Beyond composite scores in chronotype assessment: item-level predictive patterns in the Morningness-Eveningness...

Area of Science:

• Chronobiology and Sleep Science
• Cognitive Neuroscience
• Computational Psychiatry

Background:

• Chronotype, individual differences in circadian timing, impacts sleep, cognition, and health.
• Current chronotype assessment tools, like the Morningness-Eveningness-Questionnaire (MEQ), use composite scores with limitations.
• Methodological concerns exist regarding multidimensionality and unequal item weighting in traditional chronotype questionnaires.

Purpose of the Study:

• To enhance theoretical understanding and empirical precision in chronotype assessment using machine learning.
• To identify item-level predictive hierarchies and response patterns for distinguishing chronotype variations.
• To develop a more efficient and reliable method for chronotype assessment.

Main Methods:

• Applied machine learning techniques to the German version of the MEQ in the Dortmund Vital Study.
• Analyzed item-level predictive hierarchies and response patterns using prospective cohort data.
• Utilized partial dependence analysis to identify non-linear relationships in chronotype classification.

Main Results:

• Item 19 (self-assessed chronotype) showed significantly higher predictive utility than other items.
• Distinct predictive patterns and item combinations were identified for morning, neutral, and evening chronotypes.
• A six-item combination achieved robust classification, potentially reducing assessment burden by 70%.

Conclusions:

• Machine learning reveals non-linear and heterogeneous mechanisms in chronotype classification.
• A brief, self-assessment-based approach offers a promising avenue for efficient chronotype screening.
• Findings support the development of abbreviated tools for clinical and research applications, with an R tutorial provided for reproducibility.