing the potential of AI-driven chatbots to enhance accessibility of health information and chronic disease management, Chinese oncologists report significant concerns, including liability, misinformation, lack of personalization, privacy and data security risks, and patient readiness. Addressing the challenges requires comprehensive solutions, such as clear policies and guidelines, rigorous testing and validation, institutional endorsement, and robust patient and provider education. Future efforts should focus on resolving the barriers while leveraging the strengths of AI technology to support patient-centered care in a safe, effective, and ethical manner.

Background: The rapid evolution of generative artificial intelligence (GenAI) is transforming health care globally. In China, hospitals are rapidly embracing digital transformation. Senior leaders are pivotal in influencing and deciding the adoption of GenAI technologies in professional settings. However, evidence on their perceptions and attitudes and how they evolve over time is lacking. Objective: This study aims to investigate senior Chinese hospital leaders’ perceptions relating to GenAI and their attitudes toward adopting GenAI technologies in professional settings and to document how their perspectives evolve. Methods: A longitudinal, descriptive qualitative study was conducted using a phenomenological approach. Senior leaders, including hospital directors/deputies and department heads/deputies, from 3 tertiary hospitals across the Hebei, Guangdong, and Sichuan provinces were purposively sampled. Data were collected through semistructured telephone interviews at two time points (baseline and 6 months later). Interview transcriptions were analyzed using Colaizzi’s method to extract themes related to knowledge acquisition, attitudinal shifts, and evolving barriers and facilitators to GenAI adoption. Results: A total of 32 participants were interviewed in August 2024 and February 2025, including 11 (34.4%) participants from Hebei, 12 (37.5%) from Sichuan, and 9 (28.1%) from Guangdong. Their roles included 8 (25%) hospital directors/deputies and 24 (75%) department heads/deputies. The analysis of the interview transcriptions revealed three primary themes. First, participants’ understanding of GenAI improved markedly over time as they engaged with diverse information sources and gained practical experience. Second, despite widespread support for personal use, senior leaders shifted from initial reluctance to formal institutional adoption. Third, initial obstacles, such as limited technical literacy and resource constraint concerns, diminished over time, while new drivers, including peer influence and technological breakthroughs such as DeepSeek, emerged to catalyze adoption. Conclusions: Senior Chinese hospital leaders’ perceptions and attitudes toward GenAI have evolved significantly over time. Improved knowledge through diverse information channels has enhanced their comprehension and shifted their stance from cautious personal experimentation toward formal institutional adoption. The dynamic interplay between diminishing obstacles and emerging catalysts, notably the influence of peer practices and technological breakthroughs such as DeepSeek, underscores the potential for transformative change in health care management. Targeted educational initiatives, pilot projects, and robust policy frameworks are needed to facilitate GenAI integration.

Background: Myocardial infarction (MI) remains a leading cause of morbidity and mortality worldwide. Although postsurgical cardiac interventions have improved survival rates, effective management during recovery remains challenging. Traditional informational support systems often provide generic guidance that does not account for individualized medical histories or psychosocial factors. Recently, artificial intelligence (AI)–based large language models (LLM) tools have emerged as promising interventions to deliver personalized health information to post-MI patients. Objective: We aim to explore the user experiences and perceptions of an AI-based LLM tool (iflyhealth) with integrated personal health record functionality in post-MI care, assess how patients and their family members engaged with the tool during recovery, identify the perceived benefits and challenges of using the technology, and to understand the factors promoting or hindering continued use. Methods: A purposive sample of 20 participants (12 users and 8 nonusers) who underwent MI surgery within the previous 6 months was recruited between July and August 2024. Data were collected through semistructured, face-to-face interviews conducted in a private setting, using an interview guide to address participants’ first impressions, usage patterns, and reasons for adoption or nonadoption of the iflyhealth app. The interviews were audio-recorded, transcribed verbatim, and analyzed using Colaizzi method. Results: Four key themes revealed included: (1) participants’ experiences varied based on digital literacy, prior exposure to health technologies, and individual recovery needs; (2) users appreciated the app’s enhanced accessibility to professional health information, personalized advice tailored to their clinical conditions, and the tool’s responsiveness to health status changes; (3) challenges such as difficulties with digital literacy, usability concerns, and data privacy issues were significant barriers; and (4) nonusers and those who discontinued use primarily cited complexity of the interface and perceived limited relevance of the advice as major deterrents. Conclusions: iflyhealth, an LLM AI app with a built-in personal health record functionality, shows significant potential in assisting post-MI patients. The main benefits reported by iflyhealth users include improved access to personalized health information and an enhanced ability to respond to changing health conditions. However, challenges such as digital literacy, usability, and privacy and security concerns persist. Overcoming the barriers may further enhance the use of the iflyhealth app, which can play an important role in patient-centered, personalized post-MI management.

Background: Large language model (LLM) artificial intelligence (AI) tools have the potential to streamline health care administration by enhancing efficiency in document drafting, resource allocation, and communication tasks. Despite this potential, the adoption of such tools among hospital administrators remains understudied, particularly at the individual level. Objective: This study aims to explore factors influencing the adoption and use of LLM AI tools among hospital administrators in China, focusing on enablers, barriers, and practical applications in daily administrative tasks. Methods: A multicenter, cross-sectional, descriptive qualitative design was used. Data were collected through semistructured face-to-face interviews with 31 hospital administrators across 3 tertiary hospitals in Beijing, Shenzhen, and Chengdu from June 2024 to August 2024. The Colaizzi method was used for thematic analysis to identify patterns in participants’ experiences and perspectives. Results: Adoption of LLM AI tools was generally low, with significant site-specific variations. Participants with higher technological familiarity and positive early experiences reported more frequent use, while barriers such as mistrust in tool accuracy, limited prompting skills, and insufficient training hindered broader adoption. Tools were primarily used for document drafting, with limited exploration of advanced functionalities. Participants strongly emphasized the need for structured training programs and institutional support to enhance usability and confidence. Conclusions: Familiarity with technology, positive early experiences, and openness to innovation may facilitate adoption, while barriers such as limited knowledge, mistrust in tool accuracy, and insufficient prompting skills can hinder broader use. LLM AI tools are now primarily used for basic tasks such as document drafting, with limited application to more advanced functionalities due to a lack of training and confidence. Structured tutorials and institutional support are needed to enhance usability and integration. Targeted training programs, combined with organizational strategies to build trust and improve accessibility, could enhance adoption rates and broaden tool use. Future quantitative investigations should validate the adoption rate and influencing factors.