Researchers analyzed anonymous responses from four administrations of the National College Student Survey. The analytic sample included 1,015 responses from 2021, 4,222 from 2022, 3,028 from 2023 and 3,452 from 2024. Because the survey had no stable identifiers, records could not be linked across waves, and the analysis compared successive cross-sectional samples rather than following the same students over time. Annual response rates could not be calculated because the numbers of eligible or invited students were not consistently retained.

Students rated five learning-environment domains: classroom teaching quality, facilities and resources, developmental support, interpersonal support and teaching administration. Engagement was measured behaviorally, emotionally and cognitively, while learning gains covered professional competence, values and character, interest development, learning ability, communication and collaboration, and overall self-evaluation. Hierarchical linear regression, a cross-sectional parallel path model, year-specific analyses and k-means clustering were used to assess the relationships and identify student profiles.

Survey year and measured background characteristics explained 1.5% to 4.0% of the variance in the three engagement dimensions. Adding the five learning-environment domains explained an additional 22.5% for behavioral engagement, 27.6% for emotional engagement and 23.1% for cognitive engagement, producing full-model R² values of 0.265, 0.305 and 0.246, respectively. All additions were statistically significant at p<0.001, and all variance inflation factors were below 5.

Classroom teaching quality had the largest standardized coefficient for each engagement dimension in the pooled model: β=0.350 for behavioral engagement, β=0.351 for emotional engagement and β=0.287 for cognitive engagement. Developmental support was also positively associated with all three, with coefficients of 0.236, 0.160 and 0.175. Interpersonal support was associated with emotional and cognitive engagement but not behavioral engagement. Facilities and resources had no association with behavioral or cognitive engagement and a negative conditional association with emotional engagement, while teaching administration had mixed associations across the three dimensions. Because all domains entered the same model, these coefficients represent conditional associations and do not show that resources or administrative arrangements are unimportant.

Year-stratified analyses broadly reproduced the pooled pattern. Classroom teaching quality was positively and significantly associated with all three engagement dimensions in every wave, with coefficients ranging from 0.210 to 0.420. Developmental support was positive in all 12 models and statistically significant in 11; the exception was cognitive engagement in 2021. Within individual waves, the learning-environment domains added 15.5% to 31.4% to the variance explained in engagement beyond background covariates. None of six prespecified tests for interactions between selected learning-environment measures and student characteristics met the Bonferroni-adjusted significance threshold.

In the cross-sectional path model, the total association between overall learning environment and overall learning gains was 0.734 (95% bootstrap confidence interval, 0.719–0.749). The direct association was 0.500 (0.480–0.519), while the combined indirect association through behavioral, emotional and cognitive engagement was 0.234 (0.221–0.248). Engagement therefore statistically accounted for 31.93% of the contemporaneous total association. The specific indirect associations were 0.073 through behavioral engagement, 0.089 through emotional engagement and 0.072 through cognitive engagement. These estimates decompose associations measured at the same time and do not demonstrate temporal mediation.

Person-centered analysis identified three ordered profiles. A convergent-low group comprised 37.2% of respondents and scored below the sample mean on perceived learning environment, all three engagement dimensions and learning gains. An intermediate group comprised 43.2%, and a convergent-high group comprised 19.5%. The mean silhouette coefficient was 0.225, indicating modest separation. Sex, rural or urban origin, parental education, household income and program type were statistically associated with profile membership, but the associations were weak, with Cramér’s V values of 0.033 to 0.063. The profiles describe patterns within this sample rather than stable or clearly bounded student types.

The findings have several limitations. The design could not determine temporal ordering or causal effects, and the absence of linked longitudinal records meant the analysis did not measure within-student change. All principal measures were self-reported in the same survey, so common-method influences cannot be excluded despite anonymous administration and a diagnostic that did not identify a dominant single factor. Non-response bias could not be assessed directly. The data came from one university, limiting generalization to other regions, institutions and medical programs, and the k-means results depended partly on the specified number of clusters.

Subject of Research: Associations among undergraduate students’ perceived learning environment, engagement and learning gains at a western China medical university

Article Title: Evaluating the associations of learning environment with student engagement and learning gains among undergraduate students at a medical university in western China

Article References: Mo, Z., Gao, Y., Wei, Z., & Xu, H. (2026). Evaluating the associations of learning environment with student engagement and learning gains among undergraduate students at a medical university in western China. Frontiers in Public Health, 14, Article 1885067. https://doi.org/10.3389/fpubh.2026.1885067

Image Credits: AI Generated

DOI: 10.3389/fpubh.2026.1885067

Keywords: medical education, student engagement, learning environment, learning gains, undergraduate students, medical university, repeated cross-sectional study, China, quality improvement, Evaluating, associations, learning