Evaluating the construct validity of the health promotion literacy scale: a confirmatory factor analysis in Taiwan’s university social responsibility context

Abstract

Introduction

The development of a Health Promotion Literacy Scale within the USR Curriculum is crucial for addressing the existing gap in measuring the impact of USR on students’ citizenship and social engagement. This study aimed to develop and assess the psychometric properties of the Health Promotion Literacy-based Scale in USR (HPLS-USR) scale on medical university students through confirmatory factor analysis (CFA) in Taiwan.

Methods

We conducted a cross-sectional study and recruited a convenience sample of 405 students in central Taiwan. The HPLS-USR scale, developed by Hung et al. in 2022, comprises four factors—personal growth, responsibility of citizenship, social interaction, and intellectual development—that together accounted for 61.83% of the total variance. Data analysis was performed using SPSS version 26 and AMOS (version 22.0) to perform a CFA to verify the model’s goodness of fit.

Results

The modified model derived 17 items in 3 factors, with 51.02% of the total variance explained. The structural equation model displayed a good fit, suggesting that personal growth, responsibility for citizenship, and social interaction were latent variables of the Health Promotion Literacy-based Scale on a USR curriculum. The Cronbach’s alphas for the overall scale and the three subscales ranged from 0.82 to 0.93. The psychometric properties of the scale demonstrated good to excellent model fit in the CMIN/DF (χ2/df) = 3.149, GFI = 0.91, AGFI = 0.88, SRMR = 0.04, RMSEA = 0.07, NFI = 0.90, NNFI = 0.92, RFI = 0.89, IFI = 0.93, CFI = 0.93, PNFI = 0.78, and PCFI = 0.81. The Cronbach’s alphas ranged between.81 and.89; the composite alphas ranged between.87 and.94.

Conclusion

This study provides preliminary construct validity evidence for the HPLS-USR scale, a robust and psychometrically sound instrument for assessing health promotion literacy among USR students.

Introduction

Many medical schools claim to be committed to social accountability and emphasize social responsibility by recognizing the significant shortcomings in how healthcare providers are educated concerning societal obligations in medicine [1]. University social responsibility (USR) builds upon the principles of corporate social responsibility (CSR), which places a strong emphasis on seeking knowledge and the truth, raising civic consciousness, and the long-term advancement of society [2]. USR is considered an objective and tested way to show that an organization is fit to address the requirements and interests of society. This means that USR revolves around an organization’s ability to spread and utilize processes related to awareness of social issues, sources of information about such issues, requisite training, and community cooperation [3].

USR learning processes can help students master specific skills, including social networking, digital competencies, cross-cultural flexibility, and tolerance [4]. As social organizations and academic institutions, universities’ roles should not be limited to teaching, research, and cultivating students’ talent. Traditional medical education is hospital-centered and treatment-focused, prioritizing disease management over proactive community health, prevention, and health promotion. This approach often neglects broader societal responsibilities, including addressing social determinants of health and community engagement.

Medical universities can use USR to address health disparities by providing healthcare services to underserved communities to promote health-service equity for the sustainability of good health and wellbeing [1]. Medical universities can leverage USR initiatives to reduce health disparities by providing health care services, such as screening and health promotion, to underserved and underserved communities. This effort promotes equity in health care and ensures that all people have access to quality health care, thereby supporting the long-term sustainability of good health and well-being. This integration of social practice allows medical students to reflect on their learning experiences and apply their knowledge and skills through social interaction. For instance, Cooley et al. (2015) examined students who took an outdoor education course that used experiential learning to teach the importance of working in groups. After the course, skeptical university students reported a strong intention to continue using group work in the conventional university context as their attitudes about groups had changed [5]. A meta-analysis of 103 related studies concluded that learning outcomes in academic, personal, social, and civic-engagement elements are generally improved through such pedagogy [6].

The socioeconomic and educational milieu affects university training, and this training enables a university to recognize and emphasize the necessity of giving comprehensive and ideal training to university students [7]. Attitudes and values play a crucial role in guiding students toward personal, societal, and environmental wellbeing according to the OECD’s Learning Compass 2030 [8]. Universities are responsible for cultivating these values and instilling a sense of citizenship among their students, including respect, fairness, personal and social responsibility, integrity, and self-awareness. By emphasizing these core shared values, universities can help build more inclusive, fair, and sustainable economies and societies, which could strengthen trust in institutions and promote community cohesion.

This concept of USR is increasingly recognized as a critical aspect of higher education. It reflects the need for institutions to go beyond their core functions of education, research, and community outreach and to actively contribute to the well-being of society and the environment [8]. Nevertheless, there is a critical difficulty in achieving a sustainable university due to a lack of understanding of the social responsibility concept among the university community. For example, a singular definition of USR cannot address the social disparities in countries with different development levels [3]. To assess USR effectively, universities must align student evaluations with evolving societal needs. As USR varies across industries and stakeholders, tailored assessment tools are essential to measure each university’s unique impact. Recognizing this, Taiwan’s Higher Education Sprout Project launched a USR program in 2017, integrating community service learning into curricula. A key goal is to cultivate students as responsible citizens and valuable community contributors [9].

Reflection is essential to medical education and helps students develop critical thinking, problem-solving, and communication skills necessary to promote professional competence [10, 11]. Learning programs in a social context with structured reflection have shown tremendous changes, and the effects are generalized across educational levels [6]. The Health Promotion Literacy-based Scale in USR (HPLS-USR) is the first instrument developed and validated through exploratory factor analysis (EFA). The scale has four dimensions: personal growth, responsibility for citizenship, social interaction, and intellectual development. It is measured using a 22-item instrument to examine medical students’ reflections on a USR curriculum [2]. After preliminary EFA, the next step in developing the instrument is confirmatory factor analysis (CFA) [12], which can evaluate two aspects of construct validity: discriminant validity and convergent validity [13].

The primary aim of this study is to evaluating the construct validity of the Health Promotion Literacy Scale in USR (HPLS-USR) among medical university students in Taiwan using confirmatory factor analysis (CFA).

Methods

Research design and setting

Research Design: A three-phase exploratory sequential mixed-method design was employed. The first phase involved item selection and the development of an initial version of the HPLS-USR. The second phase focused on content validation, while the third phase was dedicated to validating the psychometric performance of the HPLS-USR by EFA. Finally, a cross-sectional study design was employed from September 2020 to January 2021 to further validate the scale through CFA. CFA helps establish construct validity by confirming that the observed variables appropriately measure the underlying latent constructs. This is essential for ensuring that the scale accurately reflects the theoretical constructs it is intended to measure.

Setting and Sample: The study was conducted in a medical university in Central Taiwan. We employed convenience cluster sampling to recruit medical university students participating in USR programs. While this approach allowed for efficient data collection from a large sample (N = 405), it inherently carries certain limitations and biases, particularly in terms of gender distribution. Informed consent was obtained from all subjects, and ethical approval was obtained from Chung Shan Medical University Hospital (Project No. CS2-21149). The sample size was determined using Free Statistics Calculators 4.0, with a minimum required sample of 150 to achieve the desired statistical power of 0.8 and medium effect size. Only data from students who fully completed the surveys were considered for analysis.

Measurement

The Health Promotion Literacy-based Scale in USR (HPLS-USR) was developed by Hung et al. in 2022, the scale were determined by applying the four explicit attitudes factors mentioned earlier and our teaching objectives as indicators of literacy development [2].

Factor 1 (personal growth) included seven items (items 2, 3, 4, 5, 6, 7, and 8) that primarily described the participant’s awareness of self-growth in USR, which accounted for 39.80% of the variance. Five items in Factor 2 (responsibility of citizenship) (items 1, 15, 19, 21, and 22) described participants’ concern for the community because they participated in USR and accounted for 11.50% of the variance.

Factor 3 (social interaction) consisted of five items (items 9, 10, 13, 14, and 16) describing participants’ perceptions of themselves as a result of engaging in interpersonal interactions, which accounted for 5.87% of the variance. Factor 4 (intellectual development) included five items (11, 12, 17, 18, and 20) describing the knowledge and skills acquired in the USR curriculum, which accounted for 4.66% of the variance. The four extracted factors accounted for 61.83% of the total variance in HPLS-USR. Cronbach’s alpha coefficients for each factor were 0.90, 0.79, 0.81, and 0.79, respectively. The four factors are correlated with each other.

Data analysis

Data were analyzed using the Statistical Package for Social Sciences (SPSS version 26). The third phase involved CFA to test the model formally to determine whether the chosen factors are significant. The purpose of this was to verify the model architecture. First, a path diagram of the SEM was drawn, followed by data analysis to confirm the model’s goodness of fit. If the model fits well, further examination of the research hypotheses is conducted.

SEM explores causal relationships between variables in a research model. The maximum likelihood (ML) method was employed to estimate the model parameters. The SEM analysis was done using CFA as the measurement model to test the research framework model and establish a well-fitting measurement model by adjusting the correlations between variables. To choose the model for the research architecture, many fit indices were used, including the likelihood ratio, χ2/df ratio (CMIN/DF), the goodness of fit (GFI), adjusted goodness of fit (AGFI), standardized root mean square residual (SRMR), root mean square error of approximation (RMSEA), normed fit index (NFI), non-normed fit index (NNFI), relative fit index (RFI), incremental fit index (IFI), comparative fit index (CFI), parsimonious normed fit index (PNFI), and parsimonious comparative-fit-index (PCFI).

Results

Sample characteristics

Table 1 shows the characteristics of the participants (N = 405). More than half of the respondents (60.5%) were women. A total of 55.6% were first-year students, 40.7% were second-year students, and 3.7% were third-year students, while 59% of respondents were from the College of Medicine, and 41% were from the College of Health and Management. All respondents were enrolled in the USR curriculum and had experience in service learning.

Table 1 Sociodemographic characteristics of the respondents

CFA of the HPLS-USR scale

Table 2 outlines Cronbach’s α, composite reliability (CR), and average variance extracted (AVE) for the four factors. Factor 1 has a Cronbach’s α of.89, a CR of 0.94, and an AVE of 0.49, accounting for 26.71% of the variance. Factor 2 has a Cronbach’s α of.82, a CR of 0.88, and an AVE of 0.43, accounting for 18.07% of the variance. Factor 3 has a Cronbach’s α of.84, a CR of 0.90, and an AVE of 0.52, accounting for 18.02% of the variance. Factor 4 has a Cronbach’s α of.81, a CR of 0.87, and an AVE of 0.42, accounting for 12.40% of the variance. The overall α is 0.93, and the total variance explained is 63.42%.

Table 2 Maximum likelihood of the CFA model of the HPLS-USR

CFA of the HPLS-USR scale

Specifying multiple models when conducting CFA is useful to demonstrate improved model fit before adopting a final model [12]. We performed a normal theory-based ML analysis using 22 items and the four-factor structure. Model 1 showed results of CMIN = 559.021, CMIN/DF (χ²/df) = 2.767, GFI = 0.89, AGFI = 0.86, SRMR = 0.03, RMSEA = 0.06, NFI = 0.89, NNFI = 0.91, RFI = 0.87, IFI = 0.92, CFI = 0.92, PNFI = 0.77, and PCFI = 0.79 (Table 3). According to the reported fit indices, the GFI, AGFI, NFI, and RFI values did not indicate a well-fitting model.

Table 3 Goodness-of-fit indices of the HPLS-USR

We also examined the factor loading of each item and correlation coefficients between factors. The factor loading of each item was more significant than 0.5, except for item 12. The four factors were also significantly intercorrelated, with correlation coefficients ranging from 0.3 to 0.90 (Fig. 1), indicating a certain level of interrelatedness. This was especially notable between factor 3 and factor 4, which had a correlation coefficient of 0.9. Statistically, the correlation coefficient r = 0.90 is enormous, meaning that the two factors share 81% of the variance [14].

Fig. 1

Hypothesized CFA model of HPLS-USR

We rechecked the items in factor 4, which had a similar concept related to social interaction. For example, item 11, “I focus on community health issues from multiple perspectives,” is similar to item 10, which is part of factor 3. Item 17, “When the residents …have difficulties, I will try to help …”, is similar to item 9. Item 20, “I find ways to contribute to society,” is similar to item 13. Item 18, “In the process of health promotion service, my learning motivation improved,” is similar to item 14. After reviewing the item content and structural evidence support, we deleted the five items of factor 4 and modified the model to have three factors with 17 items.

Finally, we conducted the CFA again using normal theory-based ML for the new model. The modified model (Fig. 2; Table 3) showed results of CMIN = 365.376, CMIN/DF (χ²/df) = 3.149, GFI = 0.91, AGFI = 0.88, SRMR = 0.04, RMSEA = 0.07, NFI = 0.90, NNFI = 0.92, RFI = 0.89, IFI = 0.93, CFI = 0.93, PNFI = 0.78, and PCFI = 0.81. A high GFI suggests that the hypothesized three-factor model adequately represents real-world observations, reinforcing the construct validity of the scale. The AGFI ≥ 0.85 indicates acceptable fit, which means that the scale maintains validity without unnecessary complexity, making it more applicable in real-world assessments. RMSEA ≤ 0.08 indicates acceptable fit, the low RMSEA value suggests minimal unexplained variance, confirming that the scale provides stable and generalizable measurements.

The fit of model 2 was optimal without modification, representing a good fit between the hypothetical model and the collected data. The internal consistency reliability of the HPLS-USR was calculated as 0.91, and the total variance explained was 51.02%.

Fig. 2

Revised 3-factor model of HPLS-USR

Discussion

Models of HPLS-USR

Developing good citizenship through social engagement is an essential learning objective for medical university students participating in USR. In previous research on EFA validation of the HPLS-USR, we extracted 22 items in four factors. However, the results suggested that the validation efforts for the HPLS-USR should be expanded by revising items identified as weak and examining the factor structure with a larger sample size through CFA [2]. In the present study, CFA validation of the scale model was performed with more than 400 participants, which showed slightly different results from the previous EFA. The model fit indices used in this study confirm the validity and applicability of the refined three-factor structure of the HPLS-USR scale. A GFI of 0.91 and AGFI of 0.88 suggest that the proposed model sufficiently explains the observed data while maintaining theoretical parsimony. Additionally, an RMSEA of 0.07 indicates an acceptable level of unexplained variance, suggesting that the model generalizes well beyond the study sample. These findings reinforce the construct validity of the scale and support its practical application in medical education. Given these model fit indices, the HPLS-USR scale can serve as a reliable tool for assessing students’ engagement with USR programs and their health promotion literacy. The strong model fit further suggests that the scale is suitable for use in diverse educational and institutional settings, helping universities evaluate and enhance their social responsibility curricula.

The sample comprised 58.3% female and 41.7% male participants, which does not perfectly reflect the gender ratio in all medical universities. Previous studies indicate that gender may influence perceptions of social responsibility and engagement in health promotion activities [2]. Therefore, the observed gender distribution may have introduced a response bias, where certain attitudes and experiences are more prominently reflected in the findings.

The intellectual development factor was removed due to high collinearity (r = 0.90) with the social interaction factor, indicating substantial shared variance (81%). The conceptual overlap between these factors suggests that intellectual development, as operationalized in this study, is inherently linked to social engagement rather than existing as a distinct construct. This aligns with findings from previous USR research, which emphasize that students’ intellectual growth in social responsibility settings primarily occurs through real-world interactions rather than isolated cognitive processes [2]. This high collinearity reduced the discriminatory value of the factors and could lead to biased path coefficients [15]. To address this issue, we systematically deleted inappropriate items based on the meaning of the questions and statistical data recommendations. As a result, 17 items across three factors—personal growth, responsibility for citizenship, and social interaction—were preserved.

Furthermore, justification for pairing items 16 and 12, based on conceptual alignment, statistical validity, and theoretical underpinnings. First, in the original four-factor structure, Item 16 belonged to the Social Interaction factor, while Item 12 was categorized under Intellectual Development. However, upon re-examining the content and theoretical meaning of these items, we found a strong conceptual overlap between them, supporting their reassignment within the revised three-factor model. This reassignment refer to OECD’s Learning Compass 2030, which highlights that knowledge acquisition in socially engaged learning environments occurs primarily through collaborative reflection and interaction with others, rather than in isolation [8]. Second, in the original four-factor model, Item 12 demonstrated cross-loadings on both Intellectual Development (λ = 0.496) and Social Interaction (λ = 0.695), indicating that it is more strongly associated with Social Interaction. Item 16 also had a strong loading within the Social Interaction factor (λ = 0.678), reinforcing that both items measure a similar underlying construct, finally, the modified three-factor model with Item 12 reassigned to Social Interaction showed improved fit indices. Third, there are several key theoretical perspectives support the reclassification of Item 12 within Social Interaction, including experiential learning theory.

The final three latent variables of HPLS-USR were varying with the definition of USR [3] or four elements of learning outcomes of USR [6]. While the original model included academic, personal, social, and citizenship outcomes, the revised model eliminated the academic dimension (factor 4), which was primarily related to intellectual development. Previous studies have used academic factors to evaluate students’ academic performance, such as knowledge and grade point average [6], but these elements differed from the original HPLS-USR, which focused on cognitive learning, academic growth, and mental inspiration related to health promotion through service-learning. Our study’s RMSEA of 0.07 suggests that the modified three-factor model adequately captures the data structure while balancing model complexity [13]. The CFI value of 0.93 in our study confirms that the proposed model aligns well with the observed data, indicating that the refined three-factor structure maintains theoretical and empirical robustness [14]. The revised three-factor model—personal growth, responsibility for citizenship, and social interaction—better aligns with existing USR theoretical frameworks that emphasize experiential learning and real-world engagement over traditional cognitive development.

These results indicate the scale’s reliability and applicability in measuring health promotion literacy within the USR framework, ensuring that it can be effectively used to assess medical students’ engagement in USR programs. The refined three-factor model enhances the applicability of the HPLS-USR scale in medical education by prioritizing competencies that are directly relevant to future healthcare professionals.

Reflection on the USR curriculum

Reflection is essential to a USR curriculum, especially for medical university students as future healthcare providers. Medical university students must face suffering patients in their future social responsibility practice. Participating in the USR curriculum is an excellent opportunity for students to experience learning by doing. At the same time, students could reflect on whether they have sufficient resilience to cope with the setbacks and obstacles that may be encountered in the service delivery process.

We thoroughly examined the process of deleting items from the scale one by one and discovered that some items, such as “solve difficulties” and “spend more time,” implied the trait of resilience. This insight suggests that resilience is a crucial component in students’ engagement with USR curricula, reflecting their ability to adapt and persist in challenging social practice environments. The first factor in the final model, personal growth, is based on students’ learning experiences related to USR health promotion. This factor captures their self-awareness and development in addressing health issues within the community context. The second factor, responsibility of citizenship, consists of items that concern students’ sensitivity to the community’s health needs and values of social justice. This dimension emphasizes the importance of cultivating a sense of civic duty and ethical responsibility among students, which is vital for fostering socially responsible healthcare professionals. The third factor, social interaction, highlights participation in public affairs and respect for diversity. The revised model aligns well with existing evaluations of students’ professional competence in problem-solving, communication, and critical thinking skills [10, 11]. It also supports the development of productive and responsible citizens who positively impact their communities [9] and reinforces students’ intentions to collaborate and use group work in conventional university contexts [5]. However, the role of resilience in medical university students’ engagement with USR programs requires further investigation. Understanding how resilience interacts with the identified factors and influences students’ ability to navigate the complexities of social practice is essential for developing comprehensive USR curricula that fully prepare students for their future roles as socially responsible healthcare providers. Future research should explore this dimension more deeply to provide a more nuanced understanding of students’ resilience and its impact on their learning and development within USR contexts.

As future frontline healthcare providers, medical students benefit significantly from USR participation, developing essential competencies in patient-centered care, interdisciplinary collaboration, and community-based healthcare delivery [1]. Research has shown that students actively engaged in social responsibility projects exhibit greater empathy, ethical reasoning, and a stronger commitment to addressing social determinants of health [8]. Assessing their Health Promotion Literacy (HPL) within the USR framework is therefore critical to preparing them to serve diverse populations effectively.

However, traditional medical education remains predominantly hospital-based and treatment-focused, often neglecting broader societal responsibilities. This gap has been acknowledged by global medical education bodies, including the World Health Organization (WHO) and the Association of American Medical Colleges (AAMC), which advocate for integrating social responsibility into medical curricula. A major challenge is the limited direct exposure of medical students to underserved communities, restricting their understanding of healthcare inequities [2]. Integrating USR programs into medical education provides hands-on learning experiences, fostering problem-solving skills, cultural competence, and community-based healthcare approaches. This study offers initial construct validity evidence for the HPLS-USR scale within the USR framework, helping to bridge this critical gap in medical education.

Clinical and educational relevance of the HPLS-USR scale

The Health Promotion Literacy-based Scale in University Social Responsibility (HPLS-USR) has significant implications for medical education and clinical practice. Traditional medical education primarily emphasizes hospital-centered, treatment-focused training, often neglecting broader social determinants of health and community engagement. By integrating USR-based curricula, medical students gain real-world exposure to public health challenges, enhancing their ability to address health disparities and promote health equity. The HPLS-USR scale serves as a valuable assessment tool for evaluating how well students internalize and apply health promotion literacy in USR settings. The three-factor structure—personal growth, responsibility for citizenship, and social interaction—aligns with competency-based medical education models, fostering critical thinking, ethical reasoning, and interdisciplinary collaboration. From an educational perspective, this scale provides institutions with a framework for refining USR curricula, ensuring students develop the skills necessary to become socially responsible healthcare professionals. Future research should explore longitudinal assessments to examine whether USR participation translates into long-term engagement in community healthcare and advocacy for health equity.

Conclusion

Conclusion

The Health Promotion Literacy-based Scale in USR (HPLS-USR) developed and this study offers preliminary construct validity evidence for the HPLS-USR scale using factor analysis, confirming a three-factor structure that is both theoretically and empirically supported. The structural equation model demonstrated a good fit, indicating that the three latent variables—personal growth, responsibility for citizenship, and social interaction—are robust measures of students’ reflections on a USR curriculum.

Key strengths of this study include a comprehensive validation process, employing a rigorous three-phase exploratory sequential mixed-method design, including item selection, exploratory factor analysis (EFA), and confirmatory factor analysis (CFA). The study also benefited from a large and diverse sample size of 405 students from different academic years and disciplines, achieving the desired statistical power, which enhances the generalizability of the findings. Additionally, the scale exhibited excellent psychometric properties, with high internal consistency and good to excellent model fit indices. Moreover, the scale’s focus on personal growth, responsibility for citizenship, and social interaction aligns well with the essential competencies required for future healthcare providers, emphasizing the importance of social responsibility in medical education. This highlights the relevance and necessity of incorporating USR into medical curricula to foster a sense of social responsibility among students.

Limitations

This study had limitations, and caution must be exercised in interpreting the data. First, this study employed convenience cluster sampling, which, while practical for data collection in a university setting, presents certain limitations. The sample exhibited a gender distribution of 58.3% female and 41.7% male, which may introduce response biases in attitudes toward USR engagement. Prior research suggests that gender differences may influence perceptions of social responsibility, with female students often exhibiting higher engagement in community-based health initiatives [2]. Additionally, clustering based on program enrollment may have limited participant diversity in terms of socioeconomic background or prior experience in service-learning. Future studies should aim to enhance sample representativeness by employing stratified random sampling to ensure balanced demographic distribution, quota sampling for gender representation, and multi-site data collection to capture broader variability in student experiences. These improvements would strengthen the generalizability of findings and provide a more comprehensive understanding of medical students’ engagement with USR programs.

Second, this scale is specifically designed for USR curricula related to health needs and may not fully capture the breadth of all USR courses. Convenience sampling inherently favors students who are more accessible and willing to participate, potentially excluding those with varying academic schedules, personal interests, or prior engagement in social responsibility initiatives. Furthermore, the clustering method—based on class or program enrollment—may limit sample diversity in terms of socioeconomic background, prior community involvement, or educational aspirations. Given the diverse composition of society and the unique histories and missions of each university, community needs and institutional responsibilities vary, which may influence the generalizability of the findings.

Suggestion

To enhance the applicability of the HPLS-USR scale across diverse educational and cultural settings, future research should focus on cross-cultural validation to determine whether the three-factor structure remains stable in different regions. Given that USR implementation varies between government-driven models (e.g., Taiwan) and institution-led approaches (e.g., Western contexts), comparative studies can help refine the scale for broader applicability.

Additionally, longitudinal research is needed to assess how USR participation influences students’ long-term professional behaviors, such as community healthcare engagement and advocacy for health equity. Further studies should also explore institutional factors, including faculty involvement and policy frameworks, that shape USR integration in medical education. These insights will strengthen the scale’s predictive validity and support evidence-based enhancements to USR curricula worldwide. Finally, future studies should explore the potential cultural variations in USR literacy, particularly in different educational systems and professional training contexts. Additionally, longitudinal studies could examine whether the exclusion of the intellectual development factor affects the long-term predictive validity of the scale in assessing medical students’ professional growth and community engagement.