Slicing the PAI: Creation of a Model to Predict Inpatient Rehabilitation Discharge Destination After Acute Stroke Using IRF-PAI Data

Research Objectives: To develop a robust model using Inpatient Rehabilitation Facility Patient Assessment Instrument (IRF-PAI) data to predict home versus other setting (i.e., skilled nursing, subacute care, acute hospital) discharge after acute stroke

Design: Retrospective Chart Review

Setting: Urban Academic Inpatient Rehabilitation Center

Participants: Patients admitted to acute inpatient rehabilitation with stroke from 2017 to 2023

Interventions: Not applicable

Main Outcome Measures: Discharge destination

Results: We analyzed data from 1,143 patients. 721 (63%) were discharged home and 422 (37%) were discharged to other destinations. We compared these groups using a multivariable logistic regression model that identified several variables as strong indicators of home discharge. Patients with dysphagia (p = 0.035) and patients needing assistance with self-care pre-stroke (p = 0.039) were less likely to be discharged home compared to patients without these impairments. Those dependent with: sit to laying transfers (p = 0.007), bathing (p < 0.001), and/or lower body dressing (p = 0.017) were also less likely to go home versus patients independent in these areas. Finally, patients with impaired delayed recall on neuropsych measures were less likely to go home (p value < 0.0001) relative to patients without impairments. Conversely, patients who were currently married were more likely to go home (p = 0.021) than patients who were not married. Age, sex, and race were not significant indicators of discharge.

As a post-hoc analysis, we broke down each category of significant variables to understand if there was a threshold level of disability at admission that distinguished home discharge. At the extremes, discharge home rates were highest for more independent categories (i.e., independent, set-up, supervision) and lowest for most dependent categories (i.e., dependent, maximal assistance). Interestingly, partial or moderate assist level tended to be a tipping point for many self-care and mobility variables.

Conclusions: Discharge disposition could be estimated based on information available on admission to inpatient rehabilitation. Dysphagia, pre-morbid functional impairments, and delayed recall may be indicators a patient will not be discharged home. Being currently married may be an indicator of home discharge. Demographic variables like age, sex, and race, may not be as relevant for discharge disposition. Patients requiring moderate level of assistance in self-care and mobility categories are the least predictable. This has clinical implications as it influences treatment and discharge planning. This is especially true with the tipping point group, as we theorize that in-depth interdisciplinary collaboration during inpatient rehabilitation may potentially tip this group to discharge home, rather than to alternate settings.

Author(s) Disclosures: None