A hierarchical machine learning model for predicting self-harm and suicidal behavior in hospitalized patients with schizophrenia using clinical history and nursing observations - Scorecard - MDSpire
Advertisement
A hierarchical machine learning model for predicting self-harm and suicidal behavior in hospitalized patients with schizophrenia using clinical history and nursing observations
Clinical Scorecard: A Two-Tiered Machine Learning Approach to Forecast Self-Harm and Suicidal Actions in Schizophrenic Inpatients Based on Clinical Data and Nursing Assessments
At a Glance
Category
Detail
Condition
Schizophrenia with risk of self-harm and suicidal behaviour
Key Mechanisms
Integration of admission clinical data and longitudinal nursing assessments using machine learning
Target Population
Inpatients with schizophrenia
Care Setting
Psychiatric wards
Key Highlights
Developed a two-layered machine learning framework for risk assessment
Identified independent predictors of self-harm/suicidal behaviour
Achieved an AUC of 0.9048 with the hierarchical model
Outperformed flat combined-feature models in sensitivity
Supports timely preventive strategies in psychiatric care
Guideline-Based Recommendations
Diagnosis
Utilize admission clinical information and nursing observations for risk assessment
Management
Implement individualized preventive strategies based on risk scores
Monitoring & Follow-up
Regularly assess dynamic indicators of self-harm and suicidal behaviour
Risks
Consider factors such as previous self-harm, hopelessness, and negative self-evaluation
Patient & Prescribing Data
477 patients with schizophrenia hospitalized between July 2021 and July 2024
High-risk patients identified through machine learning can benefit from targeted interventions
Clinical Best Practices
Combine static and dynamic clinical data for comprehensive risk assessment
Regularly update risk assessments based on ongoing nursing evaluations
Utilize machine learning models to enhance clinical decision-making
