<![CDATA[Railway accidents remain a persistent transportation-safety challenge in Indonesia, yet quantitative evidence linking operational factors to accident characteristics and severity is limited. This study analyzes 92 railway-accident investigation reports published by the National Transportation Safety Committee (KNKT) for 2007–2024, constituting a census of all publicly accessible KNKT railway reports meeting the inclusion criteria at the time of data collection. Chi-square tests of independence and Cramér's V examined associations between three operational factors — human, infrastructure, and rolling stock — and accident type, with analytic samples ranging from N=61 to N=72 depending on data completeness for each factor. An Equivalent Property Damage Only (EPDO) severity score (N=68, complete-case) was compared across accident types (Kruskal-Wallis) and operational factors (Mann-Whitney U). Infrastructure factors showed a strong association with accident type (χ²=26.263, df=3, p<0.0001; Monte Carlo p<0.0001; Cramér's V=0.604), and human factors a moderate-to-strong association (χ²=12.273, df=3, p=0.0065; Monte Carlo p=0.0032; V=0.449); the rolling-stock association was not significant under the primary coding (p=0.074) but reached significance in a sensitivity analysis (p=0.033). Severity differed significantly across accident types (H=20.812, df=3, p=0.0001, ε²=0.278), with train collisions showing the highest severity, significantly higher than derailments and operational disruptions in post-hoc comparisons. No statistically significant bivariate association was found between the three operational factors and severity (all p>0.05); given the modest sample sizes involved, this is interpreted as inconclusive rather than confirmatory evidence of no relationship. The findings suggest operational factors relate more strongly to the mechanism by which an accident occurs than to its resulting severity, with cautious implications for infrastructure-focused safety prioritization, interpreted in light of the study's observational design and data limitations]]>
