Machine Learning Models to Identify Clinically Significant Anxiety in Short‐Term Insomnia Using Accelerometers
Open Access
- 1 January 2025
- journal article
- Published by Wiley in Depression and Anxiety
- Vol. 2025 (1) , 3082856
Abstract
Clinically significant anxiety (CSA) is common in individuals with short‐term insomnia. This study aims to explore the relationship between CSA and the subjective and objective parameters of sleep in patients with short‐term insomnia and construct machine learning (ML) models to determine the utility of accelerometer features in identifying significant anxiety. A total of 205 short‐term insomnia participants from China were assigned to the group with CSA ( N = 33) or the group without CSA ( N = 172). Interaction analysis based on linear regression was used to estimate the possible interactive effect of accelerometer features between CSA and sleep problems. Four feature sets and eight algorithms were used to construct ML models, with Shapley Additive exPlanations (SHAP) values used to visualize feature importance and influence processes. CSA in patients with short‐term insomnia leads to more severe subjective sleep problems, and accelerometer‐measured features warrant further attention for the identification of interactive factors. A significant interaction effect was found between anxiety symptoms and longer duration of physical activity on insomnia severity ( P interaction < 0.05). Anxiety symptoms and interdaily stability had an interactive association with sleep hygiene behaviors ( P interaction < 0.01). ML can process and analyze complex accelerometer features to identify CSA in patients with short‐term insomnia. Compared with other feature sets and algorithms, the XGBoost model with accelerometer‐measured features on weekdays more effectively identified CSA with area under the curve (AUC) value of 0.777. SHAP analysis results indicated that circadian rhythm features had significant contributions. Decision plots based on SHAP were applied to visualize the personalized risk factors for each patient and provide clinicians with more easily understandable and practical explanation methods that enhance clinical decision‐making. Trial Registration: Chinese Clinical Trial Registry identifier: ChiCTR2200062910Keywords
Funding Information
- National Key Research and Development Program of China (2021YFC2501500)
- National Natural Science Foundation of China (82271525)
- National Natural Science Foundation of China (82071488)
- Natural Science Foundation of Guangdong Province (2023A1515010456)
This publication has 59 references indexed in Scilit:
- Recommendations for a Standard Research Assessment of InsomniaSLEEP, 2006
- A Brief Measure for Assessing Generalized Anxiety DisorderArchives of internal medicine (1960), 2006
- Towards a valid, reliable measure of sleep effortJournal of Sleep Research, 2005
- Vulnerability to Stress-related Sleep Disturbance and HyperarousalSLEEP, 2004
- The Role of Actigraphy in the Study of Sleep and Circadian RhythmsSLEEP, 2003
- The PHQ-9Journal of General Internal Medicine, 2001
- Validation of the Insomnia Severity Index as an outcome measure for insomnia researchSleep Medicine, 2001
- Insomniacs' attributionsJournal of Psychosomatic Research, 2000
- A Behavioral Perspective on Insomnia TreatmentPsychiatric Clinics of North America, 1987
- The phenomenology of the pre-sleep state: The development of the pre-sleep arousal scaleBehaviour Research and Therapy, 1985