The proportion of elderly persons among hospitalized patients, including intensive care unit (ICU) admissions, is rapidly growing in developed countries. In recent epidemiologic studies patients aged >85 years represented 10–15% of ICU admissions [13]. Patients admitted to the ICU for acute hypercapnic respiratory failure (AHRF) have an advanced mean age [4]. In older patients AHRF treatment may be quite complex for reasons such as reduced life expectation and an increased rate of complications. The management of critical respiratory illness in the elderly is therefore of particular importance. Controlled trials and meta-analyses have so far shown that non-invasive ventilation (NIV) administered with bilevel positive airway pressure (BPAP) devices is an effective means for the management of patients with AHRF [57]. Scarpazza et al. reported that NIV was effective for the treatment of AHRF in the elderly [8]. The reduced invasiveness of this technique in selected populations of critically ill patients leads to better outcomes than with endotracheal intubation. The choice of NIV aims to avoid complications, particularly in fragile patients [9]. Although NIV is an attractive technique for the management of AHRF in older patients, specific data for this population are limited [10, 11]. The decision to ventilate older patients becomes more of an issue in Turkey because there is no legislation about do not intubate (DNI) orders. We therefore investigated the tolerance towards and outcomes of NIV in a cohort of older patients who were admitted to the ICU with AHRF due to different disease etiologies.



A single centre, prospective, interventional study was conducted with patients aged ≥65 years who were admitted to our ICU between January 2012 and September 2016. Patients with AHRF due to chronic obstructive pulmonary disease (COPD), acute cardiogenic pulmonary edema (ACPE), community-acquired pneumonia (CAP), bronchiectasis, and kyphoscoliosis who received NIV as an initial ventilatory support measure were included in the study. The inclusion criteria were worsening dyspnea with respiratory rate >22 breaths per min, respiratory acidosis (pH <7.35), and hypercapnia (PaCO2 > 45). The exclusion criteria were urgent necessity for invasive ventilation, respiratory or cardiac arrest, persistent hemodynamic instability, unstable cardiac arrhythmia, deep hypercapnic coma that prevented NIV, alterations of the central nervous system unrelated to hypercapnic encephalopathy, inability to cooperate or to protect the airway, pneumothorax, pneumomediastinum, hemoptysis, excessive secretion and inability to clear secretions, facial deformity, upper airway obstruction due to trauma, burns and surgery involving the face, and severe upper gastrointestinal bleeding. Patients were prospectively classified according to the context in which NIV treatment was administered: COPD exacerbation, ACPE, CAP, bronchiectasis exacerbation, and kyphoscoliosis.

COPD and COPD exacerbation were defined according to the global initiative for chronic obstructive lung disease (GOLD) criteria [12]. The study included patients with COPD aged >65 years who had a history of smoking and a post-bronchodilator forced expiratory volume in the first second (FEV1) to forced vital capacity (FVC) ratio of less than 70%, and a post-bronchodilator FEV1 of less than 80% of predicted in a spirometry study performed within 1 year of joining the study. ACPE was defined based on both clinical and radiologic criteria. Bronchiectasis was diagnosed based on a high-resolution computed tomography scan of the chest [13]. Pneumonia was defined as the acute onset of symptoms suggestive of a lower respiratory tract infection and radiographic evidence of a new infiltrate [14].

All patients received empiric antimicrobial therapy at admission. All patients received prophylaxis against stress ulcers and deep venous thrombosis; patients with COPD received bronchodilators and systemic corticosteroids.

Procedures and outcome

Non-invasive ventilation

This study was conducted in Ankara University School of Medicine, Department of Chest Disease ICU. NIV was managed by four ICU physicians who were also among the authors of this manuscript. NIV sessions were applied using Hans Rudolph 6600 V2 oronasal masks (Hans Rudolph, Shawnee, KS, US) with V60 (Philips Respironics, Carlsbad, CA, US) and VENTImotion (Weinmann, Hamburg, Germany) ventilators. Pressure support (spontaneous/timed) mode was applied as the first choice. Arterial blood gases were measured 1 h after treatment to assess the response to NIV and to modify the settings accordingly. Arterial blood gas (ABG) samples were obtained through a direct vascular puncture of the radial artery, which was performed at the bedside. The samples were instantly studied using an ABL 90 Flex/Blood Gas Analyzer (Radiometer, Brønshøj, Denmark), which was available in the ICU. ABG analyses were performed immediately before, and at 1, 2, 12, and 24 h after the initiation of NIV treatment, and daily thereafter if needed. The ABG results and adjustments were registered on forms designed for the study. Pressure support was initially set at 7–10 cm H2O and gradually increased to a maximum of 30 cm H2O until the exhaled tidal volume was 8–10 ml/kg of ideal body weight. Expiratory positive airway pressure (EPAP) was set at 5 cm H2O and raised if needed to treat hypoxemia. The fraction of inspired oxygen (FiO2) of the device was adjusted to maintain arterial oxygen saturation (SpO2) above 90% in pulse oximetry and arterial blood gas analysis. With the exception of pauses for feeding and treatment administration, NIV support was provided continuously for the first 24 h, as long as it could be tolerated by the patients.

Data collection

Immediately preceding the NIV treatment, the level of consciousness was determined using the Glasgow coma scale (GCS) [15], and health status using the acute physiology and chronic health evaluation (APACHE) II score [16]. The Charlson comorbidity index (CCI) was used to calculate comorbidity scores [17]. A chest X‑ray was obtained from each subject. The following parameters were recorded at admission: age, sex, body mass index (BMI), comorbidities, serum C‑reactive protein (CRP) level, partial pressure of carbon dioxide (PaCO2), HCO3, base excess, pH, PaO2/FiO2 ratio, and the setting of the mechanical ventilator. In addition, mechanical ventilator settings and arterial blood gas results were recorded immediately before, and at 1, 2, 12, and 24 h after NIV treatment initiation, and daily thereafter if needed.

NIV success and failure

Non-invasive mechanical ventilation was considered successful when pH exceeded 7.35 during spontaneous breathing without further worsening of neurologic signs, respiratory rate was reduced to less than 24 breaths/min, heart rate decreased to less than 90 beats/min, patient awareness improved and pH exceeded 7.35 with adequate SaO2 in ambient air and a with low percentage of inspired O2 (FiO2 < 30%) for at least 48 h. The need for intubation and invasive ventilation (IV) support was based on one of the following criteria: (1) pH below 7.20, (2) pH 7.20–7.25 on two occasions 1 h apart, (3) hypercapnic coma (GCS < 8 and PaCO2 > 60 mm Hg) (4), PaO2 below 45 mmHg despite maximum tolerated FiO2, and (5) cardiorespiratory arrest. NIV failure was defined as the need for endotracheal intubation and/or ICU death.


The rate of NIV success was determined in all patients aged ≥65 years with AHRF. NIV success was assessed in patients aged 65–74 years, 75–84 years, and those aged ≥85 years. Differences between the oldest age group and the other groups were sought. We also investigated whether there was any difference between disease etiologies with respect to NIV success. Finally, factors responsible for NIV success in older patients with COPD were explored.

Statistical analysis

All statistical analyses were performed using IBM SPSS Statistics 20 (SPSS, Chicago, IL). The descriptive statistics for continuous variables included mean ± standard deviation or median (range) depending on the normality of distribution; the categoric variables are expressed as frequency (percentage). The comparison of continuous variables was performed using one-sided analysis of variance (ANOVA) when the parametric test assumptions were met, or the Kruskal-Wallis test if not. Categoric variables were compared using the χ2or Fisher’s exact test across the study groups. Statistical significance was set at p < 0.05. Several candidate variables identified in a univariate Cox regression analysis were examined using the multiple Cox model with the forward LR method to determine independent predictors of NIC success.


Of 496 patients aged ≥65 years admitted to ICU for AHRF, 162 patients (78 women, 84 men) were included in the study (Fig. 1). Among these patients, 90 (55.5%) had COPD, 31 (19.1%) had ACPE, 19 (11.7%) had CAP, 15 (9.3%) had bronchiectasis exacerbation, and 7 (4.3%) had kyphoscoliosis. During the study period, 43 (24.6%) patients were intubated and connected to a mechanical ventilator. These patients were included in the NIV failure group. The mean duration of ICU stay was 132.34 + 63.21 h. The overall mortality rate was 18%.

Fig. 1
figure 1

Flow chart of the study showing the inclusion criteria and outcomes of the study patients

The study population had a mean age of 71.81 ± 10.97 years (range, 65–92 years), 71 (43.8%) patients were aged 65–74 years, 70 (43.2%) were aged 75–84 years, and 21 (13%) patients were aged ≥85 years. No significant difference was found between the age groups with regard to disease distribution (Table 1) and there were sex differences between the age groups (p = 0.062). The mean BMI values were similar for the three groups (p = 0.863). Cumulative smoking, which was measured by pack-years was significantly lower in the oldest group (p = 0.037). pH, PaCO2, and PaO2/FiO2 values at the initiation of NIV treatment showed no significant difference between the three groups (p = 0.542, p = 0.183, p = 0.424, respectively). The three age groups had similar GCS, APACHE II score and CCI (p = 0.065, p = 0.058, p = 0.074, respectively). After initial adjustments, the mean inspiratory positive airway pressure (IPAP) and expiratory positive airway pressure (EPAP) values were similar in all age groups (Table 1).

Table 1 Summary of baseline characteristics of patients treated with non-invasive ventilation grouped by age strata

NIV treatment was successful in 119 (73.5%) patients. The rate of NIV success was not significantly different between the age groups (p = 0.803). NIV treatment was successful in patients with COPD exacerbation (n = 72, 60.5%), ACPE (n = 23, 19.3%), CAP (n = 10, 8.4%), bronchiectasis (n = 10, 8.4%), and kyphoscoliosis (n = 4, 3.4%). An analysis of disease-based NIV success revealed a significantly higher success rate in COPD and ACPE groups (p = 0.029, p = 0.035, respectively) (Table 2).

Table 2 Summary of patient characteristics, arterial blood gases, illness severity and non-invasive ventilation pressures by admission diagnosis

There was no significant difference between the different disease groups with respect to mean age, initial PH, and PaCO2 (p = 0.527, p = 0.243, p = 0.653); PaO2/FiO2 values were significantly different between the groups (p < 0.001). The PaO2/FiO2 was highest in the COPD group (202.82 ± 29.24) and the lowest in the CAP group (176.95 ± 17.23). The GCS and APACHE II score were also significantly different between the disease groups (p < 0.001) (Table 2). The IPAP values were similar across all study groups (p = 0.345), whereas EPAP values were significantly different (p = 0.045), the highest EPAP value was in the CAP group. When the treatment success and failure groups were compared GCS (p = 0.046), APACHE II (p = 0.037), pH (p = 0.041), and PaO2/FiO2 (p = 0.028) were significantly different between the groups (Table 3).

Table 3 Comparison of diagnosis, illness severity and arterial blood gases between the failed and successful treatment groups
Table 4 Logistic regression analysis addressing the main risk factors for noninvasive positive pressure ventilation failure

Univariate analysis of the factors responsible for NIV failure in patients with COPD and AHRF revealed that GCS, APACHE II, CAT, CCI, initial PH, PaO2/FiO2, and CRP were significantly correlated to NIV failure. Multivariate analysis found that GCS, CAT, and APACHE ΙΙ were significant independent predictors of NIV failure (Table 4).


With the rising life expectancy of the population, the number of older patients receiving ICU care has been increasing [18]. In severe disease, patients often have limited respiratory reserve and the resultant increased work of breathing, and subsequent exhaustion may lead to hypercapnia, hypoxia, and respiratory acidosis. Elderly patients are at increased risk of developing respiratory failure, for example as a result of limited reserves, loss of muscle mass, nutritional deficiencies, and associated comorbidities [19]. The rate of NIV use in the treatment of respiratory failure is higher among older patients compared with younger patients. Schortgen et al. reported that among 376 patients provided with ventilatory support, 163 were aged ≥85 years, in which the success rate was 60%, whereas it remained as low as 32% in the younger group (p < 0.001) [20].

Complications of mechanical ventilation increase in-hospital mortality in patients aged ≥65 years [21]. This suggests that avoiding invasive procedures might be particularly helpful in decreasing mortality in the elderly. It should be mentioned that the impact of the intensity of care on the survival of older patients is still debated [22]. Advanced age is associated with worse outcomes of mechanical ventilation in the ICU; therefore, age is an important issue to be considered. We intend to determine whether it is an absolute barrier to NIV. Our results suggest that age does not affect NIV success. Similar NIV success was observed even across age groups of patients aged over 65 years. This still held true, even in the face of the fact that the patients aged >85 years had similar GCS, APACHE II and CCI scores. We observed no differences between the age groups with respect to IPAP and EPAP pressures in the first hour; however, when an etiology-based analysis was performed, the first-hour EPAP values were significantly higher in the CAP group. This group had higher EPAP values due to increased FiO2 need. This finding suggests that patients should be evaluated independent of age, taking into consideration their personal characteristics, such as treatment adherence and tolerance, certain clinical signs (respiratory rate, PaO2/FiO2), and arterial blood gas analysis. We found a greater success rate for NIV in COPD-induced AHFR compared with the other groups. The usefulness of NIV for COPD exacerbations has been known for a long time. Our results suggest that this also applies to older patients. Scarpazza et al. reported a very high NIV success rate with a related reduction in invasive mechanical ventilation-associated complications in patients with acute hypercapnic respiratory failure due to acute exacerbations of COPD [23].

The common clinical indications for NIV in acute (usually congestive or ischemic) heart failure are dyspnea, hypoxemia, and pulmonary congestion [24]. The increased end-diastolic pressure in the left ventricle leads to increased pulmonary capillary hydrostatic pressure and extravasation of fluid into the alveoli, which in turn reduces gas exchange, produces hypoxia, and in severe cases, also leads to hypercapnia [25].

In ACPE, the use of non-invasive ventilatory support improves lung compliance, recruits previously collapsed alveoli, reduces preload and afterload, which leads to improved oxygenation and reduction of respiratory muscle workload [26]. Age does not seem to be a limiting factor for the safe use of ventilatory support because similar results were observed for various age groups [27].

In the bronchiectasis group, NIV treatment succeeded in 66.7% of subjects. The number of studies exploring the use of NIV is quite limited in bronchiectasis. Phua et al. used NIV treatment for 31 patients with acute exacerbation of bronchiectasis (11 with do-not-intubate orders), and found a success rate of 67.7% (n = 21). In that study, the APACHE II score was the only predictor of hospital mortality (OR 1.19 per point), and the PaO2/FiO2 ratio was the only predictor of NIV failure (OR 1.02 per mmHg decrease) in the logistic regression analysis [28].

Studies that investigated the efficacy of NIV treatment for CAP-induced AHRF yielded conflicting results, some studies reported quite low success rates. For instance, a study that explored the efficacy of NIV for acute respiratory failure due to CAP revealed a success rate of only 34% [29]. Another study with similar results to the present study regarding COPD and CAP included 111 patients with AHRF, 43 of whom had COPD exacerbations and 68 had other conditions. The risk of NIV failure, defined as the need for endotracheal intubation, was significantly lower with COPD than in other conditions (19% vs. 47%). The presence of pneumonia (OR 5.63), was an independent predictor of NIV failure in the non-COPD group [30]. In the present study the success rate of NIV was not high for the CAP group. This finding was attributed to a low PaO2/FiO2 ratio, high GCS and high APACHE II score in AHRF cases secondary to CAP. Adıgüzel et al. reported a success rate of 80% for NIV in patients with kyphoscoliosis [31]. Phua et al. reported that NIV proved useful in 73% of subjects with neuromusculoskeletal disorders [30]. We linked the lower success rate in the present study to the lower GCS and higher APACHE II scores of our patients with kyphoscoliosis compared with the other groups. Multivariate cox regression analysis performed to determine independent factors responsible for NIV failure in patients with AHRF due to COPD exacerbations revealed that APACHE II, GCS, and dyspnea level (as determined by the CAT score) were independently correlated with treatment failure. Confalonieri et al. recently showed that GCS < 11, APACHE > 29, respiratory rate >30 breaths per min, and pH < 7.25 predicted NIV failure [32].


First, as a single-center study, our study may not convey generalizable results. Second, a low number of subjects in the non-COPD groups limited the power of the statistical analyses. Therefore, only the patients with COPD were included in the Cox regression analysis. Third, the patients’ results only pertain to ICU stay, but not to long-term follow-up.


This cohort study showed that the efficacy of NIV, which we frequently use for selected older patients with AHRF in the ICU, was no different than that for younger patients. There was no significant difference between the oldest age group and the other groups in terms of non-invasive ventilation efficacy. NIV appeared most effective for COPD exacerbation and ACPE. Among patients with COPD exacerbation, those with more severe dyspnea level at the beginning and a worse health condition had a low success rate, independent of age.