Introduction

Oropharyngeal dysphagia (OD) is prevalent in various neurological etiologies (e.g., stroke, traumatic brain injury, Parkinson’s disease), or as a consequence of respiratory disease or structural changes (e.g., head and neck cancer, spinal cord injury) [1,2,3,4]. OD has also been acknowledged as a geriatric syndrome [5]. Physiological changes that occur in swallowing function in older healthy adults (presbyphagia) may be worsened by age-related decline in muscle mass and strength (sarcopenia), thus exacerbating OD caused by disease (e.g., stroke, Parkinson’s disease) common to the aging population [6]. Older adults who have changes in swallowing function are often unaware that they have OD or that it is treatable [7, 8].

Serious medical consequences of OD such as malnutrition, dehydration, pneumonia, and the need for enteral nutrition contribute to increased institutionalization; increased length of hospital stay, increased hospital re-admissions, and likelihood of being discharged to rehabilitation services and nursing homes instead of home [9, 10].

The complexity of OD requires a multidimensional approach to diagnosis in order to plan individually tailored intervention [5]. Identification of OD is completed by either screening, clinical non-instrumental assessment, or instrumental assessment such as videofluoroscopic swallowing study (VFSS) [11] or fiberoptic endoscopic evaluation of swallowing (FEES) [12]. A screening is a test given to distinguish between persons at risk from those that are not at risk of OD and helps to determine the need for further clinical non-instrumental or instrumental assessment [13]. In contrast, a non-instrumental assessment of OD is a more comprehensive evaluation and may include a medical history, assessment of orofacial sensorimotor and laryngeal function, and assessment of swallowing function using foods and liquids in various volumes and consistencies, determining the phase(s) of swallowing process that are deficient. These findings aid in determining dysphagia severity, possible treatment strategies and support the need for instrumental assessment [13]. Instrumental assessments, VFSS and FEES, are noted as being preferred instrumental assessments for OD in the literature although no international consensus exists on which visuo-perceptual measure to use when evaluating the radiological or endoscopic recordings of swallowing [14]. The use of patient-reported outcome measures (PROMs) provides a subjective assessment of the patients’ perspectives on the burden of living with OD [13].

Prevalence research is important as it reflects the burden of a disease or condition in a population at a particular time period. A systematic review by Kertscher and colleagues found that prevalence data on oropharyngeal dysphagia for the general population varied between 2.3 and 16% [15]. However, much of the available research reviewing the prevalence of OD is targeted toward populations according to age or diagnosis. A systematic review and meta-analysis by Madhavan et al. revealed a prevalence of OD ranging from 5 to 72% in the community-dwelling elderly population [16]. Takizawa and colleagues reviewed a broad spectrum of disorders susceptible for OD and found a prevalence of 8–45% in relation to stroke, 11–60% in Parkinson’s disease, and 27–30% in traumatic brain injury [1]. No studies were identified for prevalence of OD associated with Alzheimer’s disease [1].

Variations in reported OD prevalence can be attributed to methodological differences such as clinical setting, how dysphagia is defined, the study population, choice of measurement tools used, and time of assessment [1]. The severity of OD varies within the course of an illness or disease and can be defined as acute or chronic, and progressive or non-progressive [17]. Thus, the timing of assessment in relation to type and onset of illness or disease can impact the accuracy of OD prevalence data. Furthermore, healthcare professionals’ knowledge of OD and their routines for identification of OD have been found to be inadequate [7, 18]. Thus, OD is under-diagnosed and under-reported [1].

Existing literature on OD prevalence in different healthcare settings is mainly found in individual prevalence studies. There are systematic reviews on OD prevalence in adults with different diagnosis [1, 19, 20] and the community-dwelling elderly [16], but there is currently no overview of the prevalence of OD in the hospital, rehabilitation, nursing home, or palliative healthcare settings. Evidence about the scope of OD in adults in different healthcare settings will provide insight on the impact of OD in different settings. This evidence will increase healthcare professionals’ awareness of the likelihood of patients/residents presenting with OD and aid policy makers when assessing the allocation of interdisciplinary resources to meet the needs of persons with OD. The aim of this systematic review is thus to determine prevalence estimates of OD in adults admitted to hospitals, rehabilitation, nursing homes, and palliative care facilities using meta-analyses.

Materials and Methods

The protocol for this systematic review and meta-analysis was registered with the international prospective register of systematic reviews (PROSPERO; registration number CRD42019134585) in August 2019. The methodology and reporting of results is based on the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA), which aims to ensure complete and transparent reporting of systematic reviews and meta-analyses [21].

Eligibility Criteria

Studies were considered eligible for inclusion for this systematic review if they (1) reported on persons with oropharyngeal dysphagia (OD), (2) provided data on prevalence, frequency, or incidence, (3) described adult populations (≥ 18 years of age) ,and, (4) referred to healthcare settings (including hospital, rehabilitation, nursing home, or palliative care facilities).

Study inclusion was not limited by study design; however, only peer-reviewed original studies in English were included, and thus conference abstracts, review articles, case reports, student dissertations, and editorials were excluded. In order to minimize selection bias, prevalence estimates based on preselected groups (e.g., selected numbers of adults who already failed any previous form of OD screening or adults selected by specific comorbidity or surgical procedure) were excluded. In order to improve the level of precision in the prevalence estimates, studies with sample sizes below 30 participants were excluded.

Search Strategy and Study Selection

A literature search was completed on March 30, 2021, in two electronic databases: Embase and PubMed. Terms related to dysphagia, clinical settings (hospital, rehabilitation, nursing home, palliative care), and prevalence were entered into each electronic database to retrieve all relevant subject headings (i.e., MeSH and Thesaurus terms). In addition, free text terms were included in combination with field searches (i.e., Title/Abstract) and truncation (i.e., wildcards). Subject headings and free text terms were combined using Boolean operators to either expand searches (i.e., Boolean operator “OR”) or to restrict and combine searches (i.e., Boolean operator “AND”). All publication dates up to the search date were included. Search strategies are presented in Table 1. Two independent reviewers completed a structured assessment for eligibility. Prior to independently screening all titles and abstracts, the reviewers completed two training sessions, with a sample of 100 abstracts, in order to establish a consensus on the interpretation of the eligibility criteria. Both reviewers independently screened all titles and abstracts. The same two independent reviewers completed a full-text review of selected articles for assessment of eligibility. The reviewers also searched the references from the included articles to identify additional eligible articles. Any discrepancies of inclusion between reviewers were settled by consensus throughout the review process. When in doubt, the two main reviewers conferred with a third reviewer whom is experienced in PRISMA methodology.

Table 1 Search strategies per literature database
Full size table

Methodological Quality Assessment

A quality appraisal of included studies was completed through consensus by the two independent reviewers, using the critical appraisal tool for cross-sectional studies, AXIS [22]. The AXIS appraisal tool is comprised of 20 questions that address common methodological issues and are arranged in an order that follow the general outline of a cross-sectional paper. Examples of issues addressed in the AXIS include clearly stated study aims, study design, sample size and selection, outcome variables measured, statistical analysis, non-response bias, reporting of results, justified discussion and conclusion, limitations, and ethics. AXIS questions that were answered “yes” were scored as “1” reflecting good methodological quality and, “no” was scored as “0” reflecting lower methodological quality. Two AXIS questions were formulated such that a positive answer “yes” would reflect negatively on methodological quality. Therefore, the scoring of these two questions was reversed in order to provide a uniform scoring method. The maximum total AXIS score possible was 20 being the best possible score for good methodological quality; however, not all items were applicable to every study. As such, total scores were converted into percentage scores: total score divided by the maximum score possible and multiplied by a hundred [23]. The level of evidence of the included studies was rated using the National Health and Medical Research Council (NHMRC) Evidence Hierarchy [24].

Data Extraction

One reviewer extracted outcome data to an extraction table. Data were extracted regarding study setting and country, study population, definitions of terminology related to OD, OD screening and assessment methods, and OD prevalence data. A second reviewer performed a quality check of the extracted data. If necessary, authors of the included articles were contacted for clarification of terminology with regard to defining the setting [25,26,27] or to ask for access to raw data, when prevalence was described as a combination of different healthcare settings [28].

Data Synthesis and Risk of Bias

Data extraction and study characteristics were retrieved using comprehensive data extraction forms. Assessment of the risk of bias was completed for each individual study using the AXIS critical appraisal tool [22]. Abstract selection, final study selection and quality assessments were the result of consensus-based ratings of two reviewers. Discrepancies were resolved through consensus with a third reviewer. Bias is not expected as the reviewers are not affiliated with any of the authors of the included studies.

Meta-analysis

For the purpose of reducing heterogeneity for the meta-analysis and concerns regarding data completeness, quality, validity, reliability, and possible selection or recall bias, studies that collected prevalence data from notes in patient medical records, national databases, surveys, registries [29], or a dichotomous yes/no question to the patient or caregiver on the presence of a swallowing problem/difficulties [30] were not included in the meta-analysis. In addition, studies were excluded from the meta-analysis if it was not possible to compute proportional data results for screening or clinical assessment type and/or healthcare setting separately.

Data for subsampling were extracted from the included studies to measure the overall within- and between-group prevalence for different clinical settings: hospital, rehabilitation, and nursing home, according to the authors’ definition of the setting for each article. Overall within-group prevalence accounted for all studies with data for hospitals, rehabilitation, and nursing homes. Overall between-group prevalence was performed to determine confounding variables as a function of type of assessment method (e.g., screening, clinical non-instrumental assessment, instrumental assessment), diagnosis group, and type of hospital ward for each setting when applicable.

Meta-analysis of the prevalence of OD was completed using Comprehensive Meta-Analysis, Version 3.0 [31], providing estimates of pooled prevalence and forest plots. Due to the heterogeneity of the included studies, a random-effects model was used for summary statistics. Heterogeneity was estimated using the Q statistic to determine the spread of effect sizes about the mean and I2 to estimate the ratio of true variance to total variance. I2-values of less than 50%, 50% to 74%, and higher than 75% denote low, moderate, and high heterogeneity, respectively [32]. The classic fail-safe N test was used to assess publication bias. This test provides an estimate of the number of additional studies, with non-significant results, that would be necessary to add to the analysis in order to nullify the measured effect (N). A small N raises concern about the meta-analysis being compromised by publication bias; conversely, a large number suggests that it is unlikely that the meta-analysis is compromised by publication bias.

Results

Study Selection

The literature search and study selection results are illustrated in the PRISMA flow diagram (Fig. 1). The search resulted in 2192 records. After duplicates were removed, screening of the remaining 1956 records (abstracts and titles) resulted in 256 full-text articles assessed for eligibility. Forty articles were deemed eligible and an additional four articles were retrieved through reviewing of reference lists, resulting in inclusion of a total of forty-four articles.

Fig. 1
figure 1

Flow diagram of the review process according to PRISMA [21]

Full size image

Synthesis of Methodological Quality

The methodological quality of the included studies was assessed by the AXIS critical appraisal tool. Higher AXIS scores indicate higher methodological quality of the studies being appraised, whereas lower scores identify methodological weaknesses that may result in poor reliability and validity of study results. The mean total score and percentages for all studies was 15.3 (SD 2.2; range 10–19) and 77% (SD 11; range 50–95), respectively. Two of the 44 studies scored 50% or lower [33, 34], 18/44 studies scored above 50% and below or equal to 75% [25,26,27, 35,36,37,38,39,40,41,42,43,44,45,46,47,48,49], and 24/44 studies scored above 75% [28, 50,51,52,53,54,55,56,57,58,59,60,61,62,63,64,65,66,67,68,69,70,71,72]. AXIS scores and percentages can be found in Table 2 and AXIS scores for each question in the Online Resource 1.

Table 2 Characteristics of studies included in the systematic review
Full size table

Study Characteristics

All extracted data are summarized in Table 2. Data were recorded under eight subheadings: author, journal, study design, AXIS score, study setting, country and time period, underlying medical diagnosis, inclusion/exclusion criteria and time of assessment, sample characteristics (sample size, gender, age in years), description of OD terminology used in the study, screening/assessment tools used for prevalence calculation and what professions completed the testing, and OD prevalence data.

The included studies were published from 1986 to 2020, the majority (36/45) after 2010. The studies originated from 23 countries; 27 from Europe [25, 27, 28, 39, 41,42,43,44,45,46,47, 51, 52, 55, 57, 58, 61,62,63,64,65,66, 68,69,70,71,72], six from North America [34, 40, 48, 56, 59, 60], three from Oceania [49, 53, 54], four from Asia [26, 33, 37, 67], two from South America [36, 50], and two from Africa [35, 38]. Twenty-nine studies were prospective study designs: ten cohort and 19 cross-sectional designs. Fourteen studies were retrospective: eight cohort and six cross-sectional. One study included both retrospective and prospective cross-sectional data.

There were 32 articles providing estimates for OD prevalence from hospitals [26, 28, 34,35,36, 38,39,40, 42, 43, 45,46,47,48,49,50,51,52,53,54,55,56,57, 60,61,62,63,64,65,66, 68, 71], four from rehabilitation [26, 33, 58, 59], and 12 from nursing home settings [25,26,27,28, 37, 41, 44, 49, 67, 69, 70, 72]. Two articles provided OD prevalence estimates from both hospital and nursing home settings [28, 49] and one article reported OD prevalence estimates from all three settings: hospital, rehabilitation, and nursing home [26]. There were no studies that met the inclusion criteria from palliative care.

Healthcare Setting Description

The description of the hospital settings in this systematic review included general, tertiary, teaching, and regional hospitals. Hospital wards such as acute care, medical, neurological, and geriatric were used for participant recruitment. The rehabilitation settings included an inpatient rehabilitation clinic, rehabilitation facilities, and hospital (neuro)rehabilitation services/units. Settings that were classified as nursing homes included short-term/intermediate care, residential care, long-term care, and assisted living [73].

Participants

An estimated total of 49,436 participants were included in the 44 studies; 24,309 from hospitals, 541 from rehabilitation, and 24,586 from nursing homes. The number of participants per study ranged from 49 to 14,913, with a median participant number of 228 (25th percentile 143; 75th percentile 438). Forty-two studies, consisting of 48 datasets, included participants with a mean age of 75 years (SD 10; range 54–106 years). Two studies did not report ages, but specified the population as adult or geriatrics. The majority of studies included participants with stroke (n = 19) [34,35,36, 38, 39, 42, 48, 51,52,53,54, 56,57,58,59,60, 62, 66, 68]. Fifteen studies included patients with diverse diagnoses: e.g., post-surgery, internal medicine, geriatrics, pneumonia, trauma, gastrointestinal tract disease, cardiovascular disease, respiratory tract disorders, fractures, musculoskeletal, neurologic and neurodegenerative diseases, and head and neck cancer [25, 27, 28, 33, 40, 43, 45, 47, 50, 55, 61, 63, 69,70,71]. One study included only participants with dementia [46] and nine studies did not specify the participants’ diagnosis [26, 37, 41, 44, 49, 64, 65, 67, 72].

Type of Screening or Assessment Method

The type and combination of screening and assessment methods used to determine the prevalence for OD varied. Nearly one-third (15/44) of the studies [25, 35, 38, 39, 46, 51, 55, 57, 61,62,63,64,65, 67, 68] used a screening tool alone to identify risk of OD and three studies [36, 56, 66] used clinical non-instrumental assessments to diagnose OD. Four studies used FEES and/or VFSS either in combination with a screening [42], a clinical non-instrumental assessment [48, 58, 59]. Five studies used a patient-reported outcome measure (PROM) alone [37, 43, 49, 50, 71], three studies used a PROM together with a screen [44, 47, 69], and one used two PROMs and a clinical swallowing assessment [33]. Eight studies determined OD through a chart review [34, 40, 45, 52,53,54, 60, 72] five studies used self-formulated questions to staff and/or patient [26,27,28, 41, 70].

Over half (26/44) of studies reported OD prevalence data using screening and clinical non-instrumental assessments methods or tools that were either designed by the authors for the purpose of the study or modified versions of published tools, thus lacking information on diagnostic performance and psychometric properties [26,27,28, 34, 36, 38,39,40,41,42, 44,45,46, 48, 52,53,54, 58,59,60, 62, 64,65,66, 70, 72]. Of the screening and clinical non-instrumental assessment tools used to determine OD with known diagnostic performance and psychometric properties, the Volume-Viscosity Swallow Test (V-VST) [74] was the most commonly used screen [47, 55, 61, 63, 64, 68] and the Eating Assessment Tool-10 (EAT-10) [75] was the most frequently used PROM [33, 37, 43, 47, 49, 50, 69, 71]. The Mann Assessment of Swallowing Ability (MASA) [76] was the only clinical swallowing assessment used with known psychometric characteristics (one study) [56]. Table 2 provides an overview of the screens and assessments used in all of the included studies.

Time of Screening or Assessment

The time of screening or assessment for OD prevalence was recorded in 24/44 of the included studies. OD prevalence estimates from the hospital setting (21/32) were either reported as time post-stroke (7/21) or time from admission (14/21). Time post-stroke ranged from hyperacute phase [77]; ≤ 24 h post-stroke (1/7) [66], to acute phase; 1–7 days (4/7) [35, 42, 53, 60], to early subacute phase; 7 days–3 months (2/7) [36, 39]. Moment of screening or assessment in the rehabilitation setting was reported as hours or days from admission in three studies [33, 58, 59] and one study [26] did not specify when the participants were screened or assessed in relation to onset of disease or illness. None of the studies from the nursing home setting [25,26,27,28, 37, 41, 44, 49, 67, 69, 70, 72] specified the moment of screening or assessment for OD prevalence (Table 2).

Meta-analyses

In accordance with the pre-defined criteria, twelve studies that included data collected from medical records, national databases, surveys, or registries [26, 34, 40, 43, 45, 50,51,52,53,54, 60, 72] were excluded from meta-analyses. Further, four studies that used nurses,’ patients,’ or caregivers’ responses to a single dichotomous question about the presence of swallowing difficulties as screen for OD prevalence [27, 28, 41, 70] were also not included in the meta-analysis. In addition, six studies were excluded due to the inability to compute proportional OD prevalence data results from the datasets [33, 44, 49, 64, 65, 69]. The remaining 22 studies, 17 from the hospital, 2 from rehabilitation and 3 from nursing home settings, were included in the meta-analysis. Studies used screening [25, 35, 38, 39, 46, 55, 57, 61,62,63, 67, 68], PROM [37, 71], clinical non-instrumental assessments [36, 56, 66], or a combination of methods (screen, clinical swallowing assessment, PROM, instrumental) [42, 47, 48, 58, 59]. One study provided OD prevalence data for both screening and patient self-report for the entire study population [47]. As screening was preferred over self-report data, only prevalence estimates based on screening were included in the meta-analysis. Table 2 provides an overview of prevalence estimates as retrieved from individual studies; data used for meta-analyses have been marked.

Hospital

Meta-analysis using OD prevalence data from 17 hospital studies [35, 36, 38, 39, 42, 46,47,48, 55,56,57, 61,62,63, 66, 68, 71] resulted in an overall pooled OD prevalence estimate of 36.5% (95% confidence interval [CI] 29.9 − 43.6) (Fig. 2). Between-group analysis was computed for type of assessment (screen versus clinical assessment), diagnosis and type of ward. Twelve studies used screening [35, 38, 39, 42, 46, 47, 55, 57, 61,62,63, 68] and four studies used clinical non-instrumental assessment [36, 48, 56, 66] resulting in pooled OD prevalence estimates of 35.6% (95% CI 27.6 − 44.5) and 41.8% (95% CI 27.4 − 57.7), respectively (Fig. 3). Eleven studies included stroke diagnosis [35, 36, 38, 39, 42, 48, 56, 57, 62, 66, 68] and five studies included mixed diagnosis [47, 55, 61, 63, 71] resulting in pooled OD prevalence estimates of 37.5% (95% CI 28.7–47.2) and 34.4% (95% CI 22.5–48.6), respectively (Fig. 4). A meta-analysis for type of ward in the hospital setting revealed estimated OD prevalence of 35.3% (95% CI 27.2–44.2) for general or non-specified wards (10/17) [35, 36, 38, 39, 42, 47, 48, 61, 68, 71], 29.1% (95% CI 18.5–42.6) for stroke wards (4/17) [56, 57, 62, 66], and 51.1% (95% CI 35.0–67.0) for geriatric wards 3/17) [46, 55, 63] (Fig. 5). None of the between-group differences were significant. This meta-analysis incorporates data from 17 studies, which yield a z-value of − 12.00171 and corresponding 2-tailed p-value < 0.001. The fail-safe N is 621. This means that we would need to locate and include 621 “null” studies in order for the combined 2-tailed p-value to exceed 0.050.

Fig. 2
figure 2

Random-effects forest plot for overall pooled OD prevalence estimate in the hospital setting

Full size image
Fig. 3
figure 3

Random-effects forest plot for OD prevalence in hospital setting; between-group screen and clinical non- instrumental assessment

Full size image
Fig. 4
figure 4

Random-effects forest plot for OD prevalence in hospital setting; between-group stroke and mixed diagnoses

Full size image
Fig. 5
figure 5

Random-effects forest plot for OD prevalence in hospital setting; between-group type of ward

Full size image

Rehabilitation

The two included rehabilitation studies [58, 59] used clinical non-instrumental assessments revealing an estimated overall pooled prevalence for OD of 42.5% (95% CI 35.8–49.5) (Fig. 6). This meta-analysis of the prevalence of OD in the rehabilitation setting used data from two studies only, thus a fail-safe N analysis for publication bias was not available.

Fig. 6
figure 6

Random-effects forest plot overall OD prevalence in rehabilitation setting

Full size image

Nursing Home

Three studies from nursing homes [25, 37, 67] revealed estimated overall pooled OD prevalence of 50.2% (95% CI 33.3–67.2) (Fig. 7). Two of the three studies [25, 67] used screenings and one used a PROM, resulting in an estimated pooled OD prevalence of 58.1% (95% CI 47.3–68.2) and 35.0% (95% CI 22.8–49.5), respectively (Fig. 8). Total between-group OD prevalence estimates were significant (p = 0.012). This meta-analysis incorporates data from 3 studies, which yield a z-value of − 1.11840 and corresponding 2-tailed p-value of 0.263. Since the combined result is not statistically significant, the fail-safe N (which addresses the concern that the observed significance may be spurious) is not relevant.

Fig. 7
figure 7

Random-effects forest plot overall OD prevalence nursing home setting

Full size image
Fig. 8
figure 8

Random-effects forest plot for OD prevalence in nursing home; between-group screening and patient-report outcome measure

Full size image

Discussion

Systematic Review Findings

This systematic review and meta-analysis were conducted to establish the estimated pooled prevalence for OD in adults in different healthcare settings. The majority of the 44 included studies represented hospital (32/44) and nursing home (12/44) settings. There were few studies identified from the rehabilitation setting (4/44) and none from palliative care, revealing a knowledge gap regarding OD prevalence in these settings. Studies in the hospital and rehabilitation settings dated from 1986 to 2020, whereas the nursing home studies were from 2013 to 2020, possibly reflecting an increased awareness and exploration of OD in the elderly and nursing home population in the past decade.

Estimations of OD prevalence are dependent on (a) the definition of OD used in studies; (b) the choice of measure (screen, clinical non-instrumental assessment, instrumental assessment, or patient-reported outcome measurement); (c) the diagnostic performance and psychometric properties (including validity, reliability, responsiveness) and feasibility of the chosen measure; and (d) time of assessment (e.g., during the acute or chronic phase of the underlying disease) [13, 20, 78]. In this systematic review, terminology and definitions of oropharyngeal dysphagia varied and were conflicting. The majority of studies used the general term “dysphagia” in reporting prevalence of swallowing disorders, some studies referred to OD but lacked defining the concept of OD, while many studies provided a broad, generic definition of what constituted OD (e.g., lower capacity to swallow, generally unsafe swallow). Other studies included a definition that comprised both oropharyngeal and esophageal phases of swallowing (e.g., difficulty moving a bolus from the mouth to the stomach) or included aspects outside of OD (e.g., sitting position, difficulty transferring of food to the mouth, appetite) in addition to OD-related aspects. A consensus on a universal definition of dysphagia in the reporting of OD prevalence would support a more accurate estimation of OD prevalence [13].

In addition to lacking a universal definition for dysphagia, this systematic review revealed inconsistencies in the literature regarding what constitutes a screening and clinical non-instrumental assessment tool when estimating the prevalence of OD [14, 79, 80]. Several studies in this systematic review reported OD prevalence data from “screening”, but used measures that included administering a variation of food and liquid volumes and consistencies. This differs from common definitions of screenings: screenings for OD are designed to ensure identification of persons at risk of dysphagia and determine the need for further assessment, whereas clinical non-instrumental assessments are designed to ascertain the presence, location, severity, and possible treatment of OD [13]. The level of diagnostic performance of screening tools and psychometric robustness of clinical non-instrumental assessment methods chosen to estimate OD prevalence in the included studies is of high importance. Several systematic reviews have scrutinized the diagnostic performance and psychometric properties of available screening and clinical non-instrumental assessments [14, 79,80,81,82,83,84] indicating frequent poor methodological quality and lack of sufficient details. Instrumental assessments (VFSS or FEES) would be preferable when determining prevalence, as they have been shown to identify dysphagia in 20–30% more patients than screening and clinical non-instrumental assessments [20]. However, instrumental assessments require specialized training and equipment, thus, due to feasibility (e.g., availability, ease of administration), screenings and clinical non-instrumental assessments are the natural first choice for estimating the prevalence of OD [13, 85]. The most frequently used PROM in this systematic review used for estimating prevalence of OD was the EAT-10 (8/44) and five of these studies used the EAT-10 in isolation to screen for OD. This PROM was developed to assess symptom severity, quality of life, and treatment efficacy to be used for patients with both oropharyngeal and esophageal dysphagia [75]. However, in 2017, Cordier and colleagues evaluated the EAT-10 using Rasch analysis, challenging its diagnostic performance and psychometric properties, and recommended that it was re-developed using the Rasch model [86]. These findings were supported by other authors [80, 87, 88]. Furthermore, some studies used measurements in populations other than for which they were developed, which may affect the reliability and validity of the instrument [13]. For example, in a study by Tanigör and colleagues [33], the MDADI, developed to assess quality of life for patients with head and neck cancer [89], was used to determine the prevalence of OD in populations with neurological, musculoskeletal, and rheumatic diseases. This systematic review highlights the need for clinicians and researchers to use screening and assessment tools with optimal diagnostic performance and psychometric properties that are tailored for the population of interest when screening or assessing for OD.

Meta-analysis Findings

The meta-analysis included half (22/44) of the studies included in this systematic review: 17 from hospitals, two from rehabilitation, and three from nursing home settings. Results revealed an overall estimated pooled OD prevalence of 36.5% (95% confidence interval [CI] 29.9 − 43.6) in the hospital setting. Three studies in the meta-analysis in the hospital setting showed relatively high [36, 63] and low [47] OD estimates as compared to the other studies. Baroni et al. studied OD prevalence in stroke patients, including those with previous stroke. This inclusion criterion was an exclusion criterion in several of the included studies reporting on OD prevalence in stroke. In addition, OD was determined through a clinical evaluation using a broad definition of OD, if one or more “swallowing changes” were observed [36]. Mateos-Nozal et al. studied OD prevalence in the acute geriatric population with inclusion criteria ≥ 80 years. OD was determined if the V-VST revealed “any sign of OD.” In contrast, Spronk and colleagues studied all general hospital admissions on several wards and used both the EAT-10 and V-VST to determine OD prevalence. Participants were judged positive for OD with an EAT-10 score of ≥ 2 and “if a in any category of viscosity or multiple categories of viscosity, the maximum bolus volume was not reached” on the V-VST [47]. This study chose to apply different values to define OD than described by the original validation studies for the EAT-10 [75] and V-VST [74]. In this study, the low OD prevalence results for the V-VST were used in the meta-analysis as it is considered as having better evidence base for determining the presence of OD than the EAT-10. Although each of these individual studies did not have an impact on the estimate of overall pooled OD prevalence in the hospital setting, variations in study design, sample population, and definitions of OD provide insight into the heterogeneity of the included studies.

An overall estimated pooled OD prevalence of 42.5% (95% CI 35.8–49.5) was established from two studies in the rehabilitation setting. It might be expected that the estimated pooled OD prevalence in the hospital setting would be higher than the rehabilitation setting due to the acuteness of the underlying disease. The hospital setting included twice as many studies with stroke patients (11/17) than those with mixed diagnosis (5/17) and swallowing function tends to resolve in many patients within the first few days following stroke [90, 91]). Thus, patients with persisting OD often present with a more severe sequelae [92, 93] and require rehabilitation. In addition, the majority of studies from the hospital setting used screening tools to determine OD prevalence whereas the two studies from (neuro)rehabilitation unit/service, located within a hospital, utilized clinical non-instrumental methods and/or instrumental assessments to identify OD prevalence. Previous research has shown that use of screening methods results in lower prevalence than using clinical assessments [1, 20]. It is concerning that so few prevalence studies were identified from the rehabilitation setting. There is a need for future prevalence studies from the rehabilitation population.

As expected, nursing home settings revealed the highest OD prevalence; three studies showed an overall estimated pooled prevalence of 50.2% (95% CI 33.3–67.2) and an even higher estimated pooled OD prevalence of 58.1% for between-group difference for two of the three studies that used screening tools. Populations in the nursing home setting were older compared to other settings and suffered many medical conditions associated with OD (e.g., diseases of the circulatory and nervous systems, and cognitive disorders). Also, presbyphagia and sarcopenia may exacerbate OD resulting from comorbidities common to the aging population [6]. Results from the between-group analysis for type of ward also revealed a higher OD prevalence for the geriatric ward (51.1%), which was very similar to the nursing home setting. This systematic review reveals that overall pooled OD prevalence estimates are high for all healthcare settings, but highest in nursing homes.

The absence of studies in this systematic review included from the palliative setting raises concerns for this population. Patients receiving end of life care were excluded from six studies [33, 41, 62, 63, 65, 71] in this systematic review. This gives cause for concern regarding whether or not these patients are being screened or assessed for OD and how their OD is being managed. This systematic review has identified a need of further research in palliative healthcare setting.

Limitations

This systematic review is not without its limitations. The literature search only included two of the most relevant databases and English publications, thus, giving rise to potential publication bias. Furthermore, meta-analysis is subject to heterogeneity in study design, study population, and choice of outcome measures. Included studies differed in the definition of OD, definition of screening compared to clinical assessment, methodological study quality, and diagnostic performance and psychometric properties of outcome measures were used to determine OD prevalence. The larger number of studies included mainly stroke populations and may, therefore, limit the generalizability of calculated prevalence estimates. Consequently, although measures were taken to reduce heterogeneity for studies included in the meta-analysis, caution should be used when interpreting the results.

Conclusion

This systematic study reviewed 44 articles reporting on the prevalence of OD in different healthcare settings (hospital, rehabilitation, and nursing home). Most studies were conducted in hospital and nursing home settings, few studies in rehabilitation, and no studies were identified that reported on palliative care facilities. Future prevalence studies should provide data especially for patients in rehabilitation and palliative care. Overall, pooled prevalence estimate for OD determined by meta-analysis was high for all healthcare settings. Results revealed an overall estimated pooled OD prevalence of 36.5 (95% CI 28.8 − 44.9) for the hospital setting, 42.5% (95% CI 39.9–53.4) for the rehabilitation setting, and 50.2% (95% CI 33.3–67.2) for the nursing home setting. These high OD prevalence estimates across healthcare settings indicate that there is a large number of people at risk for malnutrition, dehydration, aspiration pneumonia, and ultimately a reduced quality of life. These findings indicate that treatment pathways including early assessment and diagnosis of OD should be a priority for healthcare professionals working in different healthcare settings with populations at risk for OD. In addition, this systematic review emphasizes a need for consensus in OD-related terminology and use of a clear operational definition when reporting OD prevalence. Further, when choosing screening and assessment tools to identify and assess OD, clinicians and researchers should take the target population into account for which a measure has been developed and validated, as well as only select screening and assessment tools with optimal diagnostic performance and psychometric properties.