Computerized Adaptive Test vs. decision trees: Development of a support decision system to identify suicidal...

Area of Science:

• Mental Health Assessment
• Clinical Decision Support Systems
• Psychometrics

Background:

• Computerized Adaptive Tests (CATs) are used for mental health assessments.
• Standard CATs often overlook valuable data, such as suicide attempt history, potentially limiting precision.
• Integrating additional patient information could enhance CAT utility.

Purpose of the Study:

• To compare the performance of a standard CAT with a decision tree system for identifying suicidal behavior.
• To evaluate the impact of incorporating suicide attempt history into assessment models.
• To develop a more precise and efficient method for suicidal risk assessment.

Main Methods:

• Utilized items from a validated suicidal risk scale.
• Compared a standard CAT approach against a decision tree model.
• Included past suicide attempt history as a key variable in the decision tree classification.

Main Results:

• The decision tree model achieved comparable accuracy to a standard CAT but required fewer items (average of 4 vs. 9).
• The decision tree demonstrated an accuracy of 81.4% after 25 cross-validations.
• A shortened test was developed for effective separation of suicidal and non-suicidal patients.

Conclusions:

• Decision tree models can enhance the precision of suicidal risk assessment by incorporating a priori information.
• Standard CATs for suicidal risk assessment could be improved by integrating additional available data.
• The developed decision tree offers a more efficient and accurate approach to identifying suicidal behavior.