Dyspnea: an unrecognized cause of discomfort in mechanically ventilated patients?

Alleviating the immediate effects of pain suffered by patients is a natural mission of caregivers in all categories. In the intensive care unit, optimizing patient comfort is a major concern and involves three steps: (1) the identification of potential discomfort, (2) the diagnosis of the reason for this discomfort, and (3) the initiation of a therapeutic response to treat this discomfort. Although mechanically ventilated ICU patients cannot communicate easily, it is possible to communicate with many of them and possible to guess at discomfort from observations in the remainder. Identifying the reason for the discomfort can be challenging because discomfort in these patients can be due to many causes (Fig. 1), and it is crucial to that the cause of the discomfort be correctly identified since each potential cause of discomfort leads to a different therapeutic response. Pain is one of the major causes of discomfort in ICU patients and has received major attention during the past decade with beneficial effects on long-term outcomes. Awareness of pain and its proactive management has resulted in improved ICU outcomes [1, 2].

Fig. 1
figure 1

 Possible reasons for discomfort in mechanically ventilated patients

Although dyspnea and pain share many similarities [35], little attention has been given to assessing and managing dyspnea; little rigorous research data are available and there are currently no clinical guidelines for managing dyspnea in the ventilated patient. However, dyspnea can be assessed, and available data do show that it frequently causes discomfort in mechanically ventilated patients admitted to the ICU.

ICU patients are exposed to many stimuli that can generate or exacerbate dyspnea. In addition to underlying cardiopulmonary abnormalities, respiratory discomfort may be caused by some of the therapeutic management strategies that have been adopted in recent years. These include lowering sedation [6], preserving spontaneous breathing activity [7], and the use of low tidal volumes [8] even in the absence of severe lung disorders [911]. It cannot be ruled out that sedation may give a falsely reassuring outward appearance of comfort in patients actually suffering from undiminished—or even increased—respiratory discomfort, as in the case of pain [12, 13]. For example, pain ratings and pain-related cortical activations in response to cutaneous pain stimuli are increased by moderate propofol sedation. Strong pain-related activations remain in some cortical regions even during heavy propofol sedation that renders subjects unresponsive.

The aim of this review is to promote awareness in ICU caregivers of the unrecognized problem of dyspnea. Herein we summarize current knowledge on the prevalence of dyspnea in mechanically ventilated ICU patients and on the corresponding risk factors. We also discuss possible approaches to detect and quantify dyspnea in these patients.

Simplified physiological basis of dyspneic sensations

The physiology of dyspnea has been extensively reviewed [1417]. The neurophysiological basis of dyspnea is more complex than that for many other sensations and involves both excitatory and inhibitory afferent inputs from sensory nerves as well as the perception of motor commands (so-called corollary discharge). In mechanically ventilated patients, it seems reasonable to focus on two main dyspnea modalities, namely “air hunger” and “excessive work/effort of breathing”.

Air hunger

Air hunger (an unpleasant, unsatisfied urge to breathe) is perhaps the most distressing dyspnea modality [18, 19] and is characterized by such verbal expressions as “I am not getting enough air”, “I feel that I am suffocating”, and “I need more air”. Experiments in paralyzed subjects (complete neuromuscular block in normal volunteers [18] or C1–C2 quadriplegic patients [20]) have shown that an acute rise in PaCO2 (arterial carbon dioxide tension) suffices to induce air hunger independent of any respiratory muscle activity. Air hunger also increases when the tidal volume is decreased under mechanical ventilation and the PaCO2 is held constant [21, 22]. Air hunger appears to arise from a “corollary discharge” or copy of the automatic efferent command from the brainstem respiratory motor center. The brainstem respiratory center is excited by various stimuli that “drive” respiration hypercapnia, hypoxia, exercise, and hyperthermia. Even when respiratory drive is held constant, air hunger is relieved by tidal expansion of the lungs [23]. The principle pathway for this volume expansion relief is pulmonary stretch receptor afferents [22], although a contribution from chest wall afferents has not been ruled out [24]. Thus, air hunger develops when there is an imbalance between the efferent and the inhibitory afferent messages. This imbalance triggers activity in interoceptive areas of the brain involved in unpleasant sensations (such as visceral pain, thirst, or hunger), particularly the insular cortex, amygdala, and anterior cingulate [21]. In contrast, when increased ventilatory drive is matched by an adequate stretch receptor return from the lungs, respiratory sensations will not necessarily be perceived as uncomfortable and will therefore not qualify as dyspnea (e.g., during moderate exercise in healthy persons).

Excessive work/effort of breathing

This dyspnea modality is characterized by such verbal expressions such as “My breathing requires effort”, “My breathing requires more work” [25], “I have difficulty breathing”, “I need to make an effort to get the air in”, and “I must concentrate on my breathing”. It arises from a sense of excessive breathing effort even when gas exchange needs are adequately met—this sense of work or effort arises in both the muscle afferents of the chest wall and the diaphragm, from volitional respiratory motor centers in the neocortex, and perhaps from automatic respiratory motor centers. Dyspnea of the excessive work/effort type characteristically occurs when there is an imbalance between the load imposed upon the respiratory muscle (respiratory impedance) and the capacity of these muscles to overcome this load [26]. This can correspond to deteriorated respiratory mechanics, a weakening of inspiratory muscles, or the combination of both. Dynamic hyperinflation in patients with chronic obstructive pulmonary disease (COPD) provides a typical example of such situations. Dynamic hyperinflation combines an increased respiratory impedance (increased lung elastance at high lung volume requiring high inspiratory muscle force) and a functional weakening of inspiratory muscles [hyperinflation-related shortening that places the muscles at disadvantage both mechanically (e.g., flattened diaphragm) and on the length–tension curve]. In addition, patients must overcome the resultant intrinsic positive expiratory pressure before beginning to move air and trigger the ventilator, thereby imposing an inspiratory threshold load. Neuromuscular disorders involving the respiratory muscles can also give rise to dyspnea of the excessive work/effort type. Adequate ventilatory support should prevent this type of dyspnea, the very purpose of ventilatory assistance being to adjust the load/capacity balance of the respiratory system.

Complex sensations

Patients mechanically ventilated for acute respiratory failure generally experience mixed respiratory sensations (Fig. 2). Increased respiratory impedance, often associated with decreased respiratory muscle strength, generates a sense of excessive respiratory work/effort, and the corresponding failure to maintain gas exchange combined with inadequate tidal volume is a source of air hunger. It is therefore expected that dyspneic ICU patients will report more than one dyspnea modality [27].

Differential affective components

Dyspnea, like pain, can be viewed as causing an immediate feeling of unpleasantness as well as having more complex emotional sequelae [28]. Recent experimental data have demonstrated that both unpleasantness and emotional response can be differentially altered with changes in the dyspnea modality or drug treatment. At a similar level of sensory intensity, air hunger elicits more unpleasantness than excessive work/effort and is associated with greater feelings of anxiety and fear [19]. Of clinical relevance is the experimental finding that an effective way to induce air hunger is to increase ventilatory demand via mild hypercapnia while hindering the normal ventilatory response [29, 30]. Likewise, therapeutic interventions can have different effects on both the sensory and affective component of dyspnea [31].

What are the main determinants of dyspnea in mechanically ventilated patients?

The determinants of dyspnea in mechanically ventilated patients are multifactorial, with contributions from the intrinsic bronchopulmonary status of the patients, ventilator settings, various healthcare activities, extrinsic physiological stimulations of the ventilatory drive, and non-respiratory factors such as anxiety or pain. Several of these factors can coincide at any particular moment, and their effects are likely to be synergistic.

Intrinsic cardiopulmonary status

Although the disease that has required ventilatory assistance is the usual cause of dyspnea, mechanical ventilation has the potential to reduce or eliminate the dyspnea. An increase in dyspnea during the course of mechanical ventilation can signal a problem with the mechanical ventilation itself or signal an adverse event, such worsening respiratory mechanics and/or the ventilation–perfusion equilibrium [pneumothorax, atelectasis, ventilator-associated pneumonia, hydrostatic pulmonary edema or acute respiratory distress syndrome (ARDS)] or a non-respiratory one (acute hemorrhage, sepsis, or worsening heart failure) [27].

Ventilator settings

The authors of the largest study dealing with dyspnea in ICU mechanically ventilated patients conducted to date [27] reported that ventilator adjustments targeted at dyspnea reduction significantly improved respiratory comfort, often dramatically, in 35 % of their mechanically ventilated patients who reported dyspnea. In addition, their data suggest that ventilator mode played a role in dyspnea since dyspneic patients were more likely to be ventilated with assist-control ventilation (ACV) than non-dyspneic patients (69 vs. 45 %), and the multivariate analysis showed that ACV was independently associated with dyspnea [odds ratio (OR) 4.77, 95 % confidence interval (CI) 1.60–14.3] [27]. This difference may be due to the fundamental characteristics of the ventilatory mode, but is more likely due to the greater ease in optimizing ventilatory parameters during support. Unfortunately, little is known about the effect of ventilator mode per se, as the studies conducted on this problem have failed to measure or control for variables known to alter dyspnea, such as tidal volume, arterial PaCO2, and inspiratory flow rate, among the different modes [32].

It is likely that tidal volume changes alone can explain the observed differences between ventilator modes. Indeed, when we re-plotted the data from two studies to show dyspnea ratings versus tidal volume [33, 34], we noted that ratings from different modes fell along the same line (Fig. 3). This is an important issue since recent publications argue that tidal volume should be minimized in mechanically ventilated patients, even in those without risk factors for ARDS [911]. When setting tidal volume the healthcare professional should also be aware of the resultant discomfort and potentially deleterious effects of the settings. Optimization of flow rate also influences respiratory comfort [27], consistent with earlier findings in healthy subjects [35, 36].

Fig. 2
figure 2

Modality of dyspnea experienced by ventilated patients (data from Schmidt et al. [27])

Finally, during inspiratory pressure support, there seems to be a U-shaped relationship between the level of assistance provided and dyspnea; if assistance is either too high or too low discomfort increases [37].

Patient–ventilator interface; respiratory and non-respiratory care activities

The patient–ventilator interface can induce discomfort, be it an endotracheal tube [38, 39] or a face mask [40]. Beyond this, the results of the small descriptive case study by Lush et al. [41] indicate that many other events can be associated with dyspnea, such as turns, transfers, or bathing, but also the presence of the physicians or external events, such as shift changes or a death elsewhere in the unit.

Extrinsic physiological stimulations of ventilatory drive

Respiratory drive in excess of achieved ventilation is the cause of air hunger (see above, physiological basis). Fever, acidosis, or anemia are frequent causes of increased ventilatory drive in ICU patients, and should therefore be looked for in the presence of apparently unexplained dyspnea.

Anxiety and pain

Anxiety and pain may increase dyspnea by stimulating ventilatory drive [42], and it is also likely that these factors can also interact with the affective dimension of dyspnea. In the dyspnea–ICU study carried out by Schmidt et al. [27], pain and anxiety were more frequent in dyspneic than in non-dyspneic mechanically ventilated patients. Anxiety was independently associated with dyspnea (OR 8.84, 95 % CI 3.26–24–14.3), and it was responsive to ventilator settings adjustments when these changes relieved dyspnea. This result points towards a bidirectional causative relationship between anxiety and dyspnea.

Prevalence of dyspnea in mechanically ventilated patients

Because dyspnea can only be perceived by the person experiencing it, it should be assessed by questioning the patient; only when communication with the patient is impossible should the clinician rely primarily on outward signs. We describe the assessment of dyspnea in more detail in following sections. There is a paucity of published data on dyspnea prevalence in mechanically ventilated patients, and clinical experience suggests that dyspnea is not routinely assessed and recorded. Studies in which patients have been asked to recall their ICU experience give insights into the magnitude of the problem. For example, Rotondi et al. [38] were able to explore the recollections of 150 patients who had survived an ICU stay during which they had been mechanically ventilated for ≥2days. Two-thirds of these patients remembered the endotracheal tube (ETT) of whom 45 % recalled “feeling choked by the ETT” and 24 % remembered “not getting enough air from the ETT”. ETT-related experiences were considered stressful and were strongly associated with the occurrence of spells of terror or “feeling nervous when left alone” [38]. In a different study involving 126 COPD patients who were interviewed at the time of their discharge from the ICU stay, 90 % recalled experiencing traumatic events during their stay in the ICU [40]. “Suffocation” was the second most frequently noted item (55 % of the respondents) just behind “sleep disorders” (63 %) [40]. To our knowledge, dyspnea in mechanically ventilated patients has been the main research focus of only eight prospective studies published to date (Table 1; [27, 41, 4348]).

Table 1 Prospective dyspnea studies in mechanically ventilated patients

Four of these studies focused on breathing comfort in the context of weaning trials [4446, 48], but they also provide pre-weaning dyspnea assessments. In this setting, dyspnea was common, although the exact proportion of patients who were dyspneic before weaning was not systematically reported. In one study [46], the authors explored the relationship between dyspnea and pre-weaning mood state in 21 mechanically ventilated patients using a shortened version of the profile of mood states (sPOMS) [49]. Weak relationships were found between dyspnea intensity [visual analog scale (VAS)] and mood states, with the “vigor” subscore of the sPOMS being the closest to reach a statistically significant negative correlation with the intensity of dyspnea (p = 0.07) [46].

The other four prospective studies evaluated dyspnea in alert mechanically ventilated patients not having entered the weaning process [27, 41, 45, 47]. The study by Lush et al. [41] has a descriptive case design and pertains to a convenience sample of five individuals. Dyspnea was systematically assessed by specially trained nurses every 4 h. All patients reported dyspnea at some time during the study, with a VAS intensity reaching 95 % of full scale in some cases. Weak correlations were found between dyspnea intensity and several physiological variables, such as PaO2, PaCO2, or ventilation. Stronger correlations were found between dyspnea intensity and the density of events and activities occurring in the ICU at the time of dyspnea evaluation—whether or not these events were directly related to patient care (examples of variables recorded include “change of shift”, “engineers in the unit”, “death next bed”, among others). The study by Karampela et al. [47] tested the “feasibility of incorporating a dyspnea evaluation protocol into bedside assessments routinely performed by respiratory therapists”. Systematic dyspnea assessment (“Are you feeling short of breath right now”; “Is your shortness of breath mild, moderate, or severe?”) was performed at 4-h intervals in 238 patients, leading to a database of 2,539 patient–respiratory therapist encounters. Dyspnea was evaluated according to protocol in 74 % of the cases, with 32.1 % of the patients being adequately alert to answer questions during encounters (n = 600). Dyspnea was assessed to be present in 11 % of the cases and was characterized as moderate to severe in one-third of these. The study by Powers et al. [45] was designed to evaluate the “test–retest reliability of five dyspnea rating scales” and to examine the “correlations between each of these 5 rating scales and physiological measures of respiratory function”. Within a convenience sample of 28 patients, 50 % reported dyspnea, with a median VAS rating of 52 % of full scale. Finally, Schmidt et al. [27] studied 96 alert mechanically ventilated patients cared for at two separate ICUs; of these patients, 47 % reported dyspnea, with a median VAS measure of 50 % of full scale. This study was the first to explore the modality of dyspnea experienced by ventilated patients. When asked to describe their dyspnea as either “air hunger” or “excessive work/effort”; 56 % of the dyspneic patients chose “air hunger” only; 16 % chose “excessive work/effort” only; 13 % chose both “air hunger” and “excessive work/effort”; the remaining patients were unable to make a choice (Fig. 2).

Fig. 3
figure 3

Relationship between dyspnea and tidal volume. Data are re-plotted from Mols et al. [34] (left) and Leung et al. [33] (right). Although the authors of both papers reported a difference in respiratory comfort dependent on ventilator mode, these plots suggest that the main effect of changing mode is to change the tidal volume delivered by the ventilator. Consequent changes in pulmonary stretch receptor activity as well as blood gasses are in the correct direction to explain the observed effects on discomfort ratings. PAV Proportional assist ventilation, PSV pressure-support ventilation, IMV intermittent mandatory volume ventilation, dyspn unassist unassisted breathing, %FS % full scale

The available set of data on dyspnea in mechanically ventilated patient is very limited in size and is heterogeneous in quality. It suggests that the frequency of this clinical issue is sufficiently high to make it worthy of attention. The next important question relates to the clinical relevance of being dyspneic under mechanical ventilation.

Clinical relevance of dyspnea in mechanically ventilated patients

Immediate suffering

Dyspnea is a noxious sensation. With VAS ratings of generally around 50 % of full scale [27], dyspnea in mechanically ventilated patients qualifies as “moderate to intense”. It can however reach unbearable levels [40]. Similar pain ratings constitute a clear indication for the administration of analgesics [50].

Dyspnea causes anxiety [51]. In the study by Schmidt et al. [27] discussed above, dyspneic patients were more likely to present with anxiety than non-dyspneic patients (71 vs. 24 %, respectively), and dyspnea was independently associated with anxiety (OR 8.84, 95 % CI 3.26–24.0, p < 0.0001). Experimental dyspnea produces anxiety even in healthy subjects who know they are safe and disease-free [19, 52]. The interplay between anxiety and dyspnea is complex, and causative relationships can exist in both directions.

Delayed psychological sequelae

As mentioned above, patients’ recollections of their ICU experience point to dyspnea during mechanical ventilation as a major ICU stressor [38, 40]. Dark “respiratory recollections” may persist for several weeks. Bergbom-Engberg and Haljamae [51] studied 158 patients who had been mechanically ventilated during an ICU stay and who could remember the treatment. When questioned 2–48 months after discharge, 47 % of these patients reported having felt anxiety and/or fear during mechanical ventilation. Respiratory difficulties (e.g. “to synchronize with the respirator…”) were among potent drivers of feelings of anxiety, fear, agony, panic, and insecurity.

Post-traumatic stress disorder (PTSD) is now recognized as a common sequel of the ICU experience [5356]. Recalled dyspnea has been found to be associated with PTSD in ICU survivors [54], and the PTSD symptom score in the post-ICU population has been found to be significantly correlated with the duration of mechanical ventilation [55].

Impact on ICU stay outcomes

Dyspnea during mechanical ventilation may be useful to predict weaning outcome and ICU length of stay. In the study of Schmidt et al. [27], successful extubation within 3 days of the assessment of dyspnea was significantly less frequent in patients whose dyspnea failed to recede following the adjustment of the ventilator settings than in the patients whose dyspnea improved after ventilator adjustment.

The impact of dyspnea on ICU outcomes and on PTSD is strongly suggested by correlative data, but needs to be clarified with interventional trials. Until this link is disproved, the sensible and compassionate approach is to identify and alleviate dyspnea in mechanically ventilated patients.

Assessing dyspnea in the mechanically ventilated patient

The 2012 statement on dyspnea by the American Thoracic Society emphasizes strongly that “…dyspnea per se can only be perceived by the person experiencing it. Perception entails conscious recognition and interpretation of sensory stimuli and their meaning. Therefore, as is the case with pain, adequate assessment of dyspnea depends on self-report” [14]. The first step in managing dyspnea is therefore to ask the patient what he or she feels. One may obtain verbal or written responses to appropriate questions and/or responses to rating scales; if these fail, indirect means should be used to assess the pain.

Direct approach: guided questioning and rating scales

Clinical practice shows that conscious mechanically ventilated patients are able to answer simple questions about their respiratory sensations, even when they are slightly confused, cognitively impaired, or unable to speak due to an endotracheal tube. It is useful to start with a qualitative approach using questions phrased to allow simple yes/no answers, such as “is your breathing OK?”, “do you have difficulties breathing?”, “is your breathing comfortable?”.

Most awake patients can respond appropriately to pain and dyspnea scales, which can provide considerably more useful information than binary responses. Rating scales provide a quantitative measure that can be tracked over time in the same patient and can be statistically tested when various parameters are being assessed, such as treatment effects, clinical unit performance, among others. In mechanically ventilated patients, dyspnea can be measured using available psychometric tools. Single-dimension dyspnea scales include the classical VAS, the modified Borg scale, a numerical ordinal scale, and the faces pain scale. Powers and Bennett [45] tested all of these scales in a small population of mechanically ventilated patients and found the Test–retest reliability and intraclass correlation to be satisfactory [45]. In patients who cannot manipulate those tools themselves for any reason, it is easy for caregivers to provide adequate help. This option was explicitly offered in the study by Schmidt et al. [27] and no problems were reported. Of note, VAS have been used in mechanically ventilated patients to assess pain, anxiety, and the sense of inspiratory effort under different ventilatory assistance modalities [57].

Indirect approach: clinical surrogates

When the patient is unable to report his or her dyspnea, clinical surrogates of dyspnea would be useful. In mechanically ventilated patients, individual clinical signs correlate poorly with dyspnea ratings [27], contrary to the case in spontaneously breathing patients with acute respiratory failure. A composite observation scale such as the Respiratory Distress Observation Scale (RDOS) may be useful [58, 59]. This scale comprises weighted measures of heart rate, respiratory rate, the use of inspiratory neck muscles to breathe, the presence of paradoxical abdominal movement during inspiration, the degree of restlessness, the presence of end-expiratory grunting and nasal flaring, and, importantly, the presence of a fearful facial display. In studies of pulmonary rehabilitation patients and palliative care patients, the RDOS has been shown to correlate with VAS ratings and exhibit internal consistency and discriminant validity with pain [58]. The RDOS has also been found to respond to treatment, including treatment in patients who had been unable to self-describe their dyspnea [59]. However, the RDOS has only been validated in spontaneously breathing patients and needs to be adapted and validated for ventilated ICU patients because several of the vital sign variables are likely to be disturbed by medications and mechanical ventilation in ICU patients.

Patient–ventilator dyssynchrony can be observed clinically and through the inspection of the ventilator-derived pressure and flow waveforms. How dyssynchrony relates to dyspnea is unclear. Indeed, patient–ventilatory asynchrony studies seldom mention dyspnea. Of note, situations typically associated with patient–ventilator asynchronies, such as over-assistance in patients with COPD, are likely to be associated with a strong central neural inhibition [60] that could, paradoxically, “protect” from dyspnea [61].

Electrophysiological surrogates

Physiological surrogates of neural respiratory drive based on flow, volume, and respiratory muscle pressure generation usually underestimate the levels of neural respiratory drive in patients with compromised respiratory mechanics [62]. Because they are not influenced by respiratory mechanics, electrophysiological indices may be better indices of the neural drive from respiratory motor centers. As mentioned above, a corollary copy of respiratory motor drive is thought to be an important excitatory input to dyspnea and, therefore, a good measure of respiratory motor drive may help to infer dyspnea. It should be remembered, however, that severe discomfort arises not from drive alone, but from a failure of ventilation to match drive.

Electromyographic approach

The electromyographic activity (EMG) of the diaphragm and of the extra-diaphragmatic inspiratory muscles has often been used as a measure of respiratory center motor drive [6365]. The results of several studies conducted in healthy subjects and in patients suggest that inspiratory muscle EMG measurements could provide a surrogate neurophysiological biomarker of dyspnea [61, 64, 6669]. These observations have been validated in ICU patients [61, 70].

Electroencephalographic approach

Normal individuals faced with external inspiratory loading exhibit a respiratory-related premotor cortical activity that can be described using electroencephalographic tools (EEG) [71]. Raux et al. also observed this activity in patients on non-invasive mechanical ventilation when respiratory discomfort was induced through the use of inappropriate ventilator settings [72]. This premotor activity was correlated with respiratory discomfort under the circumstances of these experiments [72]. Although causal relationship between this cortical activity and dyspnea remains to be established and although ICU data have not yet been published, “respiratory-related EEG” could also provide a surrogate neurophysiological biomarker of dyspnea in the ICU.

Conclusions

Dyspnea in mechanically ventilated ICU patients is an under-recognized issue. We suggest that assessment and management of dyspnea in these patients has the potential to minimize suffering, reduce the use of sedation, and reduce ICU-related PTSD. We therefore suggest that future studies should aim to (1) provide tools to evaluate dyspnea in patients who cannot respond to questioning; (2) directly evaluate the impact of dyspnea in ICU patients; (3) determine the efficacy of strategies aimed at minimizing dyspnea, ranging from improved ventilation strategies to pharmacologic interventions.