The application of Large Language Models (LLMs) in healthcare prompts critical reflection on the impact of linguistic justice on health equity. Disparities in data availability across languages during the training and evaluation phases of LLMs result in significant performance variations. These disparities risk reinforcing existing inequities in healthcare access and outcomes. This study outlines strategies for integrating linguistic justice into LLM assessment frameworks, advancing health equity and global health.

Background: The association between fine particulate matter (PM2.5) chemical constituents and blood lipid levels remains unclear. This longitudinal study, leveraging data from a comprehensive national survey across 31 provinces in mainland China, aimed to assess the associations between exposure to PM2.5 constituents and blood lipid metabolism. Methods: Utilizing data from the China National Stroke Screening Survey (CNSSS), we included 292,754 participants accounting for 683,759 visits recorded between 2013 and 2019. To investigate the longitudinal relationship between PM2.5, its five constituents (ammonium, sulfate, nitrate, black carbon (BC), and organic matter (OM)), and four blood lipid indicators (low-density lipoprotein cholesterol (LDLC), high-density lipoprotein cholesterol (HDLC), triglyceride (TG), and total cholesterol (TC)), we employed linear fixed-effects panel regression models. Results: Each interquartile range (IQR) increase in PM2.5 (lag 0-3 month) was associated with significant increases in LDLC (3.44%, 95% CI: 3.07-3.80) and TC (2.99%, 95% CI: 2.70-3.28), and a decrease in HDLC (-0.89%, 95% CI: -1.17 to -0.62). For constituents, ammonium and nitrate showed the strongest associations with LDLC (4.66% and 4.76%, respectively) and TC (3.72% and 3.60%, respectively). TG levels were positively associated with secondary inorganic aerosols, most notably sulfate (2.91%, 95% CI: 2.50-3.33). Significant negative correlations with HDLC were observed for BC (-0.46%, 95% CI: -0.72 to -0.19) and sulfate (-0.71%, 95% CI: -1.04 to -0.38). The exposure-response relationships between the constituents and blood lipid indicators were predominantly nonlinear, characterized by a plateau at higher concentration levels. Conclusion: Our results indicate that exposure to PM2.5 and its constituents, especially inorganic aerosol components, was strongly associated with adverse blood lipid levels. The findings of our study provide critical insights into policy development aimed at enhancing the management and control of PM2.5 pollution.

BACKGROUND:Negative emotions are common in women with benign breast disease (BBD) and may influence breast cancer risk, but the relationship between these emotions, cancer development, and mitigating factors remains unclear. This study aimed to assess the impact of negative emotions on breast cancer risk and identify associated contributing factors. METHODS:This prospective cohort enrolled 3,535 women with BBD from 12 Chinese hospitals. Negative emotions were assessed using the irritability, depression, and anxiety (IDA) scale, and breast pain was evaluated with the breast pain score scale. Associations between negative emotions and breast cancer risk were analyzed using Cox regression models. Factors associated with negative emotions were identified with logistic regression, and mediation analysis was used to explore the roles of life satisfaction, sleep duration, and physical exercise. RESULTS:Nearly half (47%) of participants had high negative emotions (IDA ≥19). During follow-up, 2% developed breast cancer. High negative emotion was associated with a 2.78-fold increased cancer risk (HR = 2.78; 95% confidence interval, 1.34-5.76; P = 0.006), driven by depression and introverted irritability. Negative emotions correlated with physical inactivity, short sleep, life dissatisfaction, and severe breast pain. Life satisfaction (45%), sleep (7%) and exercise (19%) partially mediated the breast pain-negative emotion relationship. CONCLUSIONS:Negative emotions significantly increase breast cancer risk in women with BBD. Addressing modifiable factors like exercise, sleep, and life satisfaction may reduce emotional distress and potentially lower cancer risk. IMPACT:Integrating psychologic assessment and lifestyle interventions into BBD management could improve emotional well-being and mitigate cancer risk.

Background: Healthy lifestyle guidance is an important component of outpatient preventive care, yet evidence is limited on whether outpatients receive satisfactory guidance across multiple lifestyle domains. This study examined receipt of satisfactory healthy lifestyle guidance and its associated factors among outpatients in China. Methods: This national multicenter cross-sectional study used data from the outpatient component of the fifth National Third-Party Evaluation of the Action Plan for Further Improving Healthcare Services. A total of 31,599 outpatients from 143 tertiary public hospitals across all 31 provincial-level administrative regions of mainland China were included. The primary outcome was receipt of satisfactory healthy lifestyle guidance, defined as receiving at least one type of guidance and reporting satisfaction or high satisfaction. Mixed-effects logistic regression models with hospital-level random intercepts were used to identify associated factors. Results: Dietary guidance was the most frequently reported domain, followed by physical activity, weight management, and smoking cessation and alcohol reduction guidance. Overall, most outpatients were classified as having received satisfactory healthy lifestyle guidance. In the fully adjusted model, male sex and hypertension were associated with higher odds of satisfactory guidance, whereas a greater number of chronic conditions and unclear physician type were associated with lower odds. Satisfaction with history taking and disease communication and perceived respect and comfort from healthcare professionals were strongly associated with satisfactory healthy lifestyle guidance. However, these contemporaneously measured patient-reported indicators may reflect overlapping dimensions of patients’ overall outpatient care experience. Sensitivity analyses yielded broadly consistent findings. Conclusions: Satisfactory healthy lifestyle guidance was common among outpatients in China’s tertiary public hospitals, but guidance was unevenly distributed across lifestyle domains. Patients’ evaluations of satisfactory healthy lifestyle guidance were closely associated with their broader communication and interpersonal care experiences. However, because these measures were self-reported and assessed contemporaneously, the observed associations should not be interpreted causally.

ObjectivesThis study aimed to generate evidence to help improve the quality of information provided by Chinese online platforms for retail of prescription medicines, as a baseline for enhancing the supervision by the regulatory authorities of internet retail of prescription medicines to offer comprehensive, accurate and reliable information to consumers.MethodsThis is an observational study to investigate the quality of the information provided by two key types of online platforms in China that sell paroxetine online, a common mental health medication. The authors utilized the DISCERN with the Likert 5 scale, a validated instrument comprising three domains and 16 items to assess the comprehensiveness, accuracy and reliability of information provided by a total of 34 Chinese online platforms. We firstly examined whether these online platforms have been diligent in their duty to provide appropriate retail service-related information (e.g. delivery, payment, return policies) for prescription medications. We also evaluated the quality of the medication information associated with treatment plans for specific prescriptions provided by the online platforms. Three trained master medical students independently conducted the evaluation. By securing the consistency of the scoring results among different evaluators and the reliability of the evaluation, we adopted the mean score of the three evaluators for each item to perform further analysis at the item and domain levels, and between two types of online platforms.ResultsAmong the 34 online platforms, 2 were evaluated as providing excellent information for retail of paroxetine (1/34, 5.88%), 6 were good (6/34, 17.65%), 25 were fair (25/34, 73.53%), and 1 was poor (1/34, 2.94%). The key identified flaws of the online platforms for providing appropriate information for retail of paroxetine included (1) inadequate provision of medication information, (2) absence of prescription validation and authentication, and (3) lack of professional pharmacy services such as mandatory pharmacist consultations or patient-specific drug counselling.ConclusionsConsumers face challenges in obtaining comprehensive, accurate and reliable information from the online platforms for common mental health prescription medications. The quality of information provided by the online platforms for retail of prescription medicines is generally suboptimal, and the compliance of prescription validation and authentication is problematic. These highlight the need for strengthening the supervision of the online platforms with novel approaches along with their rapid developments, and ensuring appropriate online information provided to consumers.

BackgroundPharmaceutical services play a critical role in safeguarding medication safety and optimizing therapeutic outcomes. Their value is increasingly recognized in China, where recent policy reforms have introduced new medical service pricing items, and several provinces have piloted an inpatient consultation fee (clinical pharmacy surcharge), namely inpatient pharmacist consultation (IPC), as a billable services.ObjectiveThis study used a discrete choice experiment (DCE) to evaluate hospitalized patients’ preferences for IPC and the factors shaping these choices, providing evidence to inform service development and policy-making.MethodsStudy attributes and levels were derived from literature review, patient surveys, and expert interviews. Attributes included hospital tier (tertiary vs. non-tertiary), pharmacist professional title (attending pharmacist, associate chief pharmacist, chief pharmacist), consultation frequency (once, three times, five times per week), duration (5, 10, 15 min), service outcome (improved efficacy vs. reduced medication errors), and unit cost (10 or 15 CNY per session; 1 CNY ≈ 0.15 USD). A DCE questionnaire was administered to inpatients across six Sichuan hospitals. Preferences were analyzed using a mixed logit model to estimate regression coefficients (β) and willingness-to-pay (WTP).ResultsOf 320 valid responses, patients preferred consultations delivered by associate chief pharmacists (β = 0.182, p < 0.001; WTP 5.00 CNY) in tertiary hospitals (β = 0.138, p < 0.001; WTP 3.80 CNY), with a primary focus on reducing medication errors (β = 0.125, p = 0.002; WTP 3.42 CNY). Patients favored three sessions per week (β = 0.156, p = 0.005; WTP 4.29 CNY) lasting 15 min each (β = 0.133, p = 0.016; WTP 3.65 CNY). Service cost had a limited negative influence on choices (β = −0.036, p < 0.001) within the narrow price range of 10 to 15 CNY tested in this study. The optimal profile (associate chief pharmacist, tertiary hospital, three times per week, 15 min/session, error reduction focus, 10 CNY/session) achieved a 68.4% predicted choice probability. Subgroup analyses showed that low-income patients prioritized consultation frequency, urban patients preferred longer duration, and patients with chronic diseases emphasized efficacy over error reduction.ConclusionInpatients demonstrated clear preferences for attributes of IPC, with variations by income, residence, and chronic disease status. Greater involvement of clinical pharmacists, tailored adjustment of consultation frequency and duration, and integration into hierarchical diagnosis and treatment and chronic disease management were encouraged. Price has limited influence on choices. We recommend including pharmaceutical consultations in health insurance reimbursement after pilot programs.

Nitrogen dioxide (NO2), a hazardous air pollutant, has been linked to various adverse health outcomes, yet its association with serum lipid profiles - a key precursor to cardiovascular diseases and metabolic diseases - remains limited and inconsistent. Therefore, we used a nationwide prospective cohort to investigate the association between ambient NO2 exposure and serum lipid profiles. A total of 296,537 adults with 691,924 repeated visits were included. High-resolution (1 × 1 km) monthly ambient NO2 concentrations were obtained by incorporating satellite observations, chemical transport model simulations and high-resolution geographical variables. We used fixed-effect model to investigate the health effects of ambient NO2 exposure on serum lipid profiles. Each 10 µg/m3 increment of ambient NO2 exposure was related to an elevated risk of hyperbetalipoproteinemia [lag 2 month, relative risk (RR) = 1.140, 95

Access to novel medications matters quality-adjusted life years and the opportunity cost associated with productivity lost. Gaps in patient access to novel medicines exist due to insufficient public funding reimbursement in emerging countries. Evidence of time from regulatory approval to reimbursement decision by public funding, referred to as time to reimbursement (TTR), remained limited in emerging countries. This study compared and analyzed public funding reimbursement of novel medicines approved in five Eurasian countries that are global leaders in pharmaceutical innovation. All of them have a centralized mechanism for reimbursement decisions on novel medicines, allowing identification of a clear date of public funding reimbursement. By exploring the facilitators of rapid application of pharmaceutical innovations, we expected to inform the public funding reimbursement decision-making in emerging countries, so as to improve patient access and contribute to addressing the global health challenge in achieving universal health coverage. This is a retrospective study which investigated the public funding reimbursement and TTR of novel medicines that obtained marketing authorization between 2018 and 2023 in China, Japan, France, the United Kingdom (UK) and Switzerland. We firstly conducted descriptive analyses of TTR across countries, followed by the pairwise comparisons using Kruskal–Wallis H tests with Bonferroni corrections. We then performed the survival analysis of time-to-event data using the multiple Cox proportional hazards regression by inclusion of country and year dummy variables. Other covariates associated with the characteristics of novel medicines and manufacturers, as well as the review and approval pathways were included in the regression. We estimated the differences of hazard ratios (HR) of novel medicines being reimbursed by public funding across countries. Subgroup analyses were conducted to assess the specific factors associated with the public funding reimbursement in different countries. Since China began to systematically publicly fund novel medicines in 2019, sensitivity analyses were conducted by removing the 2018 data and repeating the same analyses. As of July 1st, 2024, Japan had the highest proportion and fastest rate of public funding reimbursement of novel medicines, which were approved between 2018 and 2023 (HR = 11.29, [95

OBJECTIVE:Human papillomavirus (HPV) vaccination is the most effective method to prevent cervical cancer. This study aimed to investigate the status of HPV vaccination and associated factors among Chinese females. METHODS:Between January and March 2021, we conducted a large national survey among female doctors and nurses in 181 public tertiary hospitals across all 31 provinces of China. In the survey, we asked three questions: "Have you ever received an HPV vaccination? If yes, what type and in which year?" We described and compared the proportion of vaccination coverage according to occupation, age, geographic region, education, marital status, among other factors. RESULTS:Among 42,800 participants, 6185 (14.45 %) reported receiving HPV vaccination. Physicians showed a slightly higher vaccination rate (2064/13,804; 14.95 %) than nurses (4121/28,996, 14.21 %). Factors significantly associated with higher vaccination rates included younger age, being never married or divorced, higher education, better self-reported health status, residing in the western region, working in Obstetrics and Gynecology or Surgery departments, working at cancer hospitals, and being nulliparous. Regarding vaccine types, the 4-valent vaccine accounted for 50.3 %, followed by the 9-valent (33.7 %), imported 2-valent (16.0 %). CONCLUSION:Understanding the characteristics of the likelihood of receiving HPV vaccination among female physicians and nurses in China indicates their awareness of the risk for cervical cancer, which could help us better develop primary prevention strategies.

Hypertension remains a major public health challenge in China. Leisure-time physical activity (LTPA) is critical for hypertension control. Yet its social network determinants, particularly key role models, are understudied in China’s familial-centric cultural context. This sequential mixed-methods study integrated quantitative surveys with qualitative interviews among hypertensive patients in Yichang, China. A total of 2639 patients were selected from 18 primary healthcare institutions using multi-stage random sampling method. Totally, 10,550 social relationships were nominated. Quantitative data on LTPA was measured by the International Physical Activity Questionnaire Long Form, and social network characteristics (size, density, LTPA aggregation, LTPA heterogeneity, and LTPA status of specific members) was gathered through the name generator method. Generalized additive mixed models assessed nonlinear associations between social network characteristics and LTPA levels (individual-level, n = 2639); mixed-effects logistic regression analyzed member-patient LTPA linkages (relationship-level, n = 10,550). Qualitative data (n = 37) via interviews underwent grounded theory coding to contextualize mechanisms. Among participants, 70.56

Introduction The burden of gynaecological diseases is increasing globally, affecting women’s health, quality of life and fertility. Improving the quality of gynaecological care will enhance patient recovery and improve outcomes of gynaecological diseases. However, comprehensive evaluation tools and nationwide data on the quality of gynaecological care are lacking. This study aims to develop a framework for evaluating gynaecological care quality in China, analysing disparities and proposing improvement strategies.Methods and analysis This nationwide study employs a sequential explanatory mixed-methods design across three phases: (1) questionnaire development through a literature review, indepth interviews, expert consultation and pilot testing; (2) collection of cross-sectional survey data from gynaecology departments, healthcare providers and patients in public healthcare facilities across 31 provinces in mainland China; and (3) integration and interpretation of qualitative and quantitative results. The study collects data on organisational structure, care processes and outcomes. Analysis of quantitative data will include risk adjustment of key quality indicators, comprehensive evaluation of care quality and exploration of influencing factors using linear regression. Qualitative data will undergo thematic and content analyses. Joint displays will be used to integrate the mixed-method results.Ethics and dissemination The study was approved by the Peking Union Medical College Hospital Ethics Committee (I-24ZM0003). All participants provided informed consent. Data were anonymised and participants retained the right to withdraw without penalty. A risk-benefit assessment was conducted, and no vulnerable groups were included without justification. These findings will be disseminated to healthcare policymakers through academic publications, scientific conference presentations and reports. The results are expected to provide methodological guidance for the quality evaluation of gynaecological care and inform policy development in China.

BACKGROUND:Current evidence for residential greenspace and blood lipid levels was primarily based on single-center cross-sectional studies, with spatial and temporal differences possibly leading to heterogeneous results. This large multicenter community-based study aimed to investigate the longitudinal relationships of residential greenspace with blood lipids. METHODS:We included participants aged over 40 years with repeated blood lipid measurements from the China National Stroke Screening Survey (CNSSS) from 2013 to 2019. Greenspace exposure was assessed using the three-month average normalized difference vegetation index (NDVI) with a spatial resolution of 1 km. Linear fixed-effect models examined the associations between greenspace and low-density lipoprotein cholesterol (LDL-C), high-density lipoprotein cholesterol (HDL-C), triglyceride (TG), and total cholesterol (TCHO) levels, adjusting for demographic characteristics, lifestyle factors, and chronic diseases history. Air pollution and physical activity were considered as potential mediators. RESULTS AND CONCLUSION:We included 292,754 participants with 683,759 visits. A 0.1-unit increase in NDVI was significantly associated with a 1.28 % increase in HDL-C (95 % confidence interval [CI]: 1.06 %, 1.51 %), and decreases of 0.63 % in LDL-C (95 % CI: -0.88 %, -0.38 %), and 1.45 % in TCHO (95 % CI: -1.64 %, -1.26 %). These associations were modified by sex, region, residence, history of dyslipidemia, hypertension, smoking and drinking. We observed nonlinear exposure‒response curve (ERC) for the association between greenspace and lipid indicators. Air pollution (PM2.5, NO2) partially mediated the associations between residential NDVI and LDL-C, HDL-C, and TCHO. In conclusion, this longitudinal study revealed that greater residential greenspace was associated with favorable blood lipid levels. Air pollution is a significant mediator linking residential greenspace to blood lipids.

Having a first child is a contributing factor to having a second or third child; however, current research focuses on second and third children, and the present study aims to investigate the intention to have a first child and its main influencing factors among the childless population aged 15–49 in China, as well as analysing the influencing factors by age group. As part of the ‘Survey on the Psychology and Behaviour of Chinese Residents’, this survey was conducted on 6941 residents aged 15–49 in 32 provinces across China. The survey was conducted from 20 June to 31 August 2022. Out of 6941 respondents, the mean score of willingness to have a first child (score range: 0–100) was 42.51. 17.33

Cancer caregivers often face substantial psychosocial burdens, which can lead to significant anxiety and depression. This network meta-analysis aimed to evaluate the relative efficacy of various psychosocial interventions in alleviating these symptoms. Following PRISMA guidelines, we conducted a systematic literature search for randomized controlled trials (RCTs) that evaluated psychosocial interventions for cancer caregivers. Data were synthesized using a network meta-analysis to compare the effects of different interventions on anxiety and depression. A total of 35 trials involving 7 types of psychosocial interventions were included. Three interventions—meaning-centered psychotherapy (anxiety: MD = − 5.48, 95

China’s National Essential Public Health Service (NEPHS) Program was launched in 2009 to deliver population-based public health and individual health management services to all residents at 800,000 primary health-care centers nationwide. This study assessed NEPHS utilization data and evaluated usage inequities using comprehensive nationwide data. A cross-sectional study was conducted, selecting 16 indicators (out of 18) from 12 service packages to evaluate inequity. These included 4 indicators for services provided to all residents and 12 for pregnant women, new mothers, children aged 0–6 years, and patients with hypertension, diabetes, severe mental disorders, or tuberculosis. Data on service utilization and target populations for these indicators across the 31 provinces and 453 cities in mainland China were obtained from the NEPHS database and management platform for the period January 1 to December 31, 2019. Service utilization rates and bias-corrected bootstrap confidence intervals (CIs) were calculated to determine utilization. Inequities were assessed using the Gini coefficient and Sitthiyot-Holasut composite inequality index at the national and provincial levels, and the Theil index was employed to decompose overall inequity into within-region and between-region subgroups. The NEPHS collected health records for 88.25

Objective Traditional Chinese medicine (TCM) incorporates traditional diagnostic methods and several major treatment modalities including Chinese herbal medicine, Chinese patent medicine, and non-pharmacological methods such as acupuncture and tuina. Even though TCM is used daily by more than 70,000 healthcare facilities and over 700,000 clinical practitioners in China, there is a poor understanding of the extent to which TCM diagnostic methods are used, how different treatment modalities are deployed in general, and what major factors may affect the integration of TCM and Western medicine. This study aimed to fill this void in the literature. Methods In the 2021 National Healthcare Improvement Evaluation Survey, we included three questions gauging the perception and practices of TCM amongst physicians working in TCM-related facilities, investigating the frequency of their deployment of TCM diagnostic methods, and predominant TCM treatment methods. Our empirical analysis included descriptive statistics, intergroup chi-square analysis, and binary logistic regression to examine the association between different types of facilities and individual characteristics and TCM utilization patterns. Results A total of 7618 clinical physicians comprised our study sample. Among them, 84.27% have integrated TCM and Western medicine in their clinical practice, and 80.77% of TCM practitioners used the 4 diagnostic methods as a tool in their clinical practice. Chinese herbal medicine was the most widely utilized modality by Chinese TCM physicians (used by 88.49% of respondents), compared with the Chinese patent medicine and non-pharmacological TCM methods, which were used by 73.14%, and 69.39%, respectively. Herbal tea as an out-of-pocket health-maintenance intervention is also a notable practice, recommended by 29.43% of physicians. Significant variations exist across certain institutions, departments, and individual practitioners. Conclusion Given that most of the surveyed physicians integrated TCM with Western medicine in their clinical practices, the practice of “pure TCM” appears to be obsolete in China’s tertiary healthcare institutions. Notably, remarkable variation exists in the use of different TCM modalities across institutions and among individuals, which might be related to and thus limited by the practitioners’ experience. Future research focusing on the efficacy and safety of TCM interventions for specific diseases, the development of standardized clinical guidelines, and the enhancement of TCM education and training are called for to optimize TCM-Western medicine integration.

This study constructs a tripartite evolutionary game model involving healthcare data management authorities (DMAs), healthcare data operating departments (DODs), and data-related entities (DEs) within a triple principal-agent framework. We analyze dynamic interactions among these stakeholders in healthcare data governance, focusing on privacy security, moral hazard, and interest alignment. Results indicate that strategic instability arises under conditions of ambiguous data property rights and asymmetric risk responsibilities. However, the system converges to compliance-oriented equilibria when critical thresholds are surpassed. Notably, DMAs’ strong incentive strategies are pivotal in resolving regulatory paradoxes, while DODs’ risk behaviors exhibit nonlinear sensitivity to penalty intensity and revenue levels. We further demonstrate that a dual “penalty-compensation” mechanism mitigates economic losses from data breaches, and synergistic government reputation mechanisms with financial incentives reduce regulatory costs. Policy implications include a tiered dynamic regulatory system, a revenue-risk linked distribution mechanism, and a collaborative governance ecosystem driven by technology and credibility constraints.

INTRODUCTION:Retirement represents a significant life transition and is associated with individual health outcomes. Previous studies on the health effects of retirement have yielded inconsistent conclusions. This study aimed to estimate the impact of retirement on the body mass index (BMI) and BMI-defined overweight and obesity. METHODS:Between 2012 and 2015, this study included 88,471 public sector participants from the Chinese Hypertension Survey (CHS) while excluding those working in the private sector. By utilizing the fuzzy regression discontinuity design (FRDD), this research evaluated the direct impact of retirement on the BMI, as well as on the overweight and obesity rates. Additionally, the study examined variations in the effects of retirement among groups stratified by sex and educational attainment. RESULTS:The fully adjusted model suggested that retirement did not have a significant impact on the BMI or overweight and obesity rates of the overall population. Notably, retirement significantly impacted the male participants, resulting in an increase in their BMI of 2.18 (95% CI 0.23-4.13), but did not affect the BMI of the female participants. Furthermore, individuals with lower educational backgrounds experienced more pronounced BMI changes upon retirement. CONCLUSIONS:On average, retirement has no significant impact on the BMI and overweight and obesity rates. Retirement leads to an increase in the BMI among men but does not affect that in women. When considering adjustments to existing retirement policies, the differential health effects of retirement across various populations should be taken into account.

INTRODUCTION:Accurate breast cancer risk prediction is essential for early detection and personalized prevention strategies. While traditional models, such as Gail and Tyrer-Cuzick, are widely utilized, machine learning-based approaches may offer enhanced predictive performance. This systematic review and meta-analysis compare the accuracy of traditional statistical models and machine learning models in breast cancer risk prediction. METHODS:A total of 144 studies from 27 countries were systematically reviewed, incorporating genetic, clinical, and imaging data. Pooled C-statistics were calculated to assess model discrimination, while observed-to-expected (O/E) ratios were used to evaluate calibration. Subgroup and sensitivity analyses were conducted to examine heterogeneity and assess the influence of study bias across various populations. RESULTS:Machine learning-based models demonstrated superior performance, with a pooled C-statistic of 0.74, compared to 0.67 for traditional models. Models that integrated genetic and imaging data showed the highest levels of accuracy, although performance varied by population. Sensitivity analyses excluding high-bias studies showed improved discrimination in models incorporating genetic factors, with the pooled C-statistic increasing to 0.72. Traditional models, such as Gail, exhibited notably poor predictive accuracy in non-Western populations, as evidenced by a C-statistic of 0.543 in Chinese cohorts. CONCLUSION:Machine learning models provide significantly greater predictive accuracy for breast cancer risk, particularly when incorporating multidimensional data. However, issues related to model generalizability and interpretability remain, particularly in diverse populations. Future research should focus on developing more interpretable models and expanding global validation efforts to improve model applicability across different demographic groups.