Development and Internal Validation of D-DASH: A Prediction Model for Eating Difficulty at Discharge in Acute Heart Failure.

Abstract

Background: Eating difficulty is an important complication in patients with acute heart failure (AHF), but no AHF-specific tool has been established for the early prediction of eating difficulty. We aimed to develop and internally validate the Development of eating Difficulty Assessment Score in acute Heart failure (D-DASH) using variables available during early hospitalization.

Methods and results: This single-center retrospective observational study included 644 patients hospitalized for AHF who had no eating difficulty before admission. The outcome was eating difficulty at discharge, defined as a Food Intake LEVEL Scale ≤8. A multivariable logistic regression model was constructed using age, dementia, New York Heart Association class IV, albumin, and maximal tongue pressure. Eating difficulty at discharge occurred in 110 (17.1%) patients. Older age (odds ratio [OR] 1.04; 95% confidence interval [CI] 1.01-1.07), dementia (OR 2.66; 95% CI 1.44-4.85), lower albumin (OR 0.59; 95% CI 0.39-0.89), and lower maximal tongue pressure (OR 0.94; 95% CI 0.90-0.98) were independently associated with eating difficulty at discharge. The model showed moderate discrimination (area under the curve 0.749; 95% CI 0.700-0.797) with acceptable calibration. Bootstrap internal validation demonstrated an optimism-corrected C-index of 0.739 and a calibration slope of 0.940.

Conclusions: This study developed a predictive scoring system to identify patients with AHF at risk of eating difficulty, enabling timely intervention to mitigate this risk.