Development and Validation of a Clinical Risk Score for Post-extubation Dysphagia (CRISPED) in Critically Ill...

The purpose of this study was to develop and validate a clinical risk score (CRISPED) for predicting post-extubation dysphagia (PED) in critically ill children. We conducted a prospective cohort study in a pediatric intensive care unit of a tertiary hospital (2021-2025). Consecutive children aged 1 month to 15 years requiring invasive mechanical ventilation ≥ 24 h were eligible. PED was defined as Functional Oral Intake Scale < 6 at first post-extubation speech-language pathology (SLP) assessment. The candidate predictors were evaluated using LASSO logistic regression. The cohort was split into development (n = 300) and validation (n = 132) samples for model assessment. Overall, the PED incidence was 65% among 432 patients included in the study. Three independent predictors were retained: iatrogenic withdrawal syndrome, number of intubations, and intubation duration. The CRISPED score demonstrated strong discrimination in the development cohort (C-statistic 0.85, 95% CI 0.80-0.89; accuracy 75%; Brier score 0.14) with good calibration (Hosmer-Lemeshow p = 0.44). Bootstrap internal validation confirmed minimal optimism with corrected C-statistic 0.84. Performance remained robust in the validation cohort (C-statistic 0.86, 95% CI 0.79-0.92; accuracy 80%; Brier score 0.15). Risk stratification into low (≤ 13 points), intermediate (14-17), and high (≥ 18) categories showed progressive PED prevalence: 21.8%, 61.5%, and 87.5% (p < 0.001). For severe PED (tube dependent patients), discrimination remained good (C-statistic 0.80; accuracy 75.8%). The CRISPED score provides accurate risk stratification for PED using three readily available clinical variables. It enables early identification of high-risk patients for prioritized SLP interventions, potentially improving clinical outcomes. External multicenter validation is essential before widespread clinical implementation.