The main findings of our study are that we were able to perform decannulation in 86% of the whole population, and 100% of our decannulations were successful. Another important finding is our tracheostomy weaning duration (7.6 [SD: 4.6] days), which appears to be rather short compared to the current literature (19–72 days) [24,25,26,27]. Conversely, our total tracheostomy time (42.5 [SD: 24.8] days) appears to be similar to what is commonly reported (25–74 days) [24, 28, 29].

Our results suggest that our protocol may be efficient for assessing which patient is ready to be successfully decannulated, such that most will reach decannulation. Indeed, a high decannulation success rate can be associated with very conservative weaning protocols, thus with a low decannulation rate. However, this was not the case in our study, in which almost all of the participants were decannulated and all of our decannulations were successful. However, such a high success rate could also be interpreted as a consequence of unnecessary tracheostomy and easier tracheostomy weaning as a result. Although, the mean CRS-r at inclusion was 16.0 (SD: 6.7) and one-third of our 30 patients were classified as being in an unresponsive wakefulness syndrome or minimal conscious state at inclusion, almost half of our patients were tracheostomized for “neurological status incompatible with extubation”. In addition, other than having a low level of consciousness, swallowing disorders can lead to tracheostomy weaning failure [11, 15, 30]; in our study, all of our patients had impaired swallowing function at inclusion and almost half of our patients were tracheostomized for a “swallowing disorder”. Notably, all of the patients in our cohort were included exhaustively and consecutively in 2 units that are used to receive all the brain-injured patients at ICU discharge (traumatic and non-traumatic ABI). Our hospital is the largest of our state and the only one receiving severe ABI patients in ICU. This means that almost all tracheostomized ABI patients were discharged from ICU in our 2 neurosurgery units at the time of the study. Thus, we believe that our study population is representative of tracheostomized ABI patients that clinicians are used to managing.

Many of our patients could have started the tracheostomy weaning protocol earlier and even might have been decannulated in the ICU. However, delaying tracheostomy weaning after ICU discharge to the ward was imposed by the study protocol, as we wanted to assess the safety and efficiency of our tracheostomy weaning protocol outside of the ICU. Thus, our patients all benefited from “late” tracheostomy weaning (i.e. in which the mean time between tracheostomy insertion and inclusion was 35.0 [SD: 22.8] days).

Our results then clearly pose the “early vs. late” tracheostomy weaning dilemma. Generally, recent studies tend to advocate for early rehabilitation for ABI patients [31]. Are these concepts necessarily transposable to tracheostomy weaning? The ICU can arguably be considered the safest place to perform tracheostomy weaning. It provides a higher number of caregivers, monitoring facilities, and re-intubation or rescue procedures that can easily be performed by on-the-spot intensivists [2, 14]. Early tracheostomy weaning can therefore be started during the patient’s ICU stay [14, 15]. However, ABI patients’ critical statuses can be an obstacle and a cause of weaning failure in the acute phase [26]. By delaying tracheostomy weaning (i.e. mean delay was 35.0 [22.8] days in our study) at discharge from the ICU, our patients may have been able to attain respiratory, haemodynamic, and more importantly neurological stability. This may explain our very short tracheostomy weaning duration and high success rate. Moreover, before inclusion, all patients received routine physiotherapy and respiratory therapy delivered by the attending ICU physiotherapists. It is possible to consider that it has participated to the general improvement of these patients before inclusion. In a precedent study on a comparable population, early rehabilitation (motor, sensory and sometimes verticalisation) resulted in an earlier decannulation (61 vs 94 days for the delayed rehabilitation group). It has to be noted that our mean total time to decannulation remains shorter [3].

Concerns can be raised that because our protocol started only after ICU discharge, it could have been responsible of prolonged cannulation, which is thought to cause tracheal lesions such as tissue granulation, oedema, and tracheomalacia [32]. However, our total tracheostomy time is similar to what is found in the literature [24, 28, 29]. By delaying tracheostomy weaning, we did not shorten the total tracheotomy time, but tracheostomy weaning may have been easier, safer, and had a high success rate. Moreover, if tracheostomy weaning had been started in the ICU, it may have lengthened the ICU length of stay and increased hospitalization costs, as the ICU cost is higher than the ward cost (in our hospital, the cost of one day in the neurosurgery unit and in the neurological ICU are 655.86€ and 1115.39€, respectively). Our tracheostomy weaning protocol might be safe enough to be performed outside of the ICU without systematic instrumental examination, with a high success rate, and without lengthening of the total tracheostomy time. Thus, tracheostomized ABI patients could be discharged earlier from the ICU, without having completed or even started tracheostomy weaning. Counterintuitively, delayed tracheostomy weaning in ABI patients could be a potential resource that offers cost savings. However, our protocol needs to be tested against a FEES guided one in order to confirm this hypothesis.

One of the specificities of our protocol is the absence of systematic instrumental assessment such as FEES or VFSS, which are considered gold standard evaluations for swallowing disorders [33, 34]. They can be used as an effective tool to guide tracheostomy weaning; FEES, in particular, can be very useful for diagnosing vocal cords impairments or pharyngo-laryngeal lesions such as tissue granulations, all frequently associated with swallowing disorders and thus tracheostomy weaning failure [34,35,36]. Warnecke et al. have proposed a FEES-guided protocol that seems to be faster, safer, or with less false negatives [15]. Unfortunately, FEES is not always immediately available outside of the ICU or ear, nose, and throat (ENT) ward and depends almost exclusively on medical doctors in our country. VFSS for its part is not a bedside assessment and requires moving the patient to the radiology unit, which is not always appropriate at the acute or sub-acute phase. In our country (France) and in many others, relying on instrumental assessment only to manage all the tracheostomised patients with reasonable delay and to decide whether a patient should be decannulated or not would simply be impossible. Thus it might be responsible of unnecessary prolonged cannulation that are very risky too (38)(Cheung & Napolitano. 2014). One of our goals was to create a tracheostomy weaning protocol that can be used under medical supervision by a nonmedical team and that is based almost solely on clinical examination. Reverberi et al. stated that instrumental assessment is not always available or even feasible with ABI patients, and might be for selected cases only [9]. They suggested that most of the patients should be able to undergo a tracheostomy weaning protocol based only on clinical parameters [9]. Our aim was to create a protocol that is strong enough to minimise the risks (without ignoring them) for these patients and to detect which one really needs instrumental assessment. Thus, we replaced direct objective instrumental evaluation with indirect clinical assessment. For example, swallowing disorders and aspirations were revealed by respiratory signs such as an increase in suctioning and sputum, an increase in the respiratory rate, and/or an increase in the body temperature (as a consequence of lung infection). Default in airway patency (caused by tissue granulation or oedema, or by vocal cord paralysis for example) was revealed by respiratory noises such as a laboured audible breathing particularly if associated with an increased respiratory rate and/or accessory respiratory muscle involvement. Therefore, we did not consider FEES and VFSS as mandatory examinations, and management of intercurrent events was determined based directly on our clinical assessments. For example, cannula downsizing was performed at the first instance, to treat obstruction without a prior FEES. Nevertheless, in our protocol, FEES or VFSS could have been planned for patients with unresolved suspicion of tracheal stenosis or unmanageable swallowing disorders. It happened only for one patient in our cohort during tracheostomy weaning (patient 9 with FEES only). FEES was performed by an ENT resident, as we suspected a tissue granulation after an episode of respiratory distress during the night; it did not reveal tissue granulation, and a mucus plug was suspected. In one study using FEES, it was found out that the presence of tracheal lesions (tissue granulation and oedema were by far the most common) was rarely the cause of decannulation failure [32]. In this study, treatment options were tube change, laser, systemic or nebulized steroid therapy or combined therapy [32]. It is what we proposed in our protocol. The difference is that we would start with steroid therapy or tube change without prior FEES and eventually relied on FEES in case of failure.

We could also have used the blue-dye test to detect silent aspirations. However, despite having an excellent specificity (100%), it is a very low sensitive test (10%) [37].

Swallowing disorders might have been misdiagnosed or underdiagnosed because of the lack of instrumental assessment. Pulmonary infection is one of the major complications of swallowing disorders and can lead to death or delayed discharge [38]. However, in our cohort, we only had 2 pneumonia (3% of the total intercurrent events). They were successfully treated by antibiotherapy (Table 1, 2, 3). A study conducted in a population of subarachnoid haemorrhage (SAH) tracheostomized patients had a post-tracheostomy pneumonia rate over 10% which is fairly higher than ours [29].

The absence of objective assessment could be seen as a limitation of this study, as well as the absence of swallowing rehabilitation. However, we believe that tracheostomy weaning might be seen as a good way to functionally assess swallowing disorders (with cuff deflation and tube capping and a careful monitoring) at least concerning airway protection and non-alimentary swallowing, and also the best way to offer swallowing rehabilitation. We choose not to assess alimentary deglutition during tracheostomy weaning because all our patients had enteral nutrition (mostly gastrostomy). Alimentary-swallowing could be easily tested later, after decannulation. Waiting for the patient to be able to perform alimentary deglutition in order to decannulate might be responsible of unnecessary prolonged cannulation if the patient is already able to manage saliva. Moreover, there are evidences that mild dysphagia is not a strong argument against decannulation in this population (Enrichi et al. 2017). We believe that our protocol is able to detect patients unable to manage saliva and/or with severe dysphagia (as they would not be able to pass through our protocol steps) and thus prevent them to be decannulated.

We did not include cough assessment as a criteria for decannulation. Many protocols suggest that a strong cough might be a good predictor of decannulation readiness [39]. Accuracy of such classification (weak vs strong) remains questionable until you perform cough instrumental assessment. Bach et Saporito have proposed a peak cough flow (PCF) > 160 l/min as a cut-off value [40]. However, the population was composed of neuromuscular patients (e.g. amyotrophic lateral sclerosis). Despite having serious disorders these patients have few cognitive disorders and are usually able to actively participate to such testing. ABI patients are rarely able to do so. To our knowledge, only one study has described an induced peak cough flow (IPCF) suitable for ABI patients [41]. Accuracy, sensitivity, and specificity for successful decannulation were, respectively, 75%, 85,7%, and 54,7% with an optimal cut-off point of 29 l/min [41]. Here again we wanted to create an easy tracheostomy weaning protocol. The IPCF described by Chan et al. needs specific material and can be difficult to perform. Thus we choose not to include a specific cough evaluation.

Concerning the intercurrent events, aside from anxiety, the main two of them (i.e. accessory respiratory muscle involvement and a laboured audible breathing) can be associated with reduced airway patency, which can have multiple causes (e.g. vocal cord paralysis or tissue granulation) [42, 43]. The most frequent treatments and associated procedures (i.e. aerosol therapy with adrenaline and downsizing cannula) performed in our study are directly linked to these main intercurrent events. Moreover, with a mean of 2.3 per patient, the number of intercurrent events was quite low in our study. These events appear to be easily identifiable with clinical assessment, not so frequent, and quite easy to manage. This highlights the fact that tracheostomy weaning of an ABI patient outside of the ICU may not be as overwhelming as once thought.

Interestingly, anxiety, a frequent intercurrent event in our study, was not as frequently treated with medication as accessory respiratory muscle involvement or a laboured audible breathing. In fact, anxiety was probably more of a one-off state for the patients than a general one, and it was likely to be associated with intercurrent events; thus, it does not appear to require long-term drug therapy for many of our patients. A caregiver’s accompaniment or treatment for intercurrent events appears to have been sufficient to reduce anxiety in our patients. However, the use of anxiolytics may be helpful in some cases, particularly when the patient is not able to understand or to participate (e.g. in the case of comprehensive aphasia) [44, 45]. However in our study, anxiety was only clinically assessed by the whole team and decision to treat was taken collegially. This can be a cause of mis-diagnosis. But to our knowledge, there is no anxiety scale available for ABI patients with such disabilities and communication disorders.

Scopolamine has been used successfully to treat excess saliva, usually a sign of a swallowing disorder in which the patient’s swallowing frequency or efficiency is reduced. Because it thickens saliva and reduces its production, scopolamine must be used with caution to avoid mucus plug development. However, its use remains controversial and there is no clear evidence of its efficiency [46, 47].

The main limitations of our study were the relatively small number of patients in our cohort and its monocentric nature. Indeed, our protocol must be tested on a larger scale, in multiple centres and against a FEES-guided protocol. In effect, we believe that our high rate of successful decannulation does not rely only in our team’s pre-existing experience with tracheostomy weaning, or in the skills of a few clinicians. Even if team-based or multidisciplinary tracheostomy weaning had already shown its superiority over standard care ones [48], the additional strength of our protocol seems to lie in its ability to precisely drive caregivers through the use of our logigram, giving them the tools to prevent, recognize, and manage adverse events associated with tracheostomy weaning in ABI patients. Moreover, decannulation can be considered a rather stressful event for caregivers, the patients, and their families. Thus, the framework provided by our procedure secures the weaning and decannulation processes. However, even with larger-scale multicentric studies, it will never be possible to create a protocol capable of preventing every tracheostomy weaning failure. A “zero risk” tracheostomy weaning and decannulation protocol will probably never exist. However, we believe that, with our protocol, the risk of failure can be controlled to the maximum extent and is worth the effort with regard to the benefits of decannulation for these patients. In addition, team education as planned in our protocol is probably one of the keys for implementing a safe and efficient procedure.

Our death rate is 6% (2 patients). It seems to be fairly acceptable considering that tracheostomized ABI patients are usually patients with very severe disabilities and poor outcomes. Moreover, it is quite comparable with the current literature (between 4 and 21% in the decannulation failure group in Küchler et al. [49], between 5 and 6% in Huang et al. [50]). Additionally, the overall mortality in tracheostomized patients has been shown to range from 22 to 45%, which is way over our mortality rate [51].

Of the two patients who died (patients 12 and 13), the question of the imputability of our protocol remains for one (patient 13). According to our protocol and the patient’s stability parameters, tracheostomy weaning had been stopped. Although direct imputability could not have been established, it may be related to the low neurological status of this patient at inclusion (CRS-r = 3). Notably, five patients had a very low CRS-r at inclusion (below 7 and considered to be in an unresponsive wakefulness syndrome) and could have been decannulated (patients 1, 6, 11, 20, 22, see Table 4). In addition, this patient (patient 13) was considered a very severe case, with a poor recovery prognosis. Prior to inclusion, ethical decisions (limitation of active therapeutics) were considered accordingly by the medical staff and the patient’s relatives. However, we decided to include this patient considering the potential benefits of decannulation. Moreover, here again, four patients had the same ethical restriction but were decannulated successfully and were still alive at 6 months (patients 1, 22, 25, and 28, Table 4). Notably, one patient who had ethical restrictions at the inclusion died during the 6-month follow-up (patient 7); however, this patient was not able to reach decannulation and was excluded from the tracheostomy weaning procedure after 3 months, according to our protocol. Evidence of clear predictive factors is still lacking to determine which patients are able to undergo tracheostomy weaning and decannulation. Thus, we chose to include patients regardless of their neurological status. Our protocol was used as a decision-making logigram, giving every patient a chance to reach decannulation.

At 6 months, 55% of our patients were still hospitalized. Considering our national care system and the severity of these patients, our proportion of hospitalized patients is not unusual. To note, some of them are even benefiting of home hospitalisation (1 patient) or day time hospitalisation (1 patient). Moreover, if these patients would not have been decannulated, it is very likely that they would have remained in our unit.

These days, despite the growing research, tracheostomy weaning and particularly decannulation in ABI patients still resemble more of an art form than a well-established science. Due to the lack of evidence, this procedure is usually considered to be unsafe or overwhelming outside of an ICU or specialized unit. In this study, we evaluated the feasibility of such a process using a protocol tailored to patients based on the stability of their condition, allowing every patient to undergo safe, yet efficient, tracheostomy weaning. The decannulation decision was made through the use of our protocol, without the help of systematic instrumental assessment such as FEES or VFSS. Given a decannulation rate of 90% and a success rate of 100% in this study, we believe that our protocol might be used outside of the ICU or specialized unit by a pluridisciplinary non-medical team under medical supervision. The safety and efficiency of the protocol are based on team education and coordination and should be evaluated against instrumental evaluation with a randomized controlled study. Because these are results of a pilot study, we think that we must warn the reader against a misuse of our protocol especially before a controlled study has been performed.