Mechanisms of Electronic Health Record-Related Patient Safety Incidents: Errors and Risk Factors from 15 Years of Experience in Japan

• A refined analytical framework for health information technology is proposed, expanding the scope of contributing factors and enhancing the classification of human-machine interaction. • Human errors dominated EHR-related incidents, increasing notably during EHR development and transition periods. • Decision-making processes, safety culture, and rules emerged as key contributing factors, representing latent risks embedded in organisational structures. • Context-specific risks unique to healthcare settings, such as functional design and human-machine interaction, acted as direct risks triggering errors at the operational level. • Latent and direct risks operated through distinct mechanisms, highlighting the need for differentiated improvement strategies tailored to each pathway. Although electronic health records (EHRs) have improved care delivery efficiency, their use has introduced new categories of patient safety incidents. However, the risk mechanisms underlying EHR-related errors remain insufficiently understood. This study aimed to refine an existing health information technology risk framework and examine error patterns and associated risk mechanisms in Japanese EHR-related incidents, focusing on prominent factors, latent causes, and interrelationships. We analysed 621 incident reports (2010–2024) retrieved from a national database. Errors and risk factors were identified through dual independent coding, regular consensus meetings, and third-reviewer adjudication. Descriptive statistics characterised the data, multinomial logistic regression examined key factors, hierarchical clustering explored co-occurring risk factor patterns, and association rule mining analysed the most influential cluster. Human errors accounted for 91% and machine errors for 9%, Error patterns remained stable, although incident counts increased during the Development and Transition phases. Decision-making procedures were associated with all human error categories, with significant associations observed for wrong entry/retrieval errors (relative risk ratio RRR = 2.552, p < 0.05) and partial entry/retrieval errors (RRR = 3.764, p < 0.05). Rules and procedures also exhibited significant or marginal associations across human error categories. Within the core risk-factor cluster, safety culture and rules were prominent as key factors; association rule mining revealed notable co-occurrences involving decision-making procedures (lift=1.419) and inter-professional communication (lift=1.256). EHR design, implementation, and evaluation should focus on healthcare workers’ tasks and workflows to reduce direct risks, while also addressing system-level contributing factors that may generate latent risks.