Predicting feeding-tube dependence in patients following endotracheal extubation: a two-item swallowing screen

Identifying critically ill patients at risk for swallowing dysfunction is a priority after extubation. Postextubation dysphagia (PED) observed in previously dysphagia-naïve intensive care unit (ICU) patients is well recognized as a common sequela of intubation and can have serious consequences, including feeding-tube dependence [1], increased risk of pneumonia [2‐4], malnutrition [5], prolonged lengths of ICU and hospital stays [2, 6], and higher 90-day mortality [6]. Although many swallowing-screen tools are available, e.g., screens for dysphagia or aspiration, very few have been studied in ICU patients [7].

PED prevalence is, therefore, highly variable, ranging between 3 and 93% [1]. This wide range is likely exacerbated by heterogeneity in study design, including diagnostic and screening methods. In theory, PED screening comprises two approaches: screening for dysphagia or screening for risk of aspiration. At the screening level, distinguishing between dysphagia (physiologic swallowing dysfunction) and risk of aspiration (a result of physiologic swallowing dysfunction or dysphagia) is important. All three currently available PED screens (a bedside swallowing evaluation [administered only by speech therapists] [8], the Yale Swallow Protocol [9], and Postextubation Dysphagia Screening [10]) target aspiration risk and include a 3-oz water-swallow test.

However, these screening tools are not yet consistently applied clinically [11], resulting in a theory–practice gap. This underuse of screening tools is due to several reasons, including lack of guidelines to recommend their use, uncertainty about what to do with the screening results, and lack of time and specialized trained personnel to administer the screens [12‐15]. In many cases, nurses are responsible for administering these screens, i.e., introducing liquids or foods that result in clinical signs of aspiration such as coughing, choking, and wet vocal quality that usually indicate PED [16, 17]. Although the resulting clinical signs are seemingly straightforward, many nurses do not feel confident in judging them [12], especially when it has been suggested that 25% of patients aspirate silently [18]. On the other hand, nurses are concerned that aspiration triggered by water swallow tests can harm frail ICU patients even further [15].

The most efficient approach for consistently identifying cases of dysphagia may be a two-step approach in which a sensitive, user-friendly, and ultra-brief screen is administered, followed by a swallow assessment in those who “fail” the screen. Moreover, a screening tool for dysphagia that focuses more on physiologic swallow dysfunction is not yet available. Thus, the current study was designed to examine a swallowing screen that targets dysphagia and physiologic swallow dysfunction and includes tasks that are nurse-friendly and aspiration-risk free. We hypothesized that a two-item swallowing screen consisting of oral stereognosis (ability to identify shapes in the mouth only by touch, indicating oral functioning and/or cognitive ability to respond) and cough reflex (indicating airway defense) tests would predict patients with feeding-tube dependence persisting for 72 h following endotracheal extubation, a proxy for delayed return to oral feeding. By asking patients to identify three shapes (square, star, and round) of standardized acrylic resin as lollipop-like objects by oral manipulation (without seeing) and observing whether they had cough responses via citric acid inhalation using nebulizer masks, nurses could administer oral stereognosis and cough reflex tests at bedside.

As an exploratory outcome, predicted risk of aspiration was also cross-checked by the swallowing screen. Participants were offered the option of receiving the gold-standard fiberoptic endoscopic evaluation of swallowing (FEES) performed by a trained otolaryngologist. We hypothesized that participants who failed the screen were more likely to have penetration and/or aspiration during FEES. The screen and FEES were performed within 30 min of each other using standardized protocols [19].

Of the 145 enrolled participants, 123 completed both the screening and oral intake follow-ups, with 22 excluded due to reintubation (n = 2), death (n = 1), and inability to complete screens due to delirium status (n = 19). For subgroup analysis, 38 participants completed FEES within 30 min of the screen. Participants who opted to receive FEES (n = 38) and those who declined (n = 85) were not different in terms of age (P = 0.58), gender (P = 0.85), CCI score (P = 0.15), APACHE II (P = 0.67), and length of intubation (P = 0.43). For details of the study flow, see Fig. 1.

Participants’ demographic and clinical characteristics (N = 123) are in Table 1. Participants’ median age was 67 years (IQR = 54–74), with 40.7% being female. The majority of participants’ endotracheal tube sizes were 7.0 Fr. and 7.5 Fr., with a median 6 days (IQR = 4.2–10.8) of intubation. The median length of ICU stay was 10 days (IQR = 7–16). After extubation, 10.6% had increased respiratory rate (≥ 25 breaths/min), 49.6% had high suctioning demand requiring more than six suctions per day, and all had oxygen demand (82.1% received simple mask therapy, and 10.6% received nonrebreathing mask or NPPV therapy).

The most important finding of this study was that our swallowing screen, consisting of oral stereognosis and cough reflex tests, were sufficiently sensitive to identify postextubation patients with feeding-tube dependence that persisted 72 h and beyond, despite a 0.35 specificity. Although an overall accuracy of 0.68 leaves room for improvement and false positives may bring diagnostic and care burden, casting a wider net in the first stage of case-finding may enable nurses and physicians to triage patients needing further assessment before and during their recovery to oral feeding.

Three study findings warrant emphasis. First, the sufficient accuracy of our screen, supports, at least partly, the pathophysiological mechanism of impaired oral stereognostic ability and cough reflex affecting oral feeding after extubation. Although oral stereognosis has not been well-studied, it was impaired after extubation when compared to an aged-matched, non-ICU patient group [21]. In this study, we established the temporal relation between impaired oral stereognosis and feeding-tube dependence by showing that the former predicted the latter. Studies are warranted to investigate the roles of oral stereognosis in feeding-tube dependence, both in terms of test properties and magnitude of impact, to refine the screen items.

For the cough reflex, having two reflexive cough responses to 0.4% citric acid alone is highly specific (0.83), consistent with a previous study [23], to rule out feeding-tube dependence in this critically ill group. Sensitivity, however, was only 0.30 (more false negatives); thus, using this item alone is not acceptable. For improvement, ceiling effects found in our sensitivity analysis provide a direction (i.e., increasing cough counts to pass or adding cough strength to raise the ceiling). Moreover, the cough reflex test is yet to be standardized [26], and further studies could consider standardizing the protocols while titrating the passing threshold to enhance its testing performance.

Third, subgroup analysis of 38 participants receiving FEES indicated that this screen performed better in predicting feeding-tube dependence than penetration and/or aspiration during FEES (0.74 vs. 0.45 accuracy, respectively). This difference at least partially supports our aim to screen for dysphagia, rather than aspiration. The three currently available PED screens all focus on aspiration and utilize a 3-oz water swallowing test [7], whereas growing data suggest that not all nurses are confident in performing these screens, correctly identifying signs of aspiration (i.e., wet voice). One recent study revealed that 81% of nurses working in the cardiac ICU reported having no formal training in swallow screening [12]. Nearly half, 48%, of ICU nurses reported feeling “somewhat” or “not” confident in performing screens [12]. Our two-item swallowing screen provides important alternatives. Indeed, in the era of the Covid-19 pandemic, if needed, nurses could remotely test for oral stereognosis to avoid unnecessary contact; for just this item alone, accuracy was 0.65. To increase screen accuracy to 0.68, cough reflex could be added but adequate personal protection equipment and physical distancing are required to prevent disease transmission. For patients who fail the screen, a further swallow assessment is indicated to enhance the return of their oral feeding and maximize patient outcomes.

This study had important limitations. First, the swallowing screen is a two-item tool designed to be administered by nurses at the ICU bedside with minimal respondent burden. The capacity to enhance diagnostic accuracy is limited, for example by modifying criteria or cutoffs, as we decided beforehand that participants would pass the screen if they identified all three shapes and had two cough responses. Second, although our primary endpoint, feeding-tube dependence persisting over 72 h after extubation, often reflects physiologic swallowing dysfunction, it may also reflect conditions other than swallowing (i.e., gastrointestinal disease, nutritional conditions, or even practice patterns). As such, whether our screen has added value in other ICU settings remains to be tested. Third, our sample was not large enough (especially with only 38 participants receiving FEES) to allow more complex confirmatory analyses. Our results, therefore, need to be interpreted cautiously and replicated in other studies with larger samples. Fourth, our study was limited by enrolling participants from a single institution with an enrollment rate of 65.9% (145 of 220 invited patients) and completion rate of 84.8% (N = 123 of 145 enrolled; 19 unable to complete the screen, 1 death and 2 reintubations). While all eligible patients were offered enrollment, selection bias remains a limitation that will need to be addressed by replicating the findings in future studies.

Postextubation dysphagia observed in previously dysphagia-naïve ICU patients is associated with poor outcomes. An essential beginning to their care is adequate screening for dysphagia. Our ultra-brief swallowing screen, comprising oral stereognosis and cough reflex tests, is sufficiently sensitive to identify patients at high risk for feeding-tube dependence persisting over 72 h after extubation. Once identified, further assessment and care are indicated to ensure the prompt return of patients’ oral feeding.