Medication‑Related Errors Among Nurses by Unit Adaptation Levels: A Bayesian Network–Based Exploratory Study (Preprint)

Background: Medical errors occur more frequently in health care than in other industries due to challenges in patient safety education for nurses and students. To address this, it is important to identify the factors and structures underlying clinical errors and apply these insights to educational programs. Medication-related errors are highly preventable with appropriate interventions, highlighting the importance of data-driven safety education. Previous research suggests that unit adaptation, rather than clinical experience alone, plays a critical role in error occurrence. Focusing on "adaptive performance," an underexplored concept in nursing, can help identify new educational strategies and interventions. Objective: This hypothesis-generating study used Bayesian network modeling to examine how unit experience-and, secondarily, total nursing experience-relates to the structure of medication-related errors, with implications for data-driven patient safety education and nursing student instruction. Methods: This mixed methods study conducted a qualitative root cause analysis of medication error reports to identify causal factors. Bayesian network modeling, an artificial intelligence-based approach, was used to visualize error-generation flows and compare models based on years of experience within the current unit. Data were obtained from 2023 medication-related error reports submitted by nurses to the Japan Council for Quality Health Care. Results: Among 119 medication-related incidents, the largest proportion occurred in internal medicine departments (n=47 cases, 39.5%), followed by surgical departments (n=27 cases, 22.7%), the intensive care unit (n=10 cases, 8.4%), and the emergency room (n=7 cases, 5.9%). Most incidents occurred in patient rooms or wards (n=97 cases, 81.5%). Root cause analysis identified 10 types of medication-related events and 23 contributing factors. Cases were categorized into low-, moderate-, and high-adaptation groups based on years of experience in the current unit. In the low-adaptation group (n=33), 69.7% of nurses had ≥5 years of total nursing experience, indicating that the group did not consist solely of novice nurses. The average entropy of incident events across the Bayesian network models ranged from 0.37 to 0.78, suggesting moderate to relatively high uncertainty in the inferred error structures. The moderate- and high-adaptation models exhibited fewer complex error networks, with weaker chains of unsafe conditions or actions than the low-adaptation model. The extensive clinical experience did not always prevent errors; rather, it was often linked to lapses in verification behavior. Conclusions: Unit adaptation appears to play a critical role in shaping error pathways. Nurses with lower unit adaptation demonstrated more complex error structures, indicating higher vulnerability to incident chains. In contrast, total years of nursing experience did not uniformly reduce risk and may even relate to lapses in verification behaviors.

Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.