Diaphragmatic dysfunction by ultrasound and HACOR score for early prediction of noninvasive ventilation failure in patients with acute exacerbation of chronic obstructive pulmonary disease

The effect of noninvasive ventilation (NIV) in patients with acute respiratory failure (ARF) after cardiac surgery is controversial. This study identified the feasibility of NIV and assessed the risk factors of NIV failure in patients with ARF after cardiac surgery. We retrospectively reviewed data from 112 patients with ARF requiring NIV and categorized them into the NIV failure and success groups. Patient data were extracted for further analysis, the primary outcomes were the need for endotracheal intubation and NIV-related in-hospital mortality. The risk factors for NIV failure in patients with post-extubation ARF was analyzed. The median time from extubation to NIV was 11 hours. No difference in the EuroSCORE existed between the two groups. NIV failed in 38.4% of the patients. The NIV failure group had a higher in-hospital mortality and stay at the longer intensive care unit (ICU). Most cases of NIV failure occurred within 1-48 hours of the treatment. The main causes of early NIV failure were a weak cough reflex and/or excessive secretions and hemodynamic instability. A Sequential Organ Failure Assessment (SOFA) score ≥10.5, vasoactive-inotropic score ≥6, and pneumonia were predictors of NIV failure, whereas a body mass index (BMI) ≥25.0 kg/m2 predicted NIV success. NIV was effective in the study population. Multiple organ dysfunction, pneumonia, and significant inotropic drug support before NIV were associated with NIV failure, whereas a BMI ≥25 kg/m2 was a predictor of NIV success.

IntroductionPrompt recognition of impending Non-Invasive Ventilation (NIV) failure is key for mitigating mortality in patients with respiratory insufficiency. While the HACOR (Heart rate, Acidosis, Consciousness, Oxygenation, Respiratory rate) scoring system has demonstrated potential, additional evidence is required from varied patient settings.1,2 This investigation aimed to validate how well the HACOR score predicts NIV failure among individuals with hypercapnic respiratory failure.ObjectivesTo evaluate the HACOR scoring system's ability to predict Non-Invasive Ventilation (NIV) failure among patients with hypercapnic respiratory failure in a multi-facility urban healthcare setting.Materials and methodsThis retrospective study included patients hospitalized for hypercapnic respiratory failure at various facilities in the New York City area from July 2022 to June 2023. HACOR assessments were performed 1–2 hours after NIV initiation. The primary outcome was NIV failure, defined as the requirement for intubation or death. Patient characteristics were reported as means (± SD) for continuous data and frequencies for categorical variables. Group comparisons between NIV success and NIV failure cohorts were analyzed via t-tests and chi-square tests. ROC analysis quantified the scoring system's discriminatory power, and the Youden Index determined the optimum cut point for predicting NIV failure. A p-value < 0.05 indicated statistical significance.ResultsAmong the included population (mean age 66.7 ± 13.3 years; 51% female), 29.49% experienced NIV failure. Of the 41 patients who failed NIV, 36.58% did not survive. In contrast, 98 participants were classified as having successful NIV outcomes. Notably, patients who progressed to NIV failure exhibited an elevated average HACOR score of 9.89 ± 4.1, significantly exceeding the mean of 2.78 ± 1.6 observed in the successful group (t = -9.85, p = 4.34 × 10^(-18)). The ROC curve analysis produced an area under the curve (AUC) of 0.83, denoting solid predictive performance. Setting the cutoff score at 9 identified a high-risk subgroup, of whom 83.3% proceeded to NIV failure. This data aligns with the findings from our previous study, validating the utility of the HACOR score in predicting NIV failure in patients with COPD exacerbations.Discussions/ConclusionsIn this diverse urban healthcare network, elevated scores on the studied clinical measure strongly correlated with an increased likelihood of NIV failure in hypercapnic respiratory failure patients. The AUC of 0.80 supports the score's capacity to differentiate those at high risk, and a threshold of 9 emerged as particularly relevant for timely intervention. By recognizing patients with higher scores early, clinicians may be better positioned to implement strategies that enhance patient outcomes.1–3

To determine whether sleep quality helps to predict noninvasive ventilation outcome in patients with acute hypercapnic respiratory failure. Despite an initial clinical improvement, nearly one fourth of patients may fail noninvasive ventilation after several days. Because late intubation is associated with a poor prognosis, it may be useful to identify factors that may predict or explain late noninvasive ventilation failure. We prospectively studied 27 hypercapnic patients in a medical intensive care unit who required noninvasive ventilation for >48 hrs. A 17-hr sleep polysomnography (3 PM-8 AM) was recorded 2 days to 4 days after noninvasive ventilation initiation. Late noninvasive ventilation failure was defined as death, endotracheal intubation, or persistent need for noninvasive ventilation on day 6. An abnormal electroencephalographic pattern that eluded analysis by standard sleep-scoring criteria was noted in seven (50%) of the 14 patients with late noninvasive ventilation failure compared with one (8%) of the 13 patients successfully treated with noninvasive ventilation (p = .03). No clinical or laboratory variables explained the electroencephalographic differences. Patients failing noninvasive ventilation had poorer sleep quality with greater circadian sleep-cycle disruption and less nocturnal rapid eye movement sleep (6 mins [range, 0-12] vs. 26 mins [range, 6-49], p = .03), compared with patients successfully treated with noninvasive ventilation. Noninvasive ventilation failure was associated with delirium during the intensive care unit stay (64% vs. 0%). Late noninvasive ventilation failure in elderly patients with acute hypercapnic respiratory failure was associated with early sleep disturbances including an abnormal electroencephalographic pattern, disruption of the circadian sleep cycle, and decreased rapid eye movement sleep.

Risk factors for noninvasive ventilation failure in patients with acute cardiogenic pulmonary edema: A prospective, observational cohort study

ObjectiveTo explore the association between PaCO2 and noninvasive ventilation (NIV) failure in patients with hypoxemic respiratory failure.MethodsA retrospective study was performed in a respiratory ICU of a teaching hospital. Patients admitted to ICU between 2011 and 2019 were screened. We enrolled the patients with hypoxemic respiratory failure. However, patients who used NIV due to acute-on-chronic respiratory failure or heart failure were excluded. Data before the use of NIV were collected. Requirement of intubation was defined as NIV failure.ResultsA total of 1029 patients were enrolled in final analysis. The rate of NIV failure was 45% (461/1029). A nonlinear relationship between PaCO2 and NIV failure was found by restricted cubic splines (p = 0.03). The inflection point was 32 mmHg. The rate of NIV failure was 42% (224/535) in patients with PaCO2 >32 mmHg. However, it increased to 48% (237/494) in those with PaCO2 ≤ 32 mmHg. The crude and adjusted hazard ratio (HR) for NIV failure was 1.36 (95%CI:1.13–1.64) and 1.23(1.01–1.49), respectively, if the patients with PaCO2 >32 mmHg were set as reference. In patients with PaCO2 ≤ 32 mmHg, one unit increment of PaCO2 was associated with 5% reduction of NIV failure. However, it did not associate with NIV failure in patients with PaCO2 >32 mmHg.ConclusionsPaCO2 and NIV failure was nonlinear relationship. The inflection point was 32 mmHg. Below the inflection point, lower PaCO2 was associated with higher NIV failure. However, it did not associate with NIV failure above this point.

BackgroundUltrasound (US) evaluation of diaphragmatic dysfunction (DD) has proved to be a reliable technique in critical care. In this single-center prospective study, we investigated the impact of US-assessed DD on noninvasive ventilation (NIV) failure in patients with acute exacerbations of chronic obstructive pulmonary disease (AECOPD) and its correlation with the transdiaphragmatic pressure assessed using the invasive sniff maneuver (Pdi sniff).MethodsA population of 75 consecutive patients with AECOPD with hypercapnic acidosis admitted to our respiratory intensive care unit (RICU) were enrolled. Change in diaphragm thickness (ΔTdi) < 20% during tidal volume was the predefined cutoff for identifying DD+/− status. Correlations between ΔTdi < 20% NIV failure and other clinical outcomes were investigated. Correlation between ΔTdi and Pdi sniff values was analyzed in a subset of ten patients.ResultsDD+ patients had a higher risk for NIV failure than DD− patients (risk ratio, 4.4; p < 0.001), and this finding was significantly associated with higher RICU, in-hospital, and 90-day mortality rates; longer mechanical ventilation duration; higher tracheostomy rate; and longer RICU stay. Huge increases in NIV failure (HR, 6.2; p < 0.0001) and 90-day mortality (HR, 4.7; p = 0.008) in DD+ patients were found by Kaplan-Meier analysis. ΔTdi highly correlated with Pdi sniff (Pearson’s r = 0.81; p = 0.004). ΔTdi < 20% showed better accuracy in predicting NIV failure than baseline pH value and early change in both arterial blood pH and partial pressure of carbon dioxide following NIV start (AUCs 0.84 to DTdi < 20%, 0.51 to pH value at baseline, 0.56 to early change in arterial blood pH following NIV start, and 0.54 to early change in partical pressure of carbon dioxide following NIV start, respectively; p < 0.0001).ConclusionsEarly and noninvasive US assessment of DD during severe AECOPD is reliable and accurate in identifying patients at major risk for NIV failure and worse prognosis.

BackgroundThe ratio of SpO2/FiO2 to respiratory rate (ROX) index is commonly used to predict the failure of high-flow nasal cannula. However, its predictive power for noninvasive ventilation (NIV) failure is unclear.MethodsThis was a secondary analysis of a multicenter prospective observational study, intended to update risk scoring. Patients with de novo acute respiratory failure were enrolled, but hypercapnic patients were excluded. The ROX index was calculated before treatment and after 1–2, 12, and 24 h NIV. Differences in predictive power for NIV failure using the ROX index, PaO2/FiO2, and PaO2/FiO2/respiratory rate were tested.ResultsA total of 1286 patients with de novo acute respiratory failure were enrolled. Of these, 568 (44%) experienced NIV failure. Patients with NIV failure had a lower ROX index than those with NIV success. The rates of NIV failure were 92.3%, 70.5%, 55.3%, 41.1%, 35.1%, and 29.5% in patients with ROX index values calculated before NIV of ≤ 2, 2–4, 4–6, 6–8, 8–10, and > 10, respectively. Similar results were found when the ROX index was assessed after 1–2, 12, and 24 h NIV. The area under the receiver operating characteristics curve was 0.64 (95% CI 0.61–0.67) when the ROX index was used to predict NIV failure before NIV. It increased to 0.71 (95% CI 0.68–0.74), 0.74 (0.71–0.77), and 0.77 (0.74–0.80) after 1–2, 12, and 24 h NIV, respectively. The predictive power for NIV failure was similar for the ROX index and for the PaO2/FiO2. Likewise, no difference was found between the ROX index and the PaO2/FiO2/respiratory rate, except at the time point of 1–2 h NIV.ConclusionsThe ROX index has moderate predictive power for NIV failure in patients with de novo acute respiratory failure.

&lt;b&gt;Introduction: &lt;/b&gt; Noninvasive ventilation (NIV) has been developed to reduce complications associated with invasive ventilation (IV). Failure of NIV and delay in endotracheal intubation can increase patients’ morbidity and mortality. Thus early determination of patients who are unlikely to benefit from NIV is crucial for their management. We aimed in this study to identify the early predictors of success of NIV in children with acute respiratory failure (ARF).&lt;br /&gt; &lt;b&gt;Material and methods: &lt;/b&gt; Fifty patients with ARF who fulfilled the study selection criteria were ventilated non-invasively and were assessed initially for their severity of critical illness by the Pediatric Logistic Organ Dysfunction (PELOD) score. Clinical, gasometric, respiratory mechanics and oxygenation indices were assessed at 0, 30 and 60 min and 4 and 24 h from the start of NIV. The success group was identified by reduction in respiratory effort, reduction in oxygen demand, improvement in gasometric parameters, and avoidance of intubation.&lt;br /&gt; &lt;b&gt;Results&lt;/b&gt;: Sixty-two percent of patients had successful NIV. Neither type of ARF nor patients’ demographics affected the outcome of NIV. The success rate was 80% among patients with mild to moderate acute respiratory distress syndrome (ARDS), 20% with severe ARDS, and 71.8% in patients with bronchopneumonia. Multivariate analysis revealed that baseline PELOD score of less than 14.5 ±2.7, SpO2/FiO2 ratio more than 208 ±57, oxygenation index (OI) 7 ±3.4 and mean airway pressure (MAP) 8.6 ±1.3 are independent predictors for success of NIV.&lt;br /&gt; &lt;b&gt;Conclusions&lt;/b&gt;: The NIV is a promising respiratory support modality in pediatric ARF. Baseline degree of critical illness and saturation oxygenation indices together with MAP change after the 1st h from the NIV trial represented the best predictors of success of the trial in the current study.

It is unclear whether procalcitonin (PCT) is correlated with noninvasive ventilation (NIV) failure. This retrospective case–control study aimed to compare PCT levels, C-reactive protein (CRP) levels, and PaCO2 in patients (05/2014–03/2015 at the Harrison International Peace Hospital, China) with acute exacerbation of chronic obstructive pulmonary disease (AECOPD) and NIV failure/success.This was a retrospective case–control study of patients with AECOPD who required NIV between May 2014 and March 2015. All consecutive patients with AECOPD admitted at the Department of Critical Care Medicine and transferred from the general ward were included in the study. Hemogram, PCT, erythrocyte sedimentation rate (ESR), arterial blood gas (ABG), and CRP levels were measured ≤1 hour before NIV was used. NIV was considered to have failed if at least one of the following criteria was met: cardiac arrest or severe hemodynamic instability; respiratory arrest or gasping; mask intolerance; difficulty in clearing bronchial secretions; or worsening of ABGs or sensorium level during NIV. The factors associated with NIV failure were determined.A total of 376 patients were included: 286 with successful NIV and 90 wither NIV failure. The multivariate analysis showed that PCT (OR = 2.0, 95%CI: 1.2–3.2, P = .006), CRP (OR = 1.2, 95%CI: 1.1–1.3, P < .001), and PaCO2 (OR = 1.1, 95%CI: 1.1–1.2, P < .001) ≤1 hour before NIV were independently associated with NIV failure. The optimal cutoff were 0.31 ng/mL for PCT (sensitivity, 83.3%; specificity, 83.7%), 15.0 mg/mL for CRP (sensitivity, 75.6%; specificity, 93.0%), and 73.5 mm Hg for PaCO2 (sensitivity, 71.1%; specificity, 100%). The area under the curve (AUC) was 0.854 for PCT, 0.849 for CRP, and 0.828 for PaCO2. PCT, CRP, and PaCO2 were used to obtain a combined prediction factor, which achieved an AUC of 0.978 (95%CI: 0.961–0.995).High serum PCT, CRP, and PaCO2 levels predict NIV failure for patients with AECOPD. The combination of these three parameters might enable even more accurate prediction.

Dear editor, We read with interest the study by Yalcinsoy et al1 appreciating its relevance and clinical practice value. In the last decades, noninvasive ventilation (NIV) has revolutionized the management of acute respiratory failure (ARF) reducing the need for endotracheal intubation and its associated complications and also reducing the complications associated with a stay in the intensive care unit, the length of hospital stay, and mortality.2 Several studies investigated the factors associated with NIV failure in order to identify the high-risk subset of patients who are likely to fail a trial of NIV. Moreover, NIV has been proven as an effective modality in the management of ARF with a success rate significantly higher for ARF due to chronic obstructive pulmonary disease (COPD) than other causes of ARF.3 The work of Yalcinsoy et al1 reported the effectiveness of NIV in moderate and severe ARF from COPD treated on respiratory wards. As observed in other studies, the factors predicting success or failure with NIV in hypercapnic respiratory failure include pH at admission, pH after 1 hour of NIV trial, and severity of underlying illness.3–5 This study,1 as well as confirming the success of treatment with NIV for COPD patients with moderate and severe ARF, has interestingly pointed out the predictive value for NIV failure of delta pH value <0.30 and pH <7.31 after 2 hours of NIV application, rather than the initial values of pH. This result may support the need to try initially an NIV treatment in almost all patients with moderate and severe ARF and assess the effectiveness after 1 or 2 hours. We consider that some key points are needed to consider for a proper clinical extrapolation. First, a limitation of this retrospective study is the unavailability of the severity of the underlying illness as assessed by the APACHE II score or similar scoring systems, as predictors of NIV failure.5 Second, regarding the severity of acidosis and gas exchange: intriguing data not reported in the study involve the onset of hypercapnia – acute or chronic. Yalcinsoy et al report the efficacy of NIV on acidosis seems to be better in patients treated with NIV at home (~50%). However, in this study, nonresponders patients have values of pCO2 (76.8) higher than responders (69.6), compared with pH values that are similar (7.26 vs 7.27). Is it only a matter of adaptation to the NIV? In addition, initial values of bicarbonates and their change over time are not reported. It could be relevant to know the authors’ opinion about the potential role of bicarbonate and creatinine values and their variation over next time in identifying NIV success.6 Third, another discussed and interesting confirmation of the study is the relationship between reasons of ARF and NIV response. Comorbidities and reasons of ARF result are not significant in the NIV success or failure.1 We strongly agree with the authors about experienced staff being essential in achieving an NIV success in patients with ARF, evaluating failure criteria after a few hours, and applying the possible corrections in ventilation parameters for improved adaptation of the patient. In real-life setting, the use of the NIV in patients with ARF and severe acidosis is greatly increased in the respiratory ward.7 So, it is crucial to identify in a few hours patients at risk of NIV failure or intubation through the assessment of practical and fast clinical parameters. Further clinical trials need to define a solid tool for NIV applications for severe COPD in wards.

Acute respiratory failure is among the sequelae of complications that can develop in response to severe sepsis. Research into sepsis-related respiratory failure has focused on ARDS and invasive mechanical ventilation. We studied the factors associated with success and failure of noninvasive ventilation (NIV) in the treatment of sepsis-related acute respiratory failure. This retrospective study included 136 subjects with a diagnosis of acute respiratory failure and intrapulmonary or extrapulmonary sepsis who were placed on NIV. Subjects were divided into 2 groups based on the need for intubation from NIV: NIV failure (n = 70) and NIV success (n = 66). Demographic, clinical, and outcome data were collected and compared between groups, with the development of multivariate models to predict NIV failure and mortality. The overall NIV failure rate in subjects with a diagnosis of sepsis was 51%. There were no between-group differences in demographic or baseline characteristics. However, there were significant differences in clinical variables, with higher SOFA scores (NIV failure: 6.4 [± 3.0] vs NIV success: 4.9 [± 2.1]; P = .002), 2nd lactate levels (NIV failure: 2.6 [1.7 - 4.3] vs NIV success: 1.9 [1.4 - 2.6] mmol/L; P = .007), and initial NIV [Formula: see text] settings (NIV failure: 0.50 [0.40 - 0.70] vs NIV failure: 0.40 [0.35 - 0.50]; P = .003) in subjects who failed NIV. There were also more subjects in the NIV failure group who had a lactate ≥ 4 mmol/L prior to NIV start compared to those who succeeded on NIV (33% vs 15%, P = .02). At NIV start, subjects in the NIV failure group had lower mean arterial pressure (85 mm Hg [IQR 74-96] vs 91.7 mm Hg [IQR 78-108], P = .042) and Glasgow coma scale scores (14 [IQR 13-15] vs 15 [IQR 14-15], P < .002), while fewer subjects in the NIV failure group received a fluid bolus in the 24 h prior to NIV start (33% vs 53%, P = .02) or had signs of volume overload (36% vs 64%, P < .001). Multivariate analysis indicated that age (odds ratio 1.05 [95% CI 1.01-1.09], P = .02), SOFA score (odds ratio 1.49 [95% CI 1.15-1.94], P = .002), first systolic blood pressure (odds ratio 0.97 [95% CI 0.95-0.99], P = .02), signs of volume overload (odds ratio 0.23 [95% CI 0.07-0.68], P = .008], fluids prior to NIV (odds ratio 0.08 [95% CI 0.02-0.31], P < .001), and initial [Formula: see text] on NIV (odds ratio 1.04 [95% CI 1.01-1.08, P = .002) independently predicted NIV failure with an area under the curve of 0.88. Only NIV failure independently predicted death in multivariate analysis (area under the curve = 0.70). NIV failure in sepsis-related acute respiratory failure was independently predicted by patient acuity, first systolic blood pressure after sepsis alert, initial [Formula: see text] settings on NIV, fluid resuscitation, and signs of volume overload. However, only NIV failure independently predicted death in this cohort of subjects.

The current trend to manage critically ill hematologic patients admitted with acute respiratory failure is to perform noninvasive ventilation to avoid endotracheal intubation. However, failure of noninvasive ventilation may lead to an increased mortality. Retrospective study to determine the frequency of noninvasive ventilation failure and identify its determinants. Medical intensive care unit in a University hospital. All consecutive patients with hematologic malignancies admitted to the intensive care unit over a 10-yr period who received noninvasive ventilation. A total of 99 patients were studied. Simplified Acute Physiology Score II at admission was 49 (median, interquartile range, 39-57). Fifty-three patients (54%) failed noninvasive ventilation and required endotracheal intubation. Their PaO2/FiO2 ratio was significantly lower (175 [101-236] vs. 248 [134-337]) and their respiratory rate under noninvasive ventilation was significantly higher (32 breaths/min [30-36] vs. 28 [27-30]). Forty-seven patients (89%) who failed noninvasive ventilation required vasopressors. Hospital mortality was 79% in those who failed noninvasive ventilation, and 41% in those who succeeded. Patients who failed noninvasive ventilation had a significantly longer intensive care unit stay (13 days [8-23] vs. 5 [2-8]) and a significantly higher rate of intensive care unit-acquired infections (32% compared with 7%). Factors independently associated with noninvasive ventilation failure by multivariate analysis were respiratory rate under noninvasive ventilation, longer delay between admission and noninvasive ventilation first use, need for vasopressors or renal replacement therapy, and acute respiratory distress syndrome. Failure of noninvasive ventilation occurs in half the critically ill hematologic patients and is associated with an increased mortality. Predictors of noninvasive ventilation failure might be used to guide decisions regarding intubation.

Thesemodificationsoftherespiratoryfunctionoccur early after surgery and are more often transient andcould lead to ARF. The clinical result (severity of theARF) is the product of perioperative-related ventilatoryimpairment and severity of the preoperative pulmonarycondition. Maintenance of adequate oxygenation in thepostoperative period is of major importance, especiallywhen pulmonary complications such as ARF occur. Al-though invasive endotracheal mechanical ventilation hasremained the cornerstone of ventilatory strategy for manyyearsforsevereARF,severalstudieshaveshownthatmor-tality associated with pulmonary disease is largely relatedto complications of postoperative reintubation and me-chanicalventilation.Therefore,majorobjectivesforanes-thesiologists are first to prevent the occurrence of postop-erative complications and second to ensure oxygenadministration and carbon dioxide removal while avoid-ing intubation if ARF occurs. Noninvasive ventilation(NIV) does not require an artificial airway (endotrachealtube or tracheotomy), and its use is well established toprevent ARF occurrence (prophylactic treatment) or totreat ARF to avoid reintubation (curative treatment) (fig.1). Studies show that patient-related risk factors, such aschronic obstructive pulmonary disease, age older than 60yr, American Society of Anesthesiologists class of II orhigher, obesity, functional dependence, and congestiveheart failure, increase the risk for postoperative pulmo-nary complications.

Noninvasive ventilation (NIV) fails more frequently for de novo acute respiratory failure (de novo) than for cardiogenic pulmonary edema (CPE) or acute-on-chronic respiratory failure (AOC). The impact of NIV failure and success was compared between de novo and CPE or AOC after adjustment for disease severity. Patients requiring ventilatory support were enrolled in a prospective survey in 70 French ICUs. Of 1076 patients requiring ventilatory support, 524 were eligible, including 299 de novo (NIV use, 30%) and 225 CPE-AOC (NIV use, 55%). Independent risk factors associated with mortality and length of stay were identified by logistic regression analysis. The adjusted outcome of NIV success or failure was compared to that with endotracheal intubation without NIV. NIV success was independently associated with survival in both de novo, adjusted OR 0.05 (95% CI 0.01-0.42), and CPE-AOC OR 0.03 (CI 0.01-0.24). NIV failure was associated with ICU mortality in the de novo group (OR 3.24, CI 1.61-6.53) but not in the CPE-AOC group. Nosocomial pneumonia was less common in patients successful with NIV. NIV failure was associated with a longer ICU stay in CPE-AOC only. The overall use of NIV was independently associated with a better outcome only in CPE-AOC patients (OR 0.33, CI 0.15-0.73). The effect of NIV differs between de novo and CPE-AOC patients because NIV failure is associated with increased mortality for de novo patients. This finding should raise a note of caution when applying NIV in this indication.

Background Risk factors for noninvasive ventilation (NIV) failure after initial success are not fully clear in patients with acute exacerbation of chronic obstructive pulmonary disease (COPD). Methods Patients who received NIV beyond 48 h due to acute exacerbation of COPD were enrolled. However, we excluded those whose pH was higher than 7.35 or PaCO2 was less than 45 mmHg which was measured before NIV. Late failure of NIV was defined as patients required intubation or died during NIV after initial success. Results We enrolled 291 patients in this study. Of them, 48 (16%) patients experienced late NIV failure (45 received intubation and 3 died during NIV). The median time from initiation of NIV to intubation was 4.8 days (IQR: 3.4–8.1). Compared with the data collected at initiation of NIV, the heart rate, respiratory rate, pH, and PaCO2 significantly improved after 1–2 h of NIV both in the NIV success and late failure of NIV groups. Nosocomial pneumonia (odds ratio (OR) = 75, 95% confidence interval (CI): 11–537), heart rate at initiation of NIV (1.04, 1.01–1.06 beat per min), and pH at 1–2 h of NIV (2.06, 1.41–3.00 per decrease of 0.05 from 7.35) were independent risk factors for late failure of NIV. In addition, the Glasgow coma scale (OR = 0.50, 95% CI: 0.34–0.73 per one unit increase) and PaO2/FiO2 (0.992, 0.986–0.998 per one unit increase) were independent protective factors for late failure of NIV. In addition, patients with late failure of NIV had longer ICU stay (median 9.5 vs. 6.6 days) and higher hospital mortality (92% vs. 3%) compared with those with NIV success. Conclusions Nosocomial pneumonia; heart rate at initiation of NIV; and consciousness, acidosis, and oxygenation at 1–2 h of NIV were associated with late failure of NIV in patients with COPD exacerbation. And, late failure of NIV was associated with increased hospital mortality.