Development and Validation of a Risk Prediction Model for Post-Discharge Post-Traumatic Stress Disorder in EICU Survivors

Brief Title: PTSD Risk Prediction Model for EICU Survivors

Official Title: Development and Validation of a Risk Prediction Model for Post-Discharge Post-Traumatic Stress Disorder in Emergency Department-Admitted EICU Survivors

Study Type: Observational

Brief Summary: This study aims to develop and validate a risk prediction model for post-traumatic stress disorder (PTSD) at 1 month after hospital discharge in adult survivors of emergency intensive care unit (EICU) care. PTSD is a common and serious psychological condition that can occur after a life-threatening medical event. Many EICU survivors experience persistent stress responses, sleep problems, avoidance behaviors, and difficulty returning to normal life after leaving the hospital. This prospective observational cohort study will enroll approximately 800 adult patients who are admitted to the EICU through the emergency department at Sir Run Run Shaw Hospital and other participating centers. Participants will be followed up at 1 month after EICU discharge to complete the Clinician-Administered PTSD Scale for DSM-5 (CAPS-5), a diagnostic interview for PTSD. Clinical data collected during the emergency department stay and EICU course-including vital signs, laboratory tests, treatments, and severity of illness-will be used to build statistical models to predict which patients are most likely to develop PTSD. The study does not involve any experimental treatments or additional procedures beyond routine clinical care. Results from this study may help healthcare providers identify high-risk patients early and provide timely psychological support and follow-up care.

Detailed Description: Background: Advances in critical care medicine have enabled an increasing number of emergency critical care patients to survive to ICU or hospital discharge. However, a substantial proportion of survivors experience persistent psychological sequelae, including post-traumatic stress disorder (PTSD), which significantly impairs quality of life. Patients admitted to the EICU via the emergency department are exposed to multiple stressors during the early phase-sudden life-threatening illness, pain, hypoxia, altered consciousness, mechanical ventilation, sedation, and separation from family-which may serve as important risk factors for subsequent PTSD. Objectives: To develop and validate a risk prediction model for predicting 1-month post-discharge PTSD in adult patients admitted to the EICU via the emergency department. Study Design: This is a prospective observational cohort study conducted at Sir Run Run Shaw Hospital, Zhejiang University School of Medicine (coordinating center) and multiple other centers. Study Population: Adult patients (≥18 years) admitted to the emergency department and subsequently transferred to the EICU, with an ICU length of stay of at least 24 hours, who survive to EICU discharge and have a Glasgow Coma Scale score ≥12 at discharge or follow-up. Sample Size: Approximately 800 patients (approximately 200 per center if 4 centers participate), calculated based on an estimated 25% PTSD prevalence at 1 month post-discharge, with 6-8 candidate predictors, expected Nagelkerke R² of 0.15, target shrinkage of 0.90, and accounting for 15%-20% attrition. Data Collection: All variables are routine clinical observation and laboratory parameters, categorized into six domains: (1) baseline susceptibility variables (demographics, social support, psychiatric history, trauma history, prior ICU admission, substance use, chronic disease burden); (2) emergency admission phase variables (mode of admission, severity, vital signs, etiology, early labs, emergency treatments); (3) first 24 hours in EICU variables (severity scores, organ support, sedation/analgesia, neurological status, nursing exposures, early complications); (4) cumulative EICU exposure variables (duration of organ support, cumulative sedation/analgesia, mechanical ventilation duration, physical restraints, delirium, infections, major complications); (5) disease process and treatment-related variables (severe infection, respiratory failure, AKI, ARDS, transfusion, emergency surgery, intubation, CPR); and (6) 1-month post-discharge follow-up (CAPS-5 diagnostic PTSD assessment). Primary Outcome Measure: CAPS-5-diagnosed PTSD at 1 month post-EICU discharge. Statistical Analysis: Three types of prediction models will be developed: (1) an early model using variables available within the first 6 hours from emergency department presentation to EICU admission; (2) a discharge-updated model adding ICU exposure, treatment interventions, complications, and pre-discharge information; and (3) a dynamic digital model further adding post-discharge electronic follow-up and digital phenotyping data. Candidate modeling approaches include logistic regression, LASSO/ElasticNet penalized regression, random forest, gradient boosting trees, XGBoost/LightGBM, and support vector machines. Model performance will be evaluated for di