Background
Swallowing difficulties are common, and dysphagia occurs in up to 62 % of intensive care unit (ICU) patients after extubation [
1]. Postextubation dysphagia may be related to several mechanisms: (1) impaired strength and sensation of the tongue [
2]; (2) laryngeal damage [
3,
4]; (3) neuromuscular impairment [
5]; and (4) cognitive complications of critical illness, such as somnolence and sedation, contributing to discoordination of the swallowing reflex [
6]. In turn, postextubation dysphagia is associated with delayed oral food intake (potentially increasing malnutrition and dehydration) and increased hospital-acquired pneumonia, reintubation, hospitalization duration, and mortality [
5,
7‐
9]. This is particularly important in elderly ICU patients because of delayed swallowing, defective larynx elevation, and loss of cough efficacy [
10,
11]. Accurate assessment of swallowing is therefore important and often formally undertaken by speech-language pathologists (SLPs).
Yet, no guidelines on postextubation swallowing assessment exist [
12]. Also, postextubation swallowing assessment is not done regularly, as only 41 % of hospitals were found to routinely screen extubated patients for dysphagia in one study, and only 44 % of patients completed a swallowing assessment in another study [
7,
12]. Similarly, in our institution, prior to this study, patients were not receiving accurate screening postextubation, and screening for postextubation dysphagia could be improved.
One possible method of making postextubation dysphagia assessment more streamlined and consistent would be to invite nurses to perform bedside swallow screening for postextubation patients. Patients who pass the screening would then be quickly put on an oral diet, which is an important psychological boost [
13]. Conversely, patients who failed the screening would then be referred to SLPs for early management, with potentially favorable outcomes [
6].
However, few data exist on the safety (i.e., pneumonia rates, reintubation rates, mortality, hospitalization duration) and effectiveness (i.e., oral feeding rates) of dysphagia screening performed by professionals other than SLPs, especially in nonneurological patient cohorts [
7,
12]. One protocol (described by Massey) involved a 60-ml water swallow test and was implemented by SLPs, but only 25 poststroke patients were studied, and safety information was not available [
14]. Another protocol (described by Leder) involved a 90-ml water swallow test and could be implemented by either SLPs or nurses, but the results showed poor specificity (about half the patients who failed the test were actually able to safely feed orally) [
15,
16]. A third protocol, the Toronto Bedside Swallowing Screening Test, has been developed for stroke inpatients [
17], but it has not been validated for postextubation patients. We believe that accurate dysphagia screening for postextubation patients with mixed etiologies could be done by health professionals other than SLPs. Therefore, we sought to audit the practice of nurse-performed screening (NPS) for postextubation dysphagia in our medical ICU.
Results
We studied 468 patients (281 in phase I and 187 in phase II) who were ventilated for a median of 2 days prior to extubation. We excluded 104 patients in phase I: 6 who were unable to feed orally before their critical illness, 6 who developed a permanent condition that precluded oral feeding, 69 who were terminally extubated, and 23 who had undergone tracheostomy. In comparison, we excluded 67 patients in phase II: 1 who was unable to feed orally before the critical illness, 2 who developed a permanent condition that precluded oral feeding, 52 who were terminally extubated, and 12 who had undergone tracheostomy. Patients in phase II, compared with those in phase I, were more ill with higher APACHE II scores, more had congestive heart failure as a comorbidity, and more had bronchiectasis as a comorbidity (Table
1). Despite this, proportionally more patients in phase II were allowed oral feeding at ICU discharge, with fewer patients developing pneumonia postextubation (see Additional file
5). In phase I, 11 patients (3.9 %) were reintubated for the following reasons: 6 for hospital-acquired pneumonia, 3 for acute heart failure, 1 for massive gastrointestinal bleeding, and 1 for asystolic collapse of unclear cause. In phase II, 13 patients (7.0 %) were reintubated for the following reasons: 2 for hospital-acquired pneumonia, 2 for extrapulmonary sepsis, 1 for massive malignant pleural effusion, 1 for severe asthma, 5 for acute heart failure, 1 for massive gastrointestinal bleeding, and 1 for status epilepticus. No significant differences in reintubation, ICU readmission, ICU and/or hospital mortality, or ICU and/or hospital LOS were found (Table
2).
Table 1
Patient characteristics
Age, years | 60.4 ± 16.2 | 59.7 ± 17.7 | 61.4 ± 13.5 | 0.255 |
Female sex, n (%) | 172 (36.8) | 103 (36.7) | 69 (36.9) | 1.000 |
APACHE II score | 26.1 ± 8.3 | 25.4 ± 8.2 | 27.2 ± 8.2 | 0.018* |
Weight, kg | 63.7 ± 17.8 | 63.7 ± 17.8 | 63.6 ± 18.0 | 0.946 |
Ventilated days preextubation, median (IQR) | 2 (1–4) | 2 (1–4) | 3 (1–5) | 0.509 |
Main diagnosis, n (%) | | | | 0.085 |
Pneumonia | 130 (27.8) | 67 (23.8) | 63 (33.7) | |
Nonpneumonia sepsis | 72 (15.4) | 45 (16.0) | 27 (14.4) |
COPD | 16 (3.4) | 7 (2.5) | 9 (4.8) |
Asthma | 29 (6.2) | 16 (5.7) | 13 (7.0) |
Fluid overload | 32 (6.8) | 21 (7.5) | 11 (5.8) |
Stroke | 11 (2.4) | 10 (3.6) | 1 (0.5) |
Seizure | 30 (6.4) | 19 (6.8) | 11 (5.9) |
Othera
| 148 (31.6) | 96 (34.2) | 52 (27.8) |
Comorbidity, n (%) |
Diabetes mellitus | 184 (39.4) | 100 (35.7) | 84 (44.9) | 0.053 |
Hypertension | 255 (54.6) | 151 (53.9) | 104 (55.6) | 0.776 |
IHD | 111 (23.8) | 60 (21.4) | 51 (27.3) | 0.151 |
CHF | 38 (8.1) | 14 (5.0) | 24 (12.8) | 0.003* |
Asthma | 66 (14.1) | 37 (13.2) | 29 (15.5) | 0.500 |
COPD | 30 (6.4) | 14 (5.0) | 16 (8.6) | 0.129 |
Bronchiectasis | 8 (1.7) | 1 (0.4) | 7 (3.7) | 0.008* |
OSA | 23 (4.9) | 12 (4.3) | 11 (5.9) | 0.514 |
CKD | 107 (23.0) | 59 (21.2) | 48 (25.7) | 0.263 |
CLD | 26 (5.6) | 18 (6.4) | 8 (4.3) | 0.411 |
Stroke | 59 (12.7) | 40 (14.3) | 19 (10.2) | 0.204 |
Cancer | 34 (12.1) | 24 (12.8) | 24 (12.8) | 0.886 |
Oral feeding on ICU discharge, n (%) | 173 (61.6) | 144 (77.0) | 0.001* | 2.11 (1.37–3.25) | 0.001* |
Pneumonia postextubation, n (%) | 45 (16.1) | 15 (8.0) | 0.011* | 0.41 (0.22–0.77) | 0.006* |
Reintubation, n (%) | 11 (3.9) | 13 (7.0) | 0.198 | 1.65 (0.71–3.87) | 0.246 |
Readmission to ICU, n (%) | 15 (5.3) | 19 (10.2) | 0.068 | 1.72 (0.83–3.55) | 0.145 |
ICU mortality, n (%) | 3 (1.1) | 4 (2.1) | 0.445 | 2.16 (0.47–9.91) | 0.320 |
Hospital mortality, n (%) | 19 (6.8) | 13 (7.0) | 1.000 | 0.87 (0.41–1.86) | 0.726 |
ICU LOS, days, median (IQR) | 6 (4–8) | 6 (4–9) | 0.170 | 1.07 (0.95–1.21) | 0.247 |
Hospital LOS, days, median (IQR) | 16 (9–27) | 14 (9–22) | 0.170 | 0.87 (0.75–1.00) | 0.058 |
When we restricted the analysis to patients extubated after >72 h of mechanical ventilation, we found that phase I and phase II patients were not significantly different (Table
3). Again, we found that proportionally more patients in phase II were allowed oral feeding at ICU discharge, with fewer patients developing pneumonia postextubation (see Additional file
5). Furthermore, patients in phase II, compared with those in phase I, had decreased hospital LOS. No other outcome differences were found (Table
4).
Table 3
Subgroup analysis of characteristics of patients extubated after >72 h of mechanical ventilation
Age, years | 58.5 ± 15.9 | 58.4 ± 17.4 | 58.7 ± 13.7 | 0.879 |
Female sex, n (%) | 59 (36.9) | 35 (37.2) | 24 (36.4) | 1.000 |
APACHE II score | 27.6 ± 8.9 | 26.7 ± 8.9 | 29.0 ± 8.7 | 0.111 |
Weight, kg | 67.0 ± 19.8 | 68.0 ± 21.4 | 65.7 ± 17.3 | 0.472 |
Ventilator days preextubation | 6 (4–8) | 6 (4–9) | 6 (4–8) | 0.807 |
Main diagnosis, n (%) | | | | 0.128 |
Pneumonia | 53 (33.1) | 27 (28.7) | 26 (39.4) | |
Nonpneumonia sepsis | 21 (13.1) | 10 (10.6) | 11 (16.7) |
COPD | 6 (3.8) | 2 (2.1) | 4 (6.1) |
Asthma | 9 (5.6) | 5 (5.3) | 4 (6.1) |
Fluid overload | 13 (8.1) | 8 (8.5) | 5 (7.6) |
Stroke | 6 (3.8) | 6 (6.4) | 0 (0.0) |
Seizure | 10 (6.3) | 8 (8.5) | 2 (3.0) |
Othera
| 42 (26.3) | 28 (29.8) | 14 (21.2) |
Comorbidity, n (%) | | | | |
Diabetes mellitus | 66 (41.5) | 34 (36.6) | 32 (48.5) | 0.145 |
Hypertension | 86 (54.1) | 51 (54.8) | 35 (53.0) | 0.872 |
IHD | 41 (25.8) | 25 (26.9) | 16 (24.2) | 0.854 |
CHF | 8 (5.0) | 3 (3.2) | 5 (7.6) | 0.278 |
Asthma | 23 (14.5) | 11 (11.8) | 12 (18.2) | 0.360 |
COPD | 8 (5.0) | 2 (2.2) | 6 (9.1) | 0.067 |
Bronchiectasis | 3 (1.9) | 0 (0.0) | 3 (4.6) | 0.070 |
OSA | 11 (6.9) | 7 (7.5) | 4 (6.1) | 1.000 |
CKD | 36 (22.6) | 18 (19.4) | 18 (27.3) | 0.254 |
CLD | 6 (3.8) | 4 (4.3) | 2 (3.0) | 1.000 |
Stroke | 19 (12.0) | 14 (15.1) | 5 (7.6) | 0.215 |
Cancer | 17 (10.7) | 9 (9.7) | 8 (12.1) | 0.615 |
Table 4
Subgroup analysis of outcomes for patients extubated after >72 h of mechanical ventilation
Oral feeding on ICU discharge, n (%) | 46 (48.9) | 44 (66.7) | 0.035* | 2.27 (1.13–4.54) | 0.021* |
Pneumonia postextubation, n (%) | 24 (25.5) | 5 (7.6) | 0.004* | 0.20 (0.07–0.60) | 0.004* |
Reintubation, n (%) | 6 (6.4) | 9 (13.6) | 0.168 | 1.80 (0.58–5.62) | 0.306 |
Readmission to ICU, n (%) | 7 (7.5) | 10 (15.2) | 0.128 | 1.79 (0.62–5.18) | 0.285 |
ICU mortality, n (%) | 1 (1.1) | 3 (4.6) | 0.307 | 4.62 (0.46–46.2) | 0.192 |
Hospital mortality, n (%) | 7 (7.5) | 6 (9.1) | 0.773 | 0.96 (0.29–3.18) | 0.944 |
ICU LOS, days, median (IQR) | 9 (7–13) | 9 (7–12) | 0.704 | 0.97 (0.83–1.13) | 0.672 |
Hospital LOS, days, median (IQR) | 24 (17–39) | 18 (12–30.5) | 0.010* | 0.75 (0.61–0.93) | 0.009* |
The overall safety signal favored NPS even when we analyzed only patients who were allowed oral feeding at the point of transfer from the ICU to the general floor. Among all patients, 19 of 173 in phase I and 10 of 144 patients in phase II developed pneumonia postextubation (11.0 % vs. 6.9 %, P = 0.244). In this analysis, when we considered only patients who were extubated after >72 h of mechanical ventilation, 9 of 46 patients in phase I and 3 of 44 patients in phase II developed pneumonia postextubation (19.6 % vs. 6.8 %, P = 0.120).
Regarding NPS use, 98.9 % of patients received at least one NPS screen. Among patients who passed the swallowing screen, 142 (99.3 %) of 143 did so by the second screen (Table
5). Of 44 patients who failed NPS screening, 38 (86.4 %) were discharged from the ICU before three screens could be completed. Overall correlation between the swallowing screen result and oral feeding status on ICU discharge was good at 92.0 % (see Additional file
6): 136 (95.1 %) of 143 patients who passed the screen were allowed oral feeding by their attending physicians, while 36 (81.8 %) of 44 patients who failed the screen were not allowed oral feeding. In other words, 8 % of nursing recommendations were overridden by a physician.
Table 5
Results of swallowing screening (N = 187 patients)
Day 1 result only | 187 | 115 (61.5) | 67 (35.8) | 5a (2.7) |
Day 2 result only | 73 | 28 (38.4) | 16 (21.9) | 29b (39.7) |
Day 3 result only | 44 | 1 (2.3) | 6 (13.6) | 37c (84.1) |
Overall result | 187 | 143d (76.5) | 42e (22.5) | 2f (1.1) |
Discussion
NPS for dysphagia, compared with no NPS, was associated with a 111 % increase in (the odds of) oral feeding at ICU discharge and a 59 % decrease in postextubation pneumonia. Among patients extubated after >72 h of mechanical ventilation, NPS for dysphagia, compared with no NPS, was associated with a 127 % increase in oral feeding at ICU discharge, an 80 % decrease in postextubation pneumonia, and a 25 % decrease in hospital LOS. We also found relatively few instances of attending physicians overriding the NPS protocol. These results suggest that NPS is safe, likely to be superior to usual care without NPS, and acceptable to medical teams.
Our study has validated the safety of adapting the Massey Bedside Swallowing Screen for use by nurses [
14]. The results are strengthened by NPS being associated with decreased rates of postextubation pneumonia, even though our nonrandomized study design resulted in NPS patients being more ill overall (as demonstrated by APACHE II score differences). The association of NPS with decreased postextubation pneumonia was also seen in patients who did
not have pneumonia as their main diagnosis (OR 0.42, 95 % CI 0.19–0.92,
P = 0.029), after adjustment for APACHE II score, congestive heart failure (as a comorbidity), and bronchiectasis (as a comorbidity). We postulate that NPS, compared with no NPS, could improve oral feeding rates because patients were allowed additional screening after failing the first one. We additionally postulate that NPS, compared with no NPS, could better identify patients at risk of aspiration. The smaller proportion of patients with stroke or seizures in phase II could not explain the association of NPS with decreased postextubation pneumonia. Of the 29 patients with stroke and/or seizure in phase I, 6 (20.7 %) developed postextubation pneumonia. In comparison, of 12 patients with stroke and/or seizure in phase II, 2 (16.7 %) developed postextubation pneumonia, which is a nonsignificant difference (
P = 1.000 by Fisher’s exact test). The imbalance of patients with stroke also did not significantly influence the incidence of postextubation dysphagia. We did an analysis excluding these patients, and the difference in oral feeding rates remained statistically significant (168 [66.1 %] of 254 in phase I vs. 135 [77.1 %] of 175 in phase II;
P = 0.017).
Separately, the association of NPS with decreased hospital LOS in patients who were extubated after prolonged mechanical ventilation could be due to fewer patients being fed nonorally at ICU discharge. Presumably, nonoral feeding may delay hospital discharge because more time is required for the transition to oral feeding or for training caregivers to administer nutrition nonorally (usually via nasogastric feeding in our setting).
Although it was not statistically significant, we cannot completely dismiss a possible association between NPS screening and reintubation, the latter accounting for the majority of ICU readmissions. However, reintubation events appear to be due to causes other than pneumonia. Of the 13 patients in phase II who were reintubated, only 2 were reintubated because of postextubation pneumonia, with the rest being reintubated because of new-onset nonpneumonia sepsis, fluid overload, or neurological deficits. Because the reasons for reintubation were varied in phase II, we feel that the slightly higher rate of reintubation was a chance finding.
Some limitations exist in our study. First, we conducted this research with medical ICU patients, and the results may not be generalizable to surgical or cardiothoracic ICU patients. The 36 % prevalence of dysphagia on day 1 postextubation may seem high, and few comparable data are available for short mechanical ventilation duration in medical ICU patients, though there is one paper showing a dysphagia rate of 17 % for intubation durations of 24–48 h among patients following cardiovascular surgery [
23]. Second, we did not verify the screening results of NPS using instrumental techniques such as flexible endoscopic evaluation of swallowing and videofluoroscopic swallow studies. In particular, NPS would not be able to detect silent aspiration (i.e., aspirating without a protective cough response), unlike instrumental techniques. However, the validity of NPS is now supported by improved pneumonia rates and hospitalization durations, while instrumental evaluation has not been shown to affect patient outcomes [
12,
24,
25]. Third, the timing of the dysphagia screening was set to start within the first day postextubation, though recent data show that the timing does not affect the result of swallowing assessment [
26]. Fourth, no patient was kept in the ICU for NPS screening per se, and some patients who failed NPS screening were discharged from the ICU before three screens could be completed. We thus do not know if our results could have been improved if all patients had been able to receive all three NPS screens. Fifth, we acknowledge that physician determination of swallowing function may not be done in some institutions, which limits the generalizability of the results. Sixth, our median ICU LOS was relatively short at 6 days. However, this duration was similar in phases I and II. In order not to miss any postextubation pneumonia, we counted all the cases of postextubation pneumonia that occurred in the ICU or on the general floor subsequently. We also do not think that patients would develop a clinically significant but unidentified postextubation pneumonia after leaving the ICU. Seventh, our study design was not the most scientifically robust; that is, we did a retrospective cohort study rather than a randomized trial. However, it would be difficult to perform blinding and to avoid nurses’ applying dysphagia screening skills within the same ICU, potentially biasing results toward the null.
Current methods of identifying at-risk patients for SLP referral appear to rest on the duration of prior intubation, and a common definition of
prolonged intubation uses a cutoff of 48 h [
4,
6,
27]. However, if we applied this cutoff to our patient population, 50 % of all patients would have needed to be referred, which would incur substantial SLP time and resources. In place of direct SLP referral, our data suggest that screening could be done by nurses first. In our experience, the bedside swallow screen takes only 5–10 minutes and is relatively easy to perform. Nonetheless, further research is needed to check whether NPS, versus no NPS, would result in more expedient and more appropriate referrals to SLPs for swallowing dysfunction (i.e., avoiding both overuse and underuse of SLP resources).
We hope that our study can stimulate further investigation into the development of pragmatic protocols for the assessment of swallowing impairment postextubation. Our protocol appeared to be safe for medical patients and should be validated in other patient cohorts, and using a randomized trial design. Importantly, nurses can be readily trained—as we have described—to implement the NPS protocol. Extension of NPS to the general floor could conceivably be done, either by trained general floor nurses or by more specialized ICU liaison nurses [
28]. Finally, cost-effectiveness studies could be done to quantify the resource savings of NPS, which could accrue from lower treatment costs due to less postextubation pneumonia and decreased hospital LOS.
Acknowledgements
We thank and acknowledge all contributions made by the doctors, nurses, and allied health staff in the medical intensive care unit of National University Hospital.
This work was performed within the National University Health System, Singapore.