Two-hour dynamic reassessment of noninvasive ventilation failure risk in acute exacerbation of chronic obstructive pulmonary disease with hypercapnic respiratory failure: development and temporal validation

Abstract

Background Early recognition of noninvasive ventilation (NIV) failure remains difficult in acute exacerbation of chronic obstructive pulmonary disease (AECOPD) with hypercapnic respiratory failure. We developed and temporally validated a model for structured reassessment after early NIV treatment and compared it with a pre-NIV model and the heart rate, acidosis, consciousness, oxygenation, and respiratory rate (HACOR) score. Methods The hospital information system identified consecutive NIV-treated AECOPD admissions from 2021 through 2025. After prespecified data-consistency exclusions, 924 first eligible hospitalizations were divided by calendar time into development and temporal-validation cohorts. The dynamic model used the first complete physiological assessment obtained 90–150 min after NIV initiation. Discrimination, calibration, clinical utility, and robustness were evaluated; detailed analytic methods and supporting results are provided in the Supplementary material. Results NIV failure occurred in 136 of 692 patients in development and 51 of 232 in temporal validation. The dynamic model retained good discrimination and close calibration in validation. Its numerical performance was slightly better than the pre-NIV model and HACOR-derived probability, but superiority was not established. Respiratory rate and level of consciousness at reassessment carried the clearest independent prognostic information, whereas the contributions of acidosis, oxygenation, and pneumonia were more context dependent. Conclusion The model provides a reproducible probability estimate for structured early treatment-response reassessment, but it does not replace HACOR, clinical judgment, or established escalation criteria. External validation and prospective impact evaluation are required.