Importance:Despite providing an overlapping level of care, it is unknown why hospitalized older adults are transferred to long-term acute care hospitals (LTACs) vs less costly skilled nursing facilities (SNFs) for postacute care. Objective:To examine factors associated with variation in LTAC vs SNF transfer among hospitalized older adults. Design, Setting, and Participants:We conducted this retrospective observational cohort study of hospitalized older adults (≥65 years) transferred to an LTAC vs SNF during fiscal year 2012 using national 5% Medicare data. Main Outcomes and Measures:Predictors of LTAC transfer were assessed using a multilevel mixed-effects model adjusting for patient-, hospital-, and region-level factors. We estimated variation partition coefficients and adjusted hospital- and region-specific LTAC transfer rates using sequential models. Results:Among 65 525 hospitalized older adults (42 461 [64.8%] women; 39 908 [60.9%] ≥85 years) transferred to an LTAC or SNF, 3093 (4.7%) were transferred to an LTAC. We identified 29 patient-, 3 hospital-, and 5 region-level independent predictors. The strongest predictors of LTAC transfer were receiving a tracheostomy (adjusted odds ration [aOR], 23.8; 95% CI, 15.8-35.9) and being hospitalized in close proximity to an LTAC (0-2 vs >42 miles; aOR, 8.4, 95% CI, 6.1-11.5). After adjusting for case-mix, differences between patients explained 52.1% (95% CI, 47.7%-56.5%) of the variation in LTAC use. The remainder was attributable to hospital (15.0%; 95% CI, 12.3%-17.6%), and regional differences (32.9%; 95% CI, 27.6%-38.3%). Case-mix adjusted LTAC use was very high in the South (17%-37%) compared with the Pacific Northwest, North, and Northeast (<2.2%). From the full multilevel model, the median adjusted hospital LTAC transfer rate was 2.1% (10th-90th percentile, 0.24%-10.8%). Even within a region, adjusted hospital LTAC transfer rates varied substantially (intraclass correlation coefficient [ICC], 0.26; 95% CI, 0.23-0.30). Conclusions and Relevance:Although many patient-level factors were associated with LTAC use, half of the variation in LTAC vs SNF transfer is independent of patients' illness severity or clinical complexity, and is explained by where the patient was hospitalized and in what region, with far greater use in the South. Even among hospitals in regions with similar LTAC access, there was considerable variation in LTAC use. Given the higher expense associated with LTACs vs SNFs, greater attention is needed to define the optimal role of LTACs in the postacute care of older adults.