Decision-making and Optimization of Ventilatory Support for AECOPD Patients

Acute exacerbation of chronic obstructive pulmonary disease (AECOPD) is a major cause of hospitalization, respiratory failure, and poor clinical outcomes in patients with chronic obstructive pulmonary disease. Respiratory support is a key component of acute management and commonly includes conventional oxygen therapy (COT), high-flow nasal cannula oxygen therapy (HFNC), non-invasive ventilation (NIV), and invasive mechanical ventilation (IMV). In clinical practice, respiratory support decisions are not static; rather, they involve a dynamic and sequential process of initial selection, reassessment, escalation, de-escalation, switching, and withdrawal according to changes in patient condition. Inappropriate timing of escalation or unnecessary use of higher-intensity support may contribute to treatment failure and unfavorable outcomes. This study is a retrospective observational study based on routinely collected clinical data from hospitalized patients with AECOPD. Using historical data extracted from hospital clinical information systems and electronic medical records, the study will systematically characterize respiratory support pathways during hospitalization and examine how support strategies are adjusted over time in response to changes in disease status. The study will not introduce any additional tests, procedures, or interventions, and will not alter prior or current clinical care. The analytical framework of the study is based on sequential decision analysis. Patient condition will be reassessed at fixed time intervals during hospitalization using key physiological and monitoring parameters recorded in routine care. These indicators will be integrated into clinically interpretable disease-state strata, and each state will be linked to the respiratory support modality being used at that stage. The respiratory support decision at the subsequent stage will then be characterized as maintenance, escalation, de-escalation, switching, or discontinuation. Under clinically reasonable safety and feasibility constraints, retrospective data will be used to summarize the statistical patterns of disease evolution and outcomes associated with different respiratory support choices. Dynamic programming methods will be applied to explore interpretable sequential decision patterns aligned with relevant clinical goals, such as reducing treatment failure and intubation-related adverse outcomes, while avoiding unnecessary use of high-intensity support and excessive switching between modalities. Data used in this study will be derived entirely from existing inpatient records and will include demographic and baseline information, disease- and hospitalization-related information, key physiological and monitoring indicators, respiratory support records, and outcome-related information such as treatment escalation, intubation, and in-hospital outcomes. All data will be retrospectively extracted and de-identified before analysis. Only authorized research personnel will have access to the study database. Study findings will be reported in aggregate form, and no personally identifiable information will be disclosed. Because this study uses previously collected clinical data only, does not involve direct participant contact, and does not affect clinical management, a waiver of informed consent will be sought in accordance with institutional ethics requirements.