Screening Accuracy for Aspiration Using Bedside Water Swallow Tests: A Systematic Review and Meta-Analysis

Abstract

Background: Hospitalizations for aspiration pneumonia have doubled among older adults. Using a bedside water swallow test (WST) to screen for swallowing-related aspiration can be efficient and cost-effective for preventing additional comorbidities and mortality. We evaluated screening accuracy of bedside WSTs used to identify patients at risk for dysphagia-associated aspiration.

Methods: Sixteen online databases, Google Scholar, and known content experts through May 2015 were searched. Only prospective studies with patients ≥ 18 years of age given WST screenings validated against nasoendoscopy or videofluoroscopy were included. Data extraction used dual masked extraction and quality assessment following Meta-analysis of Observational Studies in Epidemiology guidelines.

Results: Airway response (eg, coughing/choking) with or without voice changes (eg, wet/gurgly voice quality) was used to identify aspiration during three different bedside WSTs. Pooled estimates for single sip volumes (1-5 mL) were 71% sensitive (95% CI, 63%-78%) and 90% specific (95% CI, 86%-93%). Consecutive sips of 90 to 100 mL trials were 91% sensitive (95% CI, 89%-93%) and 53% specific (95% CI, 51%-55%). Trials of progressively increasing volumes of water were 86% sensitive (95% CI, 76%-93%) and 65% specific (95% CI, 57%-73%). Airway response with voice change improved overall accuracy in identifying aspiration.

Conclusions: Currently used bedside WSTs offer sufficient, although not ideal, utility in screening for aspiration. Consecutive sips with large volumes in patients who did not present with overt airway responses or voice changes appropriately ruled out risk of aspiration. Small volumes with single sips appropriately ruled in aspiration when clinical signs were present. Combining these bedside approaches may offer improved screening accuracy, but further research is warranted.

Keywords: aspiration; dysphagia; pneumonia; review; screening.

Figure 1

Study selection flowchart.

Figure 2

A-C, Hierarchical SROC curves for water swallow tests: single sips (A), consecutive sips (B), and progressive volumes (C). A, Single sips of 1 to 20 mL volumes. Volumes of 1 to 5 mL are combined with volumes of 6 to 20 mL (a); references listed are Chong et al (1), Kidd et al (2), Lim et al (3), McCullough et al (4), Momosaki et al (5), and Wakasugi et al (6). B, Consecutive sips with 90 to 100 mL volumes; references listed are DePippo et al (1), Garon et al (2), Mari et al (3), McCullough et al (4), McCullough et al (5), Patterson et al (6), Suiter and Leder (7), Suiter et al (8), Wu et al (9), and Zhou et al (10). C, Single sips to consecutive sips of 2 to 90 mL volumes; references listed are Daniels et al (1), Hassan and Aboloyoun (2), Hey et al (3), Nishiwaki et al (4), Rosen et al (5), Smithard et al (6), and Somasundaram et al (7). AUC = area under the curve; SENS = sensitivity; SPEC = specificity; SROC = summary receiver operating characteristic.

Figure 3

A and B, Deeks funnel plot asymmetry test for publication bias. Studies are represented by red dots with white numbers, with their respective regression lines. The two plots have different y-axis scales, both maximized for viewing purposes. A, Funnel plot includes all 22 studies in this review and demonstrates statistical significance (P = .011). References listed are Chong et al (1), Daniels et al (2), DePippo et al (3), Garon et al (4), Hassan and Aboloyoun (5), Hey et al (6), Kidd et al (7), Lim et al (8), Mari et al (9), McCullough et al (10), McCullough et al (11), Momosaki et al (12), Nishiwaki et al (13), Patterson et al (14), Rosen et al (15), Smithard et al (16), Somasundaram et al (17), Suiter and Leder (18), Suiter et al (19), Wakasugi et al (20), Wu et al (21), and Zhou et al (22). B, Funnel plot omits one study with 3,000 patients and is not statistically significant (P = .246). References listed are Chong et al (1), Daniels et al (2), DePippo et al (3), Garon et al (4), Hassan and Aboloyoun (5); Hey et al (6), Kidd et al (7), Lim et al (8), Mari et al (9), McCullough et al (10), McCullough et al (11), Momosaki et al (12), Nishiwaki et al (13), Patterson et al (14), Rosen et al (15), Smithard et al (16), Somasundaram et al (17), Suiter et al (18), Wakasugi et al (19), Wu et al (20), and Zhou et al (21). ESS = effective sample size.