Machine learning approached a 14-item shortened version of the Positive And Negative Sleep Appraisal Measure...

Area of Science:

• Psychology
• Machine Learning
• Sleep Science

Background:

• The Positive And Negative Sleep Appraisal Measure (PANSAM) is a comprehensive tool for evaluating sleep-related beliefs.
• Streamlining assessment tools is crucial for efficient clinical practice and research.
• Machine learning offers novel approaches to optimize psychometric instruments.

Purpose of the Study:

• To develop a shortened, accurate version of the PANSAM using machine learning.
• To identify the most predictive items for each PANSAM subscale.
• To create a simplified and interpretable scoring system for sleep appraisal.

Main Methods:

• Utilized eXtreme Gradient Boosting (XGBoost) to identify key items based on R² scores from 1,000 South Korean participants.
• Employed Symbolic Regression-Based Clinical Score Generator (SymScore) for optimal weight assignment and score generation.
• Validated the predictive accuracy of selected items and the overall scoring system.

Main Results:

• Developed the PANSAM-14, comprising 14 highly representative items across four subscales.
• Achieved high predictive accuracy for subscale scores (R² ranging from 0.92 to 0.94).
• The SymScore-based PANSAM-14 demonstrated comparable performance to the XGBoost model in predicting total scores (R² ranging from 0.93 to 0.95).

Conclusions:

• The SymScore-based PANSAM-14 is a highly accurate and efficient tool for assessing dysfunctional sleep beliefs.
• This shortened version maintains reliability and validity, offering a practical alternative for sleep appraisal.
• The study highlights the potential of machine learning in refining psychological assessment instruments.