TL;DR
This paper introduces nomograms, a novel visualization tool, to predict self-regulated learning (SRL) levels in Chinese medical students. By combining machine learning and regression techniques, they identified key predictors and created user-friendly tools that can help educators identify students at risk and provide targeted interventions.
What Problem Does This Paper Address?
Self-regulated learning (SRL) is a crucial skill for medical students, impacting their academic success and future practice. While previous research has explored factors influencing SRL, there's a lack of predictive tools that can be easily used by educators in a clinical or academic setting. Existing studies often focus on descriptive analyses rather than offering practical, predictive models. This paper aims to bridge this gap by developing nomograms, which are graphical representations of statistical models, to predict SRL levels in Chinese undergraduate medical students, enabling more proactive support.
Methods
The researchers employed a multi-center, cross-sectional study design. They combined classical regression methods (logistic regression and Lasso regression) with a machine learning model (random forest) to identify significant predictors of SRL. Exploratory Factor Analysis (EFA) was used to determine the scores for SRL dimensions (motivation and strategy). Lasso regression was specifically used to prevent overfitting in the multivariable models. The final models were used to construct nomograms, which were then validated both internally and externally using receiver operating characteristic (ROC) curves and calibration curves.
Datasets
The study utilized data from 2,052 medical undergraduates across five universities in mainland China. After initial data cleaning and a propensity score matching (PSM) process to reduce heterogeneity, a training set (1,219 students) and a testing set (639 students) were created. A small external validation set of 35 undergraduates from Fudan University was also used.
Key Insights and Ideas
- Nomograms as Practical Tools: The paper highlights the potential of nomograms as user-friendly, visual tools for educators to quickly assess and predict SRL levels, moving beyond complex statistical outputs.
- Interplay of Factors: SRL is not determined by a single factor but by a complex interplay of academic performance (GPA), social-psychological environment, and time investment.
- Teaching Mode Matters: Traditional teacher-centered teaching negatively impacts learning strategy, while problem-based learning (PBL) and similar student-centered approaches are more conducive to SRL.
- Balance is Key: While time investment in learning is important, excessive extracurricular activities can be detrimental to SRL, emphasizing the need for a balanced approach.
Key Results
The constructed nomograms demonstrated good predictive performance. Internal validation showed C-indices of 0.736 for learning motivation and 0.744 for learning strategy. External validation yielded even higher C-indices of 0.986 for learning motivation and 1.000 for learning strategy, indicating strong prediction accuracy, discrimination, and generalization. Key predictors identified included GPA, growing place, primary caregiver, views on the learning atmosphere, time of learning medicine weekly, interest in medicine, and teaching mode.
Limitations and Future Work
The authors acknowledge several limitations:
- Inter-group Heterogeneity: Despite PSM, complete elimination of inter-group heterogeneity across multiple centers was challenging.
- External Validation Sample Size: The external validation set was small, suggesting a need for larger sample studies.
- Cross-sectional Design: The study's cross-sectional nature limits causal inferences.
Future work should focus on expanding the sample size, conducting more extensive external validation, and carrying out prospective intervention studies to further refine and utilize the nomograms for improving medical education.
Summary updated