To develop and evaluate machine learning models for insomnia classification using wearable-derived digital phenotypes, explicitly integrating clinical plausibility as a criterion in model selection.
Key Findings:
Wearable technologies provide objective data for sleep assessment, enhancing insomnia diagnosis.
Machine learning models must consider clinical plausibility alongside predictive performance to ensure trustworthy applications in healthcare.
Interpretation:
The study emphasizes the need for machine learning models in insomnia classification to align with established clinical knowledge, enhancing interpretability and clinical validity.
Limitations:
The study is limited to a single center, which may affect generalizability and the applicability of findings to broader insomnia populations.
Participants were selected based on specific criteria, potentially excluding individuals with diverse insomnia presentations.
Conclusion:
Integrating clinical plausibility in machine learning model selection is crucial for developing trustworthy applications in insomnia diagnosis, ensuring models are clinically relevant.
A VHA study across 11 vendors finds AI-generated primary care notes score lower than clinician-written notes, with the largest deficits in thoroughness, organization, and usefulness
