Test anxiety predictors inventory (tapi): development and initial validation of a predictor-oriented instrument for medical students

Abstract

Abstract Background Test anxiety (TA) is common among medical students and may adversely affect mental well-being and academic performance. However, existing instruments such as the Test Anxiety Inventory (TAI) predominantly quantify the severity and manifestations of anxiety (outcomes) rather than identifying modifiable, intervention-relevant antecedents (predictors). This study aimed to develop and validate the Test Anxiety Predictors Inventory (TAPI) to support concurrent identification of at-risk students and to inform domain-specific, targeted interventions in resource-constrained medical training settings. Methods A cross-sectional census survey was conducted at Can Tho University of Medicine and Pharmacy (Dec 2024–Mar 2025; 382/389 valid responses, 98.2%). A 30-item pool was piloted in 71 students, then refined through iterative EFA (3 rounds; ML extraction, Direct Oblimin) and CFA (ML and DWLS estimators). Internal consistency, convergent and discriminant validity, correlation with TAI, and concurrent classification performance (multivariable logistic regression; ROC-AUC) were assessed. Because census sampling yielded one dataset, EFA/CFA reflect internal testing pending external validation; ‘prediction’ denotes concurrent classification, not temporal forecasting. Results TA prevalence (TAI ≥ 48) was 67.0% (95% CI 62.15–71.54). The final 14-item TAPI loaded on three factors—EAS (7), AMF (4), SRP (3)—explaining 77.99% of variance (KMO = 0.935). The correlated three-factor CFA showed good fit (CFI = 0.957; RMSEA = 0.083). The second-order DWLS model showed excellent fit but produced a Heywood warning with negative latent variance for EAS; the correlated three-factor model is therefore the more defensible representation. Reliability was high (α/ω = 0.89–0.95; CR = 0.92–0.96; AVE = 0.74–0.82; HTMT < 0.85). TAPI–TAI correlation was r = 0.576; classification model AUC = 0.804, sensitivity = 0.938, specificity = 0.452. Conclusions TAPI shows promising internal structure and concurrent classification performance. High sensitivity supports first-stage screening, while modest specificity indicates it should complement, not replace, formal diagnostic procedures.