Identifying Nurses at Risk of Nursing Interruptions During Medication Administration Using Machine Learning: A...

Area of Science:

• Nursing Science
• Health Informatics
• Machine Learning in Healthcare

Background:

• Nursing interruptions during medication administration (NIMA) are a significant cause of medication errors.
• Validated tools for assessing NIMA risk are currently lacking.
• This study addresses the need for predictive models to identify nurses at high risk for NIMA.

Purpose of the Study:

• To develop and internally validate three machine learning models: logistic regression (LR), decision tree (DT), and Naive Bayes (NB).
• To identify individual nurses' risk of experiencing NIMA.
• To provide a foundation for targeted nursing interventions.

Main Methods:

• Recruited 4758 Chinese nurses from 12 tertiary hospitals.
• Developed and validated LR, DT, and NB models using training (80%) and internal validation (20%) sets.
• Evaluated model performance using AUC, accuracy, recall, specificity, precision, F1-score, and G-mean, with DeLong and Hosmer-Lemeshow tests for comparison and calibration.

Main Results:

• Over half (52.1%) of nurses experienced at least one NIMA event.
• Identified 18 predictive factors for NIMA, including department type, marital status, and self-efficacy.
• The LR model demonstrated the highest performance with an AUC of 0.748 and good calibration.

Conclusions:

• Successfully developed and internally validated three NIMA risk prediction models.
• The LR-based nomogram and web calculators show promise for risk stratification and management.
• Further external validation and feasibility assessments are recommended before clinical implementation.