New Calculator Predicts 30-Day Risk of Hospital Readmission From Inpatient Rehabilitation
Researchers at Spaulding Rehabilitation Hospital have created prediction models that can pinpoint, by type of medical condition, which patients at inpatient rehabilitation facilities are at high risk of hospital readmission within 30 days.
Takeaways
-
In this study, nationwide data on 1,849,768 patients admitted to inpatient rehabilitation facilities with 14 types of medical conditions were used to develop logistic regression models to predict 30-day hospital readmission risk for various conditions
-
The C-statistics for the prediction models (0.65–0.70) did not meet the typical threshold for good discriminative ability, but they indicated clinical utility
-
For most of the 14 impairment groups, calibration plots showed the models were accurate up to a readmission risk of 50%, and internal validation of the models revealed little bias
-
The scoring systems derived will be converted into a public online calculator that should be instrumental in identifying patients at high risk of readmission and allow resources to be focused on patients at greatest need across impairment groups
Discharge from a hospital to post-acute care, including an inpatient rehabilitation facility (IRF), may be a risk factor for hospital readmission. Now that hospitals have incentives to reduce 30-day readmissions, IRFs have become an important target of quality improvement and cost savings.
Mass General Brigham researchers have created the first calculators to help clinicians predict which patients admitted to the IRF are at higher risk of hospital readmission based on their specific type of impairment. Tawnee L. Sparling, MD, a former fellow, Erika T. Yih, MD, a pain medicine specialist, Jeffrey C. Schneider, MD, of the Department of Physical Medicine and Rehabilitation at Spaulding Rehabilitation Hospital, and colleagues report in the Journal of the American Medical Directors Association.
Methods
Using the Uniform Data System for Medical Rehabilitation, the team identified 1,849,768 adults discharged from 956 IRFs in the U.S. from October 2015 through December 2019. They fell into 14 impairment categories.
The researchers developed logistic regression models to calculate risk-standardized 30-day hospital readmission rates for each impairment group.
Readmissions
169,849 patients admitted to IRFs (9.2%) were readmitted to a hospital within 30 days (average readmission rate across impairment groups, 5.6%–13.4%). A table in the article lists the significant predictors of readmission for each group.
Model Performance
The C-statistic for each impairment group's model ranged from 0.65 to 0.70, not meeting the threshold for good discriminative ability (typically 0.8). They were better than chance, though, indicating the models are clinically useful.
For most impairment groups, calibration plots showed the models were accurate up to a readmission rate of 50%, much higher than the average rates noted in this study. Internal validation of the models revealed little bias.
Risk Scoring System
The researchers derived a point system for determining 30-day readmission risk, then an acute readmission risk scoring system for each group.
That scoring system will be converted into an online calculator for stratifying readmission risk when patients come to the IRF. Higher-risk patients can be targeted to receive specific resources during the IRF admission that might not be provided otherwise.
Learn about the Rehabilitation Outcomes Center at Spaulding >