Objective: The Postural Assessment Scale for Stroke Patients (PASS) is used to assess static and dynamic balance of stroke patients. PASS has demonstrated good measurement pro perties for reliability and validity, but its predictive effect for ambulation in stroke patients has not been investigated. The aim of this study was to investigate the predictive value of PASS for ambulation in patients with stroke after inpa tient rehabilitation. Methods: In this retrospective study, a total of 341 stroke pa tients were recruited from a rehabilitation ward of a medical university hospital. Patients were assessed at baseline using PASS and observation of rolling ability, and divided into 2 groups at discharge: independently ambulatory (n = 246) and nonambulatory (n = 95). Receiver operating character istic curve and adjusted bivariate logistic regression was ap plied to analyse the predictive value of baseline PASS scores, variables of demographic data, and rolling ability at admis sion to inpatient rehabilitation. Results: For all stroke patients, mean admission to the reha bilitation ward was 34.40 days after stroke and mean length of hospitalization 18.12 days. The receiver operating charac teristic curve was obtained with a cutoff score of 3.5 points for static PASS, 8.5 points for dynamic PASS, and 12.5 points for total PASS, demonstrating the highest percentage of accurately predicted ability of independently walking at discharge. Adjusted bivariate logistic regression found roll ing ability, static PASS and dynamic PASS to be predictors for ambulation of stroke patients at discharge. Conclusion: Initial static PASS score, dynamic PASS score and rolling can be predictors for independent ambulation of stroke patients after a course of inpatient rehabilitation.