Extubation Readiness in Critically Ill Stroke Patients

Introduction

See related article, p 1946

Endotracheal intubation and mechanical ventilation are common procedures for critically ill stroke patients. Although necessary and life-saving, timely extubation after ventilator weaning is desirable, because patients with delayed extubation experience higher pneumonia rate, increased need for tracheostomy, longer length of stay on the intensive care unit (ICU), and higher mortality.¹ However, extubation failure (EF) and subsequent need for emergent reintubation is associated with similar sequelae.^2–4 Although determining extubation readiness is paramount, this issue continues to be challenging in the neuro-ICU population as there are no widely accepted and validated standards.

Conventional criteria based on respiratory mechanics have shown little value in neurological diseases because intubation is often performed for airway protection rather than primary lung pathology.^4–7 Only recently, postextubation dysphagia (PED) became a growing concern as a major risk factor for EF and significant contributor to poor patient outcomes^8,9 with prevalence rates ranging from 12% to 69%,^10–14 being highest in neurological patients (93%).⁹ Damage to the central swallowing network itself is the primary cause of PED in cerebrovascular disease, which constitutes the leading diagnosis on neuro-ICUs.¹⁵ Further mechanisms include pharyngolaryngeal lesions caused by the tube, critical illness neuropathy and myopathy leading to muscle weakness and dyscoordination of breathing and swallowing, and an impaired sensation due to sedation, mucosal damage, or the underlying critical illness itself.⁹ As a consequence reintubation rates in neurological collectives are as high as 20% to 40%.^4,7,16

Despite the important role of PED, a recent survey confirmed there are hardly any well-established strategies for its diagnosis.¹⁷ Some authors tried to integrate neurological measurements, such as vigilance, ability to follow commands, and stroke severity into extubation decision making.^5,10,18,19 Others used parameters related to airway management like gag reflex, cough peak flow, and secretion status.^16,20 However, results were inconclusive.^1,7,21,22 In a study by Ajemian et al,¹⁰ no indirect clinical measures could be linked to PED.

There is a clear need for guidelines when and how to evaluate for PED in the neurological critically ill population to support airway management decisions. The first aim of the present trial was to identify clinical features increasing the risk of EF and in particular to clarify the role of PED in this context. Second, we intended to develop a clinical score for the prediction of EF in orally intubated stroke patients.

Patients and Methods

Supporting data are available within the article and in the online-only Data Supplement.

Study Design and Patients

This prospective observation trial was conducted on the neuro-ICU at our university hospital. Consecutive patients meeting the study criteria were included from August 2013 until February 2016. Orally intubated adult stroke patients (ischemic or hemorrhagic) were eligible to be included when they were deemed ready for primary extubation by the treating neurointensive care physician according to the extubation criteria listed below and his personal judgment. Patients were excluded from the study if extubated in a terminal setting or needing primary tracheostomy. The study was approved by the local ethics committee. Informed consent was obtained from all patients, their relatives or legal representatives.

Parameters and Procedures

All airway management decisions were at the discretion of the treating neuro-intensivist. According to local guidelines, extubation was performed if the following criteria were fulfilled: Glasgow Coma Scale >8, no signs of elevated intracranial pressure, body temperature 36°C to 38.5°C, heart rate 60 to 120 bpm, and systolic blood pressure 90 to 185 mm Hg. Additionally, the following respiratory extubation criteria were considered: spontaneous respiratory minute volume (≤12 L), positive end-expiratory pressure (≤5 mm Hg), Pao₂/Fio₂ (>200), and rapid shallow breathing index (<105).

Information on patient demographics, stroke characteristics, and treatment were collected. Before extubation, duration of ventilation, fulfillment of respiratory extubation criteria, current National Institutes of Health Stroke Scale (NIHSS) score, Glasgow Coma Scale, ability to follow simple commands, and Richmond Agitation-Sedation Scale were noted. Secretion status was assessed with a modified semiquantitative airway score as previously published (m-sqAS; Table 1).⁴ In addition, oral motor function was investigated with a score (OMF; see also Table 2) based on the literature²³ and our local practice. Respective items were scored, and a sum score was built.

Fiberoptic endoscopic examination of swallowing (FEES) was conducted within 48 hours after extubation by a speech-language pathologist and a trained neurologist. Equipment consisted of a flexible rhinolaryngoscope (11101RP2, Karl Storz, Germany) with a light source and camera (rpCam-X, rpSzene, Rehder/Partner, Germany). The secretion status was evaluated,²⁴ and spontaneous swallowing frequency per minute was observed. Pharyngeal sensitivity was tested by gently touching pharyngolaryngeal structures. Sensitivity was classified as intact (immediate swallowing response, cough, or laryngeal adduction), reduced (any weak response), or absent (no reaction at all). Following an established stepwise FEES examination protocol validated in stroke patients,^25,26 subjects were successively given standard volumes of puree consistency, liquids, and soft solid food. According to the risk of penetration or aspiration with saliva and different food consistencies, stroke-related dysphagia was classified on the 6-point Fiberoptic Endoscopic Dysphagia Severity Scale (FEDSS) with 1 scoring best and 6 being worst. A FEDSS >1 was classified as dysphagia. Moreover, a 3-ounce water swallow test²⁷ was performed within 72 hours following extubation in those patients deemed to be in the condition to swallow water by the attending physician.

End Points

Patients were followed-up for the need of reintubation until discharge. Timing and reason for reintubation were documented. Incidence of pneumonia, length of stay, functional outcome at discharge (modified Rankin Scale), and in-hospital death were recorded as secondary end points.

Statistical Analysis

Univariate comparisons of patient groups with extubation success versus EF were performed using the t test for normally distributed continuous variables, the Mann-Whitney U test for non-normally distributed continuous and ordinal variables, and the χ² or Fisher exact test for categorical parameters. Spearman correlation was used to assess the concordance between FEDSS, m-sqAS, and OMF. Optimal cutoff values were determined by receiver operator characteristics analysis with maximizing the Youden index. Independent predictors of EF were determined by multivariate binary logistic regression analysis with inclusion of those variables that were significantly related to EF in univariate comparisons.

For generation of the clinical EF prediction score multivariate regression analysis was restricted to those significant variables from the univariate analysis that had been collected before extubation. Significant variables from this model were included as items in the score and attributed points based on their odds ratio (OR), significance value, and mean values in the subgroups of successfully extubated versus reintubated patients. The final score was derived by calculating the sum of the points from all items. Afterward, we assessed the calibration of the score considering a logistic regression model for EF with the score as single predictor variable and calculating Nagelkerke R² as a goodness-of-fit measure. To assess the discrimination abilities and to determine optimal cutoff values, a receiver operator characteristics analysis was applied. A P value of <0.05 was used to indicate statistical significance. Data analysis was performed using SPSS 25.0 (IBM Corp).

Results

Reintubation rate was 24.1% of 133 patients with reintubation being performed after 24.1±43.0 hours. Reasons for reintubation were as follows: severe dysphagia: 16 (50.0%), primary respiratory complications not related to aspiration (28.1%), coma (15.6%), and need for surgery (6.3%).

Clinical Characteristics and Outcome of Successfully Extubated Versus Reintubated Patients

Patient characteristics are presented in Table 3. Patients with successful extubation had often been intubated primarily for thrombectomy, whereas patients needing reintubation had more often been intubated because of decreased protective reflexes, suffered from infratentorial stroke, and received thrombectomy for posterior circulation large vessel occlusion or in combination with later need for neurosurgery. The latter had longer ventilation duration before their extubation attempt, higher stroke severity, and lower Glasgow Coma Scale. They were less often able to follow commands. Notably, respiratory extubation criteria could not differentiate between groups. Regarding in-hospital outcome, patients needing reintubation had higher pneumonia rate, worse functional outcome, longer length of stay, and higher mortality.

Secretion Management and Swallowing Function

When looking at swallowing assessments (Table 4), the m-sqAS and OMF before extubation were significantly worse in reintubated patients. After extubation, patients needing reintubation failed significantly more often in the water swallow test. Taking FEES as a gold standard for the diagnosis of dysphagia, its prevalence in investigated patients was high with 75% in successfully extubated patients and 100% in patients needing reintubation. The latter had significantly higher FEDSS and showed worse secretion status with decreased laryngeal sensitivity, resulting in lower spontaneous swallowing frequency, impaired swallowing response, weak whiteout with incomplete inversion of the epiglottis, and insufficient or missing cough reflex.

Value of Swallowing Assessments for the Prediction of EF

OMF, m-sqAS, and FEDSS were strongly correlated (Spearman ρ: FEDSS—m-sqAS: 0.462; FEDSS—OMF: 0.425; m-sqAS—OMF: 0.556; P<0.001 for all correlations).

Using receiver operator characteristics analysis, the optimal m-sqAS cutoff to differentiate between extubated and reintubated patients was ≥3 (area under the curve, 0.700; [95% CI, 0.601–0.800], sensitivity 87.5%, specificity 45.5%). For OMF, the cutoff was ≥4 (area under the curve 0.784; [95% CI, 0.697–0.871], sensitivity 81.3%, specificity 62.4%). Ideal FEDSS cutoff was ≥5 (area under the curve 0.890; [95% CI, 0.826–0.954], sensitivity 84.6%, specificity 76.5%). A failed water swallow test predicted EF with a sensitivity of 68.4% and specificity of 82.4%.

In multivariable binary logistic regression analysis with inclusion of those variables that were significantly related to EF in univariate comparisons, that is, lesion location, thrombectomy plus neurosurgery, ventilation duration, reason for intubation, NIHSS, and Glasgow Coma Scale before extubation, ability to follow commands, m-sqAS, OMF, result from the water swallow test, and FEDSS, only the FEDSS prevailed as independent predictor of EF (adjusted OR [adjOR], 4.2; 95% CI, 1.5–11.9; P<0.007; please see Table I in the online-only Data Supplement for full results).

Determine Extubation Failure in Severe Stroke Score to Predict EF

Excluding variables FEDSS and water swallow test which were collected after extubation from the regression analysis, duration of ventilation (adjOR, 1.008), examination of OMF (adjOR, 1.593), infratentorial lesion location (adjOR, 14.002), and stroke severity before extubation (NIHSS, adjOR, 1.176) prevailed as significant independent predictors of EF (please see detailed results in Table II in the online-only Data Supplement). These items were, therefore, included in the Determine Extubation Failure In Severe Stroke (DEFISS) risk score. Points were attributed to each item based on the above-mentioned considerations. Ventilation duration was reduced to 2 categories (<24 hours: 0 points [pts] and ≥24 hours: 1 pt) according to a recent meta-analysis (7) and corroborating mean values in extubated versus reintubated patients from our study. This item received only one point because of its low OR. OMF was categorized based on the optimal cutoff value (<4: 0 pts and ≥4: 2 pts). Infratentorial lesion location received 2 pts as compared to supratentorial stroke (0 pts) because of its high OR. The NIHSS was categorized based on previously defined cut points^5,28 into: mild (<5: 0 pts), moderately severe (5–15: 1 pt), and severe/very severe impairment (>15: 2 pts). The final score was derived by summing up the points for each item (Table 5) with a maximum achievable value of 7. The score demonstrated very good calibration in regression analysis (Nagelkerkes R²=0.54) and excellent discrimination abilities in receiver operator characteristics analysis (area under the curve, 0.89; 95% CI, 0.83–0.95). An ideal cutoff ≥4 predicted EF with a sensitivity of 81.3% and a specificity of 78.2%. For cut point specific risk of EF, sensitivity, and specificity of the DEFISS, see the Figure.

• Download figure
• Download PowerPoint

Discussion

Our EF rate was within the range of what has been reported for neurological ICU collectives,⁷ with PED being the leading cause of reintubation and PED severity being strongly correlated with EF risk. Whereas passing respiratory extubation criteria could not discriminate between patients with successful extubation versus EF, all swallowing scores could differentiate between groups with sensitivities >80%. Objective examination with FEES predicted necessity of reintubation best but required trial of extubation. Therefore, we additionally propose the 4-item DEFISS-Score based on easy to collect clinical data before extubation.

The MadICU-survey revealed predominance of clinical swallowing assessment apart from neuro-ICU wards, of whom >85% use FEES.¹⁷ Our findings foster this positive trend towards objective diagnostics for PED. About clinical diagnostics, there are several screening tools validated for poststroke dysphagia but not for ICU patients and applicability of these tests is questionable. Lately, the Gugging Swallowing Screen for the ICU²⁹ was specifically developed to identify PED and guide diagnostic and feeding decisions in neuro-ICU patients but not to predict EF. The authors themselves concluded that this bedside screening tool should not replace a FEES investigation.²⁹ Previously, the water swallow test applied in our study was suggested as another method to screen patients following extubation.³⁰ In our opinion, it cannot be recommended for recently extubated neuro-ICU patients, since it performed worse in relation to FEES. Kwok et al¹² found silent aspiration in 37% of dysphagic trauma patients following extubation that would not have been detected without FEES. Silent aspiration is the main reason for the low reliability of subjective bedside evaluations.¹⁰

Within the last years, FEES, a method developed by Langmore³⁰ has evolved to an established procedure and proved to be extremely useful in the ICU setting for several reasons. A major advantage is the possibility to perform the examination repeatedly and at the bedside. Other advantages include the ability to visualize injury to laryngeal tissues, observe secretion management, and test laryngeal sensation directly.³¹ The method is associated with minimal discomfort or complications in recently extubated patients.^10,11,32 In the present study, FEES was conducted as soon as available within the first 48 hours. In line with that, most authors report examination within 48 to 72 hours post extubation.^10,13,32 However, a FEES as early as is feasible may be beneficial. Scheel et al¹³ found no difference in dysphagia frequency with the timing of swallowing assessment in the first 72 hours, opposing the anecdotal clinical practice to postpone the examination until the patient has stabilized after tube removal.²⁹ Another argument for timely FEES investigation is that the early identification of risk patients and reinstitution of ventilatory support has the potential to reduce the increased mortality associated with EF.³³ The FEDSS applied in this study has previously proven to strongly predict the outcome and intercurrent complication rate in acute stroke.²⁶ The calculated cutoff value to predict (re)intubation is concordant with earlier results.^25,26 After the evidence from these studies and the argumentation of Epstein et al³³ immediate reintubation of patients with FEDSS >5 might even be considered. However, this adverse incident may be preventable with intensified care in this high-risk group. This issue needs to be addressed in future trials.

Despite the undoubted benefits of FEES, there is a need for a valid prediction tool before an extubation trial. Two recent trials also proposed scores to guide extubation decision making, but these studies were conducted on mixed neuro-ICU collectives including traumatic brain injury and did not focus on PED, although swallowing attempts (yes/no) were significantly related to extubation success and, therefore, included as a gross measure of swallowing function in both scores.^34,35 Colonel et al²³ found use in a clinical test similar to the OMF, being able to predict a patient’s ability to cough and to eject bronchial secretions. However, no single criteria were reliably predictive of PED. In a recent meta-analysis, some items of the m-sqAS like secretion texture and gag reflex predicted EF, whereas others like secretion volume, coughing, or suctioning did not.⁷ In our study, only the OMF but not the m-sqAS prevailed as significant risk factor in multivariable analysis when excluding objective FEES results only obtainable after extubation. Besides that, we identified several further clinical predictors in critically ill stroke patients, all of which were included into the DEFISS-Score and are backed up by further scientific evidence: several authors found the duration of endotracheal intubation to be associated with PED incidence and severity,^8,9,12,36 and related EF, especially in posterior fossa strokes.³⁷ Infratentorial lesion location is known to be associated with most severe dysphagia, need for prolonged ventilation and reintubation.^37–39 Low NIHSS scores ≤15 predicted successful extubation in a study by Lioutas et al.⁵ Also in line with previous results,¹⁶ the chance for reintubation was not found to be age-dependent, and respiratory parameters were not predictive of extubation success.

While extubation seems safe in patients with a DEFISS <4, especially in combination with timely FEES, the question is how to proceed in case of a DEFISS above the cutoff. One strategy would be to reassess every 2 to 3 days, however, apart from the OMF and maybe the NIHSS, items are nonmodifiable. Another option would be to tracheostomize these patients in due time to expedite liberation from the ventilator while maintaining airway patency. Such early tracheostomy seems to be feasible and safe and may be associated with reduced sedation need in stroke patients,⁴⁰ as well as lower in-hospital morbidity and improved clinical outcomes at least in a mixed sample of brain-injured patients.⁴¹ In this context, the stroke-related early tracheostomy score including the categories neurological function, neurological lesion, and general organ function/procedure within the first 24 hours after admission has been proposed for tracheostomy decision making.⁴² However, in our preceding study conducted in collaboration with the stroke-related early tracheostomy authors, when compared with patients with unsuccessful extubation trials, primarily tracheostomized stroke patients still had longer ventilation duration and longer ICU-length of stay.⁴ This might be an argument in favor of an extubation trial before performing a tracheostomy too early. In the absence of robust medical evidence, it has been suggested to assess the need of further ventilation at the end of the first week of intensive care and proceed to tracheostomy if extubation is not feasible.⁴³ The impact of early tracheostomy versus prolonged orotracheal intubation on functional outcome of stroke patients is currently being assessed in a multicentre, prospective, randomized trial.⁴⁴

Further research should also concentrate on developing more effective treatment strategies for PED in critical illness survivors. Pharyngeal electrical stimulation which has shown to expedite decannulation in severely dysphagic critically ill stroke patients^45,46 may also hold the potential to improve PED.

Limitations

This was a preliminary single-center observational study, treatment decisions were at the discretion of the treating physician. Results may be biased by our local ICU culture, which is characterized by high awareness for PED. Our trial was unblinded, reporting, and observation biases may have influenced reintubation decisions. Finally, our study was conducted in a sample of predominantly ischemic stroke patients. Results may not be transferable to hemorrhagic stroke or the general neuro-ICU population because underlying causes for dysphagia may differ. Moreover, a sizable majority of ischemic stroke patients were intubated for thrombectomy only and thus mostly successfully extubated. Conscious sedation instead of general anesthesia, which necessitates intubation during the intervention, may in the future reduce artificial ventilation need in this subgroup. Prediction, discrimination, and calibration properties of the DEFISS-score need to be validated prospectively in a multicentric, independent patient cohort.

Conclusions

The risk of EF is strongly correlated with dysphagia severity in acute stroke. The DEFISS-score includes clinical assessment of OMF and easy to collect clinical data. It may guide extubation decision making in intubated stroke patients. If available, timely FEES is nevertheless recommended after tube removal. Patients with FEDSS ≥5 should closely be monitored for respiratory distress because of secretion aspiration with consecutive need for reintubation.

Footnotes