Exploratory Factor Analysis of the Driving Risk Checklist-25 Among Community-Dwelling Older Adults: Extraction of a Three-Factor Structure Based on Parallel Analysis and Its Clinical Interpretation

Background: Driving risk in older adults is multifactorial, and comprehensive on-road assessment is not always feasible in community practice. Although the Driving Risk Checklist-25 (DRCL-25) is easy to administer, interpretation based solely on the total score may obscure the type of driving difficulty. Objective: This study aimed to clarify the latent structure of the DRCL-25 in community-dwelling older adults and to propose a clinically interpretable three-factor solution based on parallel analysis. Methods: A total of 591 community-dwelling older drivers (mean age 75.9 ± 5.5 years), recruited through community-based long-term care prevention health checkups in Akita Prefecture, completed the DRCL-25. Internal consistency was examined using Cronbach’s alpha. The suitability of factor analysis was assessed using the Kaiser–Meyer–Olkin (KMO) statistic and Bartlett’s test of sphericity. The number of factors was examined by parallel analysis (1000 random datasets). Factors were extracted from the Pearson correlation matrix by maximum likelihood (ML) estimation and rotated using varimax rotation; an oblique (Promax) rotation was conducted as a sensitivity analysis. Loadings of |0.30| or greater were used to support interpretation. The three-factor model and the total-score model were compared with respect to their association with mild cognitive impairment (MCI) status, determined using the National Center for Geriatrics and Gerontology Functional Assessment Tool (NCGG-FAT). Results: Internal consistency was acceptable (Cronbach’s alpha = 0.767), the KMO value was 0.808, and Bartlett’s test was significant (χ2(300) = 2249.1, p < 0.001). A three-factor solution, supported by parallel analysis and interpretability, was interpreted as (1) driving burden in complex situations, (2) danger signs noticed by others, and (3) reduced attention and orientation, together explaining 21.2% of total item variance. Factor-specific reliability was acceptable for Factor 1 (α = 0.725) but low for Factor 2 (α = 0.459) and Factor 3 (α = 0.460). Oblique rotation indicated moderate-to-high inter-factor correlations (r = 0.56–0.63). The three-factor model and the total-score model showed comparable association with MCI status (AUC = 0.571 for both; this was a descriptive comparison, and no formal statistical test of the AUC difference was performed). Conclusions: The DRCL-25 showed an interpretable, although moderately inter-correlated, three-factor structure. Given its modest explained variance, the low reliability of two of the three factors, and the absence of incremental predictive validity over the total score, this structure should be regarded as provisional and hypothesis-generating, providing a conceptual and communicative framework rather than a validated basis for clinical decision-making at this stage.