Talk to the Veterans Crisis Line now
U.S. flag
An official website of the United States government

VA Health Systems Research

Go to the VA ORD website
Go to the QUERI website

RRP 07-313 – HSR Study

RRP 07-313
Ability to Cough to Predict Dysphagia in Rehabilitation Patients
Carol Smith Hammond, PhD
Durham VA Medical Center, Durham, NC
Durham, NC
Funding Period: October 2007 - October 2010
Oral pharyngeal dysphagia - a common complication in patients typically seen for rehabilitation - may affect the success of recovery. The ability to identify Veterans at-risk for oral pharyngeal dysphagia is important to preventing associated complications including aspiration pneumonia, malnutrition and hospital re-admission. Current clinical screening tests for dysphagia have poor sensitivity and specificity; however recent evidence shows objective characteristics of voluntary cough (sound pressure level and aerodynamics) are highly predictive of aspiration in patients with acute stroke. The validity of screening tests in other patient groups at-risk for oral pharyngeal dysphagia is uncertain. The goal of this study was to determine the ability of objective airflow measures of voluntary cough to predict dysphagia and aspiration in other, non-stroke patient groups at-risk for swallowing problems. A secondary goal is to determine the validity of commonly used clinical signs associated with risk for dysphagia and aspiration to detect these problems. The gold standard will be the videofluoroscopic (VSE) or fiberoptic endoscopic (FEES) evaluation of swallow to determine presence or absence of dysphagia. Clinical signs will be assessed at bedside by an experienced speech language pathologist. Objective measures of voluntary cough characteristics were assessed at bedside with the PERCI SARS system (Microtronics, Chapel Hill, NC). Test performance will be measured by sensitivity, specificity, predictive value, and ROC curve. At-risk patient groups include patients referred from the Medicine, Neurology and Surgical services. The VSE/FEE, clinical assessment and objective cough measurement will be done independently by individuals who will be blinded to the findings on the other tests.

1: Assess if there are differences in objective measures of voluntary cough. We hypothesized objective measures of voluntary cough, i.e. airflow characteristics and sound pressure level, would improve the clinician's ability to predict oral pharyngeal dysphagia, as observed on instrumental procedure of either videofluoroscopic (VSE) or fiberoptic endoscopic (FEES) evaluation, compared to findings on subjective clinical/bedside examination alone; and that the addition of these assessments to the clinical evaluation would improve the accuracy of identification of patients in need of management of swallowing problems by a speech language pathologist for evaluation and treatment. Further, we explored whether the characteristics of phonation of /i/ would also be a useful addition to the clinical bedside evaluation. A comparison was made between age matched control subjects and patients with and without oral pharyngeal dysphagia for voluntary cough and phonation measures.
2: Assess if one or multiple measures of airflow and sound pressure level of voluntary cough and/or phonation characteristics may be useful to predict oral pharyngeal dysphagia and if the prediction potential varies by the chief problem for which patients are admitted, i.e. medical, surgical, and neurological services.
3: Determine the utility of common signs of dysphagia and/or aspiration, assessed during the clinical evaluation of swallow, to predict oral pharyngeal dysphagia on instrumental evaluation (i.e. VSE or FEES).

Study Design: This study was a minimal risk pilot study that used a prospective observational design and was conducted at the Durham Veteran's Affairs Medical Center (DVAMC), Durham, NC. The study was approved by the DVAMC Research and Development and Institutional Review Boards. A total of 78 patients who were referred to the Speech Pathology Services for evaluation of oral pharyngeal dysphagia (OPD), due to concern regarding swallow function, were recruited and studied. Each patient was examined for OPD by instrumental evaluation, using either videofluoroscopic (VSE) or endoscopic (FEES) evaluation. Patients were also assessed for respiratory rate, airflow and sound pressure level measures of voluntary cough and phonation measures; clinical signs of swallow function; cognitive disorder. Their medical record was reviewed to obtain information regarding reason for referral (defined by diagnosis or presumed diagnosis). To assess if voluntary cough and phonation measures could be used to differentiate referral patients from a normal population (Objective 1), these measures were compared to a group of 39 controls, recruited in a previous study, from the GEROFIT program and from friends of the Speech Pathology Service healthcare professionals. Inclusion criteria for subjects referred for swallowing evaluation: i) Over 18 years of age; ii) Consult received for the patient by the Speech Pathology Service for evaluation of oral pharyngeal swallowing function; iii) The patient was alert, oriented and able to sign the informed consent or, if disoriented, the consent was signed by the individual designated in the patient's health care power of attorney document; and iv) The patient was able to spontaneously produce on command or imitate a voluntary cough and the phonation of 'eeee.' Exclusion critera: i) Refusal to consent to participate by patient or health care power of attorney or health care power of atorney not available; and ii) Patient is unable to spontaneously produce or imitate a voluntary cough and phonation of 'eeee.' For control subjects, the inclusion criteria were: i) Over 18 years of age; ii) iii) Provided signed consent to participate; iv) No reported medical history of neurological or respiratory problems; and v)Results of VSE indicated normal oral pharyngeal swallow function. Methods and Procedures:The following procedures were performed, when possible, for all subjects: medical record review, bedside (clinical) evaluation, instrumental objective evaluation (VSE or FEES of oral pharyngeal swallow) and cognitive screening (COGNISTAT). All procedures were completed within 24 hours of the instrumental evaluation of swallow. The control group (n=39) underwent identical procedures.
A clinical swallow evaluation was conducted at the bedside (completed for 98.8% of subjects or (n=77/78) that included assessment of level of consciousness, strength of the facial/oral/pharyngeal/laryngeal musculature, presence/absence of gag, quality and strength of a voluntary cough (cough on command) and a throat clear, phonation of 'i', palatal movement, speech and language status, and respiratory status. Ice chip or water boluses were administered during the bedside evaluation and the subject's reaction, including cervical palpation during the swallow, was recorded (i.e., coughing/ choking, upper airway signs and absent swallow).
The evaluation of aerodynamic and sound pressure level of voluntary cough and phonation of /i/ was conducted at bedside using the PERCI-SARS (Microtronics BIT, Chapel Hill, NC). A soft and pliable rubber mask, i.e., Rothenberg mask (Glottal Enterprises), was placed over the subject's nose and mouth. The mask was connected to the PERCI SARS system and to a laptop computer which displayed and saved the airflow parameters of the voluntary cough. A microphone was attached to the mask which captured phonation and the sound pressure level of the cough. Initially, the subject was asked to relax and breathe normally to capture respiratory rate. The subject was then asked to 'cough as hard as you can' and later asked to 'say /i/ and hold it as long as you can.' The signals were saved and stored for later analysis.
Cognitive screening: the Neurobehavioral Cognitive Status Examination was performed and included screening the following cognitive domains: levels of consciousness and attention, ability to describe a picture and follow single and multiple-step commands, repeat sentences, perform calculations (four arithmetic problems), memory function, and reasoning including similarities (example: Why are a hat and coat alike?) and the ability to solve every day problems (Example question: What would you do if you came home and a broken pipe was flooding your kitchen?). When a COGNISTAT was not performed, physician and/or occupational therapy notations of cognitive function were used to determine cognitive status.
Videofluoroscopic (VSE)14 or Fiberoptic Endoscopic Evaluation (FEES)15 of swallow was performed for all subjects. For the VSE, participants were evaluated while sitting upright in a chair (Video-fluoro chair, Transomotion Medical, Inc), which was designed to be used with patients seen for oral pharyngeal swallowing evaluation in the Radiology clinic, DVAMC. Food and liquid were mixed with powdered barium and the following consistencies were offered: thin, Ensure (Abbot Nutrition, Abbot Laboratories ) and pre-thickened liquid, i.e., nectar and honey thick, (Hormel ). Boluses were administered in a range of bolus volumes (5 ml, subject or clinician administered cup drinks, then straw drinks); puree solids (applesauce and pudding) in teaspoon size boluses and a cookie were offered. For all bolus types, administration was terminated at the lowest volume where aspiration was observed. The FEES was performed at the bedside, and the food and liquids were colored with green food dye. A flexible scope was inserted through the nose into the pharyngeal area below the level of the soft palate to visualize the pharyngeal and laryngeal areas before and after the swallow. Due to a change in administrative policy at the DVAMC, the FEES was not available for most of the recruitment period; therefore, the majority of instrumental swallowing evaluations were by VSE.
Participants were categorized into a dysphagia or no-dysphagia group based on the results of the VSE or FEES. Swallow function was rated as 'dysphagia' based on observed abnormal function in the oral or pharyngeal stage of the swallow. Participants were rated as having 'risk for aspiration based on a Penetration Aspiration Scale (PAS) score of >5. Evaluations were independently performed and evaluation on any exam was blinded from the outcomes of other exams. Results of VSE/FEES evaluations were reached by consensus of two certified speech language pathologists experienced in management of patients with oral pharyngeal dysphagia. Patients were also categorized by aspiration severity and diet recommendations made by a SLP following VSE or FEES evaluations.
Statistical Analysis: Measurements: Primary continuous measures obtained from the airflow studies were respiratory rate and the voluntary cough measures of the following numerical variables: respiratory inspiratory phase duration (IPD), inspiration phase volume (IPV), compression phase duration (CPD), inspiration phase peak flow (IPPF), expulsive phase peak flow (EPPF), expulsive phase reaching time (EPRT or the time from the end of the glottal closure phase to expiration phase peak flow), volume acceleration (VA, VA=EPPF/EPRT) and sound pressure level (SPL). Primary measures of phonation were the following continuous variables: duration of phonation (all phonation measures of vowel /i/), fundamental frequency, jitter factor, jitter mean percent, shimmer, pitch range, minimum and maximum pitch. Another primary measure of interest was a dichotomous variable of dysphagia (yes vs. no) based upon a patient's FEES/VSE result. This variable was also used to define a categorical variable of patient type or group (control vs. patient with dysphagia versus patient with no dysphagia) after adding the control group in some analyses. Other measures of interest included demographic and clinical characteristics of subjects, medical service, cognitive function, severity of aspiration, and recommended diet for the patients.
Statistical plan: All numerical variables were summarized by mean standard error (SE) or median (range) and categorical variables were summarized by frequency (in %) or odds ratio. Statistical analyses were performed using a SAS 9.2 software (SAS, Cary, NC) package. P-values <0.05 were considered statistically significant. Specific statistical analyses are as follows.
Analyses of Association. Each of the numerical measures of voluntary cough and phonation was considered as a dependent variable and its association to the fixed effect of patient type (control versus patient with dysphagia versus patient with no dysphagia) was assessed using a fixed effect model, after adjusting for a subject's age, gender and race. Comparisons of means between groups were performed under the fixed effect model framework. Considering most of the measures of voluntary cough and phonation except for SPL were found to be right skewed and could be corrected using a log transformation, we used log transformed variables (except for SPL) in computation of the fixed effect models. However, we report their means in original units after taking the exponential of the log transformed means. Similar fixed effect models were also used in other analyses of association when the fixed effect of patient type was replaced by other fixed effects of interest, such as severity of aspiration, diet recommendation, etc. As exploratory analyses, we assessed associations of measures of voluntary cough and phonation to the fixed effects of dysphagia (yes vs. no), managing service (medical, surgical, and neurological services) and their interaction among referred patients. Comparisons of means between dysphagia and non dysphagia patients were performed for each type of service under the fixed effect model framework. Analysis of Association, Referred Patients Only: This analysis was performed with referred patients only. The dependent variable was the dichotomous measure of dysphagia and measures of voluntary cough and phonation were considered predictors of interest. The analysis was started with a simple logistical regression model, using one predictor at a time. Both adjusted and unadjusted methods were considered in the simple model, in which patient's demographics were added as controlling covariates in the adjusted method. Predictors showing significance at a p<0.1 level were further assessed in multiple logistical regression models. Three different types of multiple models were considered. Model 1 (M1) included significant measures of voluntary cough only as predictors; Model 2 (M2) used only significant measures of phonation as predictors; and Model 3 (M3) used all significant measures of voluntary cough and phonation as predictors. Diagnostic accuracy using AUC, sensitivity and specificity. Using each logistical model proposed previously, we calculated a predicted probability (ranging 0-100%) of dysphagia for each patient. A receiver operative characteristic (ROC) curve could thus be generated corresponding to the logistic model (or predictor[s]), using the dichotomous definition of dysphagia as the reference standard. The overall diagnostic accuracy using the logistical model (or predictor[s]) could be assessed using the area under the ROC curve (AUC). Further, for each ROC curve, we provided a cut-off of the predicted probability of dysphagia such that its corresponding specificity was closest to 80%; and estimated its sensitivity under the cut off value. It should be noted that in the unadjusted simple models, the predictor probability of dysphagia was one on one related to the only predictor mathematically, indicating that the cut-off of the predictor probability of dysphagia could be converted to the original value of its corresponding predictor (or the measure of cough/phonation). Other analyses. For all other non primary measures, we used non parametric Wilcoxon rank sum tests and Fisher's exact tests to compare medians and frequencies between groups if they were numerical and categorical respectively.

Data were available on 78 referred subjects and 23 normal control subjects. Referred subjects were categorized into those with (n=69) and without (n=9) OP dysphagia based on the results of VSE or FEES. Referred patient subjects and controls were similar in age (patients median age 72.5y ; range 35-89y; control 72.0y; 41- 81y) as were patients with dysphagia (73.0y; 35-89y) and patients without dysphagia (61.0y;39- 86y). There was no difference in race between any group and all subjects were male. The distribution of referred patients based upon their chief complaint that lead to admission to the DVAMC was as follows: Cervical surgery (n=4 ;5.1% of all referred), Esophageal (n= 5; 6.4%), Pulmonary (n=23; 29.5%), Head & Neck cancer (n=5;6.4%), Cardiac (n=10;12.8%), Neurological (n=16;20.5%),Other (n=15;19.2%). Referred patients in the study were on one of 3 services: Medicine (n=36;46.2%), Surgery (n=27; 34%), Neurology (n=15;19.2%). There was no difference between patients with dysphagia (n=48; 70%) and without dysphagia (n= 6; 67%) for the presence of 'overt' signs associated with dysphagia and aspiration following ice chip or water bolus i.e., reflexive cough after water or absent swallow or not managing secretions at bedside. All control subjects took water sips without demonstrating 'overt' clinical signs of dysphagia. Results of VSE or FEES indicated a significant difference in penetration aspiration scale (PAS) score (airway protection scale) for control subjects compared to all subjects referred to the Speech Pathology Service for evaluation of swallow function (p< 0.0001) and between patients in the dysphagia and no dysphagia groups ((p< 0.0001). All PAS scores were <4 in the control and no dysphagia groups (1= no material enters laryngeal area; 2= material entered laryngeal area above the true vocal folds, TVCs, with no residue after the swallow; 3= material enters laryngeal area above the TVCs and residue remained in laryngeal area after the swallow; 4= some material enters laryngeal area above the level of the true vocal folds and no residue remains in the laryngeal area after the swallow). Among 23 control subjects, 19 (82%) scored PAS of 1; 7/9 (or 77.8%) of the 'no dysphagia' patients also scored PAS of 1. Airway protection scores for those with dysphagia are as follows: PAS of 8 for 59% (n=46; Material passes glottis into the proximal trachea absent spontaneous reflexive cough i.e. silent aspiration); PAS of 7 for 19% (n=14; Material passes the glottis into the proximal trachea and patient responds with reflexive cough); 6 for 3% (n= 2; Material passes the glottis with no subglottic residue); PAS of 5 for 9% (n=7; Material contacts the TVCs with visible residue in laryngeal area after the swallow); and 1.4% (n=1) for PAS of 3 and 2. All 9 patients without dysphagia (100%) and 35 patients (57%) with dysphagia were recommended a regular diet which consisted of all liquid levels (thin and thick liquids) and some solid food consistency. Fourteen patients (20%) of the dysphagics received a recommendation of NPO, two patients (3%) no liquids and puree solid foods; thirty one patients (45%) thickened liquid with solid food and the remaining a regular diet. Twenty seven (36.5%) dysphagia patients died, with a median time to death after the swallowing evaluation of 60 days. Only one non dysphagia patient (11%) died which occurred 180 days after evaluation. Seventeen percent (n=4) of dysphagic subjects died within 1 week; 30% within one month; 74% within 3 months; 96% within one year compared to the 'no dysphagia' group in which 1 subject died at 6 months and one 13 months after the swallowing evaluation. Although more patients with dysphagia died during this study period, there was no significant difference between patient groups for mortality.
Analyses of Association and Prediction: Referred subjects without dysphagia were similar to control subjects on all cough and phonation measures. Controls subjects were different from referred patients with dysphagia in the following measures: respiratory rate (breaths/min; mean; 95%CI) control 14.59 (11.36, 18.74) vs. dysphagic patients19.88 (15.99, 24.72) p<0.0001; cough measures of inspiration phase (IP) peak flow (L/s) 0.88 (0.49, 1.57) vs. 0.57 (0.34, 0.96) p<0.02 and IP volume (ml) 0.39 (0.20, 0.77) vs. 0.20 (0.11, 0.38) p<.003, expulsive phase (EP) peak flow (L/s 5.08 (2.73, 9.43) vs. 2.05 (1.17, 3.60) p<0.0001, EP rise time (ms) 32.78 (22.88, 46.96) vs. 58.20 (41.33, 81.95) p<0.0001 and volume acceleration (VA) 154.82 (73.61, 325.64) vs. 35.26 (17.95, 69.26) p<0.0001; and for the phonation measures of duration of phonation (s) 9.23 (4.89, 17.41) vs. 3.93 (2.24, 6.89) p<0.0001, shimmer 1.23 (0.40, 3.75) vs. 0.40 (0.15, 1.05) p< 0.005 and jitter 2.79 (0.88, 8.85) vs. 1.16 (0.43, 3.13) p< 0.03. Simple logistic regression models, using one of the cough and phonation measures at a time as the predictor, showed risk of dysphagia was positively associated with respiratory rate with an odds ratio (OR) 1.22 (95%CI 1.08, 1.38; p< 0.002), and EP rise 1.08 (1.04, 1.13; p<0.0001); and negatively associated with cough measures or IP peak flow 0.56 (0.31, 1.00; p= 0.050); IP volume 0.32 (0.13, 0.78; p= 0.011); EP peak flow 0.85 (0.77, 0.94; p< 0.001) and VA 0.99 (0.98, 0.99; p<0.0001 ); and phonation measures of duration 0.77 (0.68, 0.87; p<0.0001) and shimmer 0.65 (0.44, 0.95; p<0.03) . Three multiple logistic regression models using 1) all significant voluntary cough measures from simple models as predictors (M1), 2) all significant phonation measures (M2), and 3) combination of significant voluntary cough and phonation measures (M3) were run and the results showed EP rise time and phonation duration were the two most significant predictors among the candidate variables. In particular, the OR's (95% CI's) of dysphagia in response to one unit increase of EP rise time were 1.08 (1.03, 1.13, p=0.003) and 1.11(1.02, 1.20, p= 0.013) in M1 and M3 respectively. The OR's (95% CI's) were 0.81(0.72, 0.91, p<0.0001) and 0.78 (0.64, 0.94; p= 0.011) in response to each unit increase of phonation duration using M2 and M3 respectively. Patients were grouped by diet recommendations based on observedOP swallow function on VSE or FEES: nothing by mouth (NPO) or no liquids n=17; thickened liquids with solid foods n=31; and unrestricted liquids and solid foods n= 30). There was a difference between those recommended NPO or no liquids (severe dysphagia) and a regular diet for the following voluntary cough measures: EP peak flow (p<.04), EP rise time (p< .05), VA (p< 0.007) and SPL (p< 0.05). There was no difference between patients grouped by diet recommendations for phonation measures. When aspiration of severity (variable aspiration within a consistency vs. aspiration of one consistency vs. aspiration of > one consistency) was considered, it wasassociated with the following phonation measures: Jitter factor, Jitter mean % and Shimmer (p's<0.05) but no association to voluntary cough measures.
As part of exploratory analyses, we also compared the group means of measures of voluntary cough and phonation between the dysphagia and no dysphagia groups in each of three subtypes of patients stratified by managing clinical services. Given the small sample size of the sub-groups, most of analyses did not have reasonable statistical power.
Diagnostic Accuracy:
Simple models using significant predictors at p's<0.1 generated modest to outstanding overall accuracy with AUC's ranging from 60% to 83%; while the three multiple predictor models provided outstanding to excellent accuracy with AUC's being 88%, 83% and 94% for M1-M3 respectively. For each patient in our dataset, we calculated the probability of dysphagia using multiple logistic regression models involving the respiratory rate, IP peak flow, IP volume, EPpeak flow, EP rise time, VA, duration of phonation /i/, Jitter factor, and Shimmer. From the combined voluntary cough and phonation model (M3), a patient would be considered at risk of dysphagia (or no risk of dysphagia) if his probability of dysphagia >58.6% (or 58.6%). Using this definition, the sensitivity and specificity reached 93% and 78%, respectively.
We also evaluated diagnostic accuracy using clinical symptoms observed during the bedside/clinical evaluation for 'overt' clinical signs of dysphagia i.e. reflexive cough or upper airway secretions to predict dysphagia and aspiration. The AUC of clinical symptoms of reflexive cough after water or ice chip bolus was only 52%. While its sensitivity was modest at 71%, its specificity was poor at 33%, suggesting 67% of patients were false positives using this method.
In a scenario of applying a set of measures of cough and phonation used in combination with cough and phonation measures (M3) after pre-screening for cough symptoms, we investigated a total of 62 patients in our data who had observations available on all measures (predictors) used in the combined voluntary cough and phonation model as well as the measures of cough symptoms and dysphagia. 18 out of 62 patients (29%) did not have any cough symptoms. Among this group, 15 patients were false negative being diagnosed with dysphagia using FEES/VSE (gold standard), however, they were 100% correctly identified after using the measures of cough and phonation in the multiple predictor model of M3. The overall rate of correct identification was 100% among the 18 patients with no cough symptoms. For the remaining the 44 patients - those who had cough symptoms - the rate of correct identification was 86.4%.

We conclude that objective airflow measures of voluntary cough and phonation would be useful additions to the clinical/bedside evaluation to identify those patients in need of further evaluation of oral pharyngeal dysphagia. The combination of 'overt signs of aspiration of food or liquid into the upper airway,' i.e., reflexive cough, choking or upper airway signs, and objective measure of voluntary cough and phonation characteristics may be a preferred approach to identifying those patients that should be recommended 'nothing by mouth' pending further evaluation of oral pharyngeal swallow function by a speech language pathologist. Future studies should be focused on patient groups that are at high risk of aspiration, including those admitted for the following medical problems: cervical surgery, GI, pulmonary, head & neck, cardiac, neurological, post-extubation period for surgical patients, failure to thrive and those that are heavily sedated,

Impact/Importance to the VHA. The current policy of the Veteran's Health Administration is for all patients to be screened for oral/pharyngeal dysphagia on admission, or if a concern develops, at any time during continuation of care, to determine those patients who should be referred to the Speech Pathology Service for further evaluation. The results of this pilot project iindicate that objective measures of voluntary cough and phonation duration should be further investigated in prospective consecutively referred patients for swallowing problems, i.e., oral pharyngeal dysphagia. The goal would be to determine the accuracy of adding these objective phonation and airflow measures to the clinical/bedside evaluation of oral pharyngeal swallow.

External Links for this Project

Dimensions for VA

Dimensions for VA is a web-based tool available to VA staff that enables detailed searches of published research and research projects.

Learn more about Dimensions for VA.

VA staff not currently on the VA network can access Dimensions by registering for an account using their VA email address.
    Search Dimensions for this project


None at this time.

DRA: Aging, Older Veterans' Health and Care, Cancer, Lung Disorders
DRE: Diagnosis, Etiology, Prevention
Keywords: Nursing, Quality assessment, Screening
MeSH Terms: none

Questions about the HSR website? Email the Web Team

Any health information on this website is strictly for informational purposes and is not intended as medical advice. It should not be used to diagnose or treat any condition.