Measurement Invariance of Self-Regulated Learning Perception Scale in Medical Students

Introduction

Self-regulation is defined as an individual’s ability to control their actions (behaviors), thoughts (cognition), and surroundings (environment) [1]. Theories explaining self-regulation focus on how and why individuals exert effort to achieve their goals. It appears to be a key to life success when individuals internalize self-regulation as part of their characteristics [1]. Self-regulated learning (SRL) refers to self-regulation in the context of learning and is also known as the ability to “learn how to learn” [2]. It originates from social cognitive theory and is related to motivation, metacognition, and the environment [3]. SRL also posits that individuals are responsible for their own learning [4]. Additionally, social cognitive theory emphasizes the role of the interaction between the individual and the environment in learning. As a result of this interaction, learning occurs through (i) observing others’ experiences, (ii) evaluating these observations, and (iii) internalizing the evaluations [4].

Individuals who are competent in SRL skills are characterized by setting goals, planning and replanning when needed, using strategies, monitoring the process, and evaluating the results for specific tasks [5]. There are 3 main phases in SRL: before (forethought), during (performance), and after (self-reflection) [2]. The forethought phase involves preparation for the performance phase. In this phase, individuals set goals, plans actions, decide on strategies to use, set evaluation criteria for achievement, and motivate themselves. The performance phase is the action and monitoring phase, where individuals act according to the forethought phase and monitor their progress. In the self-reflection phase, individuals reflect on the process and evaluate whether their plans in the forethought phase contributed effectively to the performance phase [2].

A competent self-regulated learner has the ability to evaluate the next steps needed to achieve targeted goals [6]. Self-regulated learning (SRL) is one of the core competencies expected of medical students to maintain lifelong learning skills [2]. Also, high levels of SRL skills are predictive of high academic achievement in medical students [7–10]. Self-regulated learning ability/skill begins to develop with the tasks given in preclinical years and contributes to learning most in clinical years [9]. Studies show that SRL levels are higher in clinical years as a result of the learnings in preclinical years [11]. So, it is the responsibility of medical schools to develop and improve students’ SRL abilities. Since SRL is a skill that can be improved, it is crucial to have a program or module that teaches how to improve SRL skills in medical schools [12]. In medical education, measuring students’ self-regulated learning (SRL) levels is as critical as implementing a program/module designed to serve this purpose. At the outset, during the needs assessment phase of program development, utilizing an appropriate assessment tool to determine students’ perception levels is essential for obtaining valid and reliable results. In the literature, there are many tools used to assess individuals’ SRL levels, such as the Motivated Strategies for Learning Questionnaire [13,14], the Online Self-Regulated Learning Scale [15,16], the Academic Self-Regulated Learning Scale [17], and the Self-Regulated Learning Perception Scale [12]. Among these tools, the Self-Regulated Learning Perception Scale was developed specifically for medical students. The Self-Regulated Learning Perception Scale (SRLPS) is the preferred scale for assessing students’ SRL levels in Türkiye [18]. The SRLPS has been developed in the Turkish language for medical students in Türkiye in 2009 by Turan [19]. Because of its robust theoretical background (using Zimmermans’s conceptual framework), its comprehensive dimensions, and fewer items (compared to other scales, it has fewer items and is therefore easier to complete), the scale is widely used [18]. It has been adapted to different languages in different countries like Brazil [20], Pakistan [21], Portugal [22], and China [23] in recent years. The SRLPS contains 4 factors: motivation and action to learning, planning and goal setting, strategies for learning and assessment, and lack of self-directedness.

To evaluate the characteristics of SRLPS in more detail, the validity of SRLPS should be well discussed. In the process of gathering evidence for the validity of a measurement instrument, examining its measurement invariance is essential. Measurement invariance means that a scale measures its targeted constructs equally across different subgroups, independent of subgroup characteristics [24,25]. It contributes to validity and reliability by providing evidence of being error-free to varying degrees [26]. If measurement invariance cannot be ensured, it becomes impossible to determine whether observed differences between groups stem from actual differences or from different psychometric responses to scale items, rendering the results uninterpretable [27]. In all versions of the SRLPS, the analysis to determine differences between genders was performed. The differences according to the years were analyzed in the original SRLPS in Türkiye, the Portuguese version in Brazil, and the English version in Pakistan. Unfortunately, none of them analyzed any step of the measurement invariance before comparing subgroups [19–23]. Therefore, before using the SRLPS to compare preclinical and clinical year medical students’ SRL levels, measurement invariance should be analyzed in addition to standard validity and reliability analyses. In other words, measurement invariance is crucial for making valid group comparisons, and ensuring fairness and equity in selection processes based on the measure [28].

Measurement invariance comprises configural, metric, scalar, and strict invariance, which have a hierarchical structure [29]. Configural invariance is the least stringent step, showing that the model form of the scale is invariant. In this step, the constructs of the scale are tested to determine if they display the same patterns across subgroups. In other words, configural invariance means that the factors of the scale are supported in subgroups [29]. Metric invariance follows configural invariance and is moderately stringent. It indicates that each item contributes to the latent construct similarly across groups. In this step, factor loadings are tested for equivalence between subgroups. After establishing metric invariance, scalar invariance is examined. Scalar invariance ensures that the mean differences of the latent constructs capture all mean differences in the shared variance of the items, indicating that item intercepts are equivalent across subgroups [29]. Finally, strict invariance requires the equality of item residual variances between groups in addition to scalar invariance. Establishing strict invariance facilitates the comparison of both observed variable means and the variance and covariance of factors across different groups [30]. However, as the variance of the latent variable increases, the residual variance of the items also increases, making strict invariance often unattainable in practice [31].

We conducted a literature search to investigate studies related to SRL and measurement invariance. The search was conducted with the keywords “self-regulated” OR “self-regulation” AND “measurement invariance”. The search included databases such as Medline, SCOPUS, Web of Science, Cochrane Library, CINAHL Plus, PubMed, BMJ journals collection, ERIC, and the British Educational Index. A total of 14 relevant articles were found, but none focused on medical students. Nine of these studies examined measurement invariance in SRL scales in adults (eg, college students, university students, language students) [32–40], while 5 studies focused on children [41–45]. All 14 studies researched whether the scales exhibited measurement invariance for gender comparisons. Only a few investigated measurement invariance for a second subgroup, such as educational level [40] and ethnicity [43]. The SRLPS is considered important for investigating measurement invariance because it is the most widely used scale in Türkiye for assessing SRL and the only one specifically designed for medical students. Additionally, due to its robust theoretical background, it has been adapted for use in various cultures. Due to the importance of the scale and the fact that its measurement invariance has not yet been examined, this study aimed to determine whether the scale satisfies measurement invariance across genders and preclinical-clinical years of medical students.

Material and Methods

ETHICS APPROVAL:

This study was reviewed and approved by the institutional Health Sciences Research Ethics Committee (Reference Number: E-35853172-900-00002203263). The study was performed in accordance with the ethical standards of the Declaration of Helsinki. Moreover, permission was also granted by the faculty dean’s office to collect data just after the lecture sessions. Participants were provided with information regarding the study’s purpose and were assured that the data would solely be used for scientific purposes and would not be disclosed to any other person or organization. Informed consent forms were signed by all participants.

STUDY DESIGN AND SETTING:

This cross-sectional study collected data using paper-based questionnaires distributed at Hacettepe University, Faculty of Medicine (HUFM) in Ankara, Türkiye. HUFM, which employs an integrated curriculum, is one of the premier medical schools in Türkiye and an important part of Hacettepe University. The duration of HUFM’s medical education program is 6 years, and the program has been structured as preclinical (years 1–3), clinical (years 4–5), and an internship (year 6). The preclinical years emphasize basic sciences and clinical skills to prepare students for the clinical years, which focus on disease management and applying knowledge gained during preclinical training. The internship continues to focus on disease management, providing students with opportunities to enhance their competencies through direct responsibility for patient care.

PARTICIPANTS:

This study was conducted with all medical students from the first year to the sixth year. In the academic year 2022–2023 at HUFM, there were 1360 and 1278 students in preclinical years and clinical years, respectively. The students were informed about the study and asked to give their consent. Of the total sample, 60.6% (n=532) preclinical years students and 39.4% (n=346) clinical years students participated in the study (total, 943 students). The participation rate was 39.1% for preclinical years and 27% for clinical years. Due to missing data and outliers, the final data analysis included 878 students.

INSTRUMENTS:

In this study, a measurement tool consisting of 2 parts was used. The first part comprised demographic questions, including participants’ year in medical school, gender, and age. The second part consisted of items from the Self-Regulated Learning Perception Scale (SRLPS). The Self-Regulated Learning Perception Scale was originally developed in Turkish to assess self-regulated learning (SRL) levels among medical students in Türkiye [19]. The scale’s development study was conducted with first-year, second-year, and third-year medical students, with a total of 804 participants. The ratio of female to male students was 43.5% to 46.5%, respectively. Findings supporting the reliability and validity of the SRLPS scores have been reported. Expert opinions were obtained to determine the scale’s content validity, while exploratory factor analysis (EFA) was performed to assess the construct validity of the scale’s measures. The EFA yielded a construct that consisted of 41 items and 4 factors that explained 47.10% of the total variance. The reliability of the measures obtained using the SRLPS was examined via Cronbach’s alpha and the coefficients for the total scale and each factor were reported as 0.91, 0.88, 0.91, 0.83, and 0.76, respectively. The minimum score that can be obtained from the scale is 41, while the maximum score is 205. The scale uses a 5-point Likert rating, ranging from Strongly Disagree to Strongly Agree [19]. The first factor, Motivation and Action to Learning, includes 7 items and focuses on students’ willingness to engage in and proactively pursue learning. The second factor, Planning and Goal Setting, consists of 8 items and measures students’ ability to formulate goals, set objectives, and plan their learning accordingly. The third factor, Strategies for Learning and Assessment, contains 19 items and evaluates whether students select and implement appropriate learning strategies to achieve their objectives and assess their learning outcomes. The final factor, Lack of Self-Directedness, includes 7 negatively worded items that address issues students face in directing their learning process [19]. These dimensions encompass the stages theoretically defined for self-regulated learning by Zimmerman (1998) and Pintrich (2000) [19].

DATA COLLECTION:

The data collection process was carried out in paper-pencil form and on a voluntary basis, between March and April 2023. The data were collected after the lectures to reach a high number of students. The data were collected from those who volunteered and completed the consent form.

STATISTICAL ANALYSIS:

In the scope of the study, missing data rates were examined for each item, revealing rates ranging from 0.1% to 0.8%. Since these rates were less than 5% [46], indicating that the missing data were likely random, the analysis proceeded using the listwise deletion method to handle missing data. Following this, the assumptions of the statistical analyses were checked. Multivariate outliers were assessed by calculating Mahalanobis distances from the remaining 892 data points. As a result, 14 cases identified as outliers were excluded from the analysis, leaving data from 878 students. To assess the multivariate normality of these data, Mardia’s skewness and kurtosis test [47–50] was used. The results of this two-sided multivariate skewness and kurtosis test indicated p<0.01, suggesting a violation of multivariate normality.

When examining multicollinearity, it was determined that there were no issues, as the Tolerance values were greater than 0.10 and the VIF values were less than 10 [46]. Additionally, the highest inter-item correlation was found to be 0.76. Since this value is below 0.80, it indicates that there is no multicollinearity problem among the items [51]. Finally, scatter plots of standardized residual variances versus standardized predicted values were examined to assess homogeneity of residual variances. The plots showed that residual variances were homogeneously distributed, confirming that the assumption of homoscedasticity was met [46].

After examining the assumptions, the basic model was analyzed through confirmatory factor analysis (CFA) for both the entire group and subgroups (male-female, preclinical-clinical years). The CFA results included standardized factor loadings, the significance of t-values related to factor loadings, and standardized error variances for the entire group. Additionally, model-data fit was evaluated for subgroups by examining fit indices such as χ2/df, RMSEA, CFI, TLI, and SRMR. The obtained fit indices were compared with criterion values to interpret model-data fit. The criteria were as follows: χ2/df ≤2.00; RMSEA ≤0.05; CFI ≥0.95; TLI ≥0.95; SRMR ≤0.05 indicated a perfect fit; RMSEA ≤0.08; CFI ≥0.90; TLI ≥0.90; SRMR ≤0.08 indicated a good fit; and χ2/df ≤5.00; CFI ≥0.85; TLI ≥0.85; SRMR ≤0.10 indicated an acceptable/moderate fit [46,52–55].

In examining measurement invariance, multigroup confirmatory factor analysis (MGCFA) was conducted. The hypothesis tests proposed by Meredith (1993), which are considered in 4 stages, were used [56]. The basic models and all stages of measurement invariance (configural, metric, partial metric, scalar, and strict) were analyzed using MPlus 8.3 [50]. Due to the violation of multivariate normality, the maximum likelihood mean adjusted method (MLM), an estimation method that is robust to non-normality, was employed for examining measurement invariance and testing the basic model. To assess differences between models at various stages of measurement invariance, ΔSB-χ2 and ΔCFI values were reported. A significant ΔSB-χ2 and ΔCFI ≥0.01 indicate that the respective stage of invariance is not met [27]. Additionally, in the metric invariance stage, the items that cause differences were identified. A partial metric invariance model was established by allowing various estimations of these items across groups. For this partial invariance model, the most restricted possible invariance model between the 2 groups was considered in line with Byrne’s suggestion [57].

Results

DESCRIPTIVE STATISTICS FOR PARTICIPANTS’ INFORMATION:

Since SRLPS score comparisons are frequently made based on gender and preclinical/clinical years, the participants’ demographic information is also provided with respect to these 2 variables. The ages of the students ranged from 18 to 26 years (X̄=21.08, SD=1.28). The mean ages and standard deviations of the participants, categorized by year and gender, are presented in Table 1. The ages of preclinical and clinical year students, as well as those of male and female participants, were very similar.

Preclinical year students attend lectures in large amphitheaters and have high attendance rates, making it easier to reach this group and, consequently, to collect data from them. In contrast, clinical year students typically study in smaller groups and are dispersed across various clinical rotations, which made accessing this group comparatively challenging. As a result, 60.6% of the participants were preclinical year students, while 39.4% were clinical year students. Regarding the distribution by gender, the male-to-female ratio was nearly equal.

DESCRIPTIVE STATISTICS FOR THE SRLPS FACTOR AND TOTAL SCORES:

The means and standard deviations for each factor score and the total scale, calculated by averaging the corresponding items, are presented in Table 2.

Data in Table 2 show that all factors, except for the Motivation and Action to Learning factor, as well as the total scale score, were higher among female students than male students. Additionally, all factors and the total scale score were relatively higher among preclinical year students. An analysis of the scale averages for the entire group revealed that the lowest mean score was in the Lack of Self-Directedness factor (3.03), while the highest mean score was in the Motivation and Action to Learning factor.

RELIABILITY:

In this study, the stratified alpha for the total scale was 0.94, and McDonald’s omega coefficient for the total scale was 0.96. The Cronbach’s alpha coefficients for the factors were 0.88 for Motivation and Action to Learning, 0.88 for Planning and Goal Setting, 0.91 for Strategies for Learning and Assessment, and.83 for Lack of Self-Directedness. McDonald’s omega coefficients for these factors were 0.82, 0.88, 0.91, and 0.83, respectively. The estimated reliability coefficients for all 4 factors and the whole scale were acceptable.

CONFIRMATION OF THE FACTOR STRUCTURE OF THE SRLPS:

To gather evidence supporting the validity of the SRLPS, we conducted confirmatory factor analysis (CFA) to evaluate the fit of the data to the 4-factor structure, utilizing responses from 878 participants in the study. There were 4 modifications to the basic model. These modifications involve pathways indicating correlated error variances between specific items: items 1 and 2 in the Motivation and Action to Learning dimension, items 10 and 11 in the Planning and Goal Setting dimension, items 40 and 41 in the Strategies for Learning and Assessment dimension, and items 7 and 39 in the Lack of Self-Directedness dimension. Since these items are within the same factor and have similar content, it is theoretically reasonable for their error variances to be correlated. The significance of the factor loadings, along with standardized factor loadings and standardized error variances, are detailed in Table 3.

Table 3 shows that the t-values for all items’ factor loadings in the model were significant. For the dimension of Motivation and Action to Learning, factor loadings ranged from 0.51 to 0.70, with residual variances ranging from 0.53 to 0.74. In the Planning and Goal Setting dimension, factor loadings ranged from 0.62 to 0.79, and residual variances ranged from 0.37 to 0.61. For the Strategies for Learning and Assessment dimension, standardized factor loadings varied between 0.33 and 0.69, with residual variances ranging from 0.52 to 0.69. Finally, in the Lack of Self-Directedness dimension, standardized factor loadings ranged from 0.38 to 0.79, and residual variances range from 0.37 to 0.86. Notably, no item in any of the dimensions has a standardized factor loading less than 0.30 or a residual variance greater than 0.90. Thus, there are no items in the model that need to be removed.

In addition to the factor loadings and residual variances, the overall fit indices supported the 4-factor structure of the SRLPS: χ²(769)=1704.364, CFI=0.921, TLI=0.916, RMSEA=0.037 [90% CI 0.035–0.040], and SRMR=0.044, demonstrating that the model provides an adequate representation of the observed data.

FINDINGS ON MEASUREMENT INVARIANCE BY GENDER:

In this study, data from the Self-Regulated Learning Scale were subjected to confirmatory factor analysis (CFA) separately for female and male students before assessing measurement invariance by gender. The baseline model was tested through CFA, and the fit indices for this model are presented in Table 4.

Upon reviewing Table 4, it is noted that the basic model shows an excellent fit for both female and male groups, as indicated by χ2/df ≤2 and RMSEA ≤0.05. For the female group, an SRMR ≤0.05 signifies an excellent fit, while an SRMR ≤0.08 for the male group indicates a good fit. CFI values ≥0.90 suggest a good fit for both groups, with a TLI value of 0.89 for the female group indicating an acceptable fit and ≥0.90 for the male group indicating a good fit. Overall, the model fits well for both groups, with comparable fit indices between them. Findings related to measurement invariance for both groups are presented in Table 5.

Examining Table 5, it is observed that the fit indices for configural invariance in the first stage indicate an excellent fit, with RMSEA ≤0.05. Additionally, SRMR ≤0.08 and CFI and TLI values ≥0.90 suggest a good fit for the configural model. Similarly, during the metric invariance stage, the fit indices also indicate a good fit. The Satorra-Bentler χ2 difference test revealed that ΔSB-χ2=40.849 was non-significant, and ΔCFI ≤0.01, confirming that metric invariance has been achieved.

Upon examining the fit indices for scalar invariance, RMSEA ≤0.05 and χ²/df ≤2 indicate an excellent fit, while SRMR ≤0.08, CFI ≥0.90, and TLI ≥0.90 suggest a good fit. Overall, the scalar invariance model demonstrates a good level of fit. When comparing scalar invariance to metric invariance, ΔSB-χ²=49.685 was found to be non-significant, and ΔCFI ≤0.01, indicating that scalar invariance is achieved. However, despite the good fit for strict invariance (RMSEA ≤0.05, χ²/df ≤2, TLI and CFI ≥0.90, SRMR ≤0.08), the comparison of strict invariance with scalar invariance showed that ΔSB-χ²=204.630 was significant and ΔCFI ≥0.01, suggesting that strict invariance was not achieved.

FINDINGS ON MEASUREMENT INVARIANCE BY PRECLINICAL AND CLINICAL YEARS:

In this research, the measurement invariance of the “Self-Regulated Learning Perception Scale” was also examined across preclinical and clinical years. Before assessing measurement invariance, the baseline model was tested using CFA on the data obtained from both preclinical and clinical year students. The fit indices for the baseline model are presented in Table 6.

Examining Table 6 reveals that the baseline model indicates an excellent fit for both the preclinical and clinical year groups, with χ2/df ≤2 and RMSEA ≤0.05. For both groups, SRMR ≤0.08 indicates a good fit. The CFI and TLI values for the preclinical year group, being ≥0.90, also suggest a good fit. However, for the clinical year group, CFI and TLI values of ≥0.85 indicate an acceptable fit. Overall, the model shows good fit for both groups, though the fit indices differ between them. Findings related to measurement invariance for the preclinical and clinical year groups are presented in Table 7.

Table 7 shows that the fit indices for the initial stage of measurement invariance, assessing configural invariance, indicate a good model fit, as evidenced by RMSEA ≤0.05, SRMR ≤0.08, and CFI and TLI values ≥0.90. In the stage of metric invariance, RMSEA ≤0.05 also suggests an excellent fit, while the other indices indicate a good fit. However, despite the ΔCFI value being ≤0.01, the Satorra-Bentler chi-square difference test shows a significant ΔSB-χ2=119.914, indicating that metric invariance was not achieved. As metric invariance is a prerequisite for assessing scalar invariance, and scalar invariance must be established before evaluating strict invariance, the subsequent stages of the model were not examined. Instead, a partial invariance model was established. In this model, factor loadings for items 2, 11, 13, 14, 15, and 18 were allowed to vary. In the stage of partial metric invariance, RMSEA ≤0.05 suggests an excellent fit, while the other indices indicate a good fit. When comparing scalar invariance to partial metric invariance, ΔSB-χ2=40.688 was found to be non-significant, and ΔCFI ≤0.01, indicating that partial metric invariance was achieved.

Discussion

We assessed data from 804 medical students (43.5% female and 46.5% male). When the measurement invariance of the SRLPS was examined according to gender, it was seen that configural, metric and scalar invariance was provided, but strict invariance was not provided. The provision of configural, metric and scalar invariance shows that the Turkish version of the SRLPS is an adequate tool for comparing gender differences. Of the medical students to whom the scale was applied, 60.6% were in the preclinical year and 39.4% were in the clinical year. Our findings show that the Turkish version of the SRLPS is not appropriate to compare differences between preclinical and clinical year students. Although configural invariance was provided, metric invariance was not even partially provided.

This study reports on the measurement invariance of the Self-Regulated Learning Perception Scale (SRLPS) across genders and preclinical-clinical year Turkish medical students. Self-regulated learning (SRL) is essential for developing competent physicians, and medical schools strive to enhance SRL among their students [2]. To create an effective SRL program, it is important to evaluate the characteristics and SRL levels of students. Using a scale which produces valid and reliable scores is the most suitable method for assessing SRL levels in large groups. During the scale selection process, it is crucial to evaluate its measurement invariance to ensure it is appropriate for comparing different subgroups.

Comparing SRL levels between female and male participants is a common demographic analysis, although results can vary [19,21,23]. Some studies found no gender differences in SRL levels [21,58,59], while others report higher SRL levels among male students compared to female students [23,60]. None of these studies analyzed measurement invariance before comparing differences between genders [58–60]. This variability underscores the importance of establishing measurement invariance for gender subgroups. To make meaningful comparisons between groups, achieving at least scalar invariance is essential. Scalar invariance ensures that the scores obtained from different groups have similar starting points and equal intervals [61]. Our study demonstrated that the SRLPS achieved not only configural, metric, and scalar invariance, but also strict invariance across genders. Therefore, the SRLPS is validated as a reliable tool for comparing SRL levels between gender subgroups.

The structure of medical education varies significantly between preclinical and clinical years, particularly in terms of learning activities, environments, and the characteristics of medical teachers [62]. Preclinical years typically focus on theoretical knowledge, basic sciences, problem-based learning, and basic clinical skills training. In contrast, clinical years emphasize practical experience, clinical sciences, patient interaction, case-based learning, and interprofessional education [63]. Research indicates that SRL levels are generally higher in clinical years compared to preclinical years, a difference often attributes to the more hands-on nature of clinical environments [64–66]. However, in our study, the SRL levels of preclinical year students were found to be significantly higher than those of clinical year students, both in terms of factor scores and total scale scores. This outcome may be attributed to the 2-hour session on self-regulated learning strategies provided to students at the beginning of their time at the faculty.

Our study found that while the 4-factor structure of the SRLPS remains consistent across both preclinical and clinical years, metric invariance was not achieved. The partial metric invariance model established due to this result showed that some items (items 2, 11, 13, 14, 15, and 18) were not invariant across groups. This suggests that the relationship between the items and the latent variables of the SRLPS differs between preclinical and clinical year students. However, metric invariance was achieved with the remaining 35 items and the relationship between these items and the latent variables did not vary across groups. According to Byrne and Watkins, failure to achieve metric invariance can indicate item bias, meaning that the scale may be biased towards one group or that items may hold different meanings for individuals in different contexts [57]. The lack of scalar invariance further suggests that preclinical and clinical year students might interpret some items differently. Consequently, these findings imply that comparing SRLPS scores between preclinical and clinical year students is not appropriate due to these measurement inconsistencies.

A limitation of this study is that we evaluated only 1 SRL scale, the Self-Regulated Learning Perception Scale (SRLPS). It is possible that other SRL scales might demonstrate measurement invariance in comparisons between preclinical and clinical years. Future research should explore the measurement invariance of alternative SRL scales to provide a more comprehensive understanding of SRL across different stages of medical education. The number of participants can be considered as another limitation of the study, but the number of participants (878) was sufficient to perform validity and reliability analysis of the SRLPS according to the literature. The participants were included in the study from only 1 medical school, but it is an accredited medical school, so it can be considered as a typical medical school in Türkiye.

Conclusions

In conclusion, the measurement invariance should be evaluated before using a scale for subgroup comparison. Our study was conducted with medical students to evaluate measurement invariance in an SRLP scale for gender and preclinical/clinical year students. Our findings show that the 41-item SRLPS is an appropriate tool for gender comparisons but not appropriate for preclinical and clinical year comparisons in medical students. Therefore, our study is important not only because it focused on medical students, but also because it included measurement invariance by gender and preclinical-clinical years of medical education. It is anticipated that the study findings will contribute to the existing literature. We hope our study enlightens medical educators on the adequate use of scales for subgroup comparisons.