NCT06973733

Multimodal Database and Large Language Model for ARDS

Construction of ARDS Clinical Multimodal Database and ARDS-Specific Large Language Model

• Status: Not Yet Recruiting
• Phase: —
• Study type: Observational
• Enrollment: 6,500 (estimated)
• Sponsor: China-Japan Friendship Hospital · Academic / Other
• Sex: All
• Age: 18 Years
• Healthy volunteers: Not accepted

Summary

The goal of this observational study is to integrate existing clinical cohorts from the research team to establish dedicated cohorts for ARDS and high-risk ARDS patients (primarily SCAP patients), then systematically collect comprehensive clinical data and multi-omics biological samples to construct a high-quality multimodal ARDS database. Building upon this foundation, the research will develop an ARDS-specific large-scale disease model to assist clinical decision-making in early warning, diagnosis, and prognosis prediction. The main question it seeks to address is: Can the establishment of specialized ARDS cohorts and multimodal databases, combined with the development of an ARDS-specific large-scale disease model, effectively improve ARDS prediction rates, diagnostic accuracy, and reduce mortality rates, thereby enhancing overall clinical management standards?

Detailed description

ARDS is a common critical illness in ICUs with high mortality rates. Its prognostic factors are complex and multifaceted, including challenges in precise early warning, lack of early diagnostic biomarkers due to complicated pathogenesis, and difficulties in personalized precision treatment owing to high heterogeneity. Interventions targeting any single aspect are unlikely to improve overall outcomes. Only through systematic interventions addressing key aspects of ARDS - including assessment, early warning, diagnosis, phenotyping and treatment - can its mortality be significantly reduced. This study is a multicenter, retrospective and prospective observational cohort study with the following objectives: 1. To integrate existing ARDS-related cohorts from the research team (comprising 5,000 patients enrolled between January 1, 2014 and September 1, 2024) and prospectively recruit an additional 1,500 ARDS patients and ARDS high-risk individuals. This will establish a comprehensive cohort of no fewer than 6,500 cases, primarily including patient populations with conditions such as severe community-acquired pneumonia (SCAP) and other ARDS-associated disorders; 2. To collect comprehensive clinical data and multi-omics biological samples from these patients, constructing a high-quality multimodal ARDS database through rigorous data governance; 3. Based on this foundation, to develop and clinically validate an ARDS-specific large-scale disease model to assist in clinical decision-making for early warning, diagnosis, and prognosis prediction, thereby improving the overall standard of ARDS management.

Conditions

• Respiratory Distress Syndrome, Acute
• Severe Community-Acquired Pneumonia

Interventions

Type	Name	Description
BEHAVIORAL	Clinical data and biospecimen collection	Clinical Data Collection: Case report forms were utilized to systematically capture multimodal clinical data, including: demographic characteristics, clinical symptoms and physical signs, laboratory test results, chest imaging data, organ support parameters, pharmacological interventions , complications and clinical outcomes. Biospecimen Collection: ARDS patients underwent biospecimen collection at days 1, 4, and 7 post-diagnosis. High-risk ARDS cohorts provided specimens within 24 hours of ICU admission. Specimens included: peripheral blood, Sputum/BALF, stool and urine.

Timeline

• Start date: 2025-05-07
• Primary completion: 2027-12-31
• Completion: 2028-03-31
• First posted: 2025-05-15
• Last updated: 2025-05-15