Prevalence of dysphagia and risk of pneumonia and mortality in acute stroke patientsa meta-analysis

A nurse holds a cup of water to a stroke patient's lips. What happens in the next few seconds — whether the patient swallows cleanly, coughs, chokes, or silently aspirates liquid into the airway — will shape that patient's odds of surviving the coming weeks. This is the bedside swallowing screen, and it exists because stroke can damage the brain's swallowing circuitry in an instant. The condition it screens for is post-stroke dysphagia, or PSD. According to a meta-analysis by Banda and colleagues, it affects forty-two percent of acute stroke patients — nearly one in two. That figure comes from pooling forty-two observational studies covering twenty-six thousand three hundred sixty-six patients, of whom more than eight thousand seven hundred were identified with dysphagia. The range across individual studies is startling — from as low as eight percent to as high as eighty percent — which is precisely why pooling was necessary. No single study could tell you where the truth lies. Banda and colleagues used a statistical technique called the Freeman-Tukey double-arcsine transformation to combine these proportions. In plain terms, that method stabilizes the math around very small or very large percentages before averaging them to ensure that a study reporting eight percent and a study reporting eighty percent don't distort the result. After back-transforming the pooled figure, they arrived at forty-two percent, with a ninety-five percent confidence interval running from thirty-seven to forty-eight percent. That's the best available estimate. Nearly half of all acute stroke patients cannot swallow normally. Now, swallowing difficulty alone is uncomfortable and disabling. But the real danger is what comes next. When swallowing is impaired, food and liquid don't always go where they're supposed to. They slip into the airway — a process called aspiration — and they carry bacteria directly into the lungs. The result is aspiration pneumonia, and it kills stroke patients who might otherwise have recovered. Banda and colleagues quantified this risk with an odds ratio — a measure of how much more likely an outcome is in people with PSD compared to those without. The pooled odds ratio for pneumonia was four point zero eight. For mortality, it was four point zero seven. Patients with post-stroke dysphagia were roughly four times as likely to develop pneumonia, and four times as likely to die. The confidence intervals ran from about two point one to seven point eight for pneumonia, and two point two to seven point six for mortality — wide enough to reflect real uncertainty, but the direction is unambiguous. What makes those two numbers striking is how nearly identical they are. The risk of pneumonia and the risk of death are almost exactly proportional to each other. That symmetry makes biological sense: for these patients, pneumonia isn't a complication that sometimes leads to death. It is a direct step in the causal chain. Dysphagia leads to aspiration, aspiration leads to pneumonia, and pneumonia kills. The numbers trace that chain with unusual clarity. So who is most likely to develop dysphagia in the first place? Banda and colleagues identified several prognostic factors — patient and disease characteristics that predict higher or lower risk of PSD. The most striking contrast is stroke type. Patients with haemorrhagic stroke — caused by bleeding in the brain — had a pooled odds ratio of one point five two for developing dysphagia. Patients with ischaemic stroke — caused by a blocked blood vessel — had an odds ratio of just zero point five four. Haemorrhagic stroke patients are about fifty percent more likely to have swallowing problems, while ischaemic stroke patients are nearly half as likely. The paper points to lesion location and severity as the likely explanation: haemorrhagic strokes can raise intracranial pressure and damage a wider network of swallowing-control structures, including the internal capsule, sensory cortex, the insula, and the brainstem. Beyond stroke type, prior stroke history was associated with an odds ratio of one point four zero — consistent with the idea that cumulative damage reduces the brain's functional reserve and its capacity to compensate. Stroke severity, measured by a standard clinical scale called the National Institutes of Health Stroke Scale, or NIHSS, carried an odds ratio of one point three eight. More severe strokes more often impair the neural machinery of swallowing. Female sex was associated with higher odds of PSD at one point two five, while male sex showed lower odds at zero point eight two — a sex difference the authors partly attribute to age at first stroke. Diabetes mellitus carried an odds ratio of one point two four, which the team links to diabetes-related nerve damage that increases stroke morbidity more broadly. What did not emerge as significant is almost as useful as what did. Hypertension, high cholesterol, atrial fibrillation, and the specific hemisphere affected — right versus left — showed no significant association with PSD risk. The pattern sharpens the clinical picture: it's haemorrhagic type, prior stroke, severity, sex, and diabetes that should raise the alert level when screening a new patient. There's one more layer to this story, and it concerns measurement itself. The variability in PSD prevalence across studies isn't just noise — it reflects real differences in how dysphagia is detected. Banda and colleagues ran moderator analyses to pull this apart. Studies using instrumental assessments — videofluoroscopy or fiber-optic endoscopic evaluation of swallowing — reported a pooled prevalence of seventy-five percent. Studies using clinical subjective tools reported about fifty percent. Studies using clinical objective bedside tests, the kind most commonly used in routine practice, reported about thirty-eight percent. The difference between seventy-five and thirty-eight percent isn't a statistical artifact. Instrumental methods look harder. They catch aspiration that doesn't produce a visible cough — sometimes called silent aspiration — which bedside screening misses entirely. Study quality showed the same pattern. High-quality studies in this meta-analysis reported a prevalence of forty-seven percent, compared with twenty-nine percent in moderate-quality studies. The implication is uncomfortable but important: the headline figure of forty-two percent is probably conservative for clinical settings that invest in rigorous assessment. The real burden may be closer to the higher end. There's also geographic variation worth noting — pooled prevalences ranged from thirty-seven percent in Asia to sixty-one percent in South America — though the authors are careful not to over-interpret these differences given the variation in assessment methods between regions. What do you do with all of this? Banda and colleagues are direct about the clinical takeaway: screening for post-stroke dysphagia in the acute phase should be systematic, should use validated instruments, and should be stratified by the risk factors this meta-analysis identified. That means haemorrhagic stroke patients, patients with prior stroke, patients with high severity scores, women, and patients with diabetes should be flagged for particularly careful assessment. Where resources allow, instrumental evaluation will find more cases than bedside screening alone. The study's limitations deserve honest acknowledgment. Heterogeneity across the included studies remained high even after the moderator analyses — this is a pooling of very different clinical settings, patient populations, and methods. The prognostic odds ratios came largely from unadjusted primary studies, meaning confounding variables could be inflating some of those associations. Funnel plot asymmetry indicated possible publication bias, which tends to mean the literature skews toward positive findings. The four-fold risk for pneumonia and mortality should be read as an estimate, not a law. But the heterogeneity itself carries a message. The wide spread of reported prevalences — from under ten percent to over eighty percent — is not just a measurement problem. It tells you that PSD assessment is not yet standardized across the field. Different hospitals, different countries, different clinicians are making very different decisions about how and whether to screen, and with which tools. That inconsistency has consequences. Patients who aren't identified don't get interventions. The interventions — dietary modification, swallowing therapy, feeding tube placement — exist precisely to interrupt the chain that runs from dysphagia to aspiration to pneumonia to death. The nurse with the cup of water is doing something essential. But a bedside test catches only part of the picture. This meta-analysis, by pulling together data from over twenty-six thousand patients across forty-two studies, makes the case that the stakes are high enough to demand something more rigorous — and more consistent — than current practice typically delivers. This lecture was created by ennepō. Go to https://ennepo.ai to Discover, Create and Follow the latest research in your field. Read when you can. Listen when you want to.