Prediction of non-invasive mechanical ventilation (NIV) and high velocity nasal insufflation (HVNI) outcomes using lung ultrasound: an exploratory pilot cohort study

Abstract

Background
Early identification of patients likely to respond to non-invasive ventilation (NIV) or high-velocity nasal insufflation (HVNI) for acute respiratory failure (ARF) remains challenging. This exploratory pilot study evaluated the utility of lung ultrasound (LUS) in predicting clinical outcomes in this setting.


Methods
In this prospective observational cohort study, 61 consecutive patients with ARF requiring NIV or HVNI in a single intensive care unit underwent serial LUS examinations. Lung aeration profiles (A, B, and AB) were recorded at baseline and 2 h post-initiation, and the profiles were then correlated with clinical outcomes at Day 3 or discharge.


Results

Following Bonferroni correction for multiple testing, only Profile AB demonstrated a statistically significant overall change over three days (
p
 = 0.014), while Profiles A and B showed nonsignificant directional trends. Early post-initiation LUS assessment demonstrated high predictive value for clinical improvement, yielding a sensitivity of 92.5%, specificity of 87.5%, and overall accuracy of 91.8%, with substantial agreement with clinical assessment (κ = 0.690). Baseline profiles alone lacked significant predictive utility.



Conclusions
Dynamic early changes in lung aeration, rather than static baseline profiles, provide the primary predictive signal for clinical improvement in ARF. Serial LUS serves as a promising complementary bedside tool to guide timely decisions regarding the continuation or escalation of non-invasive respiratory support.