The participants of this study were medical students from five schools, including Peking University School of Medicine, Nanhua University, Tianjin Medical University, Xiangnan University and Changsha Medical College, from September to December 2023. Given the exploratory nature of this descriptive survey, a convenience sampling approach was adopted by recruiting intact class from 5 medical schools across China. Then we randomly selected 293 out of 1876 eligible students using SPSS 24.0 to ensure equal selection probability. Potential participants were contacted via email, phone calls, or in-person invitations based on their availability and institutional records. Only students who volunteered were enrolled in the study. Junior-year students who had not yet taken medical statistics courses were excluded. The survey utilized a validated self-designed scale with established reliability (α = 0.850) and construct validity (KMO = 0.74, I-CVI = 0.84, S-CVI = 0.92). Trained assistants distribute and collected questionnaires, they were available to explain the survey purpose and clarify any questions during the completion process. Questionnaires were completed through online or onsite at participants’ convenience. The sample size was determined using the formula for observation studies: n = (Zα/2)^2XP(1-P)/d^2, according to the results of pilot study, we required 245 participants, and our final sample exceeded this requirement. During the research process, 350 questionnaires were distributed and 293 valid questionnaires were collected. The questionnaire of non-medical related college students was excluded, incomplete questionnaires have been removed, and 274 questionnaires were retained. The survey was conducted anonymously.

A total of 274 medical students were included, with an average age of 23.76 ± 3.55 years old. Female students accounted for 56.93% and male students accounted for 43.07%. Undergraduate students in lower grades account for 5.84%, undergraduate students in higher grades account for 69.34%, master’s students account for 22.63%, and doctoral students account for 2.19%.

The data was analyzed using SPSS 24.0 statistical software. Descriptive statistical analysis was conducted on the questionnaire results. Quantitative data is expressed as mean ± standard deviation, while categorical variables are expressed as percentages. When comparing between groups, we used t-test or ANOVA methods.

According the results of the study, 10.22% of medical students expressed a strong interest in statistics. In terms of effort in learning statistics, 8.76% are very hard-working. 64.23% of medical students spend less than 3 h per week studying statistics. 61.31% of medical students feel anxious about learning statistics, 75.18% believe that learning statistics is difficult. In terms of understanding the importance of statistics, 66.42% believe that statistics are very important for medical students. 97.08% believe that statistics are helpful for the clinical and scientific research of medical students. 100% of students use statistics in their published papers, and 82.48% of students use statistical methods in clinical practice. In terms of the reasons for learning statistics and whether they are willing to take statistics as an elective, 70.8% of students are required to complete compulsory courses (Table 1 and Figure 1).

79.56% of medical students learn statistics through relevant school courses. In terms of learning statistical software, 71.53% of medical students learn through practical courses offered by the school. 70.07% of the statistics courses for medical students are taught in large classes. 59.12% of medical students in their schools use a lesson based approach (LBL) in statistics. The average practical course duration is 10.86 + 12.85 h. 48.91% of schools conduct statistical exams through theoretical exams. For 42.34% of students, their schools or majors do not offer practical courses in statistics. 58.39% of students mainly seek advice from non-statistical professionals when encountering problems (Table 2 and Figure 2).

83.21% reported difficulties in selecting the correct statistical methods. In terms of the statistical software used, 68.61% of students will use Excel for analysis. When answering questions about difficulties in statistical application, 71.53% of medical students expressed difficulties in interpreting statistical analysis results professionally. 69.34% of students believe that statistics is abstract for medical students. 34.31% of students have misused statistical methods in scientific research, and 28.47% of students have been affected by statistical misuse in scientific research or graduation (Table 3 and Figure 3).

56.93% of students believe that short-term specialized statistical training courses should be offered. 88.32% of students believe that statistical guidance or consultation is needed in the development of the project, while 88.32% of medical students believe that the statistics course needs to incorporate clinical and research cases to combine with reality. In terms of preferred statistical teaching methods, 79.56% of students are willing to accept blended learning (Table 4 and Figure 4).

Significant gender differences were revealed in statistics engagement among medical students (all p < 0.05). Female students reported higher clinical application of statistics (35.0% vs. 20.3% frequent use) and greater test proficiency (20.0% vs. 6.8% for chi-square), despite experiencing more difficulties in statistical design (86.3% vs. 74.6%). While 98.8% of females valued statistics for research versus 94.9% of males, more males attributed academic delays to statistical misuse (42.4% vs. 28.8%). Curriculum preferences differed markedly: 65.0% of females favored group learning (vs. 49.2%), whereas 47.5% of males advocated reducing statistics instruction time (vs. 28.8%). Assessment outcomes showed females excelled in combined exams (43.8% vs. 27.1%), while males predominated in theoretical testing (59.3% vs. 42.5%) (Table 5).

Postgraduates reported higher interest in statistics (25.71% vs. 5.77%, p < 0.001), greater belief in its importance for medical training (85.71% vs. 60.58%, p < 0.001), and more frequent use in research (91.31% vs. 55.54%, p < 0.001) and clinical practice (85.71% vs. 80.79%, p = 0.011). Despite exerting more effort (20.00% vs. 5.77% “work hard,” p < 0.001), postgraduates expressed higher anxiety (74.29% vs. 57.69%, p = 0.015). Competency gaps persisted: while postgraduates showed greater familiarity with advanced methods (e.g., logistic regression, ROC analysis, p < 0.001), both groups struggled to evaluate statistical methods in literature (≤11.54% “can” assess correctly). Demands for additional training (77.14% vs. 50.96%, p < 0.001) and dissatisfaction with teaching methods (≤22.12% “very satisfied”) highlighted unmet needs across cohorts (Table 6).

Students from key universities demonstrated higher utilization of statistics in research projects (85.19% vs. 60.02%, p < 0.001) and greater familiarity with advanced statistical methods, including t-tests (17.44% vs. 3.37% “familiar and able to use,” p = 0.001) and ANOVA (23.84% vs. 10.38%, p = 0.006). Key universities reported more frequent statistical guidance during project implementation (78.49% vs. 63.21%, p = 0.008) and stronger demand for systematic statistics training (54.07% vs. 40.57%, p = 0.036). Teaching environments differed significantly, with non-key universities more likely to employ non-specialist instructors (practical courses: 38.68% vs. 31.98%, p = 0.034) and use large-class teaching (79.25% vs. 65.12%, p = 0.015). Both groups showed limited practical competency, though key university students exhibited better understanding of multivariate techniques (principal component analysis: 4.65% vs. 1.89% “familiar,” p = 0.002) (Table 7).

This study reveals significant gaps in medical statistics education, with 83.21% of students struggling to select appropriate statistical methods and 71.53% facing interpretation challenges, leading to widespread misuse (34.31% in research projects). Didactic teaching dominates (70.07% large-class lectures; 59.12% LBL), while 42.34% of programs lack practical training, exacerbating disengagement (only 10.22% show strong interest). Assessment focuses excessively on theoretical exams (48.91%), failing to evaluate applied competencies. Despite recognizing statistics’ importance (97.08%), students report anxiety (61.31%) and inadequate preparation for research needs. Comparative analyses highlight global alignment in these challenges, though this study provides the most comprehensive multi-center evidence. Critically, 88.32% demand clinically-integrated case teaching, while 79.56% prefer blended learning. These findings underscore an urgent need for curriculum reform prioritizing active learning, longitudinal software training, and competency-based assessment. The study revealed significant gender disparities, with female students demonstrating greater statistical engagement (35.00% frequent clinical use vs. 20.34% males) and method familiarity (17.50% proficient in rank-sum test vs. 5.08% males), while postgraduate students showed substantially higher competency than undergraduates (91.31% vs. 55.54% project usage). Institutional analysis identified resource disparities, with key universities exhibiting better faculty qualifications (21.52% vs. 33.01% non-statistics instructors) and research integration (85.19% vs. 60.02% project engagement), highlighting systemic inequalities in statistical education.

The observed gender and academic-level disparities in statistical competencies likely stem from multifaceted factors: women’s greater statistical engagement may reflect evolving medical education dynamics that increasingly reward collaborative learning styles, while undergraduates’ limited research exposure explains their poorer performance compared to postgraduates – a gap exacerbated by traditional curricula delaying research training until advanced years. The institutional resource disparities between key and non-key universities mirror global patterns of unequal STEM investment, particularly problematic for statistics education requiring specialized faculty and computational infrastructure. These findings mandate three policy interventions: (1) gender-sensitive pedagogy reforms emphasizing clinical applications to bridge male students’ engagement gap, (2) longitudinal research curricula starting in undergraduate years to build gradual statistical proficiency, and (3) national standardization of statistical education resources to mitigate institutional inequities, potentially through shared digital platforms and faculty exchange programs in medical schools. The alarmingly high statistical misuse rates across all subgroups further underscore the urgent need for competency-based assessment frameworks replacing current exam-focused evaluation systems.

A study conducted among medical students of Serbia (23) indicated that implementation of blended learning approaches can be considered as and attractive, cost-effective, and efficient alternative to classroom training in medical statistics, which is consistent with the results of our study. However, the research in Serbia only compared the impact of blended learning versus traditional classroom teaching in medical students’ statistical exam scores, focusing solely on test results without addressing their practical statistical skills. MacDougall et al. (24) conducted a survey among England medical student, mainly focuses on the importance of statistics. The majority of the surveyed medical students considered learning statistics to be very important, which is consistent with the findings of this study. However, this study did not investigate the existing problems in statistical education for medical students or potential directions for improvement. A study conducted in Iran (25) attempted to incorporate game-based methods into statistics education for medical students and explored the impact of this approach on their learning outcomes. This finding indirectly suggests that traditional statistics teaching methods may have inherent limitations. The few studies conducted in other countries generally align with the findings of the present study. However, compared to our research, these researches have relatively narrower scopes, primarily focusing on students’ perceptions of statistics or specific teaching methods, lacking a comprehensive investigation into the current status and potential improvements in statistical education for medical students. Moreover, most of them were single-center investigations.

This study identifies critical gaps in medical statistics education and proposes a comprehensive reform strategy to align training with modern research needs. Our findings reveal four key areas for policy intervention: (1) curriculum modernization through integration of advanced methods with foundational theory and modular curricula (basic→advanced→research-specific); (2) pedagogical reform via blended learning (23), clinical-integrated training during rotations, and case-based active learning to address student disengagement; (3) resource development by training specialized instructor teams (26), providing open-access platforms with statistical tools/case libraries, and incentivizing EdTech partnerships for medical-statistics simulations; and (4) assessment restructuring through multimodal evaluations (research projects, peer reviews, competency badges) and national standards for hands-on training. To implement these reforms, we recommend a phased approach: pilot testing in select medical schools, faculty certification programs, and policy dialogs with accreditation bodies. These evidence-based strategies – addressing curriculum-practice gaps, pedagogical inefficiencies, and resource limitations – can transform statistical education into a cornerstone of rigorous medical research, ultimately improving research validity and clinical decision-making. The proposed reforms balance immediate feasibility (e.g., using existing clinical cases for teaching) with long-term systemic change (e.g., national competency standards), offering a actionable roadmap for policymakers to elevate statistical proficiency across medical education.

This research employed a multicenter, questionnaire-based design to comprehensively assess statistical education across China’s geographic and economic regions, enhancing the generalizability of findings. While anonymous surveys and rigorous interviewer training mitigated potential response biases. The research use questionnaire surveys to collect data, which may susceptible to response bias, including social desirability bias, acquiescence bias, courtesy bias and fear bias, which may compromise the validity and reliability of the findings. In order to mitigate these biases, we have taken the following measures: (1) the questionnaire survey was conducted anonymously. (2) A pilot study was conducted before the formal investigation to ensure that students could accurately understand the questionnaire. (3) Unified training was provided to all interviewers. Despite these limitations, the questionnaire surveys remain valuable for capturing subjective experiences and attitudes of medical students to statistics education. The percentage of undergraduate lower grades and doctoral students were relatively low, the sample size may not big enough. However, the study selected students randomly, and the selected samples were representative and roughly consistent with the proportion of students in each grade level according to the national data report.