Machine Learning Model Reveals Determinators for Admission to Acute Mental Health Wards From Emergency Department...

Area of Science:

• Emergency Medicine
• Mental Health Services Research
• Machine Learning Applications

Background:

• Emergency departments (EDs) manage a high volume of patients with mental health (MH) concerns, including suicidality.
• Accurate prediction of acute MH ward admissions is crucial for resource allocation and timely patient care.
• Current MH triage in EDs presents challenges due to the complexity of mental health assessments.

Purpose of the Study:

• To identify key factors influencing admissions to acute mental health wards for patients presenting to EDs.
• To utilize machine learning (ML) models to predict the likelihood of acute MH ward admission.
• To provide insights into clinical and financial variations in MH service delivery.

Main Methods:

• Utilized existing ED data from January 1, 2016, to December 31, 2021.
• Applied machine learning models using the Interpretable Machine Learning (InterpretML) library in Python.
• Assessed feature importance based on mean absolute score to determine impact on admission likelihood.

Main Results:

• Patient's 'Age' and 'Triage category' were identified as the most significant predictors of MH ward admission.
• 'Facility identifier', 'Presenting problem', and 'Active Client' had less impact on admission decisions.
• Suicidal ideation, despite being a common presentation, showed a negative correlation with admission.

Conclusions:

• Age and triage category are critical factors in determining acute MH ward admissions from EDs.
• The nurse's role in triage is pivotal, highlighting the need for specialized MH assessment skills.
• Further research into ML applications is recommended to support clinicians in ED MH assessments.