Non-invasive respiratory support for acute hypoxemic respiratory failure (AHRF) remains controversial. Continuous positive airway pressure (CPAP) provides sustained positive end-expiratory pressure, while helmet interface (H-CPAP) may improve tolerance and re-duce air leaks. We aimed to evaluate clinical outcomes of H-CPAP delivered in a respira-tory intermediate care unit (RICU) by pulmonologists and identify early predictors of treatment failure. We conducted a prospective single-centre observational cohort study in-cluding consecutive adults with community-acquired pneumonia (CAP)-AHRF treated with H-CPAP in a RICU. H-CPAP success was defined as hospital discharge without in-vasive mechanical ventilation (IMV). Clinical, radiological, gas exchange, laboratory var-iables were collected at baseline and after 1 hour of H-CPAP. Among 1100 screened pa-tients, 198 met inclusion criteria. H-CPAP success occurred in 158/198 patients (79.8%), while overall in-hospital mortality was 9.1%. Success rates decreased with increasing AHRF severity. Lower haematocrit, bilateral pulmonary infiltrates, lower PaO₂/FiO₂ dur-ing H-CPAP, higher alveolar–arterial oxygen gradient, and higher APACHE II score were the main predictors of treatment failure. In selected patients with CAP-AHRF, H-CPAP delivered in a RICU was associated with high rates of avoidance of IMV and low in-hospital mortality. Early physiological and radiological variables identified patients at increased risk of H-CPAP failure with acceptable predictive performance.