Sensitivity and Specificity of the Geriatric Bedside Swallowing Screen (GEBS): A Diagnostic Accuracy Pilot Study / Sensitivität und Spezifität des Geriatric Bedside Swallowing Screen (GEBS): Eine Diagnostische Pilotstudie

Participants were consecutively recruited from the Department of Geriatric Medicine, Christian Doppler University Hospital (CDK), Salzburg, Austria. To be eligible, individuals needed to be aged 75 or older and able to give informed consent. Exclusion criteria included a history of severe stroke as measured by the National Institutes of Health Stroke Scale (NIHSS). The NIHSS defines the severity of stroke by points from 0–42, where higher scores indicate more severe strokes. Scores of under 5 points indicate a minor stroke, 5–15 points a moderate, 16–20 moderate to severe, and above 20 a severe stroke. Thus, we decided on an NIHSS > 20 as cut-off for our inclusion criteria (Brott et al., 1989).

We also excluded patients with moderate to advanced cognitive impairment as calculated by the Mini Mental State Examination (MMSE), a 30-point scale with lower points indicating more advanced impairments. 0–10 points indicate severe dementia, 10–19 moderate dementia, 20–26 mild dementia or mild cognitive impairment, and points above 26 indicate no cognitive impairment. To ensure participants would be able to give informed consent and partake in the testing procedures, we decided on a cut-off value of MMSE ≤ 20 as an exclusion criterion (Folstein et al., 1975).

Lastly, we excluded patients with a tracheostoma with or without ventilatory dependence (Heidler, 2019). Newly admitted patients were screened for eligibility, and those meeting criteria provided informed consent for their participation.

The index test, GEBS, was based on the screening protocol already in use at CDK. This protocol had been reported as convenient and easy to use by hospital staff, making it a feasible starting point. It was updated to include predictors of dysphagia that were considered risk factors (table 1) and structured into a three-step protocol. The risk factors could be assessed by patient reports, chart review and observation.

After assessing risk factors and patient vigilance, a water swallow test was performed. The patient swallowed 10 sips of water from a teaspoon, with pauses for voice checks, and one sip from a cup. Clinical signs of aspiration, such as coughing, wet voice, or breathing difficulties, resulted in stopping the test, failing the screening, and referral to a speech-language pathologist (SLP). If only one risk factor was present and no clinical signs of aspiration occurred, the patient had passed the screening. If more than one risk factor was present or the water swallow test showed signs of aspiration even in the absence of any risk factors, the screening was considered to have been failed.

If the water test indicated dysphagia with multiple risk factors and no SLP was available (e.g., due to staffing limits or admission timing), a third step would be applied. This involved the patient swallowing a pureed bolus (IDDSI-4) after ensuring they were awake, upright, and had saliva control. Teaspoons of puree were administered with pauses for voice checks, and the test was stopped if there were aspiration signs or severe drooling. Dysphagia indicators include coughing, wet voice, and oral residues. To obtain full results during the study, the full protocol was administered to each participant and not stopped after the water swallow test. The three-step protocol was intended for later clinical use to make the tool more feasible and quicker to administer.

A score of 5 points allowed for oral nutrition with IDDSI Level 4 puree and Level 2 thickened liquids while awaiting SLP assessment. Scores below 5 indicated high dysphagia risk and resulted in nil-by-mouth instructions.

Following Studies of Diagnostic Accuracy (STARD) guidelines, a gold-standard reference test is essential for validating a new test (Bossuyt et al., 2015). Dysphagia can be assessed using FEES or VFSS (videofluoroscopic swallowing study) (Wirth et al., 2016). FEES, being portable, repeatable, and not time-limited, is preferred over VFSS, which requires high-resolution X-rays, transportable patients, and their compliance. FEES is considered the gold standard for dysphagia assessment in geriatric patients due to its high diagnostic accuracy (Warnecke et al., 2009), especially with standardized protocols like the Penetration-Aspiration Scale (PAS) (Rosenbek et al., 1996).

In this study, FEES was conducted within six days of the index test by two blinded, experienced physicians. This rather long time span of six days was set to enable the participation of long-term care institutions, where FEES is not available daily but rather only once a week, and the inhabitants usually living there full-time without acute events that would require hospitalization, and thus a change in overall status was less likely to occur. In fact, the average time span was 1.2 days, with FEES being administered the following day in the majority of cases (21 times), within two days (five times) or on the same day (three times) and only once with six days in between.

Colored boluses of puree, liquids and solids were administered in this order with increasing volumes, and results were recorded on the PAS. A structured examination of oral, pharyngeal, and laryngeal sensitivity and motor function was also performed. During the endoscopic assessment, both the PAS and the Bolus Observation and Dysphagia Severity (BODS) score were documented, routinely used to evaluate swallowing function, saliva management, and oral intake potential (Starrost et al., 2012). Leaking and residues were recorded in the structured assessment, but without a standardized rating score. All participants underwent testing using purees and liquids. Testing was stopped after purees and liquids in one case because of the patient’s refusal to continue. Testing was not discontinued upon the first sign of aspiration so that researchers could determine the possibility of oral intake with different strategies — e.g., changing the bolus size or integrating swallow maneuvers.

To analyze the diagnostic accuracy of the GEBS in comparison to the results of the instrumental swallowing assessment, a 2 × 2 table was used. To count true positives, the results of the PAS and BODS, along with clinical observations, were merged into two categories: those under the presence of or absence of dysphagia. PAS scores of 1–3 were considered physiological, as penetration without actual airway compromise has been reported to occur more frequently in older adults, while PAS scores of 4 or higher in any of the tested consistencies were considered to indicate dysphagia (Butler et al., 2011; Daggett et al., 2006; Jardine et al., 2021). BODS-1 scores of higher than 1, and BODS-2 scores of higher than 2, were considered pathological, with two exceptions where BODS-2 was scored 3 but no clinically relevant dysphagia was reported during the FEES.

In cases of doubt, the clinical observation results were considered in order to correctly classify results as indicating presence or absence of dysphagia. The GEBS results did not need to be pooled, as the results indicated a clear pass or fail. To evaluate whether the full protocol (including the pureed bolus test) should always be administered, two separate analyses were conducted. Firstly, the results of the GEBS according to the first two steps (risk-factor assessment and water-swallow test) were analyzed. In a second step, the results of the full protocol were analyzed again with the same calculations.

Sensitivity and specificity, as well as positive and negative likelihood, were first calculated in Microsoft Excel with the use of field functions and classic formulas (e.g. Sensitivity = True Positives ÷ True Positives + False Negatives) to get a quick impression of the results. As all participants successfully completed both tests; there was no missing data to consider. Statistical analysis of individual risk factors and their impact on dysphagia prediction, as well as confidence intervals, was performed using IBM SPSS Statistics.

To estimate a sample size, other dysphagia validation studies were taken into account. The aforementioned established tools reported recruitment rates between 50 (GUSS) and 311 (TOR-BSST) (Martino et al., 2009; Trapl et al., 2007). Since this study was planned as a pilot study, feasibility was a major factor when planning the sample size. To account for potential drop-outs due to withdrawal of consent, rapid changes in health status, or unexpected transfers, a total sample size of 100 patients was established as the recruitment goal. To form a broader picture of different geriatric patients, recruitment was planned in multiple settings: acute geriatric wards and a long-term care institution (LTC).

A total of 225 patients were admitted to the CDK, of which 30 participants (mean age 79.13 years, SD ± 5.7) fulfilled inclusion requirements and gave informed consent to participate. They were subsequently recruited between November 2022 and June 2024 at CDK. The most common reasons for non-eligibility were an age of under 75 and a history of severe stroke. The initially planned center, Haus der Barmherzigkeit in Vienna, could not recruit due to staff shortages for FEES and lockdown regulations limiting non-essential medical contact. The University Hospital Sankt Pölten also withdrew from the study due to health issues among key staff. No participants experienced adverse side effects from the tests, and there were no dropouts. Participant flow is shown in figure 1.

Figure 1:

Flow diagram of participants, f: female; m: male

All participants completed both tests. Out of the 30, 12 were diagnosed with dysphagia. The highest PAS scores were for liquids, with six patients scoring 6 or higher, indicating severe dysphagia for liquids. No severe dysphagia was observed with puree (PAS ≤ 2), while two patients had moderate dysphagia with solids (PAS = 3). The geriatric participants — all over 74 — had multiple comorbidities, with a median Charlson Comorbidity Index (CCI) of 2.5 (SD ±1.7) (Quan et al., 2011). This index evaluates the one-year mortality rate by cross-calculating the accumulated burden of the diseases. For geriatric patients, a score of < 5 indicates a 1-year mortality rate of 85%; a 3–4 point score indicates 52%; a 1–2 point score indicates 26%; and 0 points indicate 12%. Most lived at home with light dependency for activities of daily living (ADL) before admission, primarily due to neurological events (e.g., stroke) or exacerbated health conditions. Four had been admitted for acute internal events, and two for fall complications. Demographics are in Table 2 and test results of PAS and BODS are in Table 3.

Both tests were conducted within seven days, with investigators blinded to the other test’s results.

The GEBS demonstrated promising validity, with 58% sensitivity (95% CI: 0.28–0.85) and 83% specificity (95% CI: 0.59–0.96) using only the first two steps: risk factor assessment and the water swallow test. Tables 3 and 4 illustrate the distribution of cases in a 2×2 table for the combination of the risk factor assessment and water swallow test (GEBS WST) (Table 3) as well as the full GEBS protocol (Table 4).

The positive predictive value (PPV) and negative predictive value (NPV) were 70% (95% CI: 0.43–0.88) and 75% (95% CI: 0.60–0.86) respectively, while positive likelihood ratio (LR+) was 3.5 and negative likelihood ratio (LR−) was 0.5.

When the third step of pureed bolus testing was included and the full GEBS administered, the results showed a higher sensitivity of 67% (95% CI: 0.35–0.90) while maintaining a specificity of 83% (95% CI: 0.59–0.96), with a PPV of 73% (95% CI: 0.47–0.89), and an NPV of 79% (95% CI: 0.62–0.90), as well as LR+ of 4 and LR− of 0.4 (Table 5).

An overview of the results is given in Table 5.

The presence of risk factors and their significance in correlation to dysphagia were calculated using a chi-square (X²) test as well as Cramér’s V. Neurological or neuropsychological disease, recent pneumonia, and reported or observed dysphagia symptoms — such as drooling, coughing, or voice change — correlated significantly with diagnosed dysphagia assessed via FEES. In contrast, the presence of sarcopenia, unintentional weight loss and poor oral health did not show significant correlation (table 6).

The GEBS combines observational and self-reported elements — similar to tools like the V-VST, GUSS and TOR-BSST (Clavé et al., 2008; Martino et al., 2009; Trapl et al., 2007) — but this pilot study is limited by a small sample size (n = 30). This leads to very large confidence intervals, making it hard to judge the test’s actual performance. Recruitment delays due to COVID-19, and the vulnerable nature of the geriatric population, hindered statistical reliability and generalizability. The long enrollment period further contributed to possible bias regarding a consecutive sample. Though the test results of the GEBS were blinded, there is the possibility of the investigators administering the FEES not being completely blinded due to knowing from their charts or rounds that certain participants were already considered to have dysphagia or have been put on a texture-modified diet. A randomized order of test application might have also reduced bias, but for administrative reasons, this was not possible with the necessary resources for obtaining informed consent, administering the GEBS and conducting the FEES.

Another limitation is the fact that the GEBS was not only administered by nurses and physicians but also by SLPs, who might unintentionally have assessed more than simply the findings of the GEBS.

Further studies in diverse populations are needed to evaluate the GEBS’s full potential, including reliability testing, which was unfeasible here. Additionally, the tool was developed and validated in German, with the German version available in the supplementary material.