Background: Known demographic and clinical characteristics of stroke survivors that affect selection of the facility to which they are discharged after hospitalization for an acute stroke are, for the most part, not population based and therefore may be unrepresentative. We present an analysis of such characteristics using the Lehigh Valley stroke cohort which is population based. Method: We enrolled patients within 1 month of onset of their initial acute stroke who were hospitalized between 1987 and 1989 at one of the eight hospitals in the Lehigh Valley, and 662 patients were discharged alive. The facility to which they were discharged was known for 660. Data on age, sex, presence of five selected comorbidities (hypertension, myocardial infarction, cardiac arrhythmia, diabetes mellitus and transient ischemic attacks), length of hospitalization and neurologic deficits from the stroke were systematically collected on standardized forms. Polytomous logistic regression was used to determine the factors associated with not being discharged home. Relative risk (RR) associated with discharge to a nursing or rehabilitation facility for each independent predictor was calculated using as the referent, those who went home. Results: Older age was a statistically significant predictor of not being sent home (RR = 1.2 for nursing home placement), but gender and living with a spouse were not. Having a selected comorbidity did not increase the RR of not being sent home regardless of the type, nor did risk of not being sent home increase with more than one illness. Considering neurologic deficits from the stroke, patients with lower extremity weakness had a 2.6- and 3.5-fold risk of being sent to a nursing or a rehabilitation facility, respectively, compared to those without such weakness while for those with upper extremity weakness, the RR was 1.5 and 4.9, respectively. Language deficit imposed a RR of 3.1 and 2.3 of going to a nursing or rehabilitation facility, respectively. Right facial weakness also emerged as a significant risk factor for not being discharged home, perhaps because of its association with language deficit. The longer a patient was hospitalized after a stroke, the less likely it was that such a patient would go home (RR = 1.1/hospitalization day). Conclusion: Using population-based data on stroke survivors, our study showed the characteristics of patients who are less likely to be discharged home and, instead, are discharged to a nursing or a rehabilitation facility. Data like ours which were population based may be useful in discharge planning for stroke patients by policy makers and health care providers.