Abstract
Aggression and conflict between nurses, patients, and patients' relatives have become
increasingly significant challenges within Nigerian emergency departments, creating
negative consequences for healthcare delivery, workplace safety, patient satisfaction, and
institutional accountability. These conflicts are often triggered by prolonged waiting times,
overcrowding, communication breakdowns, emotional distress, and resource constraints,
which can escalate into verbal abuse, threats, physical confrontations, and legal disputes.
Despite the growing prevalence of such incidents, there remains a limited availability of
intelligent systems capable of supporting early conflict identification and mediation within
healthcare environments. This study therefore aimed to develop a Natural Language
Processing (NLP)-based framework for mediating nurse–patient aggression disputes in
Nigerian emergency departments by identifying aggression patterns and supporting proactive
conflict resolution. The study employed a machine learning and NLP-based methodology
using a study population comprising 2,000 emergency department incident records. The
dataset contained structured variables including waiting time, crowding level, staff
availability, communication clarity, pain level, relative emotional state, and nurse response
time, alongside unstructured textual incident narratives. Descriptive statistics and
exploratory data analysis (EDA) were conducted to examine aggression patterns and identify
factors influencing dispute escalation. NLP techniques based on Term Frequency–Inverse
Document Frequency (TF-IDF) were applied to transform textual incident reports into
machine-readable representations, while a Random Forest classifier was used to predict
aggression severity levels. The analytical results revealed that medium aggression incidents
constituted the largest proportion of cases (61.85%), followed by high aggression incidents
(28.90%) and low aggression incidents (9.25%). The findings further indicated that
prolonged waiting times, overcrowding, communication challenges, and staffing conditions
were strongly associated with increased aggression severity. The developed model
demonstrated the potential of NLP-driven conflict analysis in detecting aggression indicators
from emergency department narratives and operational data. From a nursing perspective,
the framework provides a proactive mechanism for improving workplace safety, reducing
emotional stress among nurses, and enhancing the quality of patient-provider interactions.
From a mediation law perspective, the system supports early dispute resolution by identifyingconflict risks before they escalate into formal complaints, workplace violence, negligence
claims, or litigation. The study concludes that integrating NLP and machine learning into
emergency healthcare environments can significantly strengthen conflict management and
institutional accountability. It is therefore recommended that healthcare institutions adopt
intelligent mediation-support systems, strengthen communication protocols, improve patient
flow management, and incorporate alternative dispute resolution mechanisms into hospital
conflict management policies to promote safer and more collaborative healthcare
environments.