Accuracy and clinical utility of comprehensive dysphagia screening assessments in acute stroke: a systematic review and meta-analysis

5 Introduction: Nurses and other non-specialists in dysphagia are often trained to screen swallowing 6 post-stroke. There are many basic tools that test water only, they are usually conservative and 7 patients that fail the test remain nil by mouth until a speech and language therapy assessment. 8 More comprehensive tests also allow non-specialists to recommend modified oral intake. Little is 9 known about the accuracy, clinical utility and cost effectiveness of these tests. 10 Methods: Following PRISMA guidelines, a systematic review was conducted to describe 11 comprehensive swallowing tests that are available for use in acute stroke by nurses or other non- 12 specialists in dysphagia. A meta-analysis was performed to evaluate accuracy and considered their 13 clinical utility. Searches and analyses, conducted by two reviewers, included MEDLINE, EMBASE, trial 14 registries, and grey literature up to December 2018. Validated studies were assessed for quality and 15 risk of bias using QUADAS-2. 16 Results: Twenty studies were included, describing five different tests, three of which had undergone 17 validation. The tests varied in content, recommendations and use. There was no test superior in 18 accuracy and clinical utility. Three studies validating the Gugging Swallow Screen provided sufficient 19 data for meta-analysis, demonstrating high sensitivity; 96% (95%CI 0.90-0.99) but low specificity, 20 65% (95%CI 0.47-0.79) in line with many water swallow tests. Results should be interpreted with 21 caution as study quality and applicability to the acute stroke population was poor. 22 Conclusions: There is no comprehensive nurse dysphagia assessment tool that has robustly 23 demonstrated good accuracy, clinical utility and cost effectiveness in acute stroke. Relevance to Clinical Practice: Nurses and other clinicians can develop competencies in screening 25 swallowing and assessing for safe oral intake in those with post stroke dysphagia. It is important to use a validated assessment tool that demonstrates good accuracy, clinical utility and cost effectiveness.  A summary and critique of the available tools for nurses and other clinicians to screen and 1 assess swallowing within the acute stroke pathway 2  An idea of how nurse-based screening and assessment of swallow post acute stroke might 3 impact on

Conclusions: There is no comprehensive nurse dysphagia assessment tool that has robustly 23 demonstrated good accuracy, clinical utility and cost effectiveness in acute stroke . 24 Relevance to Clinical Practice: Nurses and other clinicians can develop competencies in screening 25 swallowing and assessing for safe oral intake in those with post stroke dysphagia. It is important to 26 use a validated assessment tool that demonstrates good accuracy, clinical utility and cost 27 effectiveness. 28 What does this paper contribute to the wider global clinical community? 29  A description of how nurses and other clinicians are involved in screening and assessment of 30 swallowing after acute stroke 31 Post stroke dysphagia is common, affecting around 50% of acute stroke patients (Martino et al., 2005). 9 Early identification is key to reduce rates of stroke associated pneumonia and mortality (Bray et al., 10 2016;Yeh et al., 2011). Speech and Language Therapists (SLT) are, in many countries, considered to 11 be the specialists in assessment and management of dysphagia. However, swallow screening tools 12 such as water swallow tests are often used by non-specialists in dysphagia, including nurses, to identify 13 patients at risk of aspiration and refer patients for further assessment by SLT. There are a multitude 14 of screening tools described in the literature and systematic reviews have demonstrated that some of 15 the best tools have good sensitivity but often lower specificity (Schepp, Tirschwell, Miller, & 16 Longstreth, 2012). This translates to many patients unnecessarily remaining nil by mouth (NBM) for 17 prolonged periods, with or without nasogastric tube feeding, until they are assessed by a SLT, which 18 can have negative consequences (Langdon, Lee, & Binns, 2007;Langmore, Krisciunas, Miloro, Evans, 19 & Cheng, 2012). Water swallow tests have been criticised because swallowing water is not the same 20 as swallowing food (Marques, De Rosso, & Andre, 2008) and the tools have often been validated for 21 screening aspiration, one of the possible consequences of dysphagia, rather than for the presence of 22 dysphagia itself (Sasaki & Leder, 2014). Reduced efficiency or uncontrolled oral and pharyngeal transit 23 and clearance, impaired mastication and reduced sensation may result in other symptoms such as 24 choking and sub-optimal nutrition (Serra-Prat et al., 2012;Smithard et al., 1996). Aside from water 25 swallow tests, there are several more comprehensive swallowing tests that mean non-specialists can 26 screen for dysphagia and also assess various diet and fluid consistencies, so safe oral intake may be 27 commenced earlier. To In the case of  3   patients with dysphagia, clinical utility refers to how the tests improve the clinical outcomes of the  4 patients such as pneumonia rates and be more cost effective than other tools or pathways. 5

6
A systematic review was conducted to describe the comprehensive tools that are available for nurses 7 or other members of the multidisciplinary team (MDT) to screen swallowing and assess for safe oral 8 intake post stroke. The clinical utility of the tests is described, the results of a meta-analysis are 9 presented and the quality of the tools that had undergone validation is discussed. 10 Identified studies at different stages of the process were managed in folders on EndNote. 17

18
Inclusion criteria for the narrative review was broad as the number of published tools was estimated 19 to be small. Studies were included in the narrative review if they had sufficient information in 20 English or Spanish to establish that they described a comprehensive nursing or MDT assessment of 21 swallowing to screen for dysphagia in stroke patients. Comprehensive assessment was defined as a 22 screening test for dysphagia that included assessing more than one diet or fluid texture allowing for 23 recommendations of modified diet and fluids where appropriate. For the quantitative analysis, 24 studies were included if they gave data regarding the accuracy of the assessment tool such as 25 sensitivity and specificity. Studies were also included that reported the cost effectiveness or clinical 26 utility of a test. 27 Study selection 28 One reviewer (JB) searched the titles and abstracts and excluded non-relevant studies. Full text was 29 requested for relevant studies that could be included in a narrative review and, in the case of 30 validation studies, a quantitative review. Data extraction and assessment of quality were carried out 31 by the same reviewer (JB). Decisions for inclusion and exclusion, based on eligibility criteria were 32 discussed and agreed with a second reviewer (LE). The second reviewer (LE) also reviewed and 33 agreed the data extraction and quality assessments. Any disagreements were discussed with a third 1 reviewer (TE).  that to be classed as low concern for applicability to an acute stroke population, over 50% of 18 participants in the sample needed to be representative of acute stroke patients; defined as newly 19 admitted (less than one week post stroke), including all types and severities of stroke and who may 20 or may not have dysphagia. Overall quality was summarised using the GRADE guidelines 21 (Schunemann et al., 2008). 22 Identified tests 6 Five tests were identified and are summarised in Table 1. They are described as tests, screening tools 7 and assessments. They all met the criteria as a screening tool for dysphagia and included testing 8 different consistencies so that those who fail with water but can safely manage some oral intake can 9 be recommended modified diet and fluids whilst they wait for further assessment by SLT. The 2v/3t-P use a drop in oxygen saturations of >2% to detect silent aspiration. Other criteria varied 23 between tests, see Table 2 for details of the full criteria each test uses to determine aspiration or 24 dysphagia. 25  Table 4 shows the risk of bias and concern for 16 applicability of each test along with the level of quality. Reduced quality was due to concern or 17 uncertainty regarding risk of bias or applicability of index test, reference test, patient selection 18 methods or flow and timing. The study validating BESST (E. Boaden, 2011) demonstrated good study 19 design, accuracy and reliability but was scored as moderate quality due to lack of a gold standard 20 reference test and imprecise results with wide confidence intervals.. 21 Clinical utility and cost effectiveness 22 We did not find any studies evaluating the cost effectiveness of these tools over other tools or 23 pathways. However, several studies evaluated the effect of using these more comprehensive tests 24 on the clinical outcomes of patients. 25

Statistical Analysis
In a retrospective study (N=384) (Palli et al., 2017), the GUSS test was introduced into a stroke 26 service during out of hours periods where no SLTs were available to assess and manage swallowing. 27 This resulted in significantly reduced pneumonia rates from 11.6% before the introduction to 3.8% 28 after (p=0.004). Median length of hospital stay also decreased from 9 days to 8 days (p=0.033). 29 However, in another retrospective database study (N=1394) (Teuschl et al., 2018) there were no 30 differences in pneumonia rates between patients admitted with a stroke and assessed with GUSS 31 (5.0%) and those not assessed (5.5%). Due its methodological design, groups were not matched 32 therefore limited conclusions can be drawn. The 2v/3t-P test also resulted in a significant reduction 33 in pneumonia rates (6.2% before vs. 2.1% after, p = 0.05) in a prospective analysis of consecutively 1 admitted patients (N=418) to the stroke unit when it replaced a water swallow test (Cocho et al., 2 2017). A published clinical audit (N=61) described how acute patients were seen quicker and the 3 number of days they spent NBM dropped by over 30% following a fivefold increase in the number of 4 nurses trained to perform the DTNAx (M. Heritage, 2001). 5

6
Nurses and other non-specialists in dysphagia assess swallowing and recommend diet and fluid 7 intake in post stroke patients. Little is known about the content, accuracy or the way these 8 assessments are carried out. It is important that the tools used during these assessments have 9 undergone validation to ensure they are accurate in identifying dysphagia and that patients are 10 being recommended safe oral intake to prevent complications such as aspiration pneumonia, 11 choking or undernourishment. 12 We conducted a systematic review to identify and describe the available tools and compare their rates. There is also a jump between the diet and fluid consistencies tested to those recommended; 13 for example, a patient can be recommended IDDSI L1 or L2 fluids and L5 diet without having been 14 tested with any of these. In the same way, water swallow tests are also criticised for allowing normal 15 diet intake without assessment (Marques et al., 2008). Given it may not be any more accurate, safe 16 or clinically effective than water swallow tests, and training and administration time is greater, the 17 GUSS may be less cost effective. 18 The BESST was of moderate quality and had acceptable sensitivity and negative predictive value with 19 lower specificity to identifying dysphagia. However, the reference test used was a clinical bedside 20 assessment (CBA) which could be argued is not a gold standard assessment of swallowing, especially 21 because a validated CBA was not used. CBA have been shown to be less effective at describing 22 dysphagia and identifying aspiration (Splaingard, Hutchins, Sulton, & Chaudhuri, 1988) than gold 23 standard instrumental assessments and the author acknowledges this as a limitation with the BESST 24 validation. 25 The construct validity of the tests has not been reported. This pertains to how well a test is 26 constructed to identify dysphagia based on what is known about dysphagia. There are some 27 common characteristics across the tests that suggests good construct validity: all of the tests 28 evaluate liquids and solids; and they all have criteria for judging both the oral stage and pharyngeal 29 stages of swallowing. This includes specifics on identifying signs of aspiration such as cough and 30 voice change which have been shown to be the most reliable signs in water swallow tests (Brodsky 31 et al., 2016). Progressive volumes of thin fluids also increases accuracy of identifying aspiration 32 (Brodsky et al., 2016), most of the tests do this to some degree. However, there are limitations in 33 some of the tests that reduce their construct validity. Two of the tests to do not include food 34 textures that are part of regular diet (E. Boaden, 2011;Clave et al., 2008). Also, it has been 1 established that bedside assessments are limited in detecting silent aspiration (McCullough et al., 2 2005). Two of the tests have tried to address this by including pulse oximetry to measure a drop in 3 oxygen saturation, however more recently this measure has been found not to be reliable in 4 detecting aspiration(Wang, Chang, Chen, & Hsiao, 2005). These tests are designed to identify 5 dysphagia with aspiration being one aspect of that and silent aspirators may present with other signs 6 of dysphagia (Ramsey, Smithard, & Kalra, 2005). This may limit the potential of any bedside test to 7 attain high accuracy scores for identification of aspiration as to date there is no non-instrumental 8 test that has been found to identify aspiration reliably. 9 Both the VVST and the GUSS follow on from a preliminary screening component to identify those 10 who may be at risk of aspiration or dysphagia. The whole pathway (preliminary screen and test) has 11 not been validated with consecutively admitted acute stroke patients for either of these tests. 12 Perhaps this could be a more cost-effective pathway if both preliminary screening and then 13 dysphagia testing are shown to be acceptable in diagnostic accuracy in methodologically robust 14 studies. 15 Heritage 2003 argues that to manage dysphagia effectively SLTs need to share their skills, 16 responsibility and workload with nurses (Mary Heritage, 2003). Several publications suggested 17 screening tests were not designed to replace the role of the SLT (23). Instead they were meant as 18 easy-to-follow tools for those best placed (30) with the best skills (21) to identify patients with 19 dysphagia so that SLT resources could be better directed to assessment and management of those 20 most in need (20) . Training must therefore be essential if non-SLTs are assessing dysphagia. Whether training was 29 required to use the tests identified in this review appeared variable and the DTNAx is the only tool 30 that has described a training and competency assessment that meets the IDF's criteria. 31 This review only included studies published in in English or Spanish, therefore published and non-2 published studies in other languages describing assessment tools may have been missed. There are 3 likely to be many other nurse dysphagia assessments that have been developed by individual 4 services that have not been published or described in the literature and therefore have not been 5 included in this review. It is unlikely, however, that these in-house assessments have undergone 6 rigorous validation without publication. 7 Future directions 8 To make decisions around which test is superior in diagnostic accuracy, further validation using 9 robust study design is required. Information regarding clinical utility and cost effectiveness is also 10 desirable to use with accuracy data to determine which tools should be used as standard in routine 11 clinical practice. All the tests and gold standard comparators evaluate only small volumes of oral 12 intake in order to make appropriate recommendations; however, little is known about how the

19
There are several tools used by nurses and other non-specialists to screen for dysphagia and 20 recommend oral intake for acute stroke patients with mild to moderate dysphagia. Three have been 21 validated and show that they are good at identifying patients at risk of aspiration and dysphagia, but 22 often over diagnose, resulting in patients unnecessarily being kept NBM or on modified oral intake. 23 Overall, however, the quality of studies in this review was graded as poor or showing low 24 applicability for use by non-specialists to assess for dysphagia within the acute stroke setting. There 25 is limited variable quality evidence that these tests may reduce pneumonia, reduce length of time 26 patients are NBM and awaiting a swallowing assessment compared to no test. Further validation is 27 required with robust study design to discover the accuracy, clinical utility and cost effectiveness of 28 these tests so that they can be evaluated and compared. 29