||Does timing of rehabilitation after stroke affect outcomes? The association between early rehabilitation and better functional outcomes has been found by some to disappear when controlling for function on admission to rehabilitation. Very early rehabilitation might be harmful via oxidative/metabolic stress to injured brain cells. Since timing of rehabilitation is potentially modifiable, it is important to discern how it affects outcomes for certain types of patients. In this exploratory data analysis, Clinical Practice Improvement methodology was applied using the Post-Stroke Rehabilitation Outcomes Project database. Data on 1168 stroke patients were collected using chart abstraction; scannable documentation forms captured details of rehabilitation interventions. The Comprehensive Severity Index was used to measure severity in the acute hospital and rehabilitation settings. Statistical models were developed based on clinical experience and review of prior literature and were designed to control for potential interactions and confounders. Ordinary least squares and logistic regression analyses were performed using SAS. After excluding patients treated outside the U.S. and patients who died, 1033 patients were analyzed. For patients entering rehabilitation with severe functional impairment, earlier acute hospital therapy was significantly associated (p<0.0001 to p=0.0316) with better outcomes in multiple functional domains (measured by the Functional Independence Measure or FIM). For patients with moderate impairment,earlier therapy was associated with better mobility. For all but the most and least impaired, more acute hospital therapy was associated with better functional outcomes in several domains. Discharge to a skilled nursing facility prior to rehabilitation admission was associated with lower functional scores and lower likelihood of improvement. Overall, these findings support the hypothesis that earlier rehabilitation is associated with better outcomes, after controlling for severity and admission function. Patients with either minimal or maximal impairment were less likely to show benefit. Lack of initial stroke severity scores and absence of pertinent acute hospital data were important limitations. Assessing patients at discharge, rather than at predetermined endpoints, introduced additional bias related to discharge patterns. Nonetheless, this study revealed new insights into the effects of delaying poststroke rehabilitation for particular types of patients.