1. Introduction
Noninvasive ventilation (NIV) refers to ventilatory support delivered via a nasal, full- face, or helmet mask, without bypassing the upper airway, which is different from tracheal intubation or tracheostomy [
1,
2,
3,
4]. NIV is attractive when treating acute respiratory failure (RF), as it avoids complications, such as ventilator-associated pneumonia associated with invasive mechanical ventilation [
5,
6]. Although a high-flow nasal cannula (HFNC) may substitute for NIV when treating critically ill patients with acute hypoxic RF, it does not supply any inspiratory pressure, unlike NIV; therefore, NIV is preferred for patients with acute hypercapnic RF [
7,
8,
9,
10,
11].
The proportion of elderly patients in intensive care units (ICUs) is rapidly growing, because global populations are aging [
12]. If elderly patients require mechanical ventilation, weaning may be more prolonged than for younger patients, because of respiratory muscle weakness, comorbidities, and altered mental status [
13,
14,
15]. Prolonged weaning is associated with several complications that increase in-hospital mortality [
16,
17,
18]. Thus, many clinicians prefer to use a noninvasive approach, such as NIV, in elderly patients with acute RF to avoid intubations. However, in elderly patients, particularly those with pneumonia, there is concern regarding the possibility of NIV failure and the increased mortality caused by delayed intubation [
19,
20]. Thus, in elderly patients with pneumonia, NIV may be delayed, because its utility in this setting remains controversial, and excessive secretion (e.g., due to pneumonia) is a contraindication for the procedure. We hypothesized that, if we could define the factors that predict NIV failure so that it can encourage early intubation, rather than the maintenance of NIV, NIV could be safely applied to elderly patients with acute RF due to pneumonia.
The aim of this study was to investigate whether NIV was appropriate for elderly patients with pneumonia, to define factors that independently predicted NIV failure, and to make an optimal model for prediction of NIV failure.
4. Discussion
We explored whether NIV was appropriate for elderly patients with pneumonia, defined factors that independently predicted NIV failure, and built an optimal model for prediction of such failure. Pneumonia was associated with NIV failure, but the presence of pneumonia alone did not predict such failure well (C-index AUC: 0.639; 95% CI: 0.534–0.744). Therefore, NIV is not contraindicated by pneumonia alone. Improvement of heart rate and PaCO2 level measured within 2 h after NIV commencement should be considered when determining if NIV should be continued or if intubation is preferable.
Our study makes several important contributions to the literature. When elderly patients are admitted to ICUs, intensivists find it difficult to decide whether aggressive treatment is appropriate, because advanced age is associated with poorer outcomes after mechanical ventilation, and survival does not always ensure a satisfactory quality of life [
12,
13,
14,
15,
19,
20,
23,
24]. Some intensivists prefer to treat elderly patients as noninvasively as possible, often choosing NIV for those with acute RF. However, some may hesitate to use NIV in elderly patients with pneumonia on the basis of a vague concern that NIV failure is more common in elderly patients than in younger patients [
25,
26]. NIV failure is associated with poor mask-fitting, claustrophobia, excessive secretions, intolerance, agitation, and patient/ventilator asynchrony, most of which may be associated with poor respiratory muscle power [
26,
27]. A previous study also showed that weak respiratory muscle power was associated with NIV failure [
27]. Generally, older patients have weaker respiratory muscles than younger patients. Thus, concerns that NIV failure is more common in elderly patients than in younger ones may be partially right. However, some elderly patients remain active and have powerful respiratory muscles. Therefore, NIV should not be avoided simply because of patient age. In our study, age subgrouping (65–74, 75–84, and ≥85 years) did not affect the NIV failure rate (
Figure S2;
p = 0.535).
Early identification of NIV failure is important; delayed identification increases mortality. In a previous study, the timing of NIV failure was categorized as follows: (1) immediate failure (within minutes to < 1 h), (2) early failure (1–48 h), and (3) late failure (after 48 h) [
26]. The majority (~83%) of NIV failures occur between the immediate and early stages [
26]. When evaluating possible NIV failure, it is necessary to consider the overall clinical condition, but we could hesitate to perform intubation if NIV failure is not definitive. We emphasize that early variation in vital signs and arterial blood gas data measured 30–120 min after NIV commencement should be used to predict NIV possible failure, triggering intubation, thus preventing mortality attributable to delayed intubation.
We found that heart rate and PaCO
2 level independently predicted NIV failure. Although the PaO
2/FiO
2 ratio has been associated with NIV failure in patients with hypoxic respiratory failure, the arterial pH and PaCO
2 values of patients with hypercapnic respiratory failure were more closely associated with NIV failure than the PaO
2/FiO
2 ratio [
26]. In our study, 82.1% of patients were prescribed NIV, because of hypercapnic respiratory failure. Therefore, the PaCO
2 value more meaningfully predicted NIV failure than the PaO
2/FiO
2 ratio.
In this study, a low PaCO
2 level after NIV commencement was associated with NIV failure. A low PaCO
2 level was associated with severe disease and agitation, because the increased tidal volume and tachypnea caused by labored breathing led to reduction of the PaCO
2 level [
28,
29]. A PaCO
2 level after NIV commencement was related to a PaCO
2 level before NIV commencement. Therefore, patients with a low PaCO
2 level after NIV commencement initially had more severe disease, compared to patients with a high PaCO
2 level after NIV commencement. Disease severity was associated with NIV failure, so NIV failure was more likely in patients with a low PaCO
2 level than in patients with a high PaCO
2 level. A more important finding related to the evaluation of NIV failure was the variation in PaCO
2 level after NIV commencement. As shown in
Table 6, variation in the PaCO
2 level in the NIV failure group was associated with a less-improved arterial pH than that of the NIV success group. In other words, although the absolute median variation was similar in both groups, PaCO
2 improvement was less clinically effective in the NIV failure group than in the NIV success group. This finding is probably because the number of patients with hypoxic respiratory failure was higher in NIV failure group than in NIV success group. However, a PaCO
2 level reflects the clinical condition. Consequently, when predicting NIV failure, we should consider whether a PaCO
2 level has improved properly to the point where pH reaches to the normal range. The results of a subgroup analysis of patients with only hypercapnic RF were similar to the findings in the present study (
Table S1). The heart rate was also a significant independent predictor of NIV failure, because a high heart rate was associated with disease severity and agitation. Previous studies have reported that respiratory rate variation is associated with NIV failure [
26]. However, in practice, respiratory rate measurement is less accurate than heart rate measurement, because the former is not automatic. Both the respiratory and heart rates are affected by clinical condition. Therefore, heart rate variation is preferable to respiratory rate variation when predicting NIV failure.
We also explored the utility of NIV in patients with pneumonia. This has been controversial; prior studies have reported conflicting results, probably because of differences in clinical characteristics (pneumonia severity, age, or comorbidities) [
30,
31,
32,
33,
34,
35,
36,
37]. In patients with pneumonia, the median PaO
2/FiO
2 ratio and the median SOFA score at NIV commencement were 155 and 5. All pneumonia patients of the present study suffered from sepsis. In addition, 55% of patients had underlying lung parenchymal disease (COPD, asthma, interstitial lung disease, or bronchiectasis). Therefore, pneumonia was severe. We found that the C index of pneumonia alone was 0.639; thus, pneumonia was only weakly associated with NIV failure. Therefore, NIV is reasonable to apply in elderly patients with pneumonia, if we could carefully observe patients after NIV commencement.
Our study had certain limitations. First, the patient number was relatively small (n = 78). However, the number was sufficient to reveal the factors associated with NIV failure in elderly patients with acute RF, because a post hoc power analysis revealed that the power of each variable was appropriate. Second, we did not include clinical data, such as chest X-ray or laboratory findings, when determining disease severity in patients with pneumonia. However, given the SOFA scores and PaO2/FiO2 ratios, most pneumonia patients suffered from sepsis. Third, we did not set detailed indications for weaning or intubation, relying rather on respiratory intensivist discretion. However, all intensivists had extensive experience with NIV, and most decisions were based on international guidelines.