The effect of prosocial modelling on medical students’ professional identity in China: a PSM-DID analysis

Abstract

Background

Although role models significantly impact medical students’ professional identity, empirical evidence is relatively scarce, and our understanding of the mechanisms behind this influence is limited. Through the lens of prosocial modelling, we explored the effects of role models on medical students’ professional identity and attempted to elucidate the underlying reasons.

Methods

By leveraging the varying number of personnel dispatched by various provinces across China to combat the COVID-19 pandemic, we established indicators for different intensities of prosocial modelling. Using data from the two years before and after the COVID-19 pandemic, we conducted a quasi-experimental study. Employing the propensity score matching difference-in-differences method, we explored the effects of prosocial modelling on medical students’ professional identity.

Results

Prosocial modelling significantly enhanced medical students’ professional identity (β = 0.087, p < 0.01), and the effect remained significant even after controlling for economic factors and the pandemic’s severity (β = 0.067, p < 0.001). Notably, prosocial modelling more significantly impacted the professional identity of female students, those under economic pressure, those uncertain about becoming doctors during high school, and those ranked in the bottom 50% academically.

Conclusions

Prosocial modeling enhances medical students’ professional identity, especially among females, economically disadvantaged students, those initially hesitant about a medical career, and lower-performing students. This highlights the need for role models in medical education to prioritize support for these disadvantaged groups to foster professional identities.

Background

Professional identity refers to the development of a sense of belonging to one’s profession and has been widely discussed in medical education literature, as it is fundamental for medical students during their transition from laypeople to doctors [1]. Numerous factors influence the formation of professional identity, including students’ attitudes, values, and a sense of morality, while external factors are related to the complex clinical learning environment [2]. In these environments, role models play a critical role in shaping medical students’ professional identities [3]. Role modelling is defined as a process whereby faculty members demonstrate clinical skills and exhibit positive professional traits [4]. When exposed to the conscious and unconscious behaviours of these role models, students change their actions and choices through observation, judgment, and imitation [5]. Current research on role models and professional identity primarily utilises qualitative methods to examine the influence process, characteristics of positive role models, and theories that explore learner attitudes, belief systems, narratives, and clinical experiences [2, 6, 7]. However, few studies have employed quantitative analysis to investigate the impact of role models, and the existing research lacks clarity regarding the role of sociocultural factors and individual differences in the pathways through which role models exert their influence [8].

Prosocial modelling provides an opportunity to understand the influence of role models on professional identity in a broader context. Prosocial behavior refers to actions by individuals or groups that benefit others [9]. Prosocial modeling, in this context, describes role models who exhibit such behavior, influencing others through their altruistic actions [10]. The theoretical foundation of prosocial modeling traces back to Bandura’s Social Learning Theory, which posits that individuals learn new behaviors by observing others’ actions and their subsequent consequences [11]. When individuals witness others exhibiting prosocial behaviors, they may imitate these actions due to increased self-efficacy, believing in their own ability to perform similar behaviors [12]. Additionally, they might be motivated by intrinsic rewards, such as a sense of self-worth, to engage in these behaviors [13]. Prosocial modelling is widely used in various countries to cultivate student empathy and prosocial behaviour [14]. The underlying mechanisms of the prosocial modelling effect include situational pressure, modelling medium, and whether the role model receives rewards, among other factors [15]. From the theoretical and perspective standpoints of prosocial modelling, we can gain a better understanding of the mechanisms by which role models influence medical students.

COVID-19 provided a quasi-experimental [16] opportunity to study the impact of prosocial modelling on the professional identity of medical students. In early 2020, COVID-19 broke out in Hubei Province, China, where there was an extreme shortage of medical supplies. Thousands of medical workers from across the country travelled to Hubei to fight COVID-19. The actions of these medical personnel are classified as prosocial behaviours, exemplifying prosocial modelling. The effectiveness of prosocial modelling is determined by whether the observer notices the behaviour [15]. Research has indicated that due to varying levels of exposure to negative medical news across different provinces, students’ choices of medical majors also differ [17]. In our study, in early 2020, the extent of prosocial modelling promotion varied because of the different numbers of personnel dispatched by each province. The greater the number of dispatched personnel, the higher the likelihood that medical students were influenced by prosocial modelling.

The heroic conduct of the medical staff during this crisis offers a quasi-experimental avenue for investigating the impact of prosocial modelling on the professional identity of medical students. We devised a prosocial modelling intensity index based on the influx of personnel from other provinces into Hubei. An inference model utilising the propensity score matching difference-in-differences (PSM-DID) methodology was constructed by drawing on data obtained from a national survey administered two years before and after the outbreak.

Our study applied prosocial theory to the field of medical education and understand, from a broader perspective, the impact of role models on the professional identity of medical students. Using the differences in the number of people from various provinces, we constructed an index of prosocial modelling intensity. We employed quantitative research methods to precisely understand the impact of models of varying intensities on different groups of medical students by conducting a quasi-experimental study with data collected two years before and after the outbreak.

Methods

Data description

This study used self-reported data at the individual level and data on the economy and pandemic situation at the provincial level. Self-reported data were sourced from the China Medical Student Survey (CMSS) conducted in 2019 before the pandemic and in 2020 after the pandemic began. CMSS is a national survey targeting undergraduate clinical medical students, initiated each May near graduation, organised by the National Center for Health Professions Education Development and implemented by participating medical schools [18, 19].

Provincial-level data were sourced from publicly available information provided by various departments. Data on the number of medical personnel dispatched from each province were obtained from figures published by provincial health commissions.

The cumulative COVID-19 case numbers for each province were extracted from data published by the respective provincial health commissions. By March 2020, the situation in Hubei had largely been controlled, with only sporadic cases reported. On 20 March, medical workers engaged in pandemic control began to evacuate and returned to their hometowns. Consequently, we selected the cumulative number of cases in each province on 20 March 2020 as an indicator of the severity of the pandemic across the region. Data on GDP, which reflects the economic status of each province, were obtained from the 2020 China Statistical Yearbook.

Participants

This study included students from the graduating classes of 2019 and 2020, all of whom provided informed consent to participate. In the Chinese clinical medicine undergraduate program, the fifth year is the final year, during which all students participate in hospital rotations. The 2019 cohort completed the full hospital internship, while the 2020 cohort began online internships in early 2020, followed by six months of hospital rotations and six months of online rotations. Data from institutions that participated for only one year were removed. Students who lived or studied in Hubei Province were excluded because of their frontline role in fighting the pandemic. The sample comprised 16,241 students from 31 institutions (7,961 control and 8,280 intervention students).

Key variables

The outcome variable of professional identity was students’ perceptions of the doctor’s profession. The questionnaire’s corresponding question is “Doctor is my ideal career choice,” rated on a five-point Likert scale where higher scores indicate stronger agreement. Because the group with scores from 1 to 3 constituted approximately 30% of the total, this categorisation strategy was employed to mitigate data skewness, accentuate pivotal distinctions, and consequently augment the efficacy of the model. Values of 1–3 were recoded as 0, whereas values of 4 or 5 were recoded as 1. To assess result robustness and sensitivity, we reestimated the main model using the original Likert scale, yielding outcomes consistent with the binary classification results (see Appendix 1). We employed the number of medical personnel dispatched from the school’s province to Hubei Province for the pandemic response as an indicator of the intensity of prosocial modelling. Although the number of dispatched medical personnel may correlate with various factors, it is not associated with medical students’ professional identity. In other words, the number of dispatched medical personnel was an exogenous variable for medical students. It was the most crucial explanatory variable in this study. Appendix 2 shows the number of medical personnel dispatched from each province.

Sociodemographic variables

Our survey collected socioeconomic and demographic information from respondents. Sociodemographic variables included gender, willingness to become a doctor in high school, home location (urban or rural), self-reported academic performance (top 25%, 25-50%, bottom 50%), economic pressure, university/college, hometown province, and the International Socioeconomic Index (ISEI) of father and mother. Table 1 provides descriptive statistics of the students from the 2019 and 2020 surveys. The tabulated data show a substantial disparity in the datasets spanning two years, an aspect that will be the focal point of our forthcoming analytical endeavours.

Table 1 Descriptive statistics (N = 16,241)

Empirical strategy

The main estimation strategy was difference-in-differences (DID), using a continuous measure of intervention intensity (number of anti-pandemic personnel) per Chen’s method [16], capturing more variation than standard DID. Students surveyed in 2019 (pre-COVID-19, no prosocial modeling) served as the control group, while 2020 students (post-COVID-19, exposed to prosocial modeling) were the intervention group, with effects varying by province. Using repeated cross-sectional data from 2019 to 2020, we applied propensity score matching (PSM) to align cohorts, creating panel-like data.

The specific formula is as follows: Di = 1 represents the people surveyed in 2020; otherwise, it is 0. We wanted to exhaust all the covariates that were used to match the intervention and control groups. Based on the existing literature and observed variables, X included common sociodemographic characteristics, gender, home location (urban/rural), parental ISEI, home province, university/college, academic performance, dream job in high school, and economic pressure.

$$\:Logit\left[\text{Pr}\left({D}_{i}=1\right)\right]={\beta\:}_{0}+{\beta\:}_{1}{X}_{i}+{\epsilon\:}_{i}$$

(1)

After performing PSM, we applied DID to explore the effect of the number of medical personnel as a prosocial modelling proxy variable on medical student professional identity. Specifically, we regressed the following equation:

$$\:{y}_{ij}=\beta\:{lnNum}_{j}\cdot\:{I}_{i}^{Post}+{X}_{j}^{{\prime\:}}{I}_{i}^{Post}+{I}_{j}+{I}_{i}^{Post}+{\epsilon}_{ij}$$

(2)

$\:I$ represents individuals, and $\:j$ represents provinces. $\:{lnNum}_{j}$ is the natural log of the total number of anti-pandemic measures in province $\:j$, which was 0 in 2019 due to the absence of the pandemic and medical personnel deployment. $\:{I}_{i}^{Post}$ is the time dummy, and 1 represents the post-COVID-19 period, which is 2020; otherwise, it is 0. $\:{I}_{j}$ is the province-fixed effect. The outcome variable $\:{y}_{ijt}$ represents the professional identity of medical student $\:I$ in province $\:j$ at time $\:t$.

Since the number of dispatched medical personnel correlates with multiple provincial factors, it may not solely reflect prosocial modelling intensity. To control for confounders affecting professional identity, we included province-specific characteristics interacted with time-period fixed effects( $\:{X}_{j}^{{\prime\:}}{I}_{t}^{Post}$). $\:X$ comprises economic indicators and pandemic severity metrics. By accounting for economic status and pandemic impact, we aimed to isolate the true effect of prosocial modelling.

The coefficient of interest, β, estimates the effect of prosocial modelling on professional identity. The strategy follows standard DID estimation, with province-fixed effects controlling for time-invariant provincial differences and time-period fixed effects addressing common temporal fluctuations in professional identity.

Propensity score matching and the balance of covariates

To test the robustness of our results to the choice of matching strategy, we used three algorithms: nearest-neighbour matching, calliper matching, and kernel matching. All estimates were performed using the Stata17 command of psmatch2. In our study, all the algorithms could handle the balancing of covariates. Regardless of the strategy adopted, the coefficients and standard errors were very close. Finally, considering our investigation background, we chose the calliper matching method to calculate the follow-up results.

We calculated the propensity of each student to view being a doctor as an ideal career choice in combination with observable variables and matched samples from 2019 with those from 2020. The 1,102 samples that were not successfully matched were removed. The balance hypothesis of score matching is shown in Fig. 1. Finally, the control group, consisting of 7,423 individuals, was matched with the intervention group comprising 7,716 individuals.

Fig. 1

Balance of the covariates

Results

Descriptive statistics

Figure 2 presents the fundamental descriptive statistics. The horizontal axis represents the number of medical personnel dispatched by each province, and the vertical axis denotes the average value of medical students’ professional identity in each province. The graph on the left illustrates the scenario before the outbreak of the pandemic, indicating students are unaffected by prosocial modelling. Conversely, the graph on the right depicts the situation following the onset of the pandemic, thus reflecting the outcomes influenced by prosocial modelling to varying degrees. Before the pandemic, we observed no significant correlation between the score and the number of medical personnel. After the pandemic, provinces that dispatched a greater number of medical staff exhibited higher average professional identity scores among medical students.

Fig. 2

Score trend of professional identity before and after COVID-19

Notes: (1) We used the number of medical personnel to represent the intensity of prosocial behaviours. In fact, no medical team was sent out in 2019. Here, the intensity indicators for 2019 were constructed in province units to compare the scores for professional identity before COVID-19. (2) The numbers of medical teams shown in the figure are logarithmic values

DID results

We investigated the effect of the number of medical team members on medical students’ professional identities using DID in combination with a matched sample. Here, we used a double fixed-effects model, and the specific estimation results are presented in Table 2.

Table 2 Effect of number of involvements on medical students’ professional identity

Columns (1) to (2) in Table 2 show the results of calculating the estimation of professional identity as the dependent variable. Column (1) shows the estimation results without the inclusion of control variables, and Column (2) shows the estimation results with the inclusion of relevant control variables. Column (1) reports an 8.7% increase in medical students’ professional identity for every 1% increase in medical personnel (β = 0.087, p < 0.01). After controlling for sociodemographic variables (Column 2), the coefficient remained significant (β = 0.082, p < 0.001).

Columns (3) and (4) show the addition of provincial GDP and the cumulative number of confirmed cases in the province. Considering that the intensity of prosocial modelling is directly correlated with the province, factors that are also related to the province and can impact the professional identity of medical students include the level of local economic development and the severity of the pandemic in that year. To obtain the pure effects of prosocial modelling as much as possible, we controlled for the interaction terms of GDP and the cumulative number of cases at the province level. The results showed that after controlling for provincial GDP, a 1% increase in medical personnel increased medical students’ professional identity by 7.0% (p < 0.001). Furthermore, after we added the cumulative number of confirmed cases into the model, a 1% increase in the number of medical personnel increased the professional identity of medical students by an average of 6.7% (p < 0.001).

Counterfactual testing of DID

Counterfactual analyses were conducted to validate the robustness of the findings. The deployment of medical personnel during a pandemic is intrinsically linked to each province. We aimed to ascertain whether medical students’ professional identity correlated with the number of medical personnel, independent of any systematic bias attributable to provincial factors. To this end, we randomly allocated medical personnel to each province, followed by reapplication of the same DID methodology to re-estimate the outcome variable. If the observed effect were a consequence of endogenous variables related to the province, it would persist under these conditions. The insignificance of the interaction term in our results suggests that endogeneity does not compromise the integrity of the primary conclusions. Results are presented in Table 3.

Table 3 Effect of random number of involvements on medical students’ professional identity

Heterogeneity analysis

To understand differences in exposure to prosocial modelling among medical students, we conducted a heterogeneity analysis. Four dimensions were compared: gender, economic pressure, academic performance, and willingness to become a doctor in high school. Table 4 summarises the results.

Table 4 Heterogenous intervention effects by gender, status of economic pressure, and academic performance

Notably, female students were more affected by prosocial modelling than male students (Female β = 0.109, p < 0.05; Male β = 0.061, p > 0.05). In addition, students with financial pressure were more significantly affected by prosocial modelling than those without. Among students experiencing financial stress, those with lower academic rankings were more significantly influenced by prosocial modeling (Bottom 50% β = 0.201, p < 0.001; Top 25% β = 0.053, p > 0.05). Furthermore, those who were not sure about being a doctor in high school were more significantly influenced by prosocial modelling, while other students were not significantly influenced (Unsure β = 0.201, p < 0.01; Certain β = -0.018, p > 0.05).

Discussion

This study explored the effect of prosocial modelling on medical students’ professional identities. Using detailed nationwide survey data on medical students, our findings demonstrate that the prosocial modelling of medical personnel fighting COVID-19 in early 2020 positively impacted medical students’ professional identity. A series of studies conducted in China supports the contention that the professional identity of Chinese medical students was enhanced during the pandemic [20,21,22,23]. This study explored the role of prosocial modelling in this enhancement. Although prosocial modelling is a recent construct in professional identity and medical education literature, the concept of the ‘role model’ remains longstanding and influential. Positive role models are crucial for shaping the professional identities of medical students [3, 8, 24]. Based on our findings, we recommend embedding prosocial behavior into curricula through scenario-based modules, such as COVID-19 response simulations, enabling students to emulate altruistic actions. For faculty development, we suggest structured training to equip educators with skills to effectively model prosocial behaviors, potentially using reflective teaching methods to help students internalize these values [25]. Future research could explore intentional intervention designs to optimize their application in teaching. Studies show that prosocial modeling retains significant effects across subsequent stages, even after initial exposure [26]. Thus, tracking longitudinal effects is essential.

The influence of role models on medical students, from the students’ perspective, includes their learning abilities at the time, as well as their judgment and attitude. Prosocial modelling, from a sociological perspective, offers another lens for understanding medical students.

Our study found differences in prosocial modelling between male and female students. Female students were encouraged more by prosocial modelling, consistent with the findings of previous studies. The pronounced effect of prosocial modeling on female students may stem from their generally stronger empathic abilities [27]. A six-wave longitudinal study confirmed a strong correlation between empathic concern and prosocial behavior [28], and women tend to conform more readily to others’ prosocial actions than men [29]. Additionally, women’s prosocial behaviors, often more communal and relational [30], render them more susceptible to social context and emotional influences [31]. Thus, whether these effects are crisis-specific or persist long-term requires further investigation.

Additionally, we found that medical students with financial stress were more likely to be influenced by prosocial modelling than medical students without financial stress. Previous studies have suggested that observers’ imitative or matching behaviours are influenced by what happens after a model behaves [32, 33]. Observations of punished role models correspond to much less prosocial behaviour than observations of rewarded role models [34]. A meta-analysis of prosocial modelling confirmed that role models are more effective when they are rewarded [15]. The medical team on the frontlines during the COVID-19 pandemic received national accolades and appreciation. National organisations and individuals took action to show their appreciation for the heroes. Medical students, by perceiving the potential benefits of prosocial modeling, may enhance their self-efficacy through vicarious experiences [12], thereby strengthening their professional identity. Those experiencing financial stress might be particularly attuned to these benefits, fostering greater confidence and motivation to pursue a career in this profession as a result. However, people can also be extremely sensitive to the ulterior motives of others’ prosocial behaviours. When people perceive someone’s prosocial behaviour as attributable to selfish reasons, they evaluate the behaviour less positively [35].

Furthermore, this study found that prosocial modelling had a greater impact on students with lower academic performance. Lower academic performance often leads to having a weaker professional identity [36, 37]. Students may experience identity confusion owing to a lack of specific role models to learn from and unclear perceptions of the future. In addition, role models that engage in prosocial behaviour provide them with a concrete, positive role model that helps them develop a professional identity. Similar results were found for students who were unsure about becoming doctors in high school. The lack of preparation for their future career made it more difficult for them to adapt and transition to the role of ‘doctor’. The role models in this study provided them with an ideal reference, providing them with a clearer understanding of doctors, thereby enhancing their own professional identities. Thus, future role models should focus on students who are confused and uncertain about their future, as well as those from vulnerable groups. These students require positive role models to help them identify and shape their professional identities.

Limitations

Medical students’ professional identity encompasses a rich array of values, attitudes, and behaviors, yet this study relied on only a few indicators to represent it, potentially oversimplifying the construct. Future research should incorporate a broader set of indicators to enhance measurement robustness. Moreover, while medical personnel deployment serves as a practical proxy for prosocial modeling intensity, its effect may be confounded by unobservable factors, such as students’ attention to the COVID-19 pandemic, class engagement, or empathy levels, which could influence perception, exposure, or internalization of prosocial behaviors [38]. Given that staff involvement in the pandemic is a completed event, its ongoing and long-term impact on students remains unclear. Investigating the persistence of prosocial modeling over time could be a valuable next step.

Conclusion

Our findings indicate that the prosocial modelling of medical teams dealing with COVID-19 increased medical students’ professional identity. Female students were more affected by prosocial modelling than male students, and students with financial stress may have also been encouraged by seeing potential rewards for becoming a doctor. Additionally, students with lower academic performance and those who were unsure about becoming doctors in high school were more likely to be influenced by prosocial modelling.