• Shirley L Shih, Ross Zafonte, David W Bates, Paul Gerrard, Richard Goldstein, Jacqueline Mix, Paulette Niewczyk, S Ryan Greysen, Lewis Kazis, Colleen M Ryan, and Jeffrey C Schneider.
• Department of Physical Medicine and Rehabilitation, Spaulding Rehabilitation Hospital, Harvard Medical School, Boston, MA; Department of Physical Medicine and Rehabilitation, Massachusetts General Hospital, Harvard Medical School, Boston, MA.
• J Am Med Dir Assoc. 2016 Oct 1; 17 (10): 921-6.

ObjectivesFunctional status is associated with patient outcomes, but is rarely included in hospital readmission risk models. The objective of this study was to determine whether functional status is a better predictor of 30-day acute care readmission than traditionally investigated variables including demographics and comorbidities.DesignRetrospective database analysis between 2002 and 2011.Setting1158 US inpatient rehabilitation facilities.Participants4,199,002 inpatient rehabilitation facility admissions comprising patients from 16 impairment groups within the Uniform Data System for Medical Rehabilitation database.MeasurementsLogistic regression models predicting 30-day readmission were developed based on age, gender, comorbidities (Elixhauser comorbidity index, Deyo-Charlson comorbidity index, and Medicare comorbidity tier system), and functional status [Functional Independence Measure (FIM)]. We hypothesized that (1) function-based models would outperform demographic- and comorbidity-based models and (2) the addition of demographic and comorbidity data would not significantly enhance function-based models. For each impairment group, Function Only Models were compared against Demographic-Comorbidity Models and Function Plus Models (Function-Demographic-Comorbidity Models). The primary outcome was 30-day readmission, and the primary measure of model performance was the c-statistic.ResultsAll-cause 30-day readmission rate from inpatient rehabilitation facilities to acute care hospitals was 9.87%. C-statistics for the Function Only Models were 0.64 to 0.70. For all 16 impairment groups, the Function Only Model demonstrated better c-statistics than the Demographic-Comorbidity Models (c-statistic difference: 0.03-0.12). The best-performing Function Plus Models exhibited negligible improvements in model performance compared to Function Only Models, with c-statistic improvements of only 0.01 to 0.05.ConclusionReadmissions are currently used as a marker of hospital performance, with recent financial penalties to hospitals for excessive readmissions. Function-based readmission models outperform models based only on demographics and comorbidities. Readmission risk models would benefit from the inclusion of functional status as a primary predictor.Copyright © 2016 AMDA – The Society for Post-Acute and Long-Term Care Medicine. Published by Elsevier Inc. All rights reserved.

20 keywords...

hide…