This study confirms that NIV can be very trying for the patients even when clinically beneficial . Indeed, cognitive and affective attitudes toward NIV markedly deteriorated during NIV. This is in line with the findings of the “PARVENIR” study, where a significant proportion of patients regretted having been selected to receive this treatment.
NIV beliefs and attitudes
The patients generally expected NIV-derived benefits and did not describe anticipatory anxiety or fear, except in an abstract way (two patients). This raises the hypothesis that the information given by caregivers prior to NIV was mostly positive. This makes the failure of NIV to bring the anticipated benefits a source of frustration, worsening the traumatic nature of NIV through an “unpredictability” effect [8,9,10].
Prior beliefs (dyspnea relief, increased independence) were often contradicted by the actual experiences (persisting or worsening dyspnea, mask-induced pain, limited communication, or lack of independence), yet positive elements were also reported, leading to ambivalence. Ambivalence is known to make patient behavior unpredictable [20, 21]. This may lead to NIV being refused in spite of its putative benefits being acknowledged.
Our patients reported only negative emotions (anxiety, claustrophobic feelings, fear of death) likely stemming from vital threat combined with lack of control, a known source of post-traumatic manifestations: our observations are therefore perfectly consistent with reports of post-traumatic stress disorders after NIV-treated acute respiratory failure .Of particular importance, NIV awoke traumatic memories from the distant past in two of our patients, related both to breathing and to being tied and unable to move. We posit that this type of reaction can strongly contribute to the post-ICU psychological burden [22, 23]. Although positive remarks were made during the post-NIV period (improved breathing, freedom, independence and calmness), attitudes toward NIV were less positive after than before NIV. In fact, many of the positive remarks were associated with “taking the mask off”, suggesting that NIV cessation was considered, somewhat paradoxically, as a relief. All in all, our patients mentioned more barriers than benefits regarding NIV, which, according to the Health Belief Model, leads to predict NIV refusal . Only two of our patients indicated that they would refuse NIV in the future. Mentioning barriers belonging to the affective register is in line with the findings of Baxter el al.  in patients with respiratory insufficiency due to amyotrophic lateral sclerosis.
Caregivers and family members
Positive attitudes toward caregivers were associated with their perceived attention toward the needs of the patients while negative attitudes were associated with lack of dialogue. This emphasizes the importance of empathizing with the patients when implementing a treatment known as potentially stressful , something that is not always easy to achieve in the ICU high-burden environment [26, 27].
The attitudes of the patients toward the presence of their relatives during NIV sessions were generally negative, mostly because the patients feared that seeing them have breathing difficulties would distress their loved ones. This “intuitive” concern appears appropriate: being exposed to the respiratory suffering of a relative who dies during an ICU stay complicates grief and contributes to post-traumatic stress , in line with the recent demonstration that observing dyspnea in others induces a negative affect . On the same plane, the avoidance behaviors of caregivers that some patients described could proceed from the suffering induced by the observation of NIV-related difficulties, possibly exacerbated by an impression of helplessness and hence of professional failure.
Strengths and limitations
The major strength of our study lies in the novelty of the data provided: qualitative studies have addressed mechanical ventilation in the ICU [30,31,32,33,34], but none have focused on NIV in the context of acute respiratory failure. We acknowledge several limitations. Firstly, the population size is limited, and we did not formally evaluate data saturation . Yet there were enough common elements in the discourses of the 10 patients to suggest that increasing their number would not have fundamentally changed the general pattern, and therefore to propose some conclusions . Of note, the population under scrutiny was fragile and the study setting made both patient recruitment and the interviews highly demanding. In this regard, we acknowledge that the physiological conditions of the patients (e.g., intensity of gas exchange anomalies, intensity of dypnea) was bound to have an influence on their answers, and that this was not controlled in any sort of way. Secondly, our patient sample was homogeneous in some respects (hypercapnia in 8 cases, first contact with NIV in all cases), but heterogeneous in others (nature of the underlying disease, see Table 1). The majority of the patients enrolled suffered from chronic respiratory failure (COPD, ALS, left heart failure—Table 1) and only three suffered from de novo acute respiratory failure. Our data are therefore probably more pertinent to the former than to the latter situation. Thirdly, we only enrolled patients in whom NIV was considered a success after the first 24 h, namely not followed by tracheal intubation. This is evidently a possible source of bias, but it can be hypothesized that the experience of patients in whom NIV failed to avoid intubation should have been even worse. Fourthly, the interviews were conducted after NIV, so recollection biases regarding the “before” and the “after” periods are possible. It would have been methodologically superior to interview the patients before NIV and to repeat the interview after it, but this would have posed immense practical problems. Interviewing the patients during the NIV sessions is obviously impossible. However, conducting the interviews very soon after NIV initiation (rather than at the end of the ICU stay or after, as is often done in ICU qualitative studies), should have minimized recollection biases and preserved spontaneity. The delay between NIV start and the interviews could have influenced the nature of the patients’ responses, but this delay was balanced (5 patients interviewed on day 2, 5 on day 3). Finally, we did not follow the patients up and therefore cannot relate our observations to long-term physical or psychological outcomes.
Our observations help delineate propositions that may improve the NIV experience by preventing the development of negative attitudes that carry the risk of reducing patient concordance with the treatment [7, 37]. Firstly, caregivers should openly inform the patients, before the initiation of NIV, that it can provoke anxiety and has drawbacks that must be weighted against its benefits. In that, the appropriateness of the very term “noninvasive” appears questionable; “mask ventilation” probably better represents the reality of the treatment. Secondly, patients about to undergo NIV should be asked about past traumatic experiences, and informed that forgotten such experiences may be reactivated. Thirdly, the presence and availability of a caregiver able to empathize with the patient during NIV sessions seems an obvious necessity. This may call for specific training [38, 39], that may be based on simulation teaching or role play . Fourthly, caregivers should enquire about dyspnea under NIV and to try to correct it [11, 41]. Fifthly, the presence of family members at the bedside should be discussed with the patient, as it can be counterproductive. Finally, it appears important that first NIV sessions be debriefed to avoid that further sessions “consolidate” a first traumatic experience. Some of these propositions can be implemented by ICU caregivers themselves, but others would be best implemented by a trained psychologist. Given the link between NIV poor tolerance and NIV failure and the corresponding negative outcomes, such measures could be of pronostic interest in addition to improving comfort.