H-CPAP Dataset: A Clinical Dataset of COVID-19 and Non-COVID Pneumonia Patients with Respiratory Failure Treated with Helmet Continuous Positive Airway Pressure

The Helmet Continuous Positive Airway Pressure (H-CPAP) dataset contains clinical records from 685 admissions with Acute Hypoxemic Respiratory Failure (AHRF) or Acute Respiratory Distress Syndrome (ARDS) caused by pneumonia, and treated with H-CPAP. Collected at Vimercate Hospital (Italy), the data integrate a COVID-19 cohort (2020–2021) and a non-COVID cohort (2010–2025). The final dataset includes 28 variables, such as demographics, comorbidities, vital signs, and respiratory parameters. Large Language Models are employed to extract clinical information from unstructured notes.Clinical validation confirms that the data are consistent with known pathophysiology, showing expected correlations between respiratory indices and therapy failure. This resource is structured for Machine Learning applications to predict H-CPAP failure and support clinical decision-making in intensive care.