The goal of this retrospective observational study is to learn about how respiratory support is selected and adjusted in hospitalized patients with acute exacerbation of chronic obstructive pulmonary disease (AECOPD). The main questions it aims to answer are:

1. How are different respiratory support modalities selected for patients with AECOPD in real-world practice?
2. What clinical patterns are associated with escalation, de-escalation, switching, or discontinuation of respiratory support during hospitalization? Researchers will review de-identified hospital records of patients admitted with AECOPD and analyze the dynamic use of respiratory support during hospitalization, including conventional oxygen therapy, high-flow nasal cannula oxygen therapy, non-invasive ventilation, and invasive mechanical ventilation.

Participants will not receive any study-assigned intervention. The study will use existing clinical data collected during routine inpatient care and will not change diagnosis, treatment, or follow-up.

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.

Inclusion Criteria:

• Hospitalized patients with a documented diagnosis of acute exacerbation of chronic obstructive pulmonary disease.
• Patients identified retrospectively from hospital records during the predefined study period.
• Patients who received at least one form of respiratory support during the index hospitalization, including conventional oxygen therapy, high-flow nasal cannula oxygen therapy, noninvasive ventilation, or invasive mechanical ventilation.
• Patients with available clinical records containing sufficient dynamic physiological, laboratory, and treatment-related data for retrospective analysis.

Exclusion Criteria:

• Patients whose hospitalization was not primarily due to acute exacerbation of COPD, even if COPD was documented as a comorbidity.

Patients with stable COPD and no evidence of acute exacerbation during the index hospitalization.

• Patients admitted primarily for other acute conditions, including but not limited to pneumonia without documented AECOPD, acute heart failure, pulmonary embolism, pneumothorax, asthma exacerbation, trauma, or postoperative care.
• Patients with insufficient documentation to confirm AECOPD as the primary reason for admission.
• Patients with incomplete clinical records preventing retrospective analysis.
• Repeated hospitalizations of the same patient during the study period, with only the first eligible admission included.

The study population will be selected retrospectively from hospitalized patients with acute exacerbation of chronic obstructive pulmonary disease . The source population includes patients receiving routine inpatient care, with particular relevance to critically ill patients with high-quality intensive care data. Eligible participants will be identified from hospital clinical information systems and electronic medical records based on diagnosis, respiratory support records, and availability of dynamic clinical data collected during hospitalization.