Empathy is commonly defined as the capacity to perceive, understand, and share the emotions or experiences of others, and it plays a critical role in medical practice. In medical contexts, this construct is conceptualized as clinical empathy, encompassing not only emotional resonance and a cognitive understanding of patients’ experiences but also emotion regulation to maintain clinical judgment and the ability to convey understanding through caring behaviors. The Jefferson Scale of Empathy (JSE) was specifically developed to assess clinical empathy in healthcare settings, particularly its student version (JSE-S) for medical students [1, 2]. In contrast, the Interpersonal Reactivity Index (IRI) evaluates trait empathy, reflecting stable empathic tendencies in general social contexts [3, 4]. Employing both instruments provides a comprehensive evaluation, capturing context-sensitive clinical capacities alongside more enduring dispositional empathy.

Clinical empathy is a core component of professional competence in medicine, contributing not only to increased patient satisfaction, treatment adherence, and clinical outcomes [5, 6] but also to reduced physician burnout, increased job satisfaction, and improved mental and physical well-being [7–9]. The development of empathy among medical students may be influenced by demographic factors (e.g., age, gender, career aspirations, parental education) and educational factors (e.g., curriculum design, clinical exposure, role modeling) [10, 11]. Additionally, professional identity formation (PIF) has been proposed as a mechanism that reinforces the recognition of empathy’s importance and facilitates its internalization as a stable professional behavior [1, 12, 13].

Although the critical importance of empathy is widely recognized, its developmental trajectory throughout medical education remains unclear. Systematic reviews synthesized approximately 60 empirical studies conducted between 1998 and 2019 across North America, Europe, and Asia. Taken together, the findings are inconsistent: 27 studies reported a decline in empathy during medical education, 20 reported no significant change, and 13 reported an increase. In contrast, among the 13 longitudinal studies, the results were more convergent, with 10 showing decreases, 2 remaining stable, and only 1 short-term study (1 year) reporting an increase [11, 14–16]. This pattern suggests that longitudinal designs provide greater reliability in capturing developmental trends. Therefore, combining longitudinal tracking with dual-scale assessments allows for a more nuanced and robust evaluation of empathy development.

In China, despite hosting the largest population of medical students worldwide, long-term longitudinal studies examining empathy development remain scarce. To address this gap, the present study employed a longitudinal design, following the same cohort of students enrolled in the five-year clinical medicine program over the course of five years. Both the JSE-S and the IRI were administered with repeated measures analysis to depict empathy trajectories and explore potential demographic and educational predictors across the entire medical education cycle. The findings are anticipated to enrich empirical evidence on empathy development in the Chinese context and provide theoretical and practical insights for optimizing empathy cultivation in medical education.

The self-administered questionnaire consisted of three sections:

We recruited class advisors—faculty members at Chinese universities responsible for both academic guidance and student affairs—as data collectors. After standardized training to ensure procedural consistency, collectors distributed the online questionnaire via class WeChat groups and invited students to participate voluntarily during the same period each academic year. The first wave of data collection took place in October 2019, and the final wave was completed in October 2023.

An explanatory statement was provided at the beginning of the questionnaire, informing students that participation was voluntary and anonymous and that data would be reported in aggregate for research purposes only. Completion of the questionnaire was regarded as implied informed consent. The study protocol was approved by the Research Ethics Committee of the School of Basic Medical Sciences, Fudan University.

All analyses were conducted via IBM SPSS Statistics version 20.0. Descriptive statistics were performed on baseline data (i.e., each participant’s first available response), and the internal consistency of the JES-S and IRI was assessed via Cronbach’s α.

To evaluate changes in empathy over time and its associated factors, a linear mixed-effects model was constructed with the JSE-S total score as the dependent variable. Fixed effects included categorical variables—academic year, age group, career aspiration, and entry year (i.e., the year in which a student first joined the study)—as well as the continuous covariates of the total IRI score. The entry year was modeled as a categorical factor to adjust for potential cohort-specific differences due to staggered enrollment, especially considering structural disruptions such as the COVID-19 pandemic.

Participant ID was treated as a random effect to account for repeated measurements, with an autoregressive covariance structure [AR(1)] assumed. Type III sums of squares were used for hypothesis testing (α = 0.05).

A parallel model was constructed with the IRI total score as the dependent variable to compare predictors of trait empathy, using the same fixed and random effects structure.

Baseline data, including data from 104 medical students, were obtained from the first measurement. Most participants were female (63.46%) and aged ≤ 19 years (82.69%). The majority were in their first (59.62%) or second (33.65%) year of study. A high proportion (88.46%) reported a willingness to pursue a medical career, and 45.19% held student leadership roles. Additionally, 58.65% were from one-child families. The participants' hometowns spanned cities, towns, and rural areas (Table 1).

[Insert Table 1 near here.]

Both the JSE-S and the IRI demonstrated good internal consistency, with Cronbach’s α values of 0.80 and 0.78 at baseline and 0.82 and 0.79 across all valid responses, respectively.

Across the full dataset, the JSE-S total scores ranged from 72–134 (theoretical range: 20–140), with a mean of 106.84 (SD = 11.89). The IRI scores ranged from 52 to 106, with a mean of 78.32 (SD = 8.80). The skewness and kurtosis values were − 0.12 and − 0.20 for JSE-S and − 0.03 and − 0.26 for IRI, respectively. All corresponding Z values were within ± 1.96, indicating approximate normal distributions for both scales. No outliers were detected.

A linear mixed-effects model was constructed with the JSE-S total score as the dependent variable. When an AR(1) covariance structure and maximum likelihood estimation were used, the model fit was good: −2 restricted log-likelihood = 1923.91; AIC = 1929.91; and BIC = 1940.51.

The covariance parameter estimates indicated an AR (1) diagonal variance of 81.21 (SE = 14.08), an autoregressive coefficient (rho) of 0.20 (SE = 0.17), and a subject-level random intercept variance of 25.70 (SE = 13.98).

Type III tests of fixed effects revealed that academic year (F = 4.68, p = .001), age group (F = 5.71, p = .004), career aspiration (F = 4.31, p = .014), and the IRI score (F = 25.77, p < .001) were significant predictors of JES-S. Entry year was modeled as a categorical factor and did not have a significant main effect (F = 1.54, p = .207) (see Table 2).

[Insert Table 2 near here.]

Controlling for IRI, age group, career aspiration, and entry year, academic year had a significant main effect on JES-S scores (F (4, 170.60) = 4.68, p = .001). The empathy scores were highest in the first year and showed a general declining trend across subsequent academic years. Compared with those of first-year students, JES-S scores were significantly lower in the third year (B = − 6.29, p = .019, 95% CI [− 11.55, − 1.04]) and in the fifth year (B = − 11.57, p < .001, 95% CI [− 17.69, − 5.56]) (see Table 3). Bonferroni-adjusted pairwise comparisons confirmed significant differences between the 5th and 1st years (p = .002) and between the 5th and 2nd years (p = .001). Differences between the 5th and 4th years (p = .075) and between the 3rd and 2nd years (p = .088) showed marginal significance (see Fig. 1).

[Insert Table 3 near here.]

[Insert Fig. 1 near here.]

Older students reported significantly higher empathy scores (F (2, 229.71) = 5.71, p = .004). Compared with students aged ≤ 19 years, those aged 20–22 years scored higher (B = 5.65, p = .010, 95% CI [1.36, 9.93]), and those aged 23–25 years scored even higher (B = 12.27, p = .001, 95% CI [4.97, 19.56]) (see Table 3).

Bonferroni-adjusted comparisons revealed that both the 23–25 and 20–22 groups scored higher than ≤ 19 (p = .003 and p = .030, respectively). The 23–25 group also scored higher than the 20–22 group did, with marginal significance (p = .068) (see Fig. 1).

5. Effects of Career Aspiration on JSE-S scores

Career aspiration significantly influenced JES-S scores (F (2, 237.68) = 4.31, p = .014). Students who expressed a willingness to become doctors scored significantly higher than those who were uncertain did (B = − 5.67, p = .004, 95% CI [− 9.47, − 1.87]) (see Table 3).

6. Positive Predictive Effect of IRI on JES-S

Pearson correlation analysis revealed a moderate positive association between the JSE-S score and the IRI (r = .371, p < .001). In mixed linear models, IRI scores significantly predicted JSE-S (B = 0.40, p < 0.001) even after adjusting for age, grade, career aspiration, and entry year. However, the IRI scores remained stable across academic years (F (4, 171.34) = 0.57, p = .685), with no significant associations for age group, career aspiration, or entry year in the adjusted models. This stability contrasts with the significant longitudinal decline in JES-S scores (see Fig. 2).

[Insert Fig. 2 near here.]

To our knowledge, this study is the first longitudinal investigation in China examining the trajectory of empathy among undergraduate medical students. Our findings revealed a general downward trend in context-specific clinical empathy, as measured by the JSE-S, over the five years of medical education. In contrast, scores on the IRI, which reflect stable empathic traits, remained relatively constant throughout the study period. Additionally, older students were found to score higher on the JSE-S, suggesting that age has a positive association with clinical empathy. Career aspiration was also a significant predictor of empathy: students who were undecided about pursuing a medical career had significantly lower JSE-S scores than those who clearly expressed their willingness to become doctors.

In China, the five-year clinical medicine program typically follows a "4 + 1" training model, consisting of four years of coursework followed by one year of clinical internship. At Fudan University, a comprehensive university [19], the medical school curriculum is structured as follows: The first year mainly consists of general education and foundational science courses with limited medical content. The second year focuses on basic medical sciences. The first semester of the third year continues with basic medical sciences (accounting for three-quarters of weekly contact hours), whereas the second semester marks a full transition to clinical coursework and incorporates bedside teaching. The fourth year continues with clinical coursework and clerkships, and the fifth year consists of full-time clinical internships with rotations through multiple hospital departments. Given the limited clinical exposure of students in the early years, this study employed both the JSE-S, which assesses context-specific empathy in healthcare settings, and the IRI, a generic empathy scale, to provide a comprehensive evaluation of medical students’ empathy. The use of dual instruments helped offset the limitations of using a single scale and minimized the interpretive bias caused by differing measurement emphases.

A key finding of this study is the divergent trajectories of the two empathy measures: JSE-S scores showed an overall declining trend across academic years, reaching their lowest point in the fifth year during the internship, whereas IRI scores remained stable throughout the same period. This discrepancy may be explained by the fundamental differences in the measurement focus of the two scales [20]. The IRI captures trait empathy, a relatively stable personality disposition shaped by long-term life experiences, whereas the JSE-S assesses context specific, clinical empathy, particularly the cognitive and behavioral aspects of physician–patient interactions, which are subject to situational modulation. Our findings suggest that while dispositional empathy remained intact, students’ capacity for empathic expression in clinical contexts diminished over time. This underscores that empathy is not merely a personal trait but also a context-sensitive construct shaped by educational environments, psychological factors, and curricular structures.

This distinction is also supported in the literature. Studies using the IRI have often reported stable [21–23] or only minimally varying empathy levels [24] across medical education and low sensitivity to educational interventions [25]. In contrast, JSE scores are more responsive to contextual factors such as curriculum design, patient expectations, and cultural factors [26], exhibiting greater malleability in response to educational interventions [27]. For example, educational interventions aimed at enhancing empathy often yield significant improvements in JSE scores but produce negligible [28] or no changes [27] in IRI scores.

Our findings are consistent with those of prior studies in China that documented a decline rather than an increase in empathy among medical students [29, 30]. Similar trends have been observed in other countries, including the United States and Iran [31, 32]. This decline aligns with our previous cross-sectional study of medical students in an eight-year program, which revealed a significant decrease in empathy during the internship phase. Several factors may contribute to this decline. First, upon transitioning to clinical rotations, students are immersed in a high-intensity, high-workload clinical environment that may compel them to prioritize technical tasks, reducing opportunities for physician‒patient communication. This results in a task-oriented approach overshadowing humanistic care, thereby suppressing opportunities for empathic engagement [17, 33]. Second, students may resort to psychological self-protection strategies such as “dehumanizing patients” to shield themselves from the emotional burden associated with pain, suffering, and death, consequently attenuating empathic expression [14]. Third, structural limitations in medical education may also contribute: current admission criteria emphasize standardized academic examination scores with limited attention to humanistic attributes. Once enrolled, the curriculum often prioritizes biomedical knowledge and technical competencies over the cultivation of emotional competence. Upon entering clinical settings, a hospital culture dominated by professional detachment, technological orientation, and efficiency-first principles may further hinder the development of empathy [33].

In contrast, trait empathy, as measured by the IRI, a dispositional and personality-based construct, appears more resistant to contextual factors such as professional role expectations and curricular structures, showing high cross-situational stability throughout medical education. Our data further revealed a significant positive correlation between the IRI and JSE scores: each one-point increase in the IRI was associated with an average increase of 0.4 points in the JSE. This suggests that although trait empathy remains stable, it continues to play a facilitating role in students’ context-specific empathic expression in clinical settings.

We also found that age was a significant positive predictor of JSE scores: older students scored significantly higher than younger students did, whereas IRI scores remained stable across age groups. This pattern may be interpreted via Richter’s model of "multidirectional age differences in empathy" [34], which divides empathy into three core dimensions: cognitive empathy (the ability to perceive another's emotions accurately), affective empathy (the capacity to share others’ emotions), and sympathy (concern for others' predicaments and compassionate expressions). According to this theory, sympathy tends to increase steadily with age and is relatively resistant to contextual factors, whereas cognitive and affective empathy are more sensitive to situational influences. Given that many JSE items closely align with the sympathy dimension conceptualized by Richter—emphasizing physicians’ understanding, compassion, and concern for patients—which may explain the age-related increase in JSE scores observed in our study. In contrast, the IRI captures dispositional cognitive and affective empathy without being anchored to specific contexts. Richter emphasized that age-related differences in these two dimensions are more likely to emerge in emotionally salient or context-specific situations and are often absent in neutral or decontextualized settings. This may help explain why the IRI scores, which assess empathy in a generalized, nonsituational manner, remained stable across age groups in our study.

Importantly, although age had a positive effect on JSE scores, the overall trend still showed a decline across academic years. This finding indicates that the development of empathy is shaped not only by age-related factors but also by contextual influences. As students progress through medical school, increasing academic demands, pressures related to postgraduate advancement and career preparation, and increasing clinical challenges may collectively erode their capacity for context-specific empathic expression in clinical settings. In other words, the potentially beneficial effects of age on empathy may be attenuated or even overridden by the modulatory influences of the educational and clinical environment, thereby contributing to the observed downward trend in JSE scores. From this perspective, the relationship between age and empathy should be understood as a contextually modulated, multidirectional pattern whose manifestations are jointly shaped by factors such as the type of measurement tool, the empathy dimension assessed, and the educational environment. This may help explain the inconsistent findings regarding age-related empathy reported in the literature [29, 35, 36].

Our previous cross-sectional studies revealed a strong association between medical students’ career aspirations and their empathy levels [17, 18]. Specifically, students who aspired to pursue a medical career after graduation consistently presented higher JSE scores than those who were uncertain or did not plan to pursue medical careers. The present longitudinal findings are consistent with earlier results, further confirming that career aspiration is a key determinant of JSE scores. Mixed linear model analyses revealed that students who expressed a willingness to become doctors had significantly higher JSE scores than those who were undecided did, with a mean difference of 5.67 points (β = -5.67, p = 0.004). Although no significant difference in JSE scores was observed between the two groups at baseline (t (101) = -1.28, p = 0.203; Cohen's d = -0.41) and both groups exhibited similar rates of decline over time (time × career interaction effect, P > 0.05), students intending to pursue a medical career consistently demonstrated higher JSE scores throughout their medical education. This difference is unlikely to stem from divergent trajectories of empathy development but rather suggests that the divergence emerged and was sustained during the course of medical education.

Given the close association between career aspirations and professional identity formation (PIF) and the well-documented role of empathy as a core component of PIF with strong positive correlations [13], it is plausible that PIF may serve as a mediating factor in the relationship between students’ career aspirations and their empathy levels. PIF involves both the internalization of professional roles, values, and behaviors and the integration of individual characteristics with the expectations of the medical profession. Empathy acts as a key bridge in this process of integration and internalization [37]. We speculate that students with a clear intention to pursue a medical career are more likely to develop a stable and coherent professional identity throughout their training, which in turn helps to reinforce and sustain their empathic attitudes and behaviors. Although the JSE scores showed an overall decline during medical education, the positive effects of PIF may be retained and reflected in the consistently higher JSE scores observed among career-committed students. Future studies should incorporate validated PIF assessment tools to explore their mediating role in this relationship and to deepen our understanding of empathy development within the context of medical education.

Despite the strengths of its five-year longitudinal design and the use of dual measurement tools, this study has several limitations. First, the sample was drawn from a single university, which may limit the generalizability of the findings. Second, both the JSE-S and the IRI are self-reported measures that assess empathy-related attitudes rather than actual behaviors and are subject to social desirability bias. Future studies should complement self-reports with additional methods such as direct observation, peer evaluations, and feedback from standardized patients. Finally, this study followed students only through the first semester of their final year, excluding the complete internship period. Since empathy and professional identity continue to evolve during clinical training, extending the observation period through graduation would provide a more complete view of the developmental trajectory.

This study is the first longitudinal investigation of empathy development among Chinese undergraduate medical students. The findings revealed a general decline in context-specific clinical empathy (JSE) over five years, reaching its lowest point during the internship, whereas trait-based empathy (IRI) remained relatively stable. This divergence underscores the significant impact of educational context on shaping the expression of clinical empathy. Age positively predicted JSE scores, supporting the theory of multidirectional age-related empathy development. Moreover, career aspiration emerged as an important factor, with students who clearly intended to become doctors consistently achieving higher JSE scores. These results indicate that empathy is shaped by both individual dispositions and educational exposures. Medical educators should therefore prioritize curriculum design, learning environments, and clinical training that foster the sustained development of empathy throughout medical education.