Optimal cut-off point for identifying academic decline and progress in medical sciences students: a latent profile analysis-based approach

Academic education offices in Iranian medical universities commonly use a ± 2-point GPA change between two consecutive semesters to identify students with academic decline or progress. Despite being widely applied, no diagnostic validation (sensitivity, specificity, predictive values) of this criterion has been reported in the literature. This study aimed to evaluate the diagnostic accuracy of the conventional ± 2-point criterion and propose evidence-based optimal cut-off points using Latent Profile Analysis (LPA) as a reference standard. A retrospective cohort study was conducted on 1,175 medical sciences students from Mazandaran University of Medical Sciences (Fall 2022 and Spring 2023 semesters), selected by random sampling. LPA was applied as a model-based reference standard to classify students into three latent groups (academic decline, no change, academic progress) based on the GPA difference between the two consecutive semesters, stratified by quartiles of change in academic credits. The diagnostic performance of the traditional ± 2-point criterion was evaluated against this LPA classification. Receiver Operating Characteristic (ROC) curve analysis and the Youden index were used to identify the optimal cut-off points for decline and progress. Based on the ± 2-point GPA, the prevalence of academic progress was 14.7% (95% CI: 12.7–16.9%) and academic decline was 6.5% (95% CI: 5.1–8.2%). But, based on the LPA, the prevalence of academic progress was 2.47% ( n = 29) and academic decline was 2.98% ( n = 35). The traditional ≥ 2-point criterion showed specificity of 100% and PPV of 100% for both outcomes, but sensitivity was only 17.8% (95% CI: 12.3–24.5%) for progress and 53.0% (95% CI: 40.3–65.4%) for decline, with overall accuracy of 59% and 77%, respectively. ROC analysis identified an optimal overall cut-off of ≥ − 2.14 for academic decline (Se = 100%, Sp = 88.21%) and ≥ 2.41 for academic progress (Se = 100%, Sp = 91.45%). Quartile-specific cut-offs showed markedly improved specificity, reaching 100% in several subgroups. The currently used ± 2-point GPA criterion has near-perfect specificity but critically low sensitivity, missing the majority of students with genuine academic decline or progress. The LPA-derived quartile-specific cut-off points provide a superior evidence-based alternative with both high sensitivity and specificity. These findings have direct implications for the design of early-warning systems in medical education in Iran.