Validation of a short form of the German form of the classroom community scale (CCS-D-SF)
Harald Knof 1,2Victoria Huber 1,2
Thomas Shiozawa 1,2
1 Eberhard Karls University of Tuebingen, Faculty of Medicine, Department of Anatomy, Institute of Clinical Anatomy and Cell Analysis, Tuebingen, Germany
2 Eberhard Karls University of Tuebingen, Faculty of Medicine, TIME – Tuebingen Institute for Medical Education, Tuebingen, Germany
Abstract
Background: Collaborative learning in medical education relies on both social integration and cognitive engagement. The classroom community scale (CCS) measures students’ sense of community through two dimensions: Connectedness and learning. In German-speaking countries, the CCS has previously been validated in its full form (CCS-D). While an English short form (CCS-SF) exists, a validated German short form for efficient assessment was lacking. This study aimed to develop and validate a short form of the German CCS (CCS-D-SF).
Methods: The CCS-D was systematically reduced to a ten-item short form using principal component analysis (PCA) and iterative reliability analysis. Data on the short form were collected from N=171 first-semester medical students at the University of Tuebingen. Internal consistency was evaluated with Cronbach’s alpha. Construct validity was assessed using confirmatory factor analysis (CFA) and PCA.
Results: The CCS-D-SF retained a two-factor structure corresponding to connectedness (i=5) and learning (i=5). The total scale showed good internal consistency (α=.81), with reliability for the subscales connectedness (α=.86) and learning (α=.65). CFA indicated an acceptable model fit (χ2/df=2.16, RMSEA=.083, SRMR=.059, CFI=.93, TLI=.91), and PCA displayed also the factor structure.
Discussion: The CCS-D-SF is a valid, and practical instrument for assessing sense of community among German-speaking first-year medical students. Reliability ranges from marginal for learning to very good for connectedness. The overall reliability can be considered to be good. Questionnaires brevity reduces respondent burden without compromising psychometric quality and makes it suitable for educational research and classroom diagnostics. The CCS-D-SF may also inform interventions aiming to strengthen both social integration and learning engagement in collaborative medical education settings.
Keywords
medical education, interdisciplinary placement, confirmatory factor analysis, validation, classroom community
1. Introduction
Collaborative learning has increasingly gained importance in medical education as both a teaching and learning format. Such approaches promote teamwork, communication skills, and responsibility for individual as well as group performance - competencies that are essential for professional practice and the quality of healthcare delivery [1], [2]. In the following, the concepts of collaborative learning and sense of community are first elaborated, as they represent key theoretical foundations of the classroom community scale, a widely used instrument for measuring the sense of community in learning environments [3]. A more detailed conceptual framework has previously been provided by Knof and Shiozawa [4]. Subsequently, the rationale for the development of a short form of this scale is presented.
1.1. Collaborative learning in medical education
Collaborative and team-based learning approaches are increasingly used in higher education and medical education and can be implemented face-to-face, synchronously or asynchronously online [5], [6], [7].
Despite its benefits in fostering teamwork and communication, collaborative learning presents challenges. Participation may be unequal, communication ineffective, and interpersonal conflicts can arise, which can limit learning outcomes [8], [9], [10]. One key factor supporting effective collaboration is the degree to which students feel integrated into their learning environment. This construct, referred to as sense of community, is significantly associated with engagement, collaboration, and academic achievement in both face-to-face and online settings [11], [12], [13], [14], [15]. Weak classroom community is linked to disengagement and dropout risk, highlighting the importance of fostering trust, collaboration, and shared commitment [16], [17], [18].
1.2. Sense of community
A widely cited concept of sense of community by McMillan and Chavis describes it as “a feeling that members have of belonging, that members matter to one another and to the group, and a shared faith that members’ needs will be met through their commitment to be together” ([19], p.9). Essential elements include trust, mutual support, interactivity, shared values, and group cohesion [20], [21], [22], while the specific manifestation of this construct may vary depending on the context in which it occurs [23], [24], [25].
Classroom community comprises two main components: Connectedness measures feelings of interpersonal connection, while learning assesses perceptions of interaction and engagement within the classroom [21].
To measure classroom community, Rovai developed the Classroom Community Scale (CCS) based on this framework [3].
1.3. Classroom community scale
The classroom community scale (CCS) consists of 20 items across two subscales: Connectedness (10 odd-numbered items) and learning (10 even-numbered items) (see attachment 1 [Att. 1], table S1) [3]. Participants respond on a 5-point Likert scale from strongly disagree to strongly agree. Total scores range from 20 to 80, with each subscale ranging from 10 to 40, where higher scores indicate a stronger sense of classroom community [3].
The CCS has been widely assessed in various countries across face-to-face, blended, and online learning environments [26], [27], [28], [29], [30], [31]. However, no validated German version was previously available. The authors translated the CCS into German (CCS-D) and found its validity comparable to the original and other translations [4].
1.4. Rationale for developing a short form
While long questionnaires in assessments offer certain advantages, they also present additional challenges for researchers [32]. Longer questionnaires are associated with increased respondent burden, which may contribute to participant fatigue, reduced response effort, incomplete responses, and higher refusal rates [33], [34]. Furthermore, longer instruments are often associated with higher refusal rates compared to shorter surveys [33]. These issues can negatively affect data quality and may ultimately threaten the validity of research findings [35], [36]. Consequently, the development of short forms is valuable not only because it reduces the burden placed on both participants and researchers, but also because shorter questionnaires can enhance data quality by lowering the number of missing responses, reducing aberrant response patterns, and increasing response rates [37]. The demand for shorter yet psychometrically robust scales have been widely acknowledged in survey and organizational research [33], [38], [39].
The validation of the long form of the CCS revealed cross-loadings of several items across the proposed dimensions [4]. These findings are consistent with previous reports showing substantial cross-loadings and challenges in establishing strong construct validity across different CCS versions [26], [31], [40], [41].
These cross-loadings indicate that items are not exclusively linked to a single latent dimension but instead capture elements of both connectedness and learning. Such patterns tend to weaken discriminant validity and may reduce the conceptual clarity of underlying constructs [42]. As a result, items with substantial loadings on more than one factor can add additional measurement error and create uncertainty in the interpretation of scale scores [43]. If cross-loadings are ignored, the resulting factor structure may become less stable, increasing the risk of model misspecification and potentially leading to inaccurate or flawed theoretical interpretations [44].
These considerations suggest that further refinement of the measurement model may be necessary.
One approach to instrument refinement is the development of a short form of the measure.
Consistent with established principles of scale construction, a parsimonious and clearly structured factor solution is generally preferred [32], [45]. Accordingly, item retention should prioritize items that load clearly on a single intended factor, ensuring that each item represents its designated content domain as precisely as possible [45]. In line with this, exploratory factor analysis procedures typically guide the removal of items with weak primary loadings or substantial loadings on multiple factors, and items that consistently exhibit cross-loadings are therefore often considered unsuitable for retention [42], [43].
Following Barnard-Brak and Shiu’s recommendation to modify factor structure through exploratory factor analyses and item-level diagnostics to develop the instrument, Cho and Demmans Epp applied this approach using data from online learning environments [26], [41]. This work led to the development of an English short form of the CCS (CCS-SF), which has subsequently been validated in the literature [41], [46]. However, this version was derived from different datasets and a distinct contextual setting, namely online learning environments.
Given that the validation of the CCS-D already indicated differences in item loadings compared to the original CCS [4], it appeared necessary to develop a short form based on the present dataset. Because translation processes may introduce subtle semantic shifts affecting item interpretation, empirical validation within the German context is particularly important.
However, scale shortening brings methodological challenges. No methodological consensus exists regarding the optimal number of items to retain in short-form scale development [47], and excessive item reduction may compromise content validity and reliability if construct domains are insufficiently represented [48]. Therefore, item selection must follow careful and transparent procedures to minimize potential losses in measurement quality [37].
2. Methods
2.1. Protocol for systematic item reduction of the CCS-D questionnaire
Given the rationale and challenges described above, based on the already translated version of the instrument, a systematic procedure for item reduction was developed and implemented. The goal was to shorten the questionnaire to a total of ten items – five per subscale – without substantially compromising its psychometric quality, theoretical foundation, or the content coverage of the subscales.
The decision to create a ten-item short version was based on several methodological and conceptual considerations.
Methodological research suggests that less than four items per factor may result in insufficient internal reliability [47], [49]. Moreover, scale development literature emphasizes that instruments should include enough items to adequately represent the construct domain and prevent construct underrepresentation [50]. Consequently, recommendations for scale reduction stress the importance of balancing brevity with adequate construct coverage and reliability [33], [35], [37], [51].
Under classical test theory assumptions, internal consistency tends to increase with the number of items, in particular when using Cronbach’s alpha, if items measure the same underlying construct [52], [53], [54]. Retaining a slightly larger number of items in a ten-item version compared to a more shortened eight-item version therefore contributes to more stable measurement properties.
Furthermore, a scale of this length allows for broader content coverage of the underlying construct. This is essential for construct validity as item reduction may lead to insufficient representation of relevant aspects [48], [50]. Particularly in multidimensional constructs such as classroom community, overly aggressive item reduction may narrow the conceptual breadth of the instrument and weaken its ability to adequately represent the underlying construct [48], [50].
In addition, the ten-item version represents a practical compromise between measurement precision and efficiency, as it substantially reduces participants’ response time while preserving the psychometric quality of the instrument [51].
The systematic reduction process was carried out in a three-step procedure. The approach was based both on the iterative methodology of Cho and Demmans Epp in their validation of their CCS short form [41], and on the guidelines of Hinkin [45].
The custom reduction protocol was predefined prior to the refinement procedure to minimize researcher-induced selection bias. The complete CCS-D dataset from Knof & Shiozawa served as the starting point [4]. To examine the factor structure, a Principal Component Analysis (PCA) with varimax rotation was conducted, while the internal consistency of the scales was determined using Cronbach’s alpha. Subsequently, an iterative reduction process was initiated following a predefined protocol. The objective of the procedure was construct optimization while preserving content coverage rather than maximizing internal consistency coefficients.
In each iteration, the procedure began with step 1: All items that exhibited cross-loadings (secondary loadings ≥.30) or primary loadings on a non-target factor were removed simultaneously. Primary loadings on a non-target factor were defined as items loading more strongly on a factor other than the theoretically expected dimension based on the original scale structure and the framework by Rovai described above. These items do not clearly represent a single latent construct and may therefore compromise the validity of the scale [43], [53]. Whenever one or more items were removed in step 1, a new PCA was immediately conducted and Cronbach’s alpha recalculated, marking the start of a new iteration. If no items met the criteria for removal in step 1, step 2 was applied: In this step, the item whose deletion produced an improvement in internal consistency (Cronbach’s alpha if item deleted) was removed. After any removal in step 2, a PCA was performed, Cronbach’s alpha recalculated, and the procedure returned to step 1 for the next iteration. If neither step 1 nor step 2 resulted in any removal, step 3 was executed, simultaneously eliminating all items with the lowest loadings on their theoretically corresponding factors in order to preserve the overall construct validity of the scale and maintain adequate representation of the underlying theoretical dimensions [43], [53]. As in the previous steps, a PCA and recalculation of Cronbach’s alpha were performed after each removal, and the process restarted at step 1. This iterative procedure continued until no further items could be removed according to the predefined criteria, resulting in a psychometrically optimized short form with equal representation of both subscales (five items per factor), ensuring structural symmetry of the measurement model of classroom community.
The validation of the short form was conducted using an independent sample, following best-practice recommendations [33], [39], [48], [55]. This step is important because item responses may be influenced by the surrounding survey context [33]. Therefore, validating the shortened instrument on the same dataset used for item selection could introduce bias.
2.2. Data collection
At the end of the summer semester 2025, first-semester medical students at the Faculty of Medicine, University of Tuebingen, were surveyed using a paper-based questionnaire. By this time, participants had completed all courses of their first semester in the medical study program. Recruitment took place in person, conducted by the first and second author. Prior to participation, all students provided written informed consent after being informed about the purpose of the study, potential benefits and risks, their rights and responsibilities, and the voluntary nature of participation. Data were collected in pseudonymized form, and all participants consented to the publication of anonymized data. Participants were informed that they could withdraw their consent at any time without any negative consequences. Ethical approval for the study was granted by the ethics committee of Eberhard Karls University Tübingen (approval number: 086/2022BO2).
2.3. Data analysis
Analogous to the validation of the CCS-D, the same validation procedure was applied [4].
Descriptive statistical analyses were conducted to characterize the sample and to determine the scale and item properties of the CCS-D-SF. Internal consistency was evaluated using Cronbach’s alpha for each subscale as well as for the total scale. Values of alpha ≥.70 were considered satisfactory [54].
Construct validity of the two-factor structure of the CCS-D-SF was examined using confirmatory factor analysis (CFA) with maximum likelihood (ML) as used estimation method.
In the CFA, the observed variables (five connectedness items and five learning items) were specified as indicators of their respective latent variables (connectedness and learning). Model fit was evaluated according to widely accepted international standards [56], [57], [58].
The following indices and thresholds were used:
- Normed Chi-square (χ2/df): acceptable fit indicated by values <3 [59].
- Comparative fit index (CFI) and Tucker-Lewis index (TLI): good model fit indicated by values ≥.95 [60].
- Root mean square error of approximation (RMSEA): values between .05 and .08 indicate acceptable fit [61], [62], [63].
- Standardized root mean square residual (SRMR): acceptable fit indicated by values <.08 [60].
To further examine the robustness of the model and explore underlying data-driven patterns not captured by the CFA, a PCA was conducted. PCA was performed using varimax rotation with Kaiser normalization. Items with factor loadings of ≥|.30| on one or both dimensions were retained [59], [64]. This complementary approach followed recommendations emphasizing the combined use of CFA and exploratory techniques to ensure model robustness [64], [65], [66].
All data preparation and descriptive analyses were conducted using IBM SPSS Statistics 30 (IBM Corp., Armonk, NY). The CFA was performed using IBM SPSS AMOS 26 (IBM Corp., Armonk, NY).
3. Results
3.1. Sample
A total of N=178 first-semester medical students at the Faculty of Medicine Tuebingen were surveyed using a paper-based questionnaire. 205 students could have been included; this represents a response rate of 86.8%. To meet the conditions for use of AMOS in further analysis, all participants with any missing data for items of CCS-D-SF were excluded. This resulted in 171 respondents.
Study participants were 68.4% female, corresponding to the gender distribution of the admission. The age ranged from 18 to 29 years; the mean age was 20.43 years (SD 1.97). Most participants were medical students (87.1%) followed by dentistry students (12.9%).
3.2. Systematic item reduction of the CCS-D questionnaire
The validation of the stepwise item reduction for the short form was based on the PCA of the CCS-D dataset, which demonstrated good internal consistency (Cronbach’s alpha=.87; see attachment 1 [Att. 1], table S2) [4].
Based on this analysis, step 1 of the item reduction protocol was implemented. All items exhibiting substantial cross-loadings (items 9, 10, 14, and 16) were removed. Subsequently, a PCA was conducted, and Cronbach’s alpha was recalculated to provide the basis for the next iteration. This resulted in the first iteration of the CCS-D-SF with a Cronbach’s alpha=.84 (see attachment 1 [Att. 1], table S3).
In iteration 1 of the CCS-D-SF, no items displayed cross-loadings or primary loadings on a non-target factor. Therefore, step 2 was conducted, during which item 6 was removed because its exclusion increased Cronbach’s alpha. The subsequent PCA and reliability analysis produced CCS-D-SF iteration 2 (Cronbach’s alpha=.84) (see attachment 1 [Att. 1], table S4).
During CCS-D-SF iteration 2, item 5 and item 17 were removed because of cross-loadings. A subsequent PCA and recalculation of Cronbach’s alpha produced CCS-D-SF iteration 3 (Cronbach’s alpha=.82) (see attachment 1 [Att. 1], table S5).
In iteration 3 of the CCS-D-SF, item 1 was removed because it loaded on more than one factor. The subsequent PCA and reliability analysis yielded CCS-D-SF iteration 4 (Cronbach’s alpha=.80) (see attachment 1 [Att. 1], table S6).
In iteration 4, no items displayed cross-loadings or primary loadings on a non-target factor. Step 2 did not yield improvements in Cronbach’s alpha, so step 3 was conducted, leading to the removal of items 7 and 2, which had the lowest loadings on their respective factors. The resulting 10-item version contained five items per subscale. A final PCA and recalculation of Cronbach’s alpha produced CCS-D-SF iteration 5 (Cronbach’s alpha=.79) (see attachment 1 [Att. 1], table S7).
To enhance usability, items were reordered to alternate between subscales, consistent with the original CCS format, and renumbered accordingly (see table 1 [Tab. 1]). The final CCS-D-SF includes two subscales: Connectedness (five odd-numbered items) and learning (five even-numbered items). Each item is rated on a five-point Likert scale. According to the original CCS instrument for items 1, 3, 5, 7, 9, the scoring scale was: strongly agree=4, agree=3, neutral=2, disagree=1, strongly disagree=0; for items 2, 4, 6, 8, 10 the scoring scale was: strongly agree=0, agree=1, neutral=2, disagree=3, strongly disagree=4 [3]. For further analysis these item-scores must be inverted. These items are labeled with an “i” in tables and figures of this paper.
Table 1: Final version of CCS-D-SF
3.3. Validation of the CCS-D-SF
3.3.1. Scale characteristics and internal consistency
Descriptive statistics for the individual items (see table 2 [Tab. 2]) as well as the total CCS-D-SF and for each subscale are shown below (see table 3 [Tab. 3]). The total CCS-D-SF (Cronbach’s alpha=.81) and the subscale connectedness (alpha=.86) showed good internal consistency. The subscale learning is reported with a Cronbach’s alpha=.65.
Table 2: Descriptive statistics of CCS-D-SF items
Table 3: Descriptive statistics of CCS-D-SF subscales
3.3.2. Confirmatory factor analysis
The CFA-Model for the 2-factor-model is displayed in figure 1 [Fig. 1].
The two-factor CFA model demonstrated an overall acceptable model fit. The two latent factors were moderately correlated (r=.51). The normed chi-square (χ2 [34, n=171]=73.529, p<.001; χ2/df=2.163). indicated an adequate fit. The RMSEA (.083) exceeds the limit for an acceptable fit, while the SRMR (.059) indicated a good fit. The comparative fit indices, CFI (.93) and TLI (.91), were within the acceptable range but did not reach the threshold for a good fit.
3.3.3. Principal component analysis
The Kaiser-Meyer-Olkin measure was .811, indicating a relatively good suitability for factor analysis. Bartlett’s test of sphericity was significant (approx. χ2=581.215; p<.001), suggesting that the data do not form an identity matrix and are appropriate for factor analysis. Only factors with eigenvalues ≥1 were retained [68], [69]. Based on Kaiser’s criteria and the scree plot, two factors with eigenvalues greater than 1 were retained, accounting for 54.9% of the total variance.
In the two-factor model rotated using varimax (see table 4 [Tab. 4]), most items loaded primarily on a single factor, except for item 10 (“I feel that this course does not promote a desire to learn”). The underlying factor structure of the CCS-SF can thus be confirmed, with odd-numbered items loading highly on factor 1 and even-numbered items on factor 2, representing the dimensions of connectedness and learning.
4. Discussion
The CCS-D-SF shows good reliability and replicates the factor structure of the original version in the CFA. In a final PCA it could be ensured that the factor solution did not change after deleting items [51].
When considering model-fit indices, the CCS-D-SF performs better than the CCS-D.
A moderate correlation was found between the connectedness and learning subscales, suggesting that they share a substantial amount of variance while still representing distinct constructs, although the relationship indicates that the two dimensions of classroom community are not entirely independent. The marginally elevated RMSEA may partly reflect the substantial overlap between connectedness and learning. A likewise high correlation between the subscales has already been reported in the literature [30].
This relationship aligns with the community of inquiry framework, which posits that social presence (connectedness) and cognitive presence (learning engagement) interact dynamically to foster meaningful educational experiences [12], [70]. Students who feel socially integrated within a course are more likely to engage cognitively, while deep cognitive engagement can, in turn, reinforce a sense of belonging and social connection [11]. The subscale connectedness showed very good internal consistency, whereas learning exhibited only marginally acceptable reliability. Lower reliability of the learning subscale is consistent with previous studies using the CCS and its derivatives, which have frequently reported lower internal consistency for learning-related items [26], [27], [71]. The reduced number of items in the CCS-D-SF may not fully capture this conceptual breadth, thereby lowering internal consistency and potentially affecting the correlation between subscales. Additionally, learning may be more context-dependent than connectedness, as students’ perceptions of learning are influenced by instructional design, teaching style, and assessment practices.
The cross-loading of Item 10 (“I feel that this course does not promote a desire to learn”) on the connectedness subscale further supports this interpretation. This pattern may reflect the way students perceive the term “course”, which could be understood either as the content and structure of the class, the social group of participants, or a combination of both. Additionally, external factors such as the lecturer’s personality and disposition can shape students’ perceptions of social presence and classroom community, potentially affecting both connectedness and learning [72], [73]. Students’ expectations about interactions with peers in future courses may also influence their motivation to develop a sense of community, thereby contributing to the observed subscale correlation [74].
From a pedagogical perspective, the interdependence between connectedness and learning carries important implications. Teaching strategies that intentionally foster social cohesion – such as group projects, peer mentoring, or structured reflection sessions – may not only strengthen connectedness but also enhance learning outcomes [12], [24]. Similarly, curriculum designs that emphasize collaborative knowledge construction can help to reinforce both social belonging and cognitive engagement. The CCS-D-SF could thus serve as a diagnostic instrument for educators to identify imbalances between these two facets and guide targeted interventions aimed at improving classroom community.
There are several limitations to this study that should be considered. A homogenization of student communities can be observed, driven by similar social and personal challenges, including uncertainties associated with the transition to higher education [75]. Typically, first-semester students focus on building personal relationships and establishing a professional network relevant to their studies [76]. This may influence how students perceive or prioritize connectedness and learning, as social integration and network-building could temporarily outweigh academic engagement or learning-focused interactions. In contrast to the present study, in which the sense of community was measured across an entire academic program, most previous research using the CCS has focused on individual courses or classroom sessions [77].
When interpreting the results of the scale reduction process, several methodological considerations should be acknowledged.
Factor analytic approaches are commonly used in scale shortening because factor loadings indicate the strength of the relationship between items and the underlying latent dimensions [33], [78]. Accordingly, many reduction procedures rely on psychometric indicators such as factor loadings, item-total correlations, and reliability estimates [35]. In the present study, scale refinement was not based solely on factor loadings; the internal consistency of the overall instrument was considered as a complementary criterion to ensure stability of the measurement model throughout the reduction process [50], [52]. However, reliance entirely on statistical criteria may produce shortened scales that are psychometrically homogeneous but conceptually narrow and instruments that are overly restrictive in content [34], [50], or instruments that capture slightly different aspects of the construct than the long form [33]. To address these concerns, attention was given to preserving an equal representation of the two underlying dimensions of classroom community, connectedness and learning [3].
The authors acknowledge that the present approach represents one possible strategy for scale reduction among several currently discussed methodological approaches. Short-form instruments have been commonly developed by iteratively removing items from existing longer questionnaires based on psychometric indicators [32]. Although these approaches are relatively simple to implement, they often consider only a limited range of psychometric criteria and may therefore result in shortcomings such as inadequate factor structures, insufficient reliability, or lack of measurement invariance [79]. More recently, algorithmic approaches have been proposed and shown promising results in the development of short forms [79], [80]. However, heuristic optimization procedures are not guaranteed to identify a single optimal solution and may produce different item combinations across repeated analyses [32]. These developments illustrate that scale shortening should not be understood as a purely mechanical optimization procedure, but rather as a context-dependent process requiring both statistical and theoretical considerations [79], [80].
As data-driven item reduction procedures may be sensitive to sample-specific characteristics, the resulting short form was evaluated using CFA in an independent validation sample [45], [55]. This approach also allows assessment of the stability and replicability of the factor structure, acknowledging that validity evidence obtained for long-form scales may not necessarily generalize to abbreviated versions [45], [55]. Alternative strategies such as split-sample validation within the original dataset could have reduced the risk of sample-specific overfitting. However, implementing a split-sample approach would have decreased the effective sample size available for item selection and factor analysis, which is particularly problematic, given that factor analysis can become unstable in smaller samples [63], [81]. The present study did not include external validation criteria that would allow a direct comparison of the validity of the short form. Without such external anchors, it cannot be fully determined whether this instrument captures the same underlying construct [33].
Improvements in model fit following scale reduction should not be interpreted as evidence of superior measurement quality per se, as fit indices are sensitive to model complexity, instrument length, and item homogeneity [82], [83].
Future research should therefore examine the relationships between the CCS-D-SF and theoretically related constructs in order to further establish its construct validity.
5. Conclusion
This study presented the systematic development and validation of a short form of the German version of the classroom community scale (CCS-D-SF).
The CCS-D-SF demonstrated good reliability, comparable to previous findings in the literature. Confirmatory factor analysis confirmed the two-factor structure of the scale, with model-fit indices even slightly better than those reported for the long version of the CCS-D. The factor structure was also supported by the principal component analysis.
Overall, the CCS-D-SF provides a practical and psychometrically sound instrument for measuring the sense of community among first-year medical students. Its brevity allows for reduced participant burden while measurement quality was maintained or even improved.
Future research should focus on applying the CCS-D-SF across German-speaking countries and further exploring its utility in diverse educational contexts.
Conference presentation
Excerpts from this manuscript were presented at the Annual Conference of the DACH Association for Medical Education (GMA) in Dresden 2026 [84].
Authors’ ORCIDs
- Harald Knof: [0000-0002-0942-653X]
- Thomas Shiozawa: [0000-0002-7112-1016]
Acknowledgements
The authors would like to thank Prof Alfred P. Rovai from Regent University (Virginia Beach, VA, USA) for the permission to use and translate his CCS instrument.
The authors would like to thank Prof. Dr. Peter Martus for statistical consulting.
The authors are grateful to students at the Faculty of Medicine Tuebingen for participating in this study.
The authors acknowledge support from the Open Access Publication Fund of the University of Tuebingen.
Competing interests
The authors report no conflicts of interest. The authors alone are responsible for the content and writing of this article. The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
References
[1] Rosen MA, DiazGranados D, Dietz AS, Benishek LE, Thompson D, Pronovost PJ, Weaver SJ. Teamwork in healthcare: Key discoveries enabling safer, high-quality care. Am Psychol. 2018;73(4):433-450. DOI: 10.1037/amp0000298[2] Van Leuven AJ, Szymik BG, Ramsey LM, Hesse DW. A multi-year evaluation of medical student performance on and perceptions of collaborative gross anatomy laboratory examinations. Anat Sci Educ. 2023;16(3):557-566. DOI: 10.1002/ase.2223
[3] Rovai AP. Development of an instrument to measure classroom community. Internet High Educ. 2002;5(3):197-211. DOI: 10.1016/s1096-7516(02)00102-1
[4] Knof H, Shiozawa T. Validation of the German form of Classroom-Community-Scale (CCS-D). GMS J Med Educ. 2025;42(1):Doc13. DOI: 10.3205/zma001737
[5] Dillenbourg P. What do you mean by 'collaborative learning'? In: Dillenbourg P, editor. Collaborative Learning: Cognitive and Computational Approaches. Advances in learning and instruction series. 1st ed. London: Pergamon; 1999. (Advances in learning and instruction series). p.1-15.
[6] Sevenhuysen S, Haines T, Kiegaldie D, Molloy E. Implementing collaborative and peer-assisted learning. Clin Teach. 2016;13(5):325-331. DOI: 10.1111/tct.12583
[7] O'Brien BC, Forrest K, Wijnen‐Meijer M, ten Cate O. A Global View of Structures and Trends in Medical Education. In: Swanwick T, Forrest K, O’Brien BC, editors. Understanding Medical Education. 3rd ed. Hoboken: Wiley-Blackwill; 2018. p.7-22. DOI: 10.1002/9781119373780.ch2
[8] Almajed A, Winning T, Skinner V, Peterson R. Medical and Dental Students' Perceptions and Experiences of Collaborative Learning: A Systematic Review. JBI Libr Syst Rev. 2011;9(32 Suppl):1-13. DOI: 10.11124/01938924-201109321-00001
[9] Skinner VJ, Braunack-Mayer A, Winning TA. The Purpose and Value for Students of PBL Groups for Learning. Interdiscip J Problem-Based Learn. 2015;9(1). DOI: 10.7771/1541-5015.1499
[10] Pluta WJ, Richards BF, Mutnick A. PBL and beyond: trends in collaborative learning. Teach Learn Med. 2013;25 Suppl 1:S9-S16. DOI: 10.1080/10401334.2013.842917
[11] Chatterjee R, Correia A-P. Online Students’ Attitudes Toward Collaborative Learning and Sense of Community. Am J Distance Educ. 2019;34(1):53-68. DOI: 10.1080/08923647.2020.1703479
[12] Garrison DR, Anderson T, Archer W. The first decade of the community of inquiry framework: A retrospective. Internet High Educ. 2010;13(1-2):5-9. DOI: 10.1016/j.iheduc.2009.10.003
[13] Halic O, Lee D, Paulus T, Spence M. To blog or not to blog: Student perceptions of blog effectiveness for learning in a college-level course. Internet High Educ. 2010;13(4):206-213. DOI: 10.1016/j.iheduc.2010.04.001
[14] Lai KW. Knowledge construction in online learning communities: a case study of a doctoral course. Stud High Educ. 2013;40(4):561-579. DOI: 10.1080/03075079.2013.831402
[15] Rovai AP. Building Sense of Community at a Distance. Int Rev Res Open Dis. 2002;3(1):1-16. DOI: 10.19173/irrodl.v3i1.79
[16] Bruffee KA. Collaborative Learning: Higher Education, Interdependence, and the Authority of Knowledge. 1st ed. Baltimore: Johns Hopkins University Press; 1993.
[17] Tinto V. Classrooms as Communities. J High Educ. 1997;68(6):599-623. DOI: 10.1080/00221546.1997.11779003
[18] Sheets MF. Characteristics of Adult Education Students and Factors Which Determine Course Completion: A Review. New Horiz Adult Educ. 1992;6(1):3-18. DOI: 10.1002/nha3.10034
[19] McMillan D, Chavis D. Sense of Community: A Definition and Theory. J Community Psychol. 1986;14(1):6-23. DOI: 10.1002/1520-6629(198601)14:1<6::AID-JCOP2290140103>3.0.CO;2-I
[20] Hill JL. Psychological sense of community: Suggestions for future research. J Community Psychol. 1996;24(4):431-438. DOI: 10.1002/(Sici)1520-6629(199610)24:4<431::Aid-Jcop10>3.3.Co;2-Z
[21] Rovai AP. Sense of community, perceived cognitive learning, and persistence in asynchronous learning networks. Internet High Educ. 2002;5(4):319-332. DOI: 10.1016/s1096-7516(02)00130-6
[22] Shaffer C, Anundsen K. Creating community anywhere: Finding support and connection in a fragmented world. 1st ed. Los Angeles: Tarcher/Perigree; 1993.
[23] Rheingold H. The virtual community: Finding commection in a computerized world. 1st ed. Boston: Addison-Wesley Longman Publishing Co. Inc.; 1993.
[24] Strayhorn TL. College Students’ Sense of Belonging - A Key to Educational Success for All Students. 2nd ed. New York: Routlege; 2018. DOI: 10.4324/9781315297293
[25] Wang W, Gao Y, Pitts A, Dong L. A Bibliometric Analysis of Neighborhood Sense of Community. Sustainability. 2023;15(5):4183. DOI: 10.3390/su15054183
[26] Barnard-Brak L, Shiu W. Classroom Community Scale in the blended learning environment: A psychometric review. Int J E-Learn. 2010;9(3):303-311.
[27] Hur EH, Glassman M, Kim Y. Finding autonomy in activity: Development and validation of a democratic classroom survey. Educ Ass Eval Account. 2013;25(4):303-320. DOI: 10.1007/s11092-013-9173-y
[28] Ahmady S, Kohan N, Bagherzadeh R, Rakshhani T, Shahabi M. Validity testing of classroom community scale in virtual environment learning: A cross sectional study. Ann Med Surg (Lond). 2018;36:256-260. DOI: 10.1016/j.amsu.2018.08.021
[29] Yan J, Ding X, Xiong L, Liu E, Zhang Y, Luan Y, Qin L, Zhou C, Zhang W. Team-based learning: assessing the impact on anatomy teaching in People's Republic of China. Adv Med Educ Pract. 2018;9:589-594. DOI: 10.2147/AMEP.S169949
[30] Abdeldayem M, Aldulaimi SH, Aldulaimi MLA. Virtual Learning and Students’ Connectedness in the Time of Coronavirus. Int J Adv Sci. 2020;29(5):12634-12645.
[31] Perrucci V, Cacciamani S, Khanlari A, Balboni G. The Classroom Community Scale in a Face-to-Face University Context: Factorial Structure and Convergent/Divergent Validity. Test Psychom Methodol Appl Psychol. 2022;29(3):297-307. DOI: 10.4473/TPM29.3.2
[32] Franzen P, Arens AK, Greiff S, van der Westhuizen L, Fischbach A, Wollschläger R, Niepl C. Developing and Validating a Short-Form Questionnaire for the Assessment of Seven Facets of Conscientiousness in Large-Scale Assessments. J Pers Assess. 2022;104(6):759-773. DOI: 10.1080/00223891.2021.1998083
[33] Stanton JM, Sinar EF, Balzer WK, Smith PC. Issues and Strategies for Reducing the Length of Self‐Report Scales. Pers Psychol. 2002;55(1):167-194. DOI: 10.1111/j.1744-6570.2002.tb00108.x
[34] Passarelli M, Casetta L, Rizzi L, Chiorri C, Cassina F, Voi S, Rocco D. Short and sweet: Comparing strategies for the reduction of questionnaires on self-criticism and social safeness while preserving construct validity. Int J Psychol. 2024;59(6):1234-1244. DOI: 10.1002/ijop.13249
[35] Kilmen S, Bulut O. Shortening Psychological Scales: Semantic Similarity Matters. Educ Psychol Meas. 2025;85(5):910-934. DOI: 10.1177/00131644251319047
[36] Yarkoni T. The Abbreviation of Personality, or how to Measure 200 Personality Scales with 200 Items. J Res Pers. 2010;44(2):180-198. DOI: 10.1016/j.jrp.2010.01.002
[37] Alcaraz S, Viladrich C, Torregrosa M. Less time, better quality. Shortening questionnaires to assess team environment and goal orientation. Span J Psychol. 2013;16:E77. DOI: 10.1017/sjp.2013.74
[38] Fraser BJ. Development of Short Forms of Several Classroom Environment Scales. J Educ Meas. 1982;19(3):221-227. DOI: 10.1111/j.1745-3984.1982.tb00130.x
[39] Coste J, Guillemin F, Pouchot J, Fermanian J. Methodological approaches to shortening composite measurement scales. J Clin Epidemiol. 1997;50(3):247-252. DOI: 10.1016/s0895-4356(96)00363-0
[40] Lewis K, McVay-Dyche J, Chen H, Soto T. Examining Sense of Community among Medical Professionals in an Online Graduate Program. J Educ Online. 2015;12(1). DOI: 10.9743/jeo.2015.1.8
[41] Cho J, Demmans Epp C. Improving the Classroom Community Scale: Toward a Short-Form of the CCS. In: American Educational Research Association (AERA) Annual Meeting; 2019 Apr 5-9; Toronto, Canada. DOI: 10.3102/1432534
[42] Brown TA. Confirmatory factor analysis for applied research. 2nd ed. New York (NY): The Guilford Press; 2015.
[43] Hair JF, Black WC, Babin BJ, Anderson RE. Multivariate Data Analysis. 8th ed. Cheriton House: Cengage Learning; 2019.
[44] Li Y, Wen Z, Hau KT, Yuan KH, Peng Y. Effects of Cross-loadings on Determining the Number of Factors to Retain. Struct Equation Model. 2020;27(6):841-863. DOI: 10.1080/10705511.2020.1745075
[45] Hinkin TR. A Brief Tutorial on the Development of Measures for Use in Survey Questionnaires. Organ Res Method. 1998;1(1):104-121. DOI: 10.1177/109442819800100106
[46] Tackett M, Viel S, Manturuk K. A validation of the short-form classroom community scale for undergraduate mathematics and statistics students. J Uni Teach Learn Pract. 2023;20(1). DOI: 10.53761/1.20.01.08
[47] Jokovic A, Locker D, Guyatt G. Short forms of the Child Perceptions Questionnaire for 11-14-year-old children (CPQ11-14): development and initial evaluation. Health Qual Life Outcomes. 2006;4:4. DOI: 10.1186/1477-7525-4-4
[48] Smith GT, McCarthy DM, Anderson KG. On the sins of short-form development. Psychol Assess. 2000;12(1):102-111. DOI: 10.1037//1040-3590.12.1.102
[49] Watson D, Clark LA. Measurement and mismeasurement of mood: recurrent and emergent issues. J Pers Assess. 1997;68(2):267-296. DOI: 10.1207/s15327752jpa6802_4
[50] Clark LA, Watson D. Constructing validity: Basic issues in objective scale development. Psychol Assess. 1995;7(3):309-319. DOI: 10.1037/14805-012
[51] Worthington RL, Whittaker TA. Scale Development Research: A Content Analysis and Recommendations for Best Practices. Couns Psychol. 2006;34(6):806-838. DOI: 10.1177/0011000006288127
[52] Cortina JM. What is coefficient alpha? An examination of theory and applications. J Appl Psychol. 1993;78(1):98-104. DOI: 10.1037/0021-9010.78.1.98
[53] Osborne J. Best Practices in Exploratory Factor Analysis: Four Recommendations for Getting the Most From Your Analysis. Scotts Valley, CA: CreateSpace Independent Publishing; 2014.
[54] Nunnally JC. Psychometric theory. 2nd ed. New York (NY): McGraw-Hill; 1978.
[55] Kruyen PM, Emons WH, Sijtsma K. On the Shortcomings of Shortened Tests: A Literature Review. Int J Testing. 2013;13(3):223-248. DOI: 10.1080/15305058.2012.703734
[56] Crowley SL, Fan X. Structural equation modeling: basic concepts and applications in personality assessment research. J Pers Assess. 1997;68(3):508-531. DOI: 10.1207/s15327752jpa6803_4
[57] Hooper D, Coughlan J, Mullen M. Structural Equation Modeling: Guidelines for Determining Model Fit. Elect J Business Res Meth. 2007;6.
[58] Kline RB. Principles and practice of structural equation modeling. 4th ed. New York (NY): Guilford Press; 2016.
[59] Kline RB. Principles and practice of structural equation modeling. New York (NY): Guilford Press; 1998.
[60] Hu Lt, Bentler PM. Cutoff criteria for fit indexes in covariance structure analysis: Conventional criteria versus new alternatives. Struct Equat Mod. 1999;6(1):1-55. DOI: 10.1080/10705519909540118
[61] Browne MW, Cudeck R. Alternative Ways of Assessing Model Fit. Sociol Methods Res. 1992;21(2):230-258. DOI: 10.1177/0049124192021002005
[62] Jöreskog KG, Sörbom D. LISREL 8: Structural Equation Modeling with the SIMPLIS Command Language. Scientific Software International; 1993.
[63] MacCallum RC, Browne MW, Sugawara HM. Power analysis and determination of sample size for covariance structure modeling. Psychol Methods. 1996;1(2):130-149.
[64] Knekta E, Runyon C, Eddy S. One Size Doesn't Fit All: Using Factor Analysis to Gather Validity Evidence When Using Surveys in Your Research. CBE Life Sci Educ. 2019;18(1):rm1. DOI: 10.1187/cbe.18-04-0064
[65] Martin CR, Savage-McGlynn E. A ‘good practice’ guide for the reporting of design and analysis for psychometric evaluation. J Reprod Infant Psychol. 2013;31(5):449-455. DOI: 10.1080/02646838.2013.835036
[66] Fabrigar LR, Wegener DT, MacCallum RC, Strahan EJ. Evaluating the use of exploratory factor analysis in psychological research. Psychol Methods. 1999;4(3):272-299. DOI: 10.1037/1082-989x.4.3.272
[67] Comrey AL, Lee HB. A first course in factor analysis. 2nd ed. Hillsdale (NJ): L. Erlbaum Associates; 1992.
[68] Kaiser HF. The Application of Electronic Computers to Factor Analysis. Educ Psychol Meas. 1960;20(1):141-151. DOI: 10.1177/001316446002000116
[69] Guttman L. Some Necessary Conditions for Common-Factor Analysis. Psychometrika. 2025;19(2):149-161. DOI: 10.1007/bf02289162
[70] Arbaugh JB, Cleveland-Innes M, Diaz SR, Garrison DR, Ice P, Richardson JC, Swan KP. Developing a community of inquiry instrument: Testing a measure of the Community of Inquiry framework using a multi-institutional sample. Internet High Educ. 2008;11(3):133-136. DOI: 10.1016/j.iheduc.2008.06.003
[71] Zhang YY, Lin X, Xu M. Rovai’s Classroom Community Scale and Its Application in Chinese College English Class. Sino-US Engl Teach. 2011;8(9):592-598.
[72] Lowenthal PR, Trespalacios J. Classroom Community and Time: Comparing Student Perceptions in Traditional vs. Accelerated Online Courses. Online Learn. 2022;26(4):59-77. DOI: 10.24059/olj.v26i4.3498
[73] Trespalacios J, Lowenthal P. What do they really like? An investigation of students' perceptions of their coursework in a fully online educational technology program. Australas J Educ Technol. 2019;35(5):60-78. DOI: 10.14742/ajet.4364
[74] Lowenthal PR, Dunlap JC. Investigating students’ perceptions of instructional strategies to establish social presence. Distance Educ. 2018;39(3):281-298. DOI: 10.1080/01587919.2018.1476844
[75] Picton A, Greenfield S, Parry J. Why do students struggle in their first year of medical school? A qualitative study of student voices. BMC Med Educ. 2022;22(1):100. DOI: 10.1186/s12909-022-03158-4
[76] Traus A, Höffken K, Thomas S, Mangold K, Schröer W. Stu.diCo. – Studieren digital in Zeiten von Corona. Hildesheim: UVH - Universitätsverlag Hildesheim; 2020.
[77] Beeson E, Aideyan B, Shoney CO, Bowes DA, Ansell KL, Peterson HM. Predicting Sense of Community among Graduate Students in a Distance Learning Environment. Univers J Educ Res. 2019;7(3):746-753. DOI: 10.13189/ujer.2019.070314
[78] Floyd FJ, Widaman KF. Factor analysis in the development and refinement of clinical assessment instruments. Psychol Assess. 1995;7(3):286-299. DOI: 10.1037/1040-3590.7.3.286
[79] Olaru G, Witthöft M, Wilhelm O. Methods matter: Testing competing models for designing short-scale Big-Five assessments. J Res Person. 2015;59:56-68. DOI: 10.1016/j.jrp.2015.09.001
[80] Schroeders U, Wilhelm O, Olaru G. The influence of item sampling on sex differences in knowledge tests. Intelligence. 2016;58:22-32. DOI: 10.1016/j.intell.2016.06.003
[81] MacCallum R, Widaman K, Zhang S, Hong S. Sample Size in Factor Analysis. Psychol Methods. 1999;4:84-99. DOI: 10.1037/1082-989X.4.1.84
[82] Kenny DA, McCoach DB. Effect of the Number of Variables on Measures of Fit in Structural Equation Modeling. Struct Equ Modeling. 2009;10(3):333-351. DOI: 10.1207/s15328007sem1003_1
[83] Jackson DL, Gillaspy JA, Purc-Stephenson R. Reporting practices in confirmatory factor analysis: an overview and some recommendations. Psychol Methods. 2009;14(1):6-23. DOI: 10.1037/a0014694
[84] Knof H, Huber V, Shiozawa T. Validierung einer Kurzform der Deutschen Version der Classroom Community Scale (CCS-D-SF). In: Jahrestagung der Gesellschaft für Medizinische Ausbildung (GMA). Dresden, 17.-19.09.2026. Düsseldorf: German Medical Science GMS Publishing House; 2026. DocV-0701. DOI: 10.3205/26gma037
Attachments
| Attachment 1 | Supplementary tables (Attachment_1.pdf, application/pdf, 149.82 KBytes) |





