Abstract
<title>Abstract</title> <p>Background Post-stroke dysphagia is common and adds substantially to morbidity. In many tertiary centres the decision to start speech-language pathology assessment depends on the neurology team's reading of the patient's CT or MRI. We tested whether routine categorical imaging features actually predict the severity of swallowing dysfunction. We enrolled 187 consecutive adults with first-ever stroke at a tertiary stroke unit between October 2023 and February 2024. Lesion site, size, type, and phase were recorded from clinical neuroradiology reports by a radiologist blinded to swallowing scores. Each patient was assessed using a 57-item ICF-based dysphagia tool developed and content-validated by the present research team through a modified Delphi process. We analysed associations using chi-square testing with Cramer's V, one-way ANOVA, and multinomial logistic regression. Results No lesion variable was significantly associated with dysphagia severity (all p > 0.40), and lesion features together explained under 7% of severity variance (Nagelkerke R² = 0.067). Brainstem lesions showed a non-significant trend toward greater ingestion impairment (mean qualifier 3.0 ± 0.8 versus 2.5–2.7 in cortical sites; F = 1.89, p = 0.13). The Functional Oral Intake Scale could not stratify our patients because 96.3% were kept nil-by-mouth under routine intensive care protocols. The multidimensional ICF-based instrument retained complete scoring across the entire cohort. Conclusions Routine imaging features available at the bedside do not provide a useful triage signal for swallowing risk in acute and subacute stroke. Universal multidimensional functional screening of every stroke patient by the speech-language pathology team at admission, regardless of lesion profile, is the practical conclusion. Larger multi-centre prospective studies that incorporate instrumental swallowing evaluation are needed to refine the screening protocol for stroke care services in low- and middle-income settings.</p>