Objectives: Patients with acquired brain injuries (ABIs) often need tracheostomy because of dysphagia. However, many of them may recover over time and be eventually decannulated during post-acute rehabilitation. We developed the Decannulation Prediction Tool (DecaPreT) to identify, early in the post-acute course, patients with ABIs who can be safely decannulated. Design: Nonconcurrent cohort study. Setting and Participants: Patients with ABI, as well as with dysphagia and tracheostomy, were retrospectively selected from the database of a neurorehabilitation unit in Correggio, Reggio Emilia, Italy. Measures: Potential bivariate predictors of decannulation were screened from variables collected on admission during clinical examination, conducted by an expert speech therapist. Multivariable prediction was then obtained in 2 separate random subsamples to develop and validate the logistic regression model of the DecaPreT. Results: Of 463 patients with ABI (mean age 52.2 years) selected, 73.0% could be safely decannulated before discharge. After bivariate screening, multivariable predictors of decannulation were identified in the development subsample and confirmed in the validation subsample, each with its odds ratio and 95% confidence interval as follows: age tertile (1.77, 1.08-2.89; P =.024), no saliva aspiration (3.89, 1.73-8.64; P =.001), pathogenesis of ABI (trauma vs other causes vs stroke vs anoxia: 2.23, 1.41-3.54; P =.001), no vegetative status (8.47; 2.91-24.63; P <.001), and coughing score (voluntary and reflex vs voluntary vs reflex vs neither voluntary nor reflex cough: 2.62, 1.70-4.05; P <.001). In the validation subsample, the predicting equation obtained an area under the receiver operating characteristics curve of 0.836. (C) 2018 AMDA - The Society for Post-Acute and Long-Term Care Medicine.