Category: Health Services Research; Cross-Cutting
Objective : To create profiles of inpatient rehabilitation facilities (IRFs) based on the patient mix of diagnoses and describe IRFs by size, location, and ownership.
Design : Facilities were grouped based on k-means cluster analysis with patient diagnosis data from calendar year 2017.
Setting : Inpatient rehabilitation facilities in the U.S. using calendar year 2017 data from the Centers for Medicare and Medicaid Services (CMS).
Participants (or Animals, Specimens, Cadavers) :
We used publicly available data from CMS. The file lists 1,183 IRFs with data on the number of times Medicare beneficiaries who had certain medical conditions were treated in each facility. It includes data for approximately 500,000 patients discharged in calendar year 2017.
Interventions : Not applicable.
Main Outcome Measure(s) :
Profiles of IRFs with main diagnosis and other characteristics.
Results : IRFs were clustered into one of eight profiles based on the conditions most commonly treated in the facilities: 1) Hip/knee replacement, 2) Stroke and spinal cord injury, 3) Traumatic brain injury and traumatic spinal cord injury, 4) Nervous system disorder, 5) Non-traumatic brain disease or condition, 6) Non-traumatic brain disease and fewer stroke patients, 7) hip/femur fracture or other orthopedic conditions, and 8) Stroke and traumatic brain injury. Clusters were further characterized by size, geographic location, and ownership. Clusters show significant variability in IRFs in terms of conditions treated and other characteristics.
Conclusions : Understanding variation in the types of patients treated across IRFs will continue to be important as IRFs continue publicly reporting risk-adjusted quality measures. With the movement toward Alternative Payment Models, including bundled payments, understanding the differences in patient populations will also be important for understanding the potential impact of these policies on IRFs.