Predictive Modeling of Preoperative Sleep Disorder Risk in Older Adults by Using Data From Wearable Monitoring Devices: Prospective Cohort Study

BACKGROUND: Sleep disorders are common among older adults undergoing surgery and contribute significantly to postoperative complications, delayed recovery, and higher health care costs. The combined effects of age-related physiological changes and surgical stress further disrupt sleep in this vulnerable group. However, current tools for predicting surgical risk rarely account for the specific physiological, clinical, and psychological factors that affect older patients. While wearable devices are used to monitor sleep, most prediction models focus on general sleep quality in nonsurgical populations, leaving a gap in forecasting preoperative sleep disorders in older surgical candidates. Therefore, we developed and validated a tailored risk prediction model that integrates objective sleep data from wearable devices with comprehensive clinical and psychosocial evaluations for older adults preparing for surgery. OBJECTIVE: We aimed to develop and validate a risk prediction model for preoperative sleep disorders in older adult surgical patients by using data from smart wearable devices and clinical assessments, thereby facilitating early identification of the influencing factors and providing a scientific basis for personalized care planning. METHODS: We conducted a prospective study at the Second Affiliated Hospital of Zunyi Medical University. A cohort of 242 older surgical patients was monitored using smart rings on the night before surgery. We simultaneously collected data on sociodemographic factors, cognition, and psychological status. As per preoperative sleep assessments, patients were classified into sleep disorder and non-sleep disorder groups. Independent predictors of sleep disorders were identified using univariable and multivariable logistic regression. These predictors were used to build a risk prediction model, which was internally validated with 1000 bootstrap samples. The model's performance was evaluated by its ability to discriminate between groups (using receiver operating characteristic curves), its calibration, and its clinical usefulness (via decision curve analysis). RESULTS: Multifactorial logistic regression analysis showed that Hospital Anxiety and Depression Scale score (odds ratio [OR] 3.21, 95% CI 1.54-6.69; P=.002), number of awakenings (OR 3.33, 95% CI 1.82-6.12; P